Docker: Ubuntu image should be manylinux1 compatible

Fixes all of the long lines, mainly.

.coveragerc

@@ -1,9 +1,18 @@
 # Coverage isn't really compatible with subprocesses so results are unreliable
 
+[paths]
+source =
+	src
+	*/site-packages
+
 [run]
-branch = True
-#concurrency = multiprocessing
-source = ocrmypdf/
+branch = true
+parallel = true
+source =
+	src/ocrmypdf
+	tests
+omit =
+	tests/spoof/*
 
 [report]
 exclude_lines =

.docker/Dockerfile

@@ -1,50 +1,80 @@
 # OCRmyPDF
 #
-FROM      ubuntu:18.04
+FROM ubuntu:19.04 as base
 
-RUN apt-get update && apt-get install -y --no-install-recommends \
-  ocrmypdf \
-  unpaper \
-  tesseract-ocr-eng \
-  tesseract-ocr-fra \
-  tesseract-ocr-deu \
-  tesseract-ocr-spa \
-  tesseract-ocr-por \
-  tesseract-ocr-chi-sim \
-  python3-venv \
-  python3-pip
+FROM base as builder
 
 ENV LANG=C.UTF-8
 
-RUN python3 -m venv --system-site-packages /appenv
+RUN apt-get update && apt-get install -y --no-install-recommends \
+  build-essential autoconf automake libtool \
+  libleptonica-dev \
+  zlib1g-dev \
+  ocrmypdf \
+  pngquant \
+  python3-pip \
+  python3-venv \
+  tesseract-ocr \
+  unpaper \
+  wget \
+  git
+
+
+# Compile and install jbig2
+# Needs libleptonica-dev, zlib1g-dev
+RUN \
+  mkdir jbig2 \
+  && wget -q https://github.com/agl/jbig2enc/archive/0.29.tar.gz -O - | \
+  tar xz -C jbig2 --strip-components=1 \
+  && cd jbig2 \
+  && ./autogen.sh && ./configure && make && make install \
+  && cd .. \
+  && rm -rf jbig2
+
+RUN python3 -m venv /appenv
+
+COPY . /app
+
+WORKDIR /app
 
-# This installs the latest binary wheel instead of the code in the current
-# folder. Installing from source will fail, apparently because cffi needs
-# build-essentials (gcc) to do a source installation 
-# (i.e. "pip install ."). It's unclear to me why this is the case.
 RUN . /appenv/bin/activate; \
   pip install --upgrade pip \
-  && pip install ocrmypdf[fitz]
+  && pip install .
 
-# Now copy the application in, mainly to get the test suite.
-# Do this now to make the best use of Docker cache.
-COPY . /application
-RUN . /appenv/bin/activate; \
-  pip install -r /application/test_requirements.txt
+FROM base
 
-# Remove the junk, including the source version of application since it was
-# already installed
-RUN rm -rf /tmp/* /var/tmp/* /root/* /application/ocrmypdf \
-  && apt-get autoremove -y \
-  && apt-get autoclean -y
+ENV LANG=C.UTF-8
 
-RUN useradd docker \
-  && mkdir /home/docker \
-  && chown docker:docker /home/docker
+RUN apt-get update && apt-get install -y --no-install-recommends \
+  ghostscript \
+  img2pdf \
+  liblept5 \
+  libsm6 libxext6 libxrender-dev \
+  zlib1g \
+  pngquant \
+  python3 \
+  python3-venv \
+  qpdf \
+  tesseract-ocr \
+  tesseract-ocr-chi-sim \
+  tesseract-ocr-deu \
+  tesseract-ocr-eng \
+  tesseract-ocr-fra \
+  tesseract-ocr-por \
+  tesseract-ocr-spa \
+  unpaper \
+  wget
 
-USER docker
-WORKDIR /home/docker
+# Copy
+COPY --from=builder /app/misc/webservice.py /app/
 
-# Must use array form of ENTRYPOINT
-# Non-array form does not append other arguments, because that is "intuitive"
-ENTRYPOINT ["/application/.docker/docker-wrapper.sh"]
+# Copy minimal project files to get the test suite.
+COPY --from=builder /app/setup.cfg /app/setup.py /app/README.md /app/
+COPY --from=builder /app/requirements /app/requirements
+COPY --from=builder /app/tests /app/tests
+COPY --from=builder /app/src /app/src
+
+COPY --from=builder /appenv /appenv
+COPY --from=builder /usr/local /usr/local
+
+ENTRYPOINT ["/appenv/bin/ocrmypdf"]

.docker/Dockerfile.polyglot

@@ -1,17 +0,0 @@
-# OCRmyPDF polyglot
-#
-FROM      jbarlow83/ocrmypdf:latest
-
-USER root
-
-# Update system and install our dependencies
-RUN apt-get update && apt-get install -y --no-install-recommends \
-  tesseract-ocr-all 
-
-RUN apt-get autoremove -y && apt-get clean -y
-
-USER docker
-
-# Must use array form of ENTRYPOINT
-# Non-array form does not append other arguments, because that is "intuitive"
-ENTRYPOINT ["/application/.docker/docker-wrapper.sh"]

.docker/alpine.dockerfile

@@ -0,0 +1,91 @@
+FROM alpine:3.9 as base
+
+FROM base as builder
+
+ENV LANG=C.UTF-8
+
+# Normally:
+# echo '@testing http://nl.alpinelinux.org/alpine/edge/testing' >> /etc/apk/repositories
+
+RUN \
+  echo -e '@testing http://nl.alpinelinux.org/alpine/edge/testing\n@community http://nl.alpinelinux.org/alpine/edge/community'\
+  >> /etc/apk/repositories \
+  # Add runtime dependencies
+  && apk add --update \
+  python3-dev \
+  py3-setuptools \
+  jbig2enc@testing \
+  ghostscript \
+  qpdf@community \
+  qpdf-dev@community \
+  tesseract-ocr \
+  unpaper \
+  pngquant \
+  libxml2-dev \
+  libxslt-dev \
+  zlib-dev \
+  libffi-dev \
+  leptonica-dev \
+  binutils \
+  && pip3 install --upgrade pip \
+  # Install pybind11 for pikepdf
+  && pip3 install pybind11 \
+  # Install flask for the webservice
+  && pip3 install flask \
+  # Add build dependencies
+  && apk add --virtual build-dependencies \
+  build-base \
+  git
+
+COPY . /app
+
+WORKDIR /app
+
+RUN pip3 install .
+
+FROM base
+
+ENV LANG=C.UTF-8
+
+# Normally:
+# echo '@testing http://nl.alpinelinux.org/alpine/edge/testing' >> /etc/apk/repositories
+
+RUN \
+  echo -e '@testing http://nl.alpinelinux.org/alpine/edge/testing\n@community http://nl.alpinelinux.org/alpine/edge/community'\
+  >> /etc/apk/repositories \
+  # Add runtime dependencies
+  && apk add --update \
+  python3 \
+  jbig2enc@testing \
+  ghostscript \
+  qpdf@community \
+  qpdf-dev@community \
+  tesseract-ocr \
+  tesseract-ocr-data-deu \
+  tesseract-ocr-data-chi_sim \
+  unpaper \
+  pngquant \
+  libxml2 \
+  libxslt \
+  zlib \
+  libffi \
+  leptonica-dev \
+  binutils \
+  && mkdir /app
+
+WORKDIR /app
+
+# Copy build artifacts (python site-packages)
+COPY --from=builder /usr/lib/python3.6/site-packages /usr/lib/python3.6/site-packages
+COPY --from=builder /usr/bin/ocrmypdf /usr/bin/dumppdf.py /usr/bin/latin2ascii.py /usr/bin/pdf2txt.py /usr/bin/img2pdf /usr/bin/chardetect /usr/bin/
+
+# Copy
+COPY --from=builder /app/misc/webservice.py /app/
+
+# Copy minimal project files to get the test suite.
+COPY --from=builder /app/setup.cfg /app/setup.py /app/README.md /app/
+COPY --from=builder /app/requirements /app/requirements
+COPY --from=builder /app/tests /app/tests
+COPY --from=builder /app/src /app/src
+
+ENTRYPOINT ["/usr/bin/ocrmypdf"]

.docker/docker-wrapper.sh

@@ -1,5 +0,0 @@
-#!/bin/bash
-
-. /appenv/bin/activate
-cd /home/docker
-exec ocrmypdf "$@"

.dockerignore

@@ -5,7 +5,7 @@
 *.sublime*
 **/*.pyc
 .*/
-!.docker/
+!.git/
 .ruffus_history.sqlite
 bin/
 build/

.github/ISSUE_TEMPLATE

@@ -1,12 +0,0 @@
-Please include the command line and a test file with your issue report.
-
-If possible, please use a test file that we can include in future test cases (no personal information, no copyrighted material).
-
-If you wish to encrypt a test file so that only the maintainer of OCRmyPDF can view it, you may use:
-
-```bash
-
-gpg --recv-keys 4434eb74c4a35f7f --keyserver pgp.mit.edu
-gpg --output test.pdf.gpg --encrypt --receipient 4434eb74c4a35f7f test.pdf
-
-```

.github/issue_template.md

@@ -0,0 +1,33 @@
+**Describe the issue**
+A clear and concise description of what the issue is.
+
+**To Reproduce**
+What command line were you trying to run?
+
+```bash
+ocrmypdf  ...arguments... input.pdf output.pdf
+```
+
+**Example file**
+Please include an example *input* PDF (or image). The input file is more helpful.
+
+Please check any or all that apply about the test file:
+
+- [ ] This is the input file
+- [ ] The file contains no personal or confidential information
+- [ ] I am the copyright holder for this file
+- [ ] I permit this file to be included in the OCRmyPDF test suite under the CC-BY-SA 4.0 license
+- [ ] I am not the copyright holder, but this file is available under a free software license
+
+Files that are not free for inclusion in this project are quite welcome, but we like to collect free files for our test suite when possible. Please do *not* submit files with confidential information. At your option you may encrypt files for OCRmyPDF's author only.
+
+**Expected behavior**
+A clear and concise description of what you expected to happen. Include screenshots if applicable.
+
+**System:**
+
+- OS: [e.g. Linux, macOS]
+- OCRmyPDF Version: [e.g. v7.4.0]
+
+**Additional context**
+Add any other context about the problem here.

.gitignore

@@ -1,47 +1,46 @@
 # Development environment
+.bash_history
+.pylintrc
+.pytest_cache/
+.ruffus_history.sqlite
+.venv/
+.venv*/
 *.pyc
 *.sublime-*
-venv*/
-.venv/
-pyvenv.cfg
-tasks.py
-.bash_history
-.ruffus_history.sqlite
-.idea/
-.pytest_cache/
-.pylintrc
+*.DS_Store
 
 # Package building
-*.egg-info/
-.cache/
 .eggs/
+*.egg-info/
 build/
 dist/
 wheelhouse/
+pip-wheel-metadata/
 
 # Automatically generated files
-ocrmypdf/lib/_*.py
-ocrmypdf/version.py
 docs/_build/
 docs/_static/
 docs/_templates/
 docs/Makefile
+ocrmypdf/lib/_*.py
 
 # Code coverage
-.coverage
+.coverage*
 htmlcov/
 
 # Testing
-log/
+.ipynb_checkpoints/
+.vscode/
+*.ipynb
+*.profile
 /*.pdf
 /*.qdf
-*.ipynb
-.ipynb_checkpoints/
+/*.png
+/scratch.py
+IDEAS
+log/
 tests/output/
 tests/resources/private/
 tmp/
-pdfbox-app*.jar
-.vscode/
-IDEAS
-_Dockerfile.local
-scratch.py
+/debug_tests.py
+private/

.pre-commit-config.yaml

@@ -0,0 +1,6 @@
+repos:
+-   repo: https://github.com/ambv/black
+    rev: stable
+    hooks:
+    - id: black
+      language_version: python3.7

.readthedocs.yml

@@ -0,0 +1,10 @@
+build:
+  image: latest
+
+python:
+  version: 3.6
+
+formats:
+  - pdf
+
+requirements_file: requirements/main.txt

.travis.yml

@@ -1,53 +1,98 @@
-dist: trusty
-language: python
 cache:
   pip: true
   directories:
   - $HOME/Library/Caches/Homebrew
 
-addons:
-  apt:
-    update: true
-    sources:
-    - sourceline: 'ppa:alex-p/tesseract-ocr'
-    - sourceline: 'ppa:heyarje/libav-11'
-    - sourceline: 'ppa:vshn/ghostscript'
-    packages:
-    - ghostscript
-    - libavcodec56
-    - libavformat56
-    - libavutil54
-    - libexempi3
-    - libffi-dev
-    - pngquant
-    - poppler-utils
-    - qpdf
-    - tesseract-ocr
-    - tesseract-ocr-deu
-    - tesseract-ocr-eng
-    - tesseract-ocr-fra
-
-env:
-  global:
-    - secure: "hsf6MT+n2x3OiDM2fQyJZdV0/PWYmv81LdVqC6cfnHBE/8N3DloJRqQ7WfO14TxhiK9PEC7MpyCj0lSabUHEO7gSH6Vks6I1asoSkt8S9/bSMlhT4hei+pwVpeGEiU5xHVATNjY+D919VC3IFvc3XmjT74h/2SLhaZ+jhEmDggM="  # HOMEBREW_OCRMYPDF_TOKEN
-
 matrix:
   include:
     - os: linux
-      sudo: required
-      language: python
-      python: "3.5"
-    - os: linux
+      dist: trusty
       sudo: required
       language: python
       python: "3.6"
+      env:
+        - DIST=trusty
+      addons: &trusty_apt
+        apt:
+          update: true
+          sources:
+          - sourceline: 'ppa:alex-p/tesseract-ocr'
+          - sourceline: 'ppa:heyarje/libav-11'
+          - sourceline: 'ppa:vshn/ghostscript'
+          packages:
+          - ghostscript
+          - libavcodec56
+          - libavformat56
+          - libavutil54
+          - libexempi3
+          - libffi-dev
+          - pngquant
+          - poppler-utils
+          - qpdf
+          - tesseract-ocr
+          - tesseract-ocr-deu
+          - tesseract-ocr-eng
+          - tesseract-ocr-fra
     - os: linux
+      dist: xenial
       sudo: required
       language: python
-      python: "3.7-dev"
+      python: "3.7"
+      env:
+        - DIST=xenial
+      addons:
+        apt:
+          update: true
+          sources:
+          - sourceline: 'ppa:alex-p/tesseract-ocr'
+          packages:
+          - ghostscript
+          - libexempi3
+          - libffi-dev
+          - pngquant
+          - poppler-utils
+          - qpdf
+          - tesseract-ocr
+          - tesseract-ocr-deu
+          - tesseract-ocr-eng
+          - tesseract-ocr-fra
+          - unpaper
     - os: osx
-      osx_image: xcode8
+      osx_image: xcode9.2
       language: generic
+      addons:
+        homebrew:
+          update: true
+          packages:
+            - exempi
+            - ghostscript
+            - jbig2enc
+            - leptonica
+            - openjpeg
+            - pngquant
+            - python
+            - qpdf
+            - tesseract
+            - unpaper
+    - os: osx
+      osx_image: xcode9.2
+      language: generic
+      env:
+        - ADD_PDFMINER=1
+      addons:
+        homebrew:
+          update: true
+          packages:
+            - exempi
+            - ghostscript
+            - jbig2enc
+            - leptonica
+            - openjpeg
+            - pngquant
+            - python
+            - qpdf
+            - tesseract
+            - unpaper
 
 before_cache:
 - rm -f $HOME/.cache/pip/log/debug.log
@@ -57,11 +102,12 @@ before_install: |
   if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then
     pip3 install --upgrade pip
     pip3 install --upgrade wheel
-    mkdir -p packages
-    wget -q 'https://www.dropbox.com/s/vaq0kbwi6e6au80/unpaper_6.1-1.deb?raw=1' -O packages/unpaper_6.1-1.deb
-    sudo dpkg -i packages/unpaper_6.1-1.deb
+    if [[ "$DIST" == "trusty" ]]; then
+      mkdir -p packages
+      wget -q 'https://www.dropbox.com/s/vaq0kbwi6e6au80/unpaper_6.1-1.deb?raw=1' -O packages/unpaper_6.1-1.deb
+      sudo dpkg -i packages/unpaper_6.1-1.deb
+    fi
   elif [[ "$TRAVIS_OS_NAME" == "osx" ]]; then
-    brew update && brew bundle --file=.travis/Brewfile
     pip3 install --upgrade pip
     pip3 install wheel
   fi
@@ -69,9 +115,13 @@ before_install: |
 install:
 - export PATH=$PWD/bin:$PATH
 - pip3 install pycparser  # py3.7 workaround for https://github.com/eliben/pycparser/issues/251
-- pip3 install -r requirements.txt
+- pip3 install -r requirements/main.txt
 - pip3 install --no-deps .
-- pip3 install -r test_requirements.txt
+- |
+  if [[ "$ADD_PDFMINER" == "1" ]]; then
+    pip3 install --no-deps .[pdfminer]
+  fi
+- pip3 install -r requirements/test.txt
 
 script:
 - tesseract --version
@@ -94,18 +144,3 @@ deploy:
     tags: true
     condition: $TRAVIS_PYTHON_VERSION == "3.6" && $TRAVIS_OS_NAME == "linux"
   skip_upload_docs: true
-
-  # null deploy for osx
-  # we really just want to run after_deploy *after* pypi upload is done, but
-  # after_deploy on runs if a given box deployed
-- provider: script
-  script: /usr/bin/true
-  on:
-    branch: master
-    tags: true
-    condition: $TRAVIS_OS_NAME == "osx"
-
-after_deploy: |
-  if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then
-    bash .travis/osx_brew.sh
-  fi

.travis/Brewfile

@@ -1,14 +0,0 @@
-brew 'python'
-
-brew 'exempi'
-brew 'ghostscript'
-brew 'jbig2dec'
-brew 'jbig2enc'
-brew 'leptonica'
-brew 'libffi'
-brew 'libtiff'
-brew 'openjpeg'
-brew 'pngquant'
-brew 'qpdf'
-brew 'tesseract'
-brew 'unpaper'

.travis/autobrew.py

@@ -1,93 +0,0 @@
-#!/usr/bin/env python3
-# © 2017-18 James R. Barlow: github.com/jbarlow83
-
-from string import Template
-from subprocess import run, PIPE
-import re
-
-recipe_template = Template("""
-class Ocrmypdf < Formula
-  include Language::Python::Virtualenv
-
-  desc "Adds an OCR text layer to scanned PDF files"
-  homepage "https://github.com/jbarlow83/OCRmyPDF"
-  ${ocrmypdf_url}
-  ${ocrmypdf_sha256}
-
-  depends_on "pkg-config" => :build
-  depends_on "mupdf-tools" => :build  # statically links libmupdf.a
-  depends_on "freetype"
-  depends_on "ghostscript"
-  depends_on "jpeg"
-  depends_on "libpng"
-  depends_on "python"
-  depends_on "qpdf"
-  depends_on "tesseract"
-  depends_on "unpaper"
-
-${resources}
-  def install
-    venv = virtualenv_create(libexec, "python3")
-
-    resource("Pillow").stage do
-      inreplace "setup.py" do |s|
-        sdkprefix = MacOS::CLT.installed? ? "" : MacOS.sdk_path
-        s.gsub! "openjpeg.h", "probably_not_a_header_called_this_eh.h"
-        s.gsub! "ZLIB_ROOT = None", "ZLIB_ROOT = ('#{sdkprefix}/usr/lib', '#{sdkprefix}/usr/include')"
-        s.gsub! "JPEG_ROOT = None", "JPEG_ROOT = ('#{Formula["jpeg"].opt_prefix}/lib', '#{Formula["jpeg"].opt_prefix}/include')"
-        s.gsub! "FREETYPE_ROOT = None", "FREETYPE_ROOT = ('#{Formula["freetype"].opt_prefix}/lib', '#{Formula["freetype"].opt_prefix}/include')"
-      end
-
-      # avoid triggering "helpful" distutils code that doesn't recognize Xcode 7 .tbd stubs
-      ENV.append "CFLAGS", "-I#{MacOS.sdk_path}/System/Library/Frameworks/Tk.framework/Versions/8.5/Headers" unless MacOS::CLT.installed?
-      venv.pip_install Pathname.pwd
-    end
-
-    res = resources.map(&:name).to_set - ["Pillow"]
-
-    res.each do |r|
-      venv.pip_install resource(r)
-    end
-
-    venv.pip_install_and_link buildpath
-  end
-
-  test do
-    # Since we use Python 3, we require a UTF-8 locale
-    ENV["LC_ALL"] = "en_US.UTF-8"
-
-    system "#{bin}/ocrmypdf", "-f", "-q", "--deskew",
-                              test_fixtures("test.pdf"), "ocr.pdf"
-    assert_predicate testpath/"ocr.pdf", :exist?
-  end
-end
-""")
-
-def main():
-    p = run(['poet', '--single', 'ocrmypdf'],
-            encoding='utf-8', stdout=PIPE, check=True)
-
-    ocrmypdf_lines = p.stdout.splitlines()
-    ocrmypdf_url = ocrmypdf_lines[1].strip()
-    ocrmypdf_sha256 = ocrmypdf_lines[2].strip()
-
-    ocrmypdf_version = re.search(
-        r'ocrmypdf-(.+)\.tar.*', ocrmypdf_url).group(1)
-    print(f"Autobrewing {ocrmypdf_version}")
-
-    p = run(['poet', '--resources', 'ocrmypdf'],
-            encoding='utf-8', stdout=PIPE, check=True)
-
-    poet_resources = p.stdout
-
-    # Remove the duplicate "ocrmypdf" resource block
-    all_resources = poet_resources.split('resource')
-    kept_resources = [block for block in all_resources if 'ocrmypdf' not in block]
-    resources = 'resource'.join(kept_resources)
-
-    with open('ocrmypdf.rb', 'w') as out:
-        out.write(recipe_template.substitute(**locals()))
-
-
-if __name__ == '__main__':
-    main()

.travis/osx_brew.sh

@@ -1,23 +0,0 @@
-#!/bin/bash
-# © 2017 James R. Barlow: github.com/jbarlow83
-set -uo pipefail
-set -x
-
-pip3 install homebrew-pypi-poet
-python3 .travis/autobrew.py
-cat ocrmypdf.rb
-
-# brew audit crashes Travis
-#brew audit ocrmypdf.rb
-
-# Important: disable debug output so token is hidden
-set +x
-git clone https://$HOMEBREW_OCRMYPDF_TOKEN@github.com/jbarlow83/homebrew-ocrmypdf.git
-set -x
-
-pushd homebrew-ocrmypdf
-cp ../ocrmypdf.rb Formula/ocrmypdf.rb
-git add Formula/ocrmypdf.rb
-git commit -m "homebrew-ocrmypdf: automatic release $TRAVIS_BUILD_NUMBER $TRAVIS_TAG"
-git push origin master
-popd

MANIFEST.in

@@ -1,7 +1,5 @@
 # requirements
-include requirements.txt
-include test_requirements.txt
-include dev_requirements.txt
+recursive-include requirements *
 
 # git
 include .git_archival.txt
@@ -12,21 +10,20 @@ recursive-include .docker *
 
 # tests
 include .coveragerc
+recursive-include tests *.bin
 recursive-include tests *.jpg
+recursive-include tests *.jsonl
 recursive-include tests *.png
 recursive-include tests *.pdf
 recursive-include tests *.py
 recursive-include tests *.rst
 recursive-include tests *.txt
-recursive-include tests/cache *
-recursive-exclude tests/output *
-recursive-exclude tests/output_pageinfo *
 recursive-exclude tests/resources/private *
 
 # documentation
 include LICENSE
 include *.rst
-recursive-exclude .github * 
+recursive-exclude .github *
 recursive-include docs *.py
 recursive-include docs *.rst
 recursive-include docs *.svg
@@ -42,8 +39,5 @@ exclude .travis*
 
 
 # code
-recursive-include src/ocrmypdf *.py
-exclude ocrmypdf/lib/_leptonica.py
-
-
-
+exclude src/ocrmypdf/lib/_leptonica.py
+exclude scratch.py

README.md

@@ -0,0 +1,151 @@
+OCRmyPDF
+========
+
+[![Travis build status][travis]](https://travis-ci.org/jbarlow83/OCRmyPDF) [![PyPI version][pypi]](https://pypi.org/project/ocrmypdf/) ![Homebrew version][homebrew] ![ReadTheDocs][docs]
+
+[travis]: https://travis-ci.org/jbarlow83/OCRmyPDF.svg?branch=master "Travis build status"
+
+[pypi]: https://img.shields.io/pypi/v/ocrmypdf.svg "PyPI version"
+
+[homebrew]: https://img.shields.io/homebrew/v/ocrmypdf.svg "Homebrew version"
+
+[docs]: https://readthedocs.org/projects/ocrmypdf/badge/?version=latest "RTD"
+
+OCRmyPDF adds an OCR text layer to scanned PDF files, allowing them to be searched or copy-pasted.
+
+```bash
+ocrmypdf                      # it's a scriptable command line program
+   -l eng+fra                 # it supports multiple languages
+   --rotate-pages             # it can fix pages that are misrotated
+   --deskew                   # it can deskew crooked PDFs!
+   --title "My PDF"           # it can change output metadata
+   --jobs 4                   # it uses multiple cores by default
+   --output-type pdfa         # it produces PDF/A by default
+   input_scanned.pdf          # takes PDF input (or images)
+   output_searchable.pdf      # produces validated PDF output
+```
+
+[See the release notes for details on the latest changes](https://ocrmypdf.readthedocs.io/en/latest/release_notes.html).
+
+Main features
+-------------
+
+- Generates a searchable [PDF/A](https://en.wikipedia.org/?title=PDF/A) file from a regular PDF
+- Places OCR text accurately below the image to ease copy / paste
+- Keeps the exact resolution of the original embedded images
+- When possible, inserts OCR information as a "lossless" operation without disrupting any other content
+- Optimizes PDF images, often producing files smaller than the input file
+- If requested deskews and/or cleans the image before performing OCR
+- Validates input and output files
+- Distributes work across all available CPU cores
+- Uses [Tesseract OCR](https://github.com/tesseract-ocr/tesseract) engine to recognize more than [100 languages](https://github.com/tesseract-ocr/tessdata)
+- Scales properly to handle files with thousands of pages
+- Battle-tested on millions of PDFs
+
+For details: please consult the [documentation](https://ocrmypdf.readthedocs.io/en/latest/).
+
+Motivation
+----------
+
+I searched the web for a free command line tool to OCR PDF files on Linux/UNIX: I found many, but none of them were really satisfying.
+
+- Either they produced PDF files with misplaced text under the image (making copy/paste impossible)
+- Or they did not handle accents and multilingual characters
+- Or they changed the resolution of the embedded images
+- Or they generated ridiculously large PDF files
+- Or they crashed when trying to OCR
+- Or they did not produce valid PDF files
+- On top of that none of them produced PDF/A files (format dedicated for long time storage)
+
+...so I decided to develop my own tool.
+
+Installation
+------------
+
+Linux, UNIX, and macOS are supported. Windows is not directly supported but there is a Docker image available that runs on Windows.
+
+Users of Debian 9 or later or Ubuntu 16.10 or later may simply
+
+```bash
+apt-get install ocrmypdf
+```
+
+and users of Fedora 29 or later may simply
+
+```bash
+dnf install ocrmypdf
+```
+
+and macOS users with Homebrew may simply
+
+```bash
+brew install ocrmypdf
+```
+
+For everyone else, [see our documentation](https://ocrmypdf.readthedocs.io/en/latest/installation.html) for installation steps.
+
+Languages
+---------
+
+OCRmyPDF uses Tesseract for OCR, and relies on its language packs. For Linux users, you can often find packages that provide language packs:
+
+```bash
+# Display a list of all Tesseract language packs
+apt-cache search tesseract-ocr
+
+# Debian/Ubuntu users
+apt-get install tesseract-ocr-chi-sim  # Example: Install Chinese Simplified language back
+```
+
+You can then pass the `-l LANG` argument to OCRmyPDF to give a hint as to what languages it should search for. Multiple languages can be requested.
+
+Documentation and support
+-------------------------
+
+Once ocrmypdf is installed, the built-in help which explains the command syntax and options can be accessed via:
+
+```bash
+ocrmypdf --help
+```
+
+Our [documentation is served on Read the Docs](https://ocrmypdf.readthedocs.io/en/latest/index.html).
+
+If you detect an issue, please:
+
+- Check whether your issue is already known
+- If no problem report exists on github, please create one here: <https://github.com/jbarlow83/OCRmyPDF/issues>
+- Describe your problem thoroughly
+- Append the console output of the script when running the debug mode (`-v 1` option)
+- If possible provide your input PDF file as well as the content of the temporary folder (using a file sharing service like Dropbox)
+
+Requirements
+------------
+
+Runs on CPython 3.5, 3.6 and 3.7. Requires external program installations of Ghostscript, Tesseract OCR, QPDF, and Leptonica. ocrmypdf is pure Python, but uses CFFI to portably generate library bindings.
+
+Press & Media
+-------------
+
+- [Going paperless with OCRmyPDF](https://medium.com/@ikirichenko/going-paperless-with-ocrmypdf-e2f36143f46a)
+- [Converting a scanned document into a compressed searchable PDF with redactions](https://medium.com/@treyharris/converting-a-scanned-document-into-a-compressed-searchable-pdf-with-redactions-63f61c34fe4c)
+- [c't 1-2014, page 59](http://heise.de/-2279695): Detailed presentation of OCRmyPDF v1.0 in the leading German IT magazine c't
+- [heise Open Source, 09/2014: Texterkennung mit OCRmyPDF](http://heise.de/-2356670)
+
+Business enquiries
+------------------
+
+OCRmyPDF would not be the software that it is today is without companies and users choosing to provide support for feature development and consulting enquiries. We are happy to discuss all enquiries, whether for extending the existing feature set, or integrating OCRmyPDF into a larger system.
+
+License
+-------
+
+The OCRmyPDF software is licensed under the GNU GPLv3. Certain files are covered by other licenses, as noted in their source files.
+
+The license for each test file varies, and is noted in tests/resources/README.rst. The documentation is licensed under Creative Commons Attribution-ShareAlike 4.0 (CC-BY-SA 4.0).
+
+OCRmyPDF versions prior to 6.0 were distributed under the MIT License.
+
+Disclaimer
+----------
+
+The software is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

README.rst

@@ -1,152 +0,0 @@
-OCRmyPDF
-========
-
-.. image:: https://travis-ci.org/jbarlow83/OCRmyPDF.svg?branch=master
-    :target: https://travis-ci.org/jbarlow83/OCRmyPDF
-
-.. image:: https://img.shields.io/pypi/v/ocrmypdf.svg
-    :target: https://pypi.org/project/ocrmypdf/
-
-.. image:: https://img.shields.io/homebrew/v/ocrmypdf.svg
-    :alt: homebrew
-    :target: http://brewformulas.org/Ocrmypdf
-
-
-OCRmyPDF adds an OCR text layer to scanned PDF files, allowing them to
-be searched or copy-pasted.
-
-.. code-block:: bash
-
-   ocrmypdf                      # it's a scriptable command line program
-      -l eng+fra                 # it supports multiple languages
-      --rotate-pages             # it can fix pages that are misrotated
-      --deskew                   # it can deskew crooked PDFs!
-      --title "My PDF"           # it can change output metadata
-      --jobs 4                   # it uses multiple cores by default
-      --output-type pdfa         # it produces PDF/A by default
-      input_scanned.pdf          # takes PDF input (or images)
-      output_searchable.pdf      # produces validated PDF output
-
-
-Main features
--------------
-
--  Generates a searchable
-   `PDF/A <https://en.wikipedia.org/?title=PDF/A>`_ file from a regular PDF
--  Places OCR text accurately below the image to ease copy / paste
--  Keeps the exact resolution of the original embedded images
--  When possible, inserts OCR information as a "lossless" operation without rendering vector information
--  Keeps file size about the same
--  If requested deskews and/or cleans the image before performing OCR
--  Validates input and output files
--  Processes pages in parallel when more than one CPU core is
-   available
--  Uses `Tesseract OCR <https://github.com/tesseract-ocr/tesseract>`_ engine
--  Supports more than `100 languages <https://github.com/tesseract-ocr/tessdata>`_ recognized by Tesseract
--  Battle-tested on thousands of PDFs, a test suite and continuous integration
-
-For details: please consult the `documentation <https://ocrmypdf.readthedocs.io/en/latest/>`_.
-
-Motivation
-----------
-
-I searched the web for a free command line tool to OCR PDF files on
-Linux/UNIX: I found many, but none of them were really satisfying.
-
--  Either they produced PDF files with misplaced text under the image (making copy/paste impossible)
--  Or they did not handle accents and multilingual characters
--  Or they changed the resolution of the embedded images
--  Or they generated ridiculously large PDF files
--  Or they crashed when trying to OCR some of my PDF files
--  Or they did not produce valid PDF files (even though they were readable with my current PDF reader)
--  On top of that none of them produced PDF/A files (format dedicated for long time storage)
-
-...so I decided to develop my own tool (using various existing scripts
-as an inspiration).
-
-Installation
-------------
-
-Linux, UNIX, and macOS are supported. Windows is not directly supported but there is a Docker image available that runs on Windows.
-
-Users of Debian 9 or later or Ubuntu 16.10 or later may simply
-
-.. code-block:: bash
-
-   apt-get install ocrmypdf
-
-and macOS users with Homebrew may simply
-
-.. code-block:: bash
-
-   brew install ocrmypdf
-
-For everyone else, `see our documentation <https://ocrmypdf.readthedocs.io/en/latest/installation.html>`_ for installation steps.
-
-Languages
----------
-
-OCRmyPDF uses Tesseract for OCR, and relies on its language packs. For Linux users,
-you can often find packages that provide language packs:
-
-.. code-block:: bash
-
-   # Display a list of all Tesseract language packs
-   apt-cache search tesseract-ocr
-
-   # Debian/Ubuntu users
-   apt-get install tesseract-ocr-chi-sim  # Example: Install Chinese Simplified language back
-
-You can then pass the ``-l LANG`` argument to OCRmyPDF to give a hint as to what languages it should search for. Multiple
-languages can be requested.
-
-Documentation and support
--------------------------
-
-Once ocrmypdf is installed, the built-in help which explains the command syntax and options can be accessed via:
-
-.. code-block:: bash
-
-   ocrmypdf --help
-
-Our `documentation is served on Read the Docs <https://ocrmypdf.readthedocs.io/en/latest/index.html>`_.
-
-If you detect an issue, please:
-
--  Check whether your issue is already known
--  If no problem report exists on github, please create one here:
-   https://github.com/jbarlow83/OCRmyPDF/issues
--  Describe your problem thoroughly
--  Append the console output of the script when running the debug mode
-   (``-v 1`` option)
--  If possible provide your input PDF file as well as the content of the
-   temporary folder (using a file sharing service like Dropbox)
-
-Requirements
-------------
-
-Runs on CPython 3.5, 3.6 and 3.7. Requires external program installations of Ghostscript, Tesseract OCR, QPDF, and Leptonica. ocrmypdf is pure Python, but uses CFFI to portably generate library bindings.
-
-Press & Media
--------------
-
--  `c't 1-2014, page 59 <http://heise.de/-2279695>`_:
-   Detailed presentation of OCRmyPDF v1.0 in the leading German IT
-   magazine c't
--  `heise Open Source, 09/2014: Texterkennung mit
-   OCRmyPDF <http://heise.de/-2356670>`_
-
-License
--------
-
-The OCRmyPDF software is licensed under the GNU GPLv3. Certain files are covered by other licenses, as noted in their source files.
-
-The license for each test file varies, and is noted in tests/resources/README.rst. The documentation is licensed under Creative Commons Attribution-ShareAlike 4.0 (CC-BY-SA 4.0).
-
-OCRmyPDF versions prior to 6.0 were licensed under the MIT License.
-
-Disclaimer
-----------
-
-The software is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
-CONDITIONS OF ANY KIND, either express or implied.

docs/advanced.rst

@@ -1,29 +1,107 @@
+=================
 Advanced features
 =================
 
-Control of OCR options
-----------------------
+Control of unpaper
+==================
 
-OCRmyPDF provides many features to control the behavior of the OCR engine, Tesseract.
+OCRmyPDF uses ``unpaper`` to provide the implementation of the
+``--clean`` and ``--clean-final`` arguments.
+`unpaper <https://github.com/Flameeyes/unpaper/blob/master/doc/basic-concepts.md>`__
+provides a variety of image processing filters to improve images.
+
+By default, OCRmyPDF uses only ``unpaper`` arguments that were found to
+be safe to use on almost all files without having to inspect every page
+of the file afterwards. This is particularly true when only ``--clean``
+is used, since that instructs OCRmyPDF to only clean the image before
+OCR and not the final image.
+
+However, if you wish to use the more aggressive options in ``unpaper``,
+you may use ``--unpaper-args '...'`` to override the OCRmyPDF's defaults
+and forward other arguments to unpaper. This option will forward
+arguments to ``unpaper`` without any knowledge of what that program
+considers to be valid arguments. The string of arguments must be quoted
+as shown in the examples below. No filename arguments may be included.
+OCRmyPDF will assume it can append input and output filename of
+intermediate images to the ``--unpaper-args`` string.
+
+In this example, we tell ``unpaper`` to expect two pages of text on a
+sheet (image), such as occurs when two facing pages of a book are
+scanned. ``unpaper`` uses this information to deskew each independently
+and clean up the margins of both.
+
+.. code-block:: bash
+
+    ocrmypdf --clean --clean-final --unpaper-args '--layout double' input.pdf output.pdf
+    ocrmypdf --clean --clean-final --unpaper-args '--layout double --no-noisefilter' input.pdf output.pdf
+
+.. warning::
+
+   Some ``unpaper`` features will reposition text within the image.
+   ``--clean-final`` is recommended to avoid this issue.
+
+.. warning::
+
+   Some ``unpaper`` features cause multiple input or output files to be
+   consumed or produced. OCRmyPDF requires ``unpaper`` to consume one
+   file and produce one file. An deviation from that condition will
+   result in errors.
+
+.. note::
+
+   ``unpaper`` uses uncompressed PBM/PGM/PPM files for its intermediate
+   files. For large images or documents, it can take a lot of temporary
+   disk space.
+
+Control of OCR options
+======================
+
+OCRmyPDF provides many features to control the behavior of the OCR
+engine, Tesseract.
 
 When OCR is skipped
-"""""""""""""""""""
+-------------------
 
-If a page in a PDF seems to have text, by default OCRmyPDF will exit without modifying the PDF. This is to ensure that PDFs that were previously OCRed or were "born digital" rather than scanned are not processed.
+If a page in a PDF seems to have text, by default OCRmyPDF will exit
+without modifying the PDF. This is to ensure that PDFs that were
+previously OCRed or were "born digital" rather than scanned are not
+processed.
 
-If ``--skip-text`` is issued, then no OCR will be performed on pages that already have text. The page will be copied to the output. This may be useful for documents that contain both "born digital" and scanned content, or to use OCRmyPDF to normalize and convert to PDF/A regardless of their contents.
+If ``--skip-text`` is issued, then no OCR will be performed on pages
+that already have text. The page will be copied to the output. This may
+be useful for documents that contain both "born digital" and scanned
+content, or to use OCRmyPDF to normalize and convert to PDF/A regardless
+of their contents.
 
-If ``--force-ocr`` is issued, then all pages will be rasterized to images, discarding any hidden OCR text, and rasterizing any printable text. This is useful for redoing OCR, for fixing OCR text with a damaged character map (text is selectable but not searchable), and destroying redacted information.
+If ``--redo-ocr`` is issued, then a detailed text analysis is performed.
+Text is categorized as either visible or invisible. Invisible text (OCR)
+is stripped out. Then an image of each page is created with visible text
+masked out. The page image is sent for OCR, and any additional text is
+inserted as OCR. If a file contains a mix of text and bitmap images that
+contain text, OCRmyPDF will locate the additional text in images without
+disrupting the existing text.
 
+If ``--force-ocr`` is issued, then all pages will be rasterized to
+images, discarding any hidden OCR text, and rasterizing any printable
+text. This is useful for redoing OCR, for fixing OCR text with a damaged
+character map (text is selectable but not searchable), and destroying
+redacted information. Any forms and vector graphics will be rasterized
+as well.
 
 Time and image size limits
-""""""""""""""""""""""""""
+--------------------------
 
-By default, OCRmyPDF permits tesseract to run for three minutes (180 seconds) per page. This is usually more than enough time to find all text on a reasonably sized page with modern hardware.
+By default, OCRmyPDF permits tesseract to run for three minutes (180
+seconds) per page. This is usually more than enough time to find all
+text on a reasonably sized page with modern hardware.
 
-If a page is skipped, it will be inserted without OCR. If preprocessing was requested, the preprocessed image layer will be inserted.
+If a page is skipped, it will be inserted without OCR. If preprocessing
+was requested, the preprocessed image layer will be inserted.
 
-If you want to adjust the amount of time spent on OCR, change ``--tesseract-timeout``.  You can also automatically skip images that exceed a certain number of megapixels with ``--skip-big``. (A 300 DPI, 8.5×11" page is 8.4 megapixels.)
+If you want to adjust the amount of time spent on OCR, change
+``--tesseract-timeout``. You can also automatically skip images that
+exceed a certain number of megapixels with ``--skip-big``. (A 300 DPI,
+8.5×11" page is 8.4 megapixels.)
 
 .. code-block:: bash
 
@@ -31,50 +109,62 @@ If you want to adjust the amount of time spent on OCR, change ``--tesseract-time
     ocrmypdf --tesseract-timeout 300 --skip-big 50 bigfile.pdf output.pdf
 
 Overriding default tesseract
-""""""""""""""""""""""""""""
+----------------------------
 
 OCRmyPDF checks the system ``PATH`` for the ``tesseract`` binary.
 
-Some relevant environment variables that influence Tesseract's behavior include:
+Some relevant environment variables that influence Tesseract's behavior
+include:
 
 .. envvar:: TESSDATA_PREFIX
 
-    Overrides the path to Tesseract's data files. This can allow simultaneous installation of the "best" and "fast" training data sets. OCRmyPDF does not manage this environment variable.
+   Overrides the path to Tesseract's data files. This can allow
+   simultaneous installation of the "best" and "fast" training data
+   sets. OCRmyPDF does not manage this environment variable.
 
 .. envvar:: OMP_THREAD_LIMIT
 
-    Controls the number of threads Tesseract will use. OCRmyPDF will manage this environment if it is not already set. (Currently, it will set it to 1 because this gives the best results in testing.)
+   Controls the number of threads Tesseract will use. OCRmyPDF will
+   manage this environment if it is not already set. (Currently, it will
+   set it to 1 because this gives the best results in testing.)
 
-For example, if you are testing tesseract 4.00 and don't wish to use an existing tesseract 3.04 installation, you can launch OCRmyPDF as follows:
+For example, if you have a development build of Tesseract don't wish to
+use the system installation, you can launch OCRmyPDF as follows:
 
 .. code-block:: bash
 
     env \
-        PATH=/home/user/src/tesseract4/api:$PATH \
-        TESSDATA_PREFIX=/home/user/src/tesseract4 \
-        ocrmypdf --tesseract-oem 2 input.pdf output.pdf
-
-In this example ``TESSDATA_PREFIX`` directs Tesseract 4.0 to use LSTM training data. ``--tesseract-oem 1`` requests tesseract 4.0's new LSTM engine. (Tesseract 4.0 only.)
+        PATH=/home/user/src/tesseract/api:$PATH \
+        TESSDATA_PREFIX=/home/user/src/tesseract \
+        ocrmypdf input.pdf output.pdf
 
+In this example ``TESSDATA_PREFIX`` is required to redirect Tesseract to
+an alternate folder for its "tessdata" files.
 
 Overriding other support programs
-"""""""""""""""""""""""""""""""""
+---------------------------------
 
 In addition to tesseract, OCRmyPDF uses the following external binaries:
 
-* ``gs`` (Ghostscript)
-* ``unpaper``
-* ``qpdf``
-
-In each case OCRmyPDF will search the ``PATH`` environment variable to locate the binaries.
+-  ``gs`` (Ghostscript)
+-  ``unpaper``
+-  ``qpdf``
 
+In each case OCRmyPDF will search the ``PATH`` environment variable to
+locate the binaries.
 
 Changing tesseract configuration variables
-""""""""""""""""""""""""""""""""""""""""""
+------------------------------------------
 
-You can override tesseract's default `control parameters <https://github.com/tesseract-ocr/tesseract/wiki/ControlParams>`_ with a configuration file.
+You can override tesseract's default `control
+parameters <https://github.com/tesseract-ocr/tesseract/wiki/ControlParams>`__
+with a configuration file.
 
-As an example, this configuration will disable Tesseract's dictionary for current language. Normally the dictionary is helpful for interpolating words that are unclear, but it may interfere with OCR if the document does not contain many words (for example, a list of part numbers).
+As an example, this configuration will disable Tesseract's dictionary
+for current language. Normally the dictionary is helpful for
+interpolating words that are unclear, but it may interfere with OCR if
+the document does not contain many words (for example, a list of part
+numbers).
 
 Create a file named "no-dict.cfg" with these contents:
 
@@ -92,11 +182,11 @@ then run ocrmypdf as follows (along with any other desired arguments):
 
 .. warning::
 
-    Some combinations of control parameters will break Tesseract or break assumptions that OCRmyPDF makes about Tesseract's output.
-
+   Some combinations of control parameters will break Tesseract or break
+   assumptions that OCRmyPDF makes about Tesseract's output.
 
 Changing the PDF renderer
--------------------------
+=========================
 
 rasterizing
   Converting a PDF to an image for display.
@@ -104,44 +194,63 @@ rasterizing
 rendering
   Creating a new PDF from other data (such as an existing PDF).
 
-
-OCRmyPDF has these PDF renderers: ``sandwich`` and ``hocr``. The renderer may be selected using ``--pdf-renderer``. The default is ``auto`` which lets OCRmyPDF select the renderer to use. Currently, ``auto`` selects ``sandwich`` for Tesseract 3.05.01 or newer, or ``hocr`` for older versions of Tesseract.
+OCRmyPDF has these PDF renderers: ``sandwich`` and ``hocr``. The
+renderer may be selected using ``--pdf-renderer``. The default is
+``auto`` which lets OCRmyPDF select the renderer to use. Currently,
+``auto`` always selects ``sandwich``.
 
 The ``sandwich`` renderer
-"""""""""""""""""""""""""
+-------------------------
 
-The ``sandwich`` renderer uses Tesseract's new text-only PDF feature, which produces a PDF page that lays out the OCR in invisible text. This page is then "sandwiched" onto the original PDF page, allowing lossless application of OCR even to PDF pages that contain other vector objects.
+The ``sandwich`` renderer uses Tesseract's new text-only PDF feature,
+which produces a PDF page that lays out the OCR in invisible text. This
+page is then "sandwiched" onto the original PDF page, allowing lossless
+application of OCR even to PDF pages that contain other vector objects.
 
-Currently this is the best renderer for most uses, however it is implemented in Tesseract so OCRmyPDF cannot influence it. Currently some problematic PDF viewers like Mozilla PDF.js and macOS Preview have problems with segmenting its text output, and mightrunseveralwordstogether.
+Currently this is the best renderer for most uses, however it is
+implemented in Tesseract so OCRmyPDF cannot influence it. Currently some
+problematic PDF viewers like Mozilla PDF.js and macOS Preview have
+problems with segmenting its text output, and
+mightrunseveralwordstogether.
 
-When image preprocessing features like ``--deskew`` are used, the original PDF will be rendered as a full page and the OCR layer will be placed on top.
-
-If a PDF created with this renderer using Tesseract versions older than 3.05.00 is then passed through Ghostscript's pdfwrite feature, the OCR text *may* be corrupted. The ``--output-type=pdfa`` argument will produce a warning in this situation.  For this reason, OCRmyPDF automatically selects the ``hocr`` for older Tesseract versions.
+When image preprocessing features like ``--deskew`` are used, the
+original PDF will be rendered as a full page and the OCR layer will be
+placed on top.
 
 The ``hocr`` renderer
-"""""""""""""""""""""
+---------------------
 
-The ``hocr`` renderer works with older versions of Tesseract. The image layer is copied from the original PDF page if possible, avoiding potentially lossy transcoding or loss of other PDF information. If preprocessing is specified, then the image layer is a new PDF.
+The ``hocr`` renderer works with older versions of Tesseract. The image
+layer is copied from the original PDF page if possible, avoiding
+potentially lossy transcoding or loss of other PDF information. If
+preprocessing is specified, then the image layer is a new PDF.
 
-Unlike ``sandwich`` this renderer is implemented within OCRmyPDF; anyone looking to customize how OCR is presented should look here. A major disadvantage of this renderer is it not capable of correctly handling text outside the Latin alphabet. Pull requests to improve the situation are welcome.
+Unlike ``sandwich`` this renderer is implemented within OCRmyPDF; anyone
+looking to customize how OCR is presented should look here. A major
+disadvantage of this renderer is it not capable of correctly handling
+text outside the Latin alphabet. Pull requests to improve the situation
+are welcome.
 
-Currently, this renderer has the best compatibility with Mozilla's PDF.js viewer.
+Currently, this renderer has the best compatibility with Mozilla's
+PDF.js viewer.
 
 This works in all versions of Tesseract.
 
 The ``tesseract`` renderer
-""""""""""""""""""""""""""
+--------------------------
 
-The ``tesseract`` renderer was removed. OCRmyPDF's new approach to text layer grafting makes it functionally equivalent to ``sandwich``.
+The ``tesseract`` renderer was removed. OCRmyPDF's new approach to text
+layer grafting makes it functionally equivalent to ``sandwich``.
 
 Return code policy
-------------------
+==================
 
-OCRmyPDF writes all messages to ``stderr``.  ``stdout`` is reserved for piping
-output files.  ``stdin`` is reserved for piping input files.
+OCRmyPDF writes all messages to ``stderr``. ``stdout`` is reserved for
+piping output files. ``stdin`` is reserved for piping input files.
 
-The return codes generated by the OCRmyPDF are considered part of the stable
-user interface.  They may be imported from ``ocrmypdf.exceptions``.
+The return codes generated by the OCRmyPDF are considered part of the
+stable user interface. They may be imported from
+``ocrmypdf.exceptions``.
 
 .. list-table:: Return codes
     :widths: 5 35 60
@@ -189,3 +298,39 @@ user interface.  They may be imported from ``ocrmypdf.exceptions``.
     *	- 130
         - ``ExitCode.ctrl_c``
         - The program was interrupted by pressing Ctrl+C.
+
+
+Debugging the intermediate files
+================================
+
+OCRmyPDF normally saves its intermediate results to a temporary folder
+and deletes this folder when it exits, whether it succeeded or failed.
+
+If the ``-k`` argument is issued on the command line, OCRmyPDF will keep
+the temporary folder and print the location, whether it succeeded or
+failed (provided the Python interpreter did not crash). An example
+message is:
+
+.. code-block:: none
+
+    Temporary working files retained at:
+    /tmp/com.github.ocrmypdf.u20wpz07
+
+The organization of this folder is an implementation detail and subject
+to change between releases. However the general organization is that
+working files on a per page basis have the page number as a prefix
+(starting with page 1), an infix indicates the processing stage, and a
+suffix indicates the file type. Some important files include:
+
+-  ``.page.png`` - what the input page looks like
+-  ``.image`` - the image we will show the user if we are in a mode that
+   changes the final appearance; may be in one of several image formats
+-  ``.text.pdf`` - the OCR file; this will load as a blank page but
+   should have visible text if checked with a tool like pdftotext or
+   pdfminder.six
+-  ``.ocr.png`` - the file that is sent to Tesseract for OCR; depending
+   on arguments this may differ from the presentation image
+-  ``layers.rendered.pdf`` - the composite PDF, before metadata repair
+   and optimization
+-  ``images/*`` - images extracted during the optimization process; here
+   the prefix indicates a PDF object ID not a page number

docs/api.rst

@@ -0,0 +1,101 @@
+======================
+Using the OCRmyPDF API
+======================
+
+OCRmyPDF originated as a command line program and continues to have this
+legacy, but parts of it can be imported and used in other Python
+applications.
+
+Some applications may want to consider running ocrmypdf from a
+subprocess call anyway, as this provides isolation of its activities.
+
+Example
+=======
+
+OCRmyPDF one high-level function to run its main engine from an
+application. The parameters are symmetric to the command line arguments
+and largely have the same functions.
+
+.. code-block:: python
+
+    import ocrmypdf
+
+    ocrmypdf.run('input.pdf', 'output.pdf', deskew=True)
+
+With a few exceptions, all of the command line arguments are available
+and may be passed as equivalent keywords.
+
+A few differences are that ``verbose`` and ``quiet`` are not available.
+Instead, output should be managed by configuring logging.
+
+Parent process requirements
+---------------------------
+
+The :func:`ocrmypdf.run` function runs OCRmyPDF similar to command line
+execution. To do this, it will: - create a monitoring thread - create
+worker processes (forking itself) - manage the signal flags of worker
+processes 0 execute other subprocesses (forking and executing other
+programs)
+
+The Python process that calls ``ocrmypdf.run()`` must be sufficiently
+privileged to perform these actions. If it is not, ``ocrmypdf()`` will
+fail.
+
+There is no currently no option to manage how jobs are scheduled other
+than the argument ``jobs=`` which will limit the number of worker
+processes.
+
+Forking a child process to call ``ocrmypdf.run()`` is suggested. That
+way your application will survive and remain interactive even if
+OCRmyPDF does not.
+
+Logging
+-------
+
+OCRmyPDF will log under loggers named ``ocrmypdf``. In addition, it
+imports ``pdfminer`` and ``PIL``, both of which post log messages under
+those logging namespaces.
+
+You can configure the logging as desired for your application or call
+:func:`ocrmypdf.configure_logging` to configure logging the same way
+OCRmyPDF itself does. The command line parameters such as ``--quiet``
+and ``--verbose`` have no equivalents in the API; you must use the
+provided configuration function or do configuration in a way that suits
+your use case.
+
+Progress monitoring
+-------------------
+
+OCRmyPDF uses the ``tqdm`` package to implement its progress bars.
+:func:`ocrmypdf.configure_logging` will set up logging output to
+``sys.stderr`` in a way that is compatible with the display of the
+progress bar.
+
+Exceptions
+----------
+
+OCRmyPDF may throw standard Python exceptions, ``ocrmypdf.exceptions.*``
+exceptions, some exceptions related to multiprocessing, and
+``KeyboardInterrupt``. The parent process should provide an exception
+handler. OCRmyPDF will clean up its temporary files and worker processes
+automatically when an exception occurs.
+
+Programs that call OCRmyPDF should consider trapping KeyboardInterrupt
+so that they allow OCR to terminate with the whole program terminating.
+
+When OCRmyPDF succeeds conditionally, it returns an integer exit code.
+
+Reference
+---------
+
+.. autofunction:: ocrmypdf.run
+
+.. autoclass:: ocrmypdf.Verbosity
+    :members:
+    :undoc-members:
+
+.. autoclass:: ocrmypdf.ExitCode
+    :members:
+    :undoc-members:
+
+.. autofunction:: ocrmypdf.configure_logging

docs/batch.rst

@@ -1,214 +1,263 @@
+================
 Batch processing
 ================
 
-This article provides information about running OCRmyPDF on multiple files or configuring it as a service triggered by file system events.
+This article provides information about running OCRmyPDF on multiple
+files or configuring it as a service triggered by file system events.
 
 Batch jobs
-----------
+==========
 
-Consider using the excellent `GNU Parallel <https://www.gnu.org/software/parallel/>`_ to apply OCRmyPDF to multiple files at once.
+Consider using the excellent `GNU
+Parallel <https://www.gnu.org/software/parallel/>`__ to apply OCRmyPDF
+to multiple files at once.
 
-Both ``parallel`` and ``ocrmypdf`` will try to use all available processors. To maximize parallelism without overloading your system with processes, consider using ``parallel -j 2`` to limit parallel to running two jobs at once.
+Both ``parallel`` and ``ocrmypdf`` will try to use all available
+processors. To maximize parallelism without overloading your system with
+processes, consider using ``parallel -j 2`` to limit parallel to running
+two jobs at once.
 
-This command will run all ocrmypdf all files named ``*.pdf`` in the current directory and write them to the previous created ``output/`` folder. It will not search subdirectories.
+This command will run all ocrmypdf all files named ``*.pdf`` in the
+current directory and write them to the previous created ``output/``
+folder. It will not search subdirectories.
 
-The ``--tag`` argument tells parallel to print the filename as a prefix whenever a message is printed, so that one can trace any errors to the file that produced them.
+The ``--tag`` argument tells parallel to print the filename as a prefix
+whenever a message is printed, so that one can trace any errors to the
+file that produced them.
 
 .. code-block:: bash
 
-	parallel --tag -j 2 ocrmypdf '{}' 'output/{}' ::: *.pdf
+   parallel --tag -j 2 ocrmypdf '{}' 'output/{}' ::: *.pdf
 
-OCRmyPDF automatically repairs PDFs before parsing and gathering information from them.  If you are already repairing PDFs with ``qpdf`` prior to attempting OCR, or you can use ``--skip-repair`` to skip this step.  It may improve performance for large files, since repairing PDFs is single-threaded.
+OCRmyPDF automatically repairs PDFs before parsing and gathering
+information from them.
 
 Directory trees
----------------
+===============
 
-This will walk through a directory tree and run OCR on all files in place, printing the output in a way that makes 
+This will walk through a directory tree and run OCR on all files in
+place, printing the output in a way that makes
 
 .. code-block:: bash
 
-	find . --printf '%p' -name '*.pdf' -exec ocrmypdf '{}' '{}' \;
+   find . -printf '%p' -name '*.pdf' -exec ocrmypdf '{}' '{}' \;
 
-This only runs one ``ocrmypdf`` process at a time. This variation uses ``find`` to create a directory list and ``parallel`` to parallelize runs of ``ocrmypdf``, again updating files in place.
+Alternatively, with a docker container (mounts a volume to the container
+where the PDFs are stored):
 
 .. code-block:: bash
 
-	find . -name '*.pdf' | parallel --tag -j 2 ocrmypdf '{}' '{}'
+   find . -printf '%p' -name '*.pdf' -exec docker run --rm -v <host dir>:<container dir> jbarlow83/ocrmypdf-alpine '<container dir>/{}' '<container dir>/{}' \;
 
+This only runs one ``ocrmypdf`` process at a time. This variation uses
+``find`` to create a directory list and ``parallel`` to parallelize runs
+of ``ocrmypdf``, again updating files in place.
+
+.. code-block:: bash
+
+   find . -name '*.pdf' | parallel --tag -j 2 ocrmypdf '{}' '{}'
 
 Sample script
-"""""""""""""
+-------------
 
-This user contributed script also provides an example of batch processing.
+This user contributed script also provides an example of batch
+processing.
 
 .. code-block:: python
 
-	#!/usr/bin/env python3
-	# Walk through directory tree, replacing all files with OCR'd version
-	# Contributed by DeliciousPickle@github
+   #!/usr/bin/env python3
+   # Walk through directory tree, replacing all files with OCR'd version
+   # Contributed by DeliciousPickle@github
 
-	import logging
-	import os
-	import subprocess
-	import sys
+   import logging
+   import os
+   import subprocess
+   import sys
 
-	script_dir = os.path.dirname(os.path.realpath(__file__))
-	print(script_dir + '/ocr-tree.py: Start')
+   script_dir = os.path.dirname(os.path.realpath(__file__))
+   print(script_dir + '/ocr-tree.py: Start')
 
-	if len(sys.argv) > 1:
-	    start_dir = sys.argv[1]
-	else:
-	    start_dir = '.'
+   if len(sys.argv) > 1:
+       start_dir = sys.argv[1]
+   else:
+       start_dir = '.'
 
-	if len(sys.argv) > 2:
-	    log_file = sys.argv[2]
-	else:
-	    log_file = script_dir + '/ocr-tree.log'
+   if len(sys.argv) > 2:
+       log_file = sys.argv[2]
+   else:
+       log_file = script_dir + '/ocr-tree.log'
 
-	logging.basicConfig(
-			level=logging.INFO, format='%(asctime)s %(message)s', 
-			filename=log_file, filemode='w')
-
-	for dir_name, subdirs, file_list in os.walk(start_dir):
-	    logging.info('\n')
-	    logging.info(dir_name + '\n')
-	    os.chdir(dir_name)
-	    for filename in file_list:
-	        file_ext = os.path.splitext(filename)[1]
-	        if file_ext == '.pdf':
-	            full_path = dir_name + '/' + filename
-	            print(full_path)
-	            cmd = ["ocrmypdf",  "--deskew", filename, filename]
-	            logging.info(cmd)
-	            proc = subprocess.Popen(
-	            	cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
-	            result = proc.stdout.read()
-	            if proc.returncode == 6:
-	                print("Skipped document because it already contained text")
-	            elif proc.returncode == 0:
-	                print("OCR complete")
-	            logging.info(result)
-
-API
-"""
-
-OCRmyPDF is currently supported as a command line interface. This means that even if you are using OCRmyPDF in a Python script, you should run it in a subprocess rather importing the ocrmypdf package.
-
-The reason for this limitation is that the `ruffus <https://github.com/bunbun/ruffus/>`_ library that OCRmyPDF depends on is unfortunately not reentrant. OCRmyPDF works by defining each operation it does as a ruffus task that takes one or more files as input and generates one or more files as output. As such ruffus is fairly fundamental.
-
-(If you find individual functions implemented in OCRmyPDF useful (such as ``ocrmypdf.pdfinfo``), you can use these if you wish to.)
+   logging.basicConfig(
+           level=logging.INFO, format='%(asctime)s %(message)s',
+           filename=log_file, filemode='w')
 
+   for dir_name, subdirs, file_list in os.walk(start_dir):
+       logging.info('\n')
+       logging.info(dir_name + '\n')
+       os.chdir(dir_name)
+       for filename in file_list:
+           file_ext = os.path.splitext(filename)[1]
+           if file_ext == '.pdf':
+               full_path = dir_name + '/' + filename
+               print(full_path)
+               cmd = ["ocrmypdf",  "--deskew", filename, filename]
+               logging.info(cmd)
+               proc = subprocess.run(
+                   cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
+               result = proc.stdout
+               if proc.returncode == 6:
+                   print("Skipped document because it already contained text")
+               elif proc.returncode == 0:
+                   print("OCR complete")
+               logging.info(result)
 
 Synology DiskStations
-"""""""""""""""""""""
+---------------------
 
-Synology DiskStations (Network Attached Storage devices) can run the Docker image of OCRmyPDF if the Synology `Docker package <https://www.synology.com/en-global/dsm/packages/Docker>`_ is installed. Attached is a script to address particular quirks of using OCRmyPDF on one of these devices.
+Synology DiskStations (Network Attached Storage devices) can run the
+Docker image of OCRmyPDF if the Synology `Docker
+package <https://www.synology.com/en-global/dsm/packages/Docker>`__ is
+installed. Attached is a script to address particular quirks of using
+OCRmyPDF on one of these devices.
 
-This is only possible for x86-based Synology products. Some Synology products use ARM or Power processors and do not support Docker. Further adjustments might be needed to deal with the Synology's relatively limited CPU and RAM.
+This is only possible for x86-based Synology products. Some Synology
+products use ARM or Power processors and do not support Docker. Further
+adjustments might be needed to deal with the Synology's relatively
+limited CPU and RAM.
 
 .. code-block:: python
 
-	#!/bin/env python3
-	# Contributed by github.com/Enantiomerie
+   #!/bin/env python3
+   # Contributed by github.com/Enantiomerie
 
-	# script needs 2 arguments
-	# 1. source dir with *.pdf - default is location of script
-	# 2. move dir where *.pdf and *_OCR.pdf are moved to
+   # script needs 2 arguments
+   # 1. source dir with *.pdf - default is location of script
+   # 2. move dir where *.pdf and *_OCR.pdf are moved to
 
-	import logging
-	import os
-	import subprocess
-	import sys
-	import time
-	import shutil
+   import logging
+   import os
+   import subprocess
+   import sys
+   import time
+   import shutil
 
-	script_dir = os.path.dirname(os.path.realpath(__file__))
-	timestamp = time.strftime("%Y-%m-%d-%H%M_")
-	log_file = script_dir + '/' + timestamp + 'ocrmypdf.log'
-	logging.basicConfig(level=logging.INFO, format='%(asctime)s %(message)s', filename=log_file, filemode='w')
-
-	if len(sys.argv) > 1:
-	    start_dir = sys.argv[1]
-	else:
-	    start_dir = '.'
-
-	for dir_name, subdirs, file_list in os.walk(start_dir):
-	    logging.info('\n')
-	    logging.info(dir_name + '\n')
-	    os.chdir(dir_name)
-	    for filename in file_list:
-	        file_ext = os.path.splitext(filename)[1]
-	        if file_ext == '.pdf':
-	            full_path = dir_name + '/' + filename
-	            file_noext = os.path.splitext(filename)[0]
-	            timestamp_OCR = time.strftime("%Y-%m-%d-%H%M_OCR_")
-	            filename_OCR = timestamp_OCR + file_noext + '.pdf'
-	            docker_mount = dir_name + ':/home/docker'
-	# create string for pdf processing 
-	# diskstation needs a user:group docker:docker. find uid:gid of your diskstation docker:docker with id docker.
-	# use this uid:gid in -u flag
-	# rw rights for docker:docker at source dir are also necessary
-	# the script is processed as root user via chron 
-	            cmd = ['docker', 'run', '--rm', '-v', docker_mount, '-u=1030:65538', 'jbarlow83/ocrmypdf', , '--deskew' , filename, filename_OCR]
-	            logging.info(cmd)
-	            proc = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
-	            result = proc.stdout.read()
-	            logging.info(result)
-	            full_path_OCR = dir_name + '/' + filename_OCR
-	            os.chmod(full_path_OCR, 0o666)
-	            os.chmod(full_path, 0o666)
-	            full_path_OCR_archive = sys.argv[2]
-	            full_path_archive = sys.argv[2] + '/no_ocr'
-	            shutil.move(full_path_OCR,full_path_OCR_archive)
-	            shutil.move(full_path, full_path_archive)
-	logging.info('Finished.\n')
+   script_dir = os.path.dirname(os.path.realpath(__file__))
+   timestamp = time.strftime("%Y-%m-%d-%H%M_")
+   log_file = script_dir + '/' + timestamp + 'ocrmypdf.log'
+   logging.basicConfig(level=logging.INFO, format='%(asctime)s %(message)s', filename=log_file, filemode='w')
 
+   if len(sys.argv) > 1:
+       start_dir = sys.argv[1]
+   else:
+       start_dir = '.'
 
+   for dir_name, subdirs, file_list in os.walk(start_dir):
+       logging.info('\n')
+       logging.info(dir_name + '\n')
+       os.chdir(dir_name)
+       for filename in file_list:
+           file_ext = os.path.splitext(filename)[1]
+           if file_ext == '.pdf':
+               full_path = dir_name + '/' + filename
+               file_noext = os.path.splitext(filename)[0]
+               timestamp_OCR = time.strftime("%Y-%m-%d-%H%M_OCR_")
+               filename_OCR = timestamp_OCR + file_noext + '.pdf'
+               docker_mount = dir_name + ':/home/docker'
+   # create string for pdf processing
+   # diskstation needs a user:group docker:docker. find uid:gid of your diskstation docker:docker with id docker.
+   # use this uid:gid in -u flag
+   # rw rights for docker:docker at source dir are also necessary
+   # the script is processed as root user via chron
+               cmd = ['docker', 'run', '--rm', '-v', docker_mount, '-u=1030:65538', 'jbarlow83/ocrmypdf', , '--deskew' , filename, filename_OCR]
+               logging.info(cmd)
+               proc = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
+               result = proc.stdout.read()
+               logging.info(result)
+               full_path_OCR = dir_name + '/' + filename_OCR
+               os.chmod(full_path_OCR, 0o666)
+               os.chmod(full_path, 0o666)
+               full_path_OCR_archive = sys.argv[2]
+               full_path_archive = sys.argv[2] + '/no_ocr'
+               shutil.move(full_path_OCR,full_path_OCR_archive)
+               shutil.move(full_path, full_path_archive)
+   logging.info('Finished.\n')
 
 Huge batch jobs
-"""""""""""""""
-
-If you have thousands of files to work with, contact the author. Consulting work related to OCRmyPDF helps fund this open source project and all inquiries are appreciated.
+---------------
 
+If you have thousands of files to work with, contact the author.
+Consulting work related to OCRmyPDF helps fund this open source project
+and all inquiries are appreciated.
 
 Hot (watched) folders
----------------------
+=====================
 
-To set up a "hot folder" that will trigger OCR for every file inserted, use a program like Python `watchdog <https://pypi.python.org/pypi/watchdog>`_ (supports all major OS).
+To set up a "hot folder" that will trigger OCR for every file inserted,
+use a program like Python
+`watchdog <https://pypi.python.org/pypi/watchdog>`__ (supports all major
+OS).
 
-One could then configure a scanner to automatically place scanned files in a hot folder, so that they will be queued for OCR and copied to the destination.
+One could then configure a scanner to automatically place scanned files
+in a hot folder, so that they will be queued for OCR and copied to the
+destination.
 
 .. code-block:: bash
 
-	pip install watchdog
+   pip install watchdog
 
-watchdog installs the command line program ``watchmedo``, which can be told to run ``ocrmypdf`` on any .pdf added to the current directory (``.``) and place the result in the previously created ``out/`` folder.
+watchdog installs the command line program ``watchmedo``, which can be
+told to run ``ocrmypdf`` on any .pdf added to the current directory
+(``.``) and place the result in the previously created ``out/`` folder.
 
 .. code-block:: bash
 
-	cd hot-folder
-	mkdir out
-	watchmedo shell-command \
-		--patterns="*.pdf" \
-		--ignore-directories \
-		--command='ocrmypdf "${watch_src_path}" "out/${watch_src_path}" ' \
-		.  # don't forget the final dot
+   cd hot-folder
+   mkdir out
+   watchmedo shell-command \
+       --patterns="*.pdf" \
+       --ignore-directories \
+       --command='ocrmypdf "${watch_src_path}" "out/${watch_src_path}" ' \
+       .  # don't forget the final dot
 
-For more complex behavior you can write a Python script around to use the watchdog API.
+For more complex behavior you can write a Python script around to use
+the watchdog API.
 
-On file servers, you could configure watchmedo as a system service so it will run all the time.
+On file servers, you could configure watchmedo as a system service so it
+will run all the time.
 
 Caveats
-"""""""
+-------
 
-* ``watchmedo`` may not work properly on a networked file system, depending on the capabilities of the file system client and server.
-* This simple recipe does not filter for the type of file system event, so file copies, deletes and moves, and directory operations, will all be sent to ocrmypdf, producing errors in several cases. Disable your watched folder if you are doing anything other than copying files to it.
-* If the source and destination directory are the same, watchmedo may create an infinite loop.
-* On BSD, FreeBSD and older versions of macOS, you may need to increase the number of file descriptors to monitor more files, using ``ulimit -n 1024`` to watch a folder of up to 1024 files.
+-  ``watchmedo`` may not work properly on a networked file system,
+   depending on the capabilities of the file system client and server.
+-  This simple recipe does not filter for the type of file system event,
+   so file copies, deletes and moves, and directory operations, will all
+   be sent to ocrmypdf, producing errors in several cases. Disable your
+   watched folder if you are doing anything other than copying files to
+   it.
+-  If the source and destination directory are the same, watchmedo may
+   create an infinite loop.
+-  On BSD, FreeBSD and older versions of macOS, you may need to increase
+   the number of file descriptors to monitor more files, using
+   ``ulimit -n 1024`` to watch a folder of up to 1024 files.
 
 Alternatives
-""""""""""""
+------------
 
-* `Watchman <https://facebook.github.io/watchman/>`_ is a more powerful alternative to ``watchmedo``.
+-  `Watchman <https://facebook.github.io/watchman/>`__ is a more
+   powerful alternative to ``watchmedo``.
 
+macOS Automator
+===============
 
+You can use the Automator app with macOS, to create a Workflow or Quick
+Action. Use a *Run Shell Script* action in your workflow. In the context
+of Automator, the ``PATH`` may be set differently your Terminal's
+``PATH``; you may need to explicitly set the PATH to include
+``ocrmypdf``. The following example may serve as a starting point:
+
+|Example macOS Automator script|
+
+You may customize the command sent to ocrmypdf.
+
+.. |Example macOS Automator script| image:: images/macos-workflow.png

docs/conf.py

@@ -30,9 +30,7 @@
 # Add any Sphinx extension module names here, as strings. They can be
 # extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
 # ones.
-extensions = [
-    # 'sphinx.ext.mathjax',
-]
+extensions = ['sphinx.ext.napoleon']
 
 # Add any paths that contain templates here, relative to this directory.
 templates_path = ['_templates']
@@ -52,7 +50,9 @@ master_doc = 'index'
 
 # General information about the project.
 project = 'ocrmypdf'
-copyright = '2018, James R. Barlow. Licensed under Creative Commons Attribution-ShareAlike 4.0.'
+copyright = (
+    '2019, James R. Barlow. Licensed under Creative Commons Attribution-ShareAlike 4.0.'
+)
 author = 'James R. Barlow'
 
 # The version info for the project you're documenting, acts as replacement for
@@ -60,13 +60,35 @@ author = 'James R. Barlow'
 # built documents.
 #
 # The short X.Y version.
-from ocrmypdf import __version__ as OCRMYPDF_VERSION
 
-_version_parts = OCRMYPDF_VERSION.split('.')
+import os
+
+on_rtd = os.environ.get('READTHEDOCS') == 'True'
+
+if on_rtd:
+    # Help ReadTheDocs avoid having to install any binary extension modules
+    import sys
+    from unittest.mock import MagicMock
+
+    class Mock(MagicMock):
+        @classmethod
+        def __getattr__(cls, name):
+            return MagicMock()
+
+    MOCK_MODULES = [
+        'pikepdf',
+        'pikepdf.models',
+        'pikepdf.models.metadata',
+        'ocrmypdf.leptonica',
+    ]
+    sys.modules.update((mod_name, Mock()) for mod_name in MOCK_MODULES)
+
+
+from pkg_resources import get_distribution, DistributionNotFound
 
-version = '.'.join(_version_parts[0:2])
 # The full version, including alpha/beta/rc tags.
-release = OCRMYPDF_VERSION
+release = get_distribution('ocrmypdf').version
+version = '.'.join(release.split('.')[:2])
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.
@@ -135,9 +157,7 @@ html_theme_path = [sphinx_rtd_theme.get_html_theme_path()]
 # further.  For a list of options available for each theme, see the
 # documentation.
 #
-html_theme_options = {
-    'display_version': False
-}
+html_theme_options = {'display_version': False}
 
 # Add any paths that contain custom themes here, relative to this directory.
 # html_theme_path = []
@@ -250,29 +270,25 @@ htmlhelp_basename = 'ocrmypdfdoc'
 # -- Options for LaTeX output ---------------------------------------------
 
 latex_elements = {
-     # The paper size ('letterpaper' or 'a4paper').
-     #
-     # 'papersize': 'letterpaper',
-
-     # The font size ('10pt', '11pt' or '12pt').
-     #
-     # 'pointsize': '10pt',
-
-     # Additional stuff for the LaTeX preamble.
-     #
-     # 'preamble': '',
-
-     # Latex figure (float) alignment
-     #
-     # 'figure_align': 'htbp',
+    # The paper size ('letterpaper' or 'a4paper').
+    #
+    # 'papersize': 'letterpaper',
+    # The font size ('10pt', '11pt' or '12pt').
+    #
+    # 'pointsize': '10pt',
+    # Additional stuff for the LaTeX preamble.
+    #
+    # 'preamble': '',
+    # Latex figure (float) alignment
+    #
+    # 'figure_align': 'htbp',
 }
 
 # Grouping the document tree into LaTeX files. List of tuples
 # (source start file, target name, title,
 #  author, documentclass [howto, manual, or own class]).
 latex_documents = [
-    (master_doc, 'ocrmypdf.tex', 'ocrmypdf Documentation',
-     'James R. Barlow', 'manual'),
+    (master_doc, 'ocrmypdf.tex', 'ocrmypdf Documentation', 'James R. Barlow', 'manual')
 ]
 
 # The name of an image file (relative to this directory) to place at the top of
@@ -312,10 +328,7 @@ latex_documents = [
 
 # One entry per manual page. List of tuples
 # (source start file, name, description, authors, manual section).
-man_pages = [
-    (master_doc, 'ocrmypdf', 'ocrmypdf Documentation',
-     [author], 1)
-]
+man_pages = [(master_doc, 'ocrmypdf', 'ocrmypdf Documentation', [author], 1)]
 
 # If true, show URL addresses after external links.
 #
@@ -328,9 +341,15 @@ man_pages = [
 # (source start file, target name, title, author,
 #  dir menu entry, description, category)
 texinfo_documents = [
-    (master_doc, 'ocrmypdf', 'ocrmypdf Documentation',
-     author, 'ocrmypdf', 'One line description of project.',
-     'Miscellaneous'),
+    (
+        master_doc,
+        'ocrmypdf',
+        'ocrmypdf Documentation',
+        author,
+        'ocrmypdf',
+        'One line description of project.',
+        'Miscellaneous',
+    )
 ]
 
 # Documents to append as an appendix to all manuals.

docs/cookbook.rst

@@ -1,11 +1,12 @@
+========
 Cookbook
 ========
 
 Basic examples
---------------
+==============
 
 Help!
-"""""
+-----
 
 ocrmypdf has built-in help.
 
@@ -13,30 +14,29 @@ ocrmypdf has built-in help.
 
     ocrmypdf --help
 
-
 Add an OCR layer and convert to PDF/A
-"""""""""""""""""""""""""""""""""""""
+-------------------------------------
 
 .. code-block:: bash
 
     ocrmypdf input.pdf output.pdf
 
 Add an OCR layer and output a standard PDF
-""""""""""""""""""""""""""""""""""""""""""
+------------------------------------------
 
 .. code-block:: bash
 
     ocrmypdf --output-type pdf input.pdf output.pdf
 
 Create a PDF/A with all color and grayscale images converted to JPEG
-""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
+--------------------------------------------------------------------
 
 .. code-block:: bash
 
     ocrmypdf --output-type pdfa --pdfa-image-compression jpeg input.pdf output.pdf
 
 Modify a file in place
-""""""""""""""""""""""
+----------------------
 
 The file will only be overwritten if OCRmyPDF is successful.
 
@@ -45,36 +45,45 @@ The file will only be overwritten if OCRmyPDF is successful.
     ocrmypdf myfile.pdf myfile.pdf
 
 Correct page rotation
-"""""""""""""""""""""
+---------------------
 
-OCR will attempt to automatic correct the rotation of each page. This can help fix a scanning job that contains a mix of landscape and portrait pages.
+OCR will attempt to automatic correct the rotation of each page. This
+can help fix a scanning job that contains a mix of landscape and
+portrait pages.
 
 .. code-block:: bash
 
     ocrmypdf --rotate-pages myfile.pdf myfile.pdf
 
-You can increase (decrease) the parameter ``--rotate-pages-threshold`` to make page rotation more (less) aggressive.
-
-If the page is "just a little off horizontal", like a crooked picture, then you want ``--deskew``. ``--rotate-pages`` is for when the cardinal angle is wrong.
+You can increase (decrease) the parameter ``--rotate-pages-threshold``
+to make page rotation more (less) aggressive.
 
+If the page is "just a little off horizontal", like a crooked picture,
+then you want ``--deskew``. ``--rotate-pages`` is for when the cardinal
+angle is wrong.
 
 OCR languages other than English
-""""""""""""""""""""""""""""""""
+--------------------------------
 
-By default OCRmyPDF assumes the document is English. 
+OCRmyPDF assumes the document is in English unless told otherwise. OCR
+quality may be poor if the wrong language is used.
 
 .. code-block:: bash
 
     ocrmypdf -l fra LeParisien.pdf LeParisien.pdf
     ocrmypdf -l eng+fra Bilingual-English-French.pdf Bilingual-English-French.pdf
 
-Language packs must be installed for all languages specified. See :ref:`Installing additional language packs <lang-packs>`.
+Language packs must be installed for all languages specified. See
+:ref:`Installing additional language packs <lang-packs>`.
 
+Unfortunately, the Tesseract OCR engine has no ability to detect the
+language when it is unknown.
 
 Produce PDF and text file containing OCR text
-"""""""""""""""""""""""""""""""""""""""""""""
+---------------------------------------------
 
-This produces a file named "output.pdf" and a companion text file named "output.txt".
+This produces a file named "output.pdf" and a companion text file named
+"output.txt".
 
 .. code-block:: bash
 
@@ -83,7 +92,11 @@ This produces a file named "output.pdf" and a companion text file named "output.
 OCR images, not PDFs
 --------------------
 
-If you are starting with images, you can just use Tesseract 3.04 or later directly to convert images to PDFs:
+Option: use Tesseract
+~~~~~~~~~~~~~~~~~~~~~
+
+If you are starting with images, you can just use Tesseract directly to
+convert images to PDFs:
 
 .. code-block:: bash
 
@@ -94,55 +107,88 @@ If you are starting with images, you can just use Tesseract 3.04 or later direct
     # When there are multiple images
     tesseract text-file-containing-list-of-image-filenames.txt output-prefix pdf
 
-Tesseract's PDF output is quite good – OCRmyPDF uses it by internally by default. However, OCRmyPDF has many features not available in Tesseract like like image processing, metadata control, and PDF/A generation.
+Tesseract's PDF output is quite good – OCRmyPDF uses it internally, in
+some cases. However, OCRmyPDF has many features not available in
+Tesseract like image processing, metadata control, and PDF/A generation.
 
-Use a program like `img2pdf <https://gitlab.mister-muffin.de/josch/img2pdf>`_ to convert your images to PDFs, and then pipe the results to run ocrmypdf.  The `-` tells ocrmypdf to read standard input.
+Option: use img2pdf
+~~~~~~~~~~~~~~~~~~~
+
+You can also use a program like
+`img2pdf <https://gitlab.mister-muffin.de/josch/img2pdf>`__ to convert
+your images to PDFs, and then pipe the results to run ocrmypdf. The
+``-`` tells ocrmypdf to read standard input.
 
 .. code-block:: bash
 
     img2pdf my-images*.jpg | ocrmypdf - myfile.pdf
 
-``img2pdf`` is recommended because it does an excellent job at generating PDFs without transcoding images.
+``img2pdf`` is recommended because it does an excellent job at
+generating PDFs without transcoding images.
 
-For convenience, OCRmyPDF can also convert single images to PDFs on its own. If the resolution (dots per inch, DPI) of an image is not set or is incorrect, it can be overridden with ``--image-dpi``. (As 1 inch is 2.54 cm, 1 dpi = 0.39 dpcm).
+Option: use OCRmyPDF (single images only)
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+For convenience, OCRmyPDF can also convert single images to PDFs on its
+own. If the resolution (dots per inch, DPI) of an image is not set or is
+incorrect, it can be overridden with ``--image-dpi``. (As 1 inch is 2.54
+cm, 1 dpi = 0.39 dpcm).
 
 .. code-block:: bash
 
     ocrmypdf --image-dpi 300 image.png myfile.pdf
 
-If you have multiple images, you must use ``img2pdf`` to convert the images to PDF.
+If you have multiple images, you must use ``img2pdf`` to convert the
+images to PDF.
 
-.. note::
-
-    ImageMagick ``convert`` can also convert a group of images to PDF, but in the author's experience it takes a long time, transcodes unnecessarily and gives poor results.
+Not recommended
+~~~~~~~~~~~~~~~
 
+We caution against using ImageMagick or Ghostscript to convert images to
+PDF, since they may transcode images or produce downsampled images,
+sometimes without warning.
 
 Image processing
-----------------
+================
 
-OCRmyPDF perform some image processing on each page of a PDF, if desired.  The same processing is applied to each page.  It is suggested that the user review files after image processing as these commands might remove desirable content, especially from poor quality scans.
+OCRmyPDF perform some image processing on each page of a PDF, if
+desired. The same processing is applied to each page. It is suggested
+that the user review files after image processing as these commands
+might remove desirable content, especially from poor quality scans.
 
-* ``--rotate-pages`` attempts to determine the correct orientation for each page and rotates the page if necessary.
-
-* ``--remove-background`` attempts to detect and remove a noisy background from grayscale or color images.  Monochrome images are ignored. This should not be used on documents that contain color photos as it may remove them.
-
-* ``--deskew`` will correct pages were scanned at a skewed angle by rotating them back into place.  Skew determination and correction is performed using `Postl's variance of line sums <http://www.leptonica.com/skew-measurement.html>`_ algorithm as implemented in `Leptonica <http://www.leptonica.com/index.html>`_.
-  
-* ``--clean`` uses `unpaper <https://www.flameeyes.eu/projects/unpaper>`_ to clean up pages before OCR, but does not alter the final output.  This makes it less likely that OCR will try to find text in background noise.
-
-* ``--clean-final`` uses unpaper to clean up pages before OCR and inserts the page into the final output.  You will want to review each page to ensure that unpaper did not remove something important.
+-  ``--rotate-pages`` attempts to determine the correct orientation for
+   each page and rotates the page if necessary.
+-  ``--remove-background`` attempts to detect and remove a noisy
+   background from grayscale or color images. Monochrome images are
+   ignored. This should not be used on documents that contain color
+   photos as it may remove them.
+-  ``--deskew`` will correct pages were scanned at a skewed angle by
+   rotating them back into place. Skew determination and correction is
+   performed using `Postl's variance of line
+   sums <http://www.leptonica.com/skew-measurement.html>`__ algorithm as
+   implemented in `Leptonica <http://www.leptonica.com/index.html>`__.
+-  ``--clean`` uses
+   `unpaper <https://www.flameeyes.eu/projects/unpaper>`__ to clean up
+   pages before OCR, but does not alter the final output. This makes it
+   less likely that OCR will try to find text in background noise.
+-  ``--clean-final`` uses unpaper to clean up pages before OCR and
+   inserts the page into the final output. You will want to review each
+   page to ensure that unpaper did not remove something important.
 
 .. note::
 
-    In many cases image processing will rasterize PDF pages as images, potentially losing quality.
+   In many cases image processing will rasterize PDF pages as images,
+   potentially losing quality.
 
 .. warning::
 
-    ``--clean-final`` and ``-remove-background`` may leave undesirable visual artifacts in some images where their algorithms have shortcomings. Files should be visually reviewed after using these options.
+   ``--clean-final`` and ``-remove-background`` may leave undesirable
+   visual artifacts in some images where their algorithms have
+   shortcomings. Files should be visually reviewed after using these
+   options.
 
-
-OCR and correct document skew (crooked scan)
-""""""""""""""""""""""""""""""""""""""""""""
+Example: OCR and correct document skew (crooked scan)
+-----------------------------------------------------
 
 Deskew:
 
@@ -150,51 +196,113 @@ Deskew:
 
     ocrmypdf --deskew input.pdf output.pdf
 
-Image processing commands can be combined. The order in which options are given does not matter. OCRmyPDF always applies the steps of the image processing pipeline in the same order (rotate, remove background, deskew, clean).
+Image processing commands can be combined. The order in which options
+are given does not matter. OCRmyPDF always applies the steps of the
+image processing pipeline in the same order (rotate, remove background,
+deskew, clean).
 
 .. code-block:: bash
 
     ocrmypdf --deskew --clean --rotate-pages input.pdf output.pdf
 
-
 Don't actually OCR my PDF
-"""""""""""""""""""""""""
+=========================
 
-If you set ``--tesseract-timeout 0`` OCRmyPDF will apply its image processing without performing OCR, if all you want to is to apply image processing or PDF/A conversion.
+If you set ``--tesseract-timeout 0`` OCRmyPDF will apply its image
+processing without performing OCR, if all you want to is to apply image
+processing or PDF/A conversion.
 
 .. code-block:: bash
 
     ocrmypdf --tesseract-timeout=0 --remove-background input.pdf output.pdf
 
+Redo existing OCR
+=================
 
-Redo OCR
-""""""""
+To redo OCR on a file OCRed with other OCR software or a previous
+version of OCRmyPDF and/or Tesseract, you may use the ``--redo-ocr``
+argument. (Normally, OCRmyPDF will exit with an error if asked to modify
+a file with OCR.)
 
-To redo OCR on a file OCRed with other OCR software or a previous version of OCRmyPDF and/or Tesseract, you may use the ``--force-ocr`` argument. Normally, OCRmyPDF does not modify files that already appear to contain OCR text.
+This may be helpful for users who want to take advantage of accuracy
+improvements in Tesseract 4.0 for files they previously OCRed with an
+earlier version of Tesseract and OCRmyPDF.
 
 .. code-block:: bash
 
-    ocrmypdf --force-ocr input.pdf output.pdf
+    ocrmypdf --redo-ocr input.pdf output.pdf
 
-Note that the method above will force rasterization of all pages, potentially reducing quality or losing vector content. 
+This method will replace OCR without rasterizing, reducing quality or
+removing vector content. If a file contains a mix of pure digital text
+and OCR, digital text will be ignored and OCR will be replaced. As such
+this mode is incompatible with image processing options, since they
+alter the appearance of the file.
 
-To ensure quality is preserved, one could extract all of the images and rebuild the PDF for a lossless transformation. This recipe does not work when PDFs contain multiple images per page, as many do in practice. It will also lose any page rotation information.
-
-.. code-block:: bash
-
-    pdfimages -all old-ocr.pdf prefix  # extract all images
-    img2pdf -o temp.pdf prefix*        # construct new PDF from the images
-    # review the new PDF to ensure it visually matches the old one
-    ocrmypdf --output-type pdf temp.pdf new-ocr.pdf
-
-``--output-type pdf`` is used here to avoid using Ghostscript which will also rasterize images.
+In some cases, existing OCR cannot be detected or replaced. Files
+produced by OCRmyPDF v2.2 or earlier, for example, are internally
+represented as having visible text with an opaque image drawn on top.
+This situation cannot be detected.
 
+If ``--redo-ocr`` does not work, you can use ``--force-ocr``, which will
+force rasterization of all pages, potentially reducing quality or losing
+vector content.
 
 Improving OCR quality
----------------------
+=====================
 
-The `Image processing`_ features can improve OCR quality.
+The `Image processing <#image-processing>`__ features can improve OCR
+quality.
 
-Rotating pages and deskewing helps to ensure that the page orientation is correct before OCR begins. Removing the background and/or cleaning the page can also improve results. The ``--oversample DPI`` argument can be specified to resample images to higher resolution before attempting OCR; this can improve results as well.
+Rotating pages and deskewing helps to ensure that the page orientation
+is correct before OCR begins. Removing the background and/or cleaning
+the page can also improve results. The ``--oversample DPI`` argument can
+be specified to resample images to higher resolution before attempting
+OCR; this can improve results as well.
 
-OCR quality will suffer if the resolution of input images is not correct (since the range of pixel sizes that will be checked for possible fonts will also be incorrect).
+OCR quality will suffer if the resolution of input images is not correct
+(since the range of pixel sizes that will be checked for possible fonts
+will also be incorrect).
+
+PDF optimization
+================
+
+By default OCRmyPDF will attempt to perform lossless optimizations on
+the images inside PDFs after OCR is complete. Optimization is performed
+even if no OCR text is found.
+
+The ``--optimize N`` (short form ``-O``) argument controls optimization,
+where ``N`` ranges from 0 to 3 inclusive, analogous to the optimization
+levels in the GCC compiler.
+
+.. list-table::
+    :widths: auto
+    :header-rows: 1
+
+    *   - Level
+        - Comments
+    *   - ``--optimize 0``
+        - Disables optimization.
+    *   - ``--optimize 1``
+        - Enables lossless optimizations, such as transcoding images to more
+          efficient formats. Also compress other uncompressed objects in the
+          PDF and enables the more efficient "object streams" within the PDF.
+    *   - ``--optimize 2``
+        - All of the above, and enables lossy optimizations and color quantization.
+    *   - ``--optimize 3``
+        - All of the above, and enables more aggressive optimizations and targets lower image quality.
+
+Optimization is improved when a JBIG2 encoder is available and when
+``pngquant`` is installed. If either of these components are missing,
+then some types of images cannot be optimized.
+
+The types of optimization available may expand over time. By default,
+OCRmyPDF compresses data streams inside PDFs, and will change
+inefficient compression modes to more modern versions. A program like
+``qpdf`` can be used to change encodings, e.g. to inspect the internals
+fo a PDF.
+
+.. code-block:: bash
+
+    ocrmypdf --optimize 3 in.pdf out.pdf  # Make it small
+
+Some users may consider enabling lossy JBIG2. See: :ref:`jbig2-lossy`.

docs/docker.rst

@@ -0,0 +1,175 @@
+=====================
+OCRmyPDF Docker image
+=====================
+
+OCRmyPDF is also available in a Docker image that packages recent
+versions of all dependencies.
+
+For users who already have Docker installed this may be an easy and
+convenient option. However, it is less performant than a system
+installation and may require Docker engine configuration.
+
+OCRmyPDF needs a generous amount of RAM, CPU cores, temporary storage
+space, whether running in a Docker container or on its own. It may be
+necessary to ensure the container is provisioned with additional
+resources.
+
+.. _docker-install:
+
+Installing the Docker image
+===========================
+
+If you have `Docker <https://docs.docker.com/>`__ installed on your
+system, you can install a Docker image of the latest release.
+
+The recommended OCRmyPDF Docker image is currently named
+``ocrmypdf-alpine``:
+
+.. code-block:: bash
+
+   docker pull jbarlow83/ocrmypdf-alpine
+
+Follow the Docker installation instructions for your platform. If you
+can run this command successfully, your system is ready to download and
+execute the image:
+
+.. code-block:: bash
+
+   docker run hello-world
+
+OCRmyPDF will use all available CPU cores. By default, the VirtualBox
+machine instance on Windows and macOS has only a single CPU core
+enabled. Use the VirtualBox Manager to determine the name of your Docker
+engine host, and then follow these optional steps to enable multiple
+CPUs:
+
+.. code-block:: bash
+
+   # Optional step for Mac OS X users
+   docker-machine stop "yourVM"
+   VBoxManage modifyvm "yourVM" --cpus 2  # or whatever number of core is desired
+   docker-machine start "yourVM"
+   eval $(docker-machine env "yourVM")
+
+Using the Docker image on the command line
+==========================================
+
+**Unlike typical Docker containers**, in this mode we are using the
+OCRmyPDF Docker container is intended to be emphemeral – it runs for one
+OCR job and then terminates, just like a command line program. We are
+using Docker as a way of delivering an application, not a server.
+
+To start a Docker container (instance of the image):
+
+.. code-block:: bash
+
+   docker tag jbarlow83/ocrmypdf-alpine ocrmypdf
+   docker run --rm -i ocrmypdf (... all other arguments here...)
+
+For convenience, create a shell alias to hide the Docker command. It is
+easier to send the input file to file stdin and read the output from
+stdout – this avoids the occasionally messy permission issues with
+Docker entirely.
+
+.. code-block:: bash
+
+   alias ocrmypdf='docker run --rm -i ocrmypdf'
+   ocrmypdf --version  # runs docker version
+   ocrmypdf <input.pdf >output.pdf
+
+Or in the wonderful `fish shell <https://fishshell.com/>`__:
+
+.. code-block:: fish
+
+   alias ocrmypdf 'docker run --rm ocrmypdf'
+   funcsave ocrmypdf
+
+Alternately, you could mount the local current working directory as a
+Docker volume:
+
+.. code-block:: bash
+
+   docker run --rm -v $(pwd):/data ocrmypdf /data/input.pdf /data/output.pdf
+
+.. _docker-lang-packs:
+
+Adding languages to the Docker image
+====================================
+
+By default the Docker image includes English, German and Simplified
+Chinese, the most popular languages for OCRmyPDF users based on
+feedback. You may add other languages by creating a new Dockerfile based
+on the public one:
+
+.. code-block:: dockerfile
+
+   FROM jbarlow83/ocrmypdf-alpine
+
+   # Add French
+   RUN apk add tesseract-ocr-data-fra
+
+You can also copy training data to ``/usr/share/tessdata``.
+
+Executing the test suite
+========================
+
+The OCRmyPDF test suite is installed with image. To run it:
+
+.. code-block:: bash
+
+   docker run --entrypoint python3 jbarlow83/ocrmypdf-alpine setup.py test
+
+Accessing the shell
+===================
+
+``bash`` is not installed in the image. To use the busybox shell in the
+Docker image:
+
+.. code-block:: bash
+
+   docker run -it --entrypoint busybox  jbarlow83/ocrmypdf-alpine sh
+
+Using the OCRmyPDF web service wrapper
+======================================
+
+The OCRmyPDF Docker image includes an example, barebones HTTP web
+service. The webservice may be launched as follows:
+
+.. code-block:: bash
+
+   docker run --entrypoint python3 -p 5000:5000 jbarlow83/ocrmypdf-alpine webservice.py
+
+Unlike command line usage this program will open a socket and wait for
+connections.
+
+.. warning::
+
+   The OCRmyPDF web service wrapper is intended for demonstration or
+   development. It provides no security, no authentication, no
+   protection against denial of service attacks, and no load balancing.
+   The default Flask WSGI server is used, which is intended for
+   development only. The server is single-threaded and so can respond to
+   only one client at a time. While running OCR, it cannot respond to
+   any other clients.
+
+Clients must keep their open connection while waiting for OCR to
+complete. This may entail setting a long timeout; this interface is more
+useful for internal HTTP API calls.
+
+Unlike the rest of OCRmyPDF, this web service is licensed under the
+Affero GPLv3 (AGPLv3) since Ghostscript, a dependency of OCRmyPDF, is
+also licensed in this way.
+
+In addition to the above, please read our
+:ref:`general remarks on using OCRmyPDF as a service <ocr-service>`.
+
+Ubuntu-based Docker image
+=========================
+
+A Ubuntu-based OCRmyPDF image is also available. The main advantage this
+image offers is that it supports manylinux Python wheels (which are not
+supported on Alpine Linux). This may be useful for plugins.
+
+.. code-block:: bash
+
+   docker pull jbarlow83/ocrmypdf

docs/errors.rst

@@ -1,33 +1,47 @@
+=====================
 Common error messages
 =====================
 
 Page already has text
----------------------
+=====================
 
-.. code::
+.. code-block::
 
-	ERROR -    1: page already has text! – aborting (use --force-ocr to force OCR)
+   ERROR -    1: page already has text! – aborting (use --force-ocr to force OCR)
 
-You ran ocrmypdf on a file that already contains printable text or a hidden OCR text layer (it can't quite tell the difference). You probably don't want to do this, because the file is already searchable.
+You ran ocrmypdf on a file that already contains printable text or a
+hidden OCR text layer (it can't quite tell the difference). You probably
+don't want to do this, because the file is already searchable.
 
 As the error message suggests, your options are:
 
-- ``ocrmypdf --force-ocr`` to :ref:`rasterize <raster-vector>` all vector content and run OCR on the images. This is useful if a previous OCR program failed, or if the document contains a text watermark.
-
-- ``ocrmypdf --skip-text`` to skip OCR and other processing on any pages that contain text. Text pages will be copied into the output PDF without modification.
-
+-  ``ocrmypdf --force-ocr`` to :ref:`rasterize <raster-vector>` all
+   vector content and run OCR on the images. This is useful if a
+   previous OCR program failed, or if the document contains a text
+   watermark.
+-  ``ocrmypdf --skip-text`` to skip OCR and other processing on any
+   pages that contain text. Text pages will be copied into the output
+   PDF without modification.
 
 Input file 'filename' is not a valid PDF
-----------------------------------------
+========================================
 
-OCRmyPDF passes files through qpdf, a program that fixes errors in PDFs, before it tries to work on them. In most cases this happens because the PDF is corrupt and
-truncated (incomplete file copying) and not much can be done.
+OCRmyPDF passes files through qpdf, a program that fixes errors in PDFs,
+before it tries to work on them. In most cases this happens because the
+PDF is corrupt and truncated (incomplete file copying) and not much can
+be done.
 
-You can try rewriting the file with Ghostscript or pdftk:
+You can try rewriting the file with Ghostscript:
 
-- ``gs -o output.pdf -dSAFER -sDEVICE=pdfwrite input.pdf``
+.. code-block:: bash
 
-- ``pdftk input.pdf cat output output.pdf``
+    gs -o output.pdf -dSAFER -sDEVICE=pdfwrite input.pdf
 
-Sometimes Acrobat can repair PDFs with its `Preflight tool <https://helpx.adobe.com/acrobat/using/correcting-problem-areas-preflight-tool.html>`_.
+``pdftk`` can also rewrite PDFs:
 
+.. code-block:: bash
+
+    pdftk input.pdf cat output output.pdf
+
+Sometimes Acrobat can repair PDFs with its `Preflight
+tool <https://helpx.adobe.com/acrobat/using/correcting-problem-areas-preflight-tool.html>`__.

docs/index.rst

@@ -18,13 +18,17 @@ PDF is the best format for storing and exchanging scanned documents.  Unfortunat
    release_notes
    installation
    languages
+   jbig2
 
 .. toctree::
    :caption: Usage
    :maxdepth: 2
 
    cookbook
+   docker
    advanced
+   api
+   plugins
    batch
    security
    errors
@@ -36,4 +40,3 @@ Indices and tables
 * :ref:`genindex`
 * :ref:`modindex`
 * :ref:`search`
-

docs/installation.rst

@@ -1,155 +1,353 @@
-Installation
-============
+===================
+Installing OCRmyPDF
+===================
 
-The easiest way to install OCRmyPDF to follow the steps for your operating system/platform.
+.. |latest| image:: https://img.shields.io/pypi/v/ocrmypdf.svg
+    :alt: OCRmyPDF latest released version on PyPI
 
-If you want to use the latest version of OCRmyPDF, your best bet is to install the most recent version your platform provides, and then upgrade that version by installing the Python binary wheels.
+|latest|
+
+The easiest way to install OCRmyPDF is to follow the steps for your operating
+system/platform, although sometimes this version may be out of date.
+
+If you want to use the latest version of OCRmyPDF, your best bet is to install
+the most recent version your platform provides, and then upgrade that version by
+installing the Python binary wheels.
 
 .. contents:: Platform-specific steps
-    :depth: 1
+    :depth: 2
     :local:
 
-Installing on Debian and Ubuntu 16.10 or newer
-----------------------------------------------
+Installing on Linux
+===================
 
-Users of Debian 9 ("stretch") or later or Ubuntu 16.10 or later may simply
+Debian and Ubuntu 16.10 or newer
+--------------------------------
+
+.. |deb-stable| image:: https://repology.org/badge/version-for-repo/debian_stable/ocrmypdf.svg
+    :alt: Debian 9 stable ("stretch")
+
+.. |deb-testing| image:: https://repology.org/badge/version-for-repo/debian_testing/ocrmypdf.svg
+    :alt: Debian 10 testing ("buster")
+
+.. |deb-unstable| image:: https://repology.org/badge/version-for-repo/debian_unstable/ocrmypdf.svg
+    :alt: Debian unstable
+
+.. |ubu-1710| image:: https://repology.org/badge/version-for-repo/ubuntu_17_10/ocrmypdf.svg
+    :alt: Ubuntu 17.10
+
+.. |ubu-1804| image:: https://repology.org/badge/version-for-repo/ubuntu_18_04/ocrmypdf.svg
+    :alt: Ubuntu 18.04 LTS
+
+.. |ubu-1810| image:: https://repology.org/badge/version-for-repo/ubuntu_18_10/ocrmypdf.svg
+    :alt: Ubuntu 18.10
+
+
++-------------------------------------------+
+| **OCRmyPDF versions in Debian & Ubuntu**  |
++-------------------------------------------+
+| |latest|                                  |
++-------------------------------------------+
+| |deb-stable| |deb-testing| |deb-unstable| |
++-------------------------------------------+
+| |ubu-1710| |ubu-1804| |ubu-1810|          |
++-------------------------------------------+
+
+Users of Debian 9 ("stretch") or later or Ubuntu 16.10 or later may
+simply
 
 .. code-block:: bash
 
     apt-get install ocrmypdf
 
-To see what versions are available, check the `Debian Package Tracker <https://tracker.debian.org/pkg/ocrmypdf>`_ or `Ubuntu launchpad.net <https://launchpad.net/ocrmypdf>`_.
+As indicated in the table above, Debian and Ubuntu releases may lag
+behind the latest version. If the version available for your platform is
+out of date, you could opt to install the latest version from source.
+See `Installing HEAD revision from
+sources <#installing-head-revision-from-sources>`__.
 
-Installing on macOS with Homebrew
----------------------------------
+For full details on version availability for your platform, check the
+`Debian Package Tracker <https://tracker.debian.org/pkg/ocrmypdf>`__ or
+`Ubuntu launchpad.net <https://launchpad.net/ocrmypdf>`__.
+
+.. note::
+
+   OCRmyPDF for Debian and Ubuntu currently omit the JBIG2 encoder.
+   OCRmyPDF works fine without it but will produce larger output files.
+   If you build jbig2enc from source, ocrmypdf 7.0.0 and later will
+   automatically detect it (specifically the ``jbig2`` binary) on the
+   ``PATH``. To add JBIG2 encoding, see :ref:`jbig2`.
+
+Fedora 29 or newer
+------------------
+
+.. |fedora-29| image:: https://repology.org/badge/version-for-repo/fedora29/ocrmypdf.svg
+    :alt: Fedora 29
+
+.. |fedora-rawhide| image:: https://repology.org/badge/version-for-repo/fedora_rawhide/ocrmypdf.svg
+    :alt: Fedore Rawhide
+
+
++------------------------------+
+| **OCRmyPDF version**         |
++------------------------------+
+| |latest|                     |
++------------------------------+
+| |fedora-29| |fedora-rawhide| |
++------------------------------+
+
+Users of Fedora 29 later may simply
+
+.. code-block:: bash
+
+    dnf install ocrmypdf
+
+For full details on version availability, check the `Fedora Package
+Tracker <https://apps.fedoraproject.org/packages/ocrmypdf>`__.
+
+If the version available for your platform is out of date, you could opt
+to install the latest version from source. See `Installing HEAD revision
+from sources <#installing-head-revision-from-sources>`__.
+
+.. note::
+
+   OCRmyPDF for Fedora currently omits the JBIG2 encoder due to patent
+   issues. OCRmyPDF works fine without it but will produce larger output
+   files. If you build jbig2enc from source, ocrmypdf 7.0.0 and later
+   will automatically detect it on the ``PATH``. To add JBIG2 encoding,
+   see `Installing the JBIG2 encoder <jbig2>`__.
+
+Installing the latest version on Ubuntu 18.04 LTS
+-------------------------------------------------
+
+Ubuntu 18.04 includes ocrmypdf 6.1.2. To install a more recent version,
+first install the system version to get most of the dependencies:
+
+.. code-block:: bash
+
+    sudo apt-get update
+    sudo apt-get install \
+        ocrmypdf \
+        python3-pip
+
+There are a few system dependency changes since ocrmypdf 6.1.2. Let's
+get these, too.
+
+.. code-block:: bash
+
+    sudo apt-get install \
+        libxml2 \
+        pngquant
+
+Then install the most recent ocrmypdf for the local user and set the
+user's ``PATH`` to check for the user's Python packages.
+
+.. code-block:: bash
+
+    export PATH=$HOME/.local/bin:$PATH
+    pip3 install --user ocrmypdf
+
+To add JBIG2 encoding, see :ref:`jbig2`.
+
+Ubuntu 16.04 LTS
+----------------
+
+No package is available for Ubuntu 16.04. OCRmyPDF 8.0 and newer require
+Python 3.6. Ubuntu 16.04 ships Python 3.5, but you can install Python
+3.6 on it. Or, you can skip Python 3.6 and install OCRmyPDF 7.x or older
+- for that procedure, please see the installation documentation for the
+version of OCRmyPDF you plan to use.
+
+**Install system packages for OCRmyPDF**
+
+.. code-block:: bash
+
+    sudo apt-get update
+    sudo apt-get install -y software-properties-common python-software-properties
+    sudo add-apt-repository -y \
+        ppa:jonathonf/python-3.6 \
+        ppa:alex-p/tesseract-ocr
+    sudo apt-get update
+    sudo apt-get install -y \
+        ghostscript \
+        libexempi3 \
+        libffi6 \
+        pngquant \
+        python3.6 \
+        qpdf \
+        tesseract-ocr \
+        unpaper
+
+This will install a Python 3.6 binary at ``/usr/bin/python3.6``
+alongside the system's Python 3.5. Do not remove the system Python. This
+will also install Tesseract 4.0 from a PPA, since the version available
+in Ubuntu 16.04 is too old for OCRmyPDF.
+
+Now install pip for Python 3.6. This will install the Python 3.6 version
+of ``pip`` at ``/usr/local/bin/pip``.
+
+.. code-block:: bash
+
+    curl https://bootstrap.pypa.io/get-pip.py | sudo python3.6
+
+**Install OCRmyPDF**
+
+OCRmyPDF requires the locale to be set for UTF-8. **On some minimal
+Ubuntu installations systems**, it may be necessary to set the locale.
+
+.. code-block:: bash
+
+    # Optional: Only need to set these if they are not already set
+    export LC_ALL=C.UTF-8
+    export LANG=C.UTF-8
+
+Now install OCRmyPDF for the current user, and ensure that the ``PATH``
+environment variable contains ``$HOME/.local/bin``.
+
+.. code-block:: bash
+
+    export PATH=$HOME/.local/bin:$PATH
+    pip3 install --user ocrmypdf
+
+To add JBIG2 encoding, see :ref:`jbig2`.
+
+Ubuntu 14.04 LTS
+----------------
+
+Installing on Ubuntu 14.04 LTS (trusty) is more difficult than some
+other options, because of its age. Several backports are required. For
+explanations of some steps of this procedure, see the similar steps for
+Ubuntu 16.04.
+
+Install system dependencies:
+
+.. code-block:: bash
+
+    sudo apt-get update
+    sudo apt-get install \
+        software-properties-common python-software-properties \
+        zlib1g-dev \
+        libexempi3 \
+        libjpeg-dev \
+        libffi-dev \
+        pngquant \
+        qpdf
+
+We will need backports of Ghostscript 9.16, libav-11 (for unpaper 6.1),
+Tesseract 4.00 (alpha), and Python 3.6. This will replace Ghostscript
+and Tesseract 3.x on your system. Python 3.6 will be installed alongside
+the system Python 3.4.
+
+If you prefer to not modify your system in this matter, consider using a
+Docker container.
+
+.. code-block:: bash
+
+    sudo add-apt-repository ppa:vshn/ghostscript -y
+    sudo add-apt-repository ppa:heyarje/libav-11 -y
+    sudo add-apt-repository ppa:alex-p/tesseract-ocr -y
+    sudo add-apt-repository ppa:jonathonf/python-3.6 -y
+
+    sudo apt-get update
+
+    sudo apt-get install \
+        python3.6-dev \
+        ghostscript \
+        tesseract-ocr \
+        tesseract-ocr-eng \
+        libavformat56 libavcodec56 libavutil54 \
+        wget
+
+Now we need to install ``pip`` and let it install ocrmypdf:
+
+.. code-block:: bash
+
+    curl https://bootstrap.pypa.io/ez_setup.py -o - | python3.6 && python3.6 -m easy_install pip
+    pip3.6 install ocrmypdf
+
+These installation instructions omit the optional dependency
+``unpaper``, which is only available at version 0.4.2 in Ubuntu 14.04.
+The author could not find a backport of ``unpaper``, and created a .deb
+package to do the job of installing unpaper 6.1 (for x86 64-bit only):
+
+.. code-block:: bash
+
+    wget -q 'https://www.dropbox.com/s/vaq0kbwi6e6au80/unpaper_6.1-1.deb?raw=1' -O unpaper_6.1-1.deb
+    sudo dpkg -i unpaper_6.1-1.deb
+
+To add JBIG2 encoding, see :ref:`jbig2`.
+
+ArchLinux (AUR)
+---------------
+
+.. image:: https://repology.org/badge/version-for-repo/aur/ocrmypdf.svg
+    :alt: ArchLinux
+    :target: https://repology.org/metapackage/ocrmypdf
+
+There is an `ArchLinux User Repository package for
+ocrmypdf <https://aur.archlinux.org/packages/ocrmypdf/>`__. You can use
+the following command.
+
+.. code-block:: bash
+
+    yaourt -S ocrmypdf
+
+If you have any difficulties with installation, check the repository
+package page.
+
+Other Linux packages
+--------------------
+
+See the
+`Repology <https://repology.org/metapackage/ocrmypdf/versions>`__ page.
+
+In general, first install the OCRmyPDF package for your system, then
+optionally use the procedure `Installing with Python
+pip <#installing-with-python-pip>`__ to install a more recent version.
+
+Installing on macOS
+===================
+
+Homebrew
+--------
 
 .. image:: https://img.shields.io/homebrew/v/ocrmypdf.svg
     :alt: homebrew
     :target: http://brewformulas.org/Ocrmypdf
 
-OCRmyPDF is now a standard `Homebrew <https://brew.sh>`_ formula. To install on macOS:
+OCRmyPDF is now a standard `Homebrew <https://brew.sh>`__ formula. To
+install on macOS:
 
 .. code-block:: bash
 
     brew install ocrmypdf
 
-.. note::
+This will include only the English language pack. If you need other
+languages you can optionally install them all:
 
-    Users who previously installed OCRmyPDF on macOS using ``pip install ocrmypdf`` should remove the pip version (``pip3 uninstall ocrmypdf``) before switching to the Homebrew version.
+.. code-block:: bash
+
+    brew install tesseract-lang  # Optional: Install all language packs
 
 .. note::
 
-    Users who previously installed OCRmyPDF from the private tap should switch to the mainline version (``brew untap jbarlow83/ocrmypdf``) and install from there.
-
-.. _Docker-install:
-
-Installing the Docker image
----------------------------
-
-For many users, installing the Docker image will be easier than installing all of OCRmyPDF's dependencies. For Windows, it is the only option.
-
-If you have `Docker <https://docs.docker.com/>`_ installed on your system, you can install a Docker image of the latest release.
-
-Follow the Docker installation instructions for your platform.  If you can run this command successfully, your system is ready to download and execute the image:
-
-.. code-block:: bash
-
-    docker run hello-world
-
-OCRmyPDF will use all available CPU cores.  By default, the VirtualBox machine instance on Windows and macOS has only a single CPU core enabled. Use the VirtualBox Manager to determine the name of your Docker engine host, and then follow these optional steps to enable multiple CPUs:
-
-.. code-block:: bash
-
-    # Optional step for Mac OS X users
-    docker-machine stop "yourVM"
-    VBoxManage modifyvm "yourVM" --cpus 2  # or whatever number of core is desired
-    docker-machine start "yourVM"
-    eval $(docker-machine env "yourVM")
-
-Assuming you have a Docker engine running, you can download one of the three available images:
-
-.. list-table::
-    :widths: auto
-    :header-rows: 1
-
-    *   - Image name
-        - Download command
-        - Notes
-    *   - ocrmypdf
-        - ``docker pull jbarlow83/ocrmypdf``
-        - Latest ocrmypdf with Tesseract 4.0.0-beta1 on Ubuntu 18.04. Includes English, French, German, Spanish, Portugeuse and Simplified Chinese.
-    *   - ocrmypdf-polyglot
-        - ``docker pull jbarlow83/ocrmypdf-polyglot``
-        - As above, with all available language packs.
-
-For example:
-
-.. code-block:: bash
-
-    docker pull jbarlow83/ocrmypdf
-
-Then tag it to give a more convenient name, just ocrmypdf:
-
-.. code-block:: bash
-
-    docker tag jbarlow83/ocrmypdf ocrmypdf
-
-.. _docker-polyglot:
-
-The alternative "polyglot" image provides `all available language packs <https://github.com/tesseract-ocr/tesseract/blob/master/doc/tesseract.1.asc#languages>`_.
-
-You can then run ocrmypdf using the command:
-
-.. code-block:: bash
-
-    docker run --rm ocrmypdf --help
-
-To execute the OCRmyPDF on a local file, you must `provide a writable volume to the Docker image <https://docs.docker.com/userguide/dockervolumes/>`_, and both the input and output file must be inside the writable volume.  This example command uses the current working directory as the writable volume:
-
-.. code-block:: bash
-
-    docker run --rm -v "$(pwd):/home/docker" <other docker arguments>   ocrmypdf <your arguments to ocrmypdf>
-
-In this worked example, the current working directory contains an input file called ``test.pdf`` and the output will go to ``output.pdf``:
-
-.. code-block:: bash
-
-    docker run --rm -v "$(pwd):/home/docker"   ocrmypdf --skip-text test.pdf output.pdf
-
-.. note:: The working directory should be a writable local volume or Docker may not have permission to access it.
-
-Note that ``ocrmypdf`` has its own separate ``-v VERBOSITYLEVEL`` argument to control debug verbosity. All Docker arguments should before the ``ocrmypdf`` image name and all arguments to ``ocrmypdf`` should be listed after.
-
-In some environments the permissions associated with Docker can be complex to configure. The process that executes Docker may end up not having the permissions to write the specified file system. In that case one can stream the file into and out of the Docker process and avoid all permission hassles, using ``-`` as the input and output filename:
-
-.. code-block:: bash
-
-    docker run --rm -i   ocrmypdf <other arguments to ocrmypdf> - - <input.pdf >output.pdf
-
-For convenience, a shell alias can hide the docker command:
-
-.. code-block:: bash
-
-    alias ocrmypdf='docker run --rm -v "$(pwd):/home/docker" ocrmypdf'
-    ocrmypdf --version  # runs docker version
-
-Or in the wonderful `fish shell <https://fishshell.com/>`_:
-
-.. code-block:: fish
-
-    alias ocrmypdf 'docker run --rm -v (pwd):/home/docker ocrmypdf'
-    funcsave ocrmypdf
+   Users who previously installed OCRmyPDF on macOS using
+   ``pip install ocrmypdf`` should remove the pip version
+   (``pip3 uninstall ocrmypdf``) before switching to the Homebrew
+   version.
 
 .. note::
 
-    The ocrmypdf Docker containers are designed to be used for a single OCR job. The ``docker run --rm`` argument tells Docker to delete temporary storage associated with container when it is done executing.
+   Users who previously installed OCRmyPDF from the private tap should
+   switch to the mainline version (``brew untap jbarlow83/ocrmypdf``)
+   and install from there.
 
 Manual installation on macOS
 ----------------------------
 
 These instructions probably work on all macOS supported by Homebrew.
 
-If it's not already present, `install Homebrew <http://brew.sh/>`_.
+If it's not already present, `install Homebrew <http://brew.sh/>`__.
 
 Update Homebrew:
 
@@ -157,28 +355,22 @@ Update Homebrew:
 
     brew update
 
-Install or upgrade the required Homebrew packages, if any are missing:
+Install or upgrade the required Homebrew packages, if any are missing.
+To do this, download the ``Brewfile`` that lists all of the dependencies
+to the current directory, and run ``brew bundle`` to process them
+(installing or upgrading as needed). ``Brewfile`` is a plain text file.
 
 .. code-block:: bash
 
-    brew install libpng openjpeg jbig2dec libtiff     # image libraries
-    brew install qpdf
-    brew install ghostscript
-    brew install python3
-    brew install libxml2 libffi leptonica
-    brew install unpaper   # optional
+    wget https://github.com/jbarlow83/OCRmyPDF/raw/master/.travis/Brewfile
+    brew bundle
 
-Python 3.5, 3.6 and 3.7 are supported.
-
-Install the required Tesseract OCR engine with the language packs you plan to use:
-
-.. code-block:: bash
-
-    brew install tesseract                       # Option 1: for English, French, German, Spanish
+This will include the English, French, German and Spanish language
+packs. If you need other languages you can optionally install them all:
 
 .. _macos-all-languages:
 
-.. code-block:: bash
+   .. code-block:: bash
 
     brew install tesseract --with-all-languages  # Option 2: for all language packs
 
@@ -206,218 +398,148 @@ The command line program should now be available:
 
     ocrmypdf --help
 
-Installing the latest version on Ubuntu 18.04 LTS
--------------------------------------------------
+Installing on FreeBSD
+=====================
 
-Ubuntu 18.04 includes ocrmypdf 6.1.2. To install a more recent version, first
-install the system version to get all the dependencies:
+FreeBSD 11.2 is known to work. Other versions likely work but have not
+been tested.
 
-.. code-block:: bash
+In general it should work to:
 
-    sudo apt-get update
-    sudo apt-get install \
-        ocrmypdf \
-        python3-pip
+#. `Install and build
+   pikepdf <https://pikepdf.readthedocs.io/en/latest/installation.html#installing-on-freebsd-11-2>`__.
+#. Install the equivalent list of dependencies for Linux.
 
-Then install the most recent ocrmypdf for the local user and set the user's ``PATH`` to check for the user's Python packages.
+Installing the Docker image
+===========================
 
-.. code-block:: bash
-
-    export PATH=$HOME/.local/bin:$PATH
-    pip3 install --user ocrmypdf
-
-
-Installing on Ubuntu 16.04 LTS
-------------------------------
-
-No package is currently available for Ubuntu 16.04, but you can install the dependencies manually:
-
-.. code-block:: bash
-
-    sudo apt-get update
-    sudo apt-get install \
-        unpaper \
-        ghostscript \
-        tesseract-ocr \
-        qpdf \
-        python3-pip \
-        python3-cffi
-
-If you wish install OCRmyPDF for the current user:
-
-.. code-block:: bash
-
-    pip3 install --user ocrmypdf
-
-Alternately, system-wide. Note that this may modify the system Python environment:
-
-.. code-block:: bash
-
-    sudo pip3 install ocrmypdf
-
-If you wish to install OCRmyPDF to a virtual environment to isolate the system Python, you can follow these steps.
-
-.. code-block:: bash
-
-    python3 -m venv venv-ocrmypdf
-    source venv-ocrmypdf/bin/activate
-    pip3 install ocrmypdf
-
-Installing on Ubuntu 14.04 LTS
-------------------------------
-
-Installing on Ubuntu 14.04 LTS (trusty) is more difficult than some other options, because it is older and does not provide ``pip``.
-
-Update apt-get:
-
-.. code-block:: bash
-
-    sudo apt-get update
-
-Install system dependencies:
-
-.. code-block:: bash
-
-    sudo apt-get install \
-        software-properties-common python-software-properties \
-        zlib1g-dev \
-        libjpeg-dev \
-        libffi-dev \
-        qpdf
-
-We will need backports of Ghostscript 9.16, libav-11 (for unpaper 6.1), Tesseract 4.00 (alpha), and Python 3.6. This will replace Ghostscript and Tesseract 3.x on your system. Python 3.6 will be installed alongside the system Python 3.
-
-If you prefer to not modify your system in this matter, consider using a Docker container.
-
-.. code-block:: bash
-
-    sudo add-apt-repository ppa:vshn/ghostscript -y
-    sudo add-apt-repository ppa:heyarje/libav-11 -y
-    sudo add-apt-repository ppa:alex-p/tesseract-ocr -y
-    sudo add-apt-repository ppa:jonathonf/python-3.6 -y
-
-    sudo apt-get update
-
-    sudo apt-get install \
-        python3.6-dev \
-        ghostscript \
-        tesseract-ocr \
-        tesseract-ocr-eng \
-        libavformat56 libavcodec56 libavutil54 \
-        wget
-
-Now we need to install ``pip`` and let it install ocrmypdf:
-
-.. code-block:: bash
-
-    curl https://bootstrap.pypa.io/ez_setup.py -o - | python3.6 && python3.6 -m easy_install pip
-    pip3.6 install ocrmypdf
-
-The ``wget`` command will download a program and run it.
-
-These installation instructions omit the optional dependency ``unpaper``, which is only available at version 0.4.2 in Ubuntu 14.04. The author could not find a backport of ``unpaper``, and created a .deb package to do the job of installing unpaper 6.1 (for x86 64-bit only):
-
-.. code-block:: bash
-
-    wget -q 'https://www.dropbox.com/s/vaq0kbwi6e6au80/unpaper_6.1-1.deb?raw=1' -O unpaper_6.1-1.deb
-    sudo dpkg -i unpaper_6.1-1.deb
-
-
-Installing on ArchLinux
------------------------
-
-The author is aware of an `ArchLinux package for ocrmypdf <https://aur.archlinux.org/packages/ocrmypdf/>`_. It seems like the following command might work.
-
-.. code-block:: bash
-
-    pacman -S ocrmypdf
+For some users, installing the Docker image will be easier than
+installing all of OCRmyPDF's dependencies. For Windows, it is the only
+option.
 
+See `OCRmyPDF Docker Image <docker>`__ for more information.
 
 Installing on Windows
----------------------
+=====================
 
-Direct installation on Windows is not possible.  Install the _`Docker` container as described above.  Ensure that your command prompt can run the docker "hello world" container.
-
-It would probably not be too difficult to run on Windows.  The main reason this has been avoided is the difficulty of packaging and installing the various non-Python dependencies: Tesseract, QPDF, Ghostscript, Leptonica.  Pull requests to add or improve Windows support would be quite welcome.
-
-
-Running on Windows
-~~~~~~~~~~~~~~~~~~
-
-The command line syntax to run ocrmypdf from a command prompt will resemble:
-
-.. code-block:: bat
-
-    docker run -v /c/Users/sampleuser:/home/docker ocrmypdf --skip-text test.pdf output.pdf
-
-where /c/Users/sampleuser is a Unix representation of the Windows path C:\\Users\\sampleuser, assuming a user named "sampleuser" is running ocrmypdf on a file in their home directory, and the files "test.pdf" and "output.pdf" are in the sampleuser folder. The Windows user must have read and write permissions.
-
-`Bash on Ubuntu on Windows <https://github.com/Microsoft/BashOnWindows>`_ should also be a viable route for running the OCRmyPDF Docker container.
+Direct installation on Windows is not possible. `Install the
+Docker <docker-install>`__ container as described above. Ensure that
+your command prompt can run the docker "hello world" container.
 
+It would probably not be too difficult to port on Windows. The main
+reason this has been avoided is the difficulty of packaging and
+installing the various non-Python dependencies: Tesseract, QPDF,
+Ghostscript, Leptonica. Pull requests to add or improve Windows support
+would be quite welcome.
 
 Installing with Python pip
---------------------------
+==========================
 
-First, install `your platform's version <https://repology.org/metapackage/ocrmypdf/versions>`_ of ``ocrmypdf``, if available, as a way of ensuring that external dependencies are (mostly) satisified, even though the platform version may be out of date. Use ``ocrmypdf --version`` to confirm what version was installed.
+OCRmyPDF is delivered by PyPI because it is a convenient way to install
+the latest version. However, PyPI and ``pip`` cannot address the fact
+that ``ocrmypdf`` depends on certain non-Python system libraries and
+programs being instsalled.
 
-Then you can install the latest OCRmyPDF from the Python wheels. First try:
+For best results, first install `your platform's
+version <https://repology.org/metapackage/ocrmypdf/versions>`__ of
+``ocrmypdf``, using the instructions elsewhere in this document. Then
+you can use ``pip`` to get the latest version if your platform version
+is out of date. Chances are that this will satisfy most dependencies.
+
+Use ``ocrmypdf --version`` to confirm what version was installed.
+
+Then you can install the latest OCRmyPDF from the Python wheels. First
+try:
 
 .. code-block:: bash
 
     pip3 install --user ocrmypdf
 
-You should then be able to run ``ocrmypdf --version`` and see that the latest version was located.
+You should then be able to run ``ocrmypdf --version`` and see that the
+latest version was located.
 
-Since ``pip3 install --user`` does not work correctly on some platforms, notably Ubuntu 16.04 and older, and the Homebrew version of Python, instead use this for a system wide installation:
+Since ``pip3 install --user`` does not work correctly on some platforms,
+notably Ubuntu 16.04 and older, and the Homebrew version of Python,
+instead use this for a system wide installation:
 
 .. code-block:: bash
 
     pip3 install ocrmypdf
 
 Requirements for pip and HEAD install
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-
-OCRmyPDF currently requires these external programs to be installed:
-
-- Python 3.5 or newer
-- Tesseract 3.04 or newer
-- Ghostscript 9.15 or newer
-- qpdf 7.0.0 or newer
-
-The following dependencies are recommended:
-
-- Python 3.6
-- Tesseract 4.00 or newer
-- Ghostscript 9.22 or newer
-- qpdf 8.0.2 or newer
-- unpaper 6.1
-- PyMuPDF 1.12.5 or newer
-
-These are in addition to the Python packaging dependencies, meaning that unfortunately, the ``pip install`` command cannot satisfy all of them.
-
-Python 3.6 and Tesseract 4.0.0-beta.1 are recommended for best OCR results and best performance.
-
-The library PyMuPDF is not widely available in platform distributions, and it improves OCRmyPDF in certain conditions. Consider installing OCRmyPDF from the Python binary wheels, which include a precompiled version of this library.
-
-
-Installing HEAD revision from sources
 -------------------------------------
 
-If you have ``git`` and Python 3.5 or newer installed, you can install from source. When the ``pip`` installer runs, it will alert you if dependencies are missing.
+OCRmyPDF currently requires these external programs and libraries to be
+installed, and must be satisfied using the operating system package
+manager. ``pip`` cannot provide them.
 
-To install the HEAD revision from sources in the current Python 3 environment:
+-  Python 3.6 or newer
+-  Ghostscript 9.15 or newer
+-  qpdf 8.1.0 or newer
+-  Tesseract 4.0.0-alpha or newer
+
+As of ocrmypdf 7.2.1, the following versions are recommended:
+
+-  Python 3.7
+-  Ghostscript 9.23 or newer
+-  qpdf 8.2.1
+-  Tesseract 4.0.0 or newer
+-  jbig2enc 0.29 or newer
+-  pngquant 2.5 or newer
+-  unpaper 6.1
+
+jbig2enc, pngquant, and unpaper are optional. If missing certain
+features are disabled. OCRmyPDF will discover them as soon as they are
+available.
+
+**jbig2enc**, if present, will be used to optimize the encoding of
+monochrome images. This can significantly reduce the file size of the
+output file. It is not required.
+`jbig2enc <https://github.com/agl/jbig2enc>`__ is not generally
+available for Ubuntu or Debian due to lingering concerns about patent
+issues, but can easily be built from source. To add JBIG2 encoding, see
+:ref:`jbig2`.
+
+**pngquant**, if present, is optionally used to optimize the encoding of
+PNG-style images in PDFs (actually, any that are that losslessly
+encoded) by lossily quantizing to a smaller color palette. It is only
+activated then the ``--optimize`` argument is ``2`` or ``3``.
+
+**unpaper**, if present, enables the ``--clean`` and ``--clean-final``
+command line options.
+
+These are in addition to the Python packaging dependencies, meaning that
+unfortunately, the ``pip install`` command cannot satisfy all of them.
+
+Installing HEAD revision from sources
+=====================================
+
+If you have ``git`` and Python 3.6 or newer installed, you can install
+from source. When the ``pip`` installer runs, it will alert you if
+dependencies are missing.
+
+If you prefer to build every from source, you will need to `build
+pikepdf from
+source <https://pikepdf.readthedocs.io/en/latest/installation.html#building-from-source>`__.
+First ensure you can build and install pikepdf.
+
+To install the HEAD revision from sources in the current Python 3
+environment:
 
 .. code-block:: bash
 
     pip3 install git+https://github.com/jbarlow83/OCRmyPDF.git
 
-Or, to install in `development mode <https://pythonhosted.org/setuptools/setuptools.html#development-mode>`_,  allowing customization of OCRmyPDF, use the ``-e`` flag:
+Or, to install in `development
+mode <https://pythonhosted.org/setuptools/setuptools.html#development-mode>`__,
+allowing customization of OCRmyPDF, use the ``-e`` flag:
 
 .. code-block:: bash
 
     pip3 install -e git+https://github.com/jbarlow83/OCRmyPDF.git
 
-You may find it easiest to install in a virtual environment, rather than system-wide:
+You may find it easiest to install in a virtual environment, rather than
+system-wide:
 
 .. code-block:: bash
 
@@ -427,7 +549,7 @@ You may find it easiest to install in a virtual environment, rather than system-
     cd OCRmyPDF
     pip3 install .
 
-However, ``ocrmypdf`` will only be accessible on the system PATH after
+However, ``ocrmypdf`` will only be accessible on the system PATH when
 you activate the virtual environment.
 
 To run the program:
@@ -441,8 +563,34 @@ need to be installed. The script requires specific versions of the
 dependencies. Older version than the ones mentioned in the release notes
 are likely not to be compatible to OCRmyPDF.
 
+For development
+---------------
 
-Other Linux packages
---------------------
+To install all of the development and test requirements:
 
-See the `Repology <https://repology.org/metapackage/ocrmypdf/versions>`_ page.
+.. code-block:: bash
+
+    git clone -b master https://github.com/jbarlow83/OCRmyPDF.git
+    python3 -m venv
+    source venv/bin/activate
+    cd OCRmyPDF
+    pip install -e .
+    pip install -r requirements/dev.txt -r requirements/test.txt
+
+To add JBIG2 encoding, see :ref:`jbig2`.
+
+Shell completions
+=================
+
+Completions for ``bash`` and ``fish`` are available in the project's
+``misc/completion`` folder. The ``bash`` completions are likely ``zsh``
+compatible but this has not been confirmed. Package maintainers, please
+install these at the appropriate locations for your system.
+
+To manually install the ``bash`` completion, copy
+``misc/completion/ocrmypdf.bash`` to ``/etc/bash_completion.d/ocrmypdf``
+(rename the file).
+
+To manually install the ``fish`` completion, copy
+``misc/completion/ocrmypdf.fish`` to
+``~/.config/fish/completions/ocrmypdf.fish``.

docs/introduction.rst

@@ -1,115 +1,228 @@
+============
 Introduction
 ============
 
 OCRmyPDF is a Python 3 package that adds OCR layers to PDFs.
 
-
 About OCR
----------
+=========
 
-`Optical character recognition <https://en.wikipedia.org/wiki/Optical_character_recognition>`_ is technology that converts images of typed or handwritten text, such as in a scanned document, to computer text that can be searched and copied.
+`Optical character
+recognition <https://en.wikipedia.org/wiki/Optical_character_recognition>`__
+is technology that converts images of typed or handwritten text, such as
+in a scanned document, to computer text that can be searched and copied.
 
-OCRmyPDF uses `Tesseract <https://github.com/tesseract-ocr/tesseract>`_, the best available open source OCR engine, to perform OCR.
+OCRmyPDF uses
+`Tesseract <https://github.com/tesseract-ocr/tesseract>`__, the best
+available open source OCR engine, to perform OCR.
 
 .. _raster-vector:
 
 About PDFs
-----------
+==========
 
-PDFs are page description files that attempts to preserve a layout exactly. They  contain `vector graphics <http://vector-conversions.com/vectorizing/raster_vs_vector.html>`_ that can contain raster objects such as scanned images. Because PDFs can contain multiple pages (unlike many image formats) and can contain fonts and text, it is a good formats for exchanging scanned documents.
+PDFs are page description files that attempts to preserve a layout
+exactly. They contain `vector
+graphics <http://vector-conversions.com/vectorizing/raster_vs_vector.html>`__
+that can contain raster objects such as scanned images. Because PDFs can
+contain multiple pages (unlike many image formats) and can contain fonts
+and text, it is a good formats for exchanging scanned documents.
 
-.. image:: bitmap_vs_svg.svg
+|image|
 
-A PDF page might contain multiple images, even if it only appears to have one image.  Some scanners or scanning software will segment pages into monochromatic text and color regions for example, to improve the compression ratio and appearance of the page.
-
-Rasterizing a PDF is the process of generating an image suitable for display or analyzing with an OCR engine.  OCR engines like Tesseract work with images, not vector objects.
+A PDF page might contain multiple images, even if it only appears to
+have one image. Some scanners or scanning software will segment pages
+into monochromatic text and color regions for example, to improve the
+compression ratio and appearance of the page.
 
+Rasterizing a PDF is the process of generating an image suitable for
+display or analyzing with an OCR engine. OCR engines like Tesseract work
+with images, not vector objects.
 
 About PDF/A
------------
+===========
 
-`PDF/A <https://en.wikipedia.org/wiki/PDF/A>`_ is an ISO-standardized subset of the full PDF specification that is designed for archiving (the 'A' stands for Archive).  PDF/A differs from PDF primarily by omitting features that would make it difficult to read the file in the future, such as embedded Javascript, video, audio and references to external fonts.  All fonts and resources needed to interpret the PDF must be contained within it. Because PDF/A disables Javascript and other types of embedded content, it is probably more secure.
+`PDF/A <https://en.wikipedia.org/wiki/PDF/A>`__ is an ISO-standardized
+subset of the full PDF specification that is designed for archiving (the
+'A' stands for Archive). PDF/A differs from PDF primarily by omitting
+features that would make it difficult to read the file in the future,
+such as embedded Javascript, video, audio and references to external
+fonts. All fonts and resources needed to interpret the PDF must be
+contained within it. Because PDF/A disables Javascript and other types
+of embedded content, it is probably more secure.
 
 There are various conformance levels and versions, such as "PDF/A-2b".
 
-Generally speaking, the best format for scanned documents is PDF/A. Some governments and jurisdictions, US Courts in particular, `mandate the use of PDF/A <https://pdfblog.com/2012/02/13/what-is-pdfa/>`_ for scanned documents.
+Generally speaking, the best format for scanned documents is PDF/A. Some
+governments and jurisdictions, US Courts in particular, `mandate the use
+of PDF/A <https://pdfblog.com/2012/02/13/what-is-pdfa/>`__ for scanned
+documents.
 
-Since most people who scan documents are interested in reading them indefinitely into the future, OCRmyPDF generates PDF/A-2b by default.
-
-PDF/A has a few drawbacks.  Some PDF viewers include an alert that the file is a PDF/A, which may confuse some users.  It also tends to produce larger files than PDF, because it embeds certain resources even if they are commonly available. PDF/A files can be digitally signed, but may not be encrypted, to ensure they can be read in the future.  Fortunately, converting from PDF/A to a regular PDF is trivial, and any PDF viewer can view PDF/A.
+Since most people who scan documents are interested in reading them
+indefinitely into the future, OCRmyPDF generates PDF/A-2b by default.
 
+PDF/A has a few drawbacks. Some PDF viewers include an alert that the
+file is a PDF/A, which may confuse some users. It also tends to produce
+larger files than PDF, because it embeds certain resources even if they
+are commonly available. PDF/A files can be digitally signed, but may not
+be encrypted, to ensure they can be read in the future. Fortunately,
+converting from PDF/A to a regular PDF is trivial, and any PDF viewer
+can view PDF/A.
 
 What OCRmyPDF does
-------------------
+==================
 
-OCRmyPDF analyzes each page of a PDF to determine the colorspace and resolution (DPI) needed to capture all of the information on that page without losing content.  It uses `Ghostscript <http://ghostscript.com/>`_ to rasterize the page, and then performs on OCR on the rasterized image to create an OCR "layer". The layer is then grafted back onto the original PDF.
+OCRmyPDF analyzes each page of a PDF to determine the colorspace and
+resolution (DPI) needed to capture all of the information on that page
+without losing content. It uses
+`Ghostscript <http://ghostscript.com/>`__ to rasterize the page, and
+then performs on OCR on the rasterized image to create an OCR "layer".
+The layer is then grafted back onto the original PDF.
 
-While one can use a program like Ghostscript or ImageMagick to get an image and put the image through Tesseract, that actually creates a new PDF and many details may be lost. OCRmyPDF can produce a minimally changed PDF as output.
+While one can use a program like Ghostscript or ImageMagick to get an
+image and put the image through Tesseract, that actually creates a new
+PDF and many details may be lost. OCRmyPDF can produce a minimally
+changed PDF as output.
 
-OCRmyPDF also some image processing options like deskew which improve the appearance of files and quality of OCR. When these are used, the OCR layer is grafted onto the processed image instead.
-
-By default, OCRmyPDF produces archival PDFs – PDF/A, which are a stricter subset of PDF features designed for long term archives. If regular PDFs are desired, this can be disabled with ``--output-type pdf``.
+OCRmyPDF also some image processing options like deskew which improve
+the appearance of files and quality of OCR. When these are used, the OCR
+layer is grafted onto the processed image instead.
 
+By default, OCRmyPDF produces archival PDFs – PDF/A, which are a
+stricter subset of PDF features designed for long term archives. If
+regular PDFs are desired, this can be disabled with
+``--output-type pdf``.
 
 Why you shouldn't do this manually
-----------------------------------
+==================================
 
-A PDF is similar to an HTML file, in that it contains document structure along with images.  Sometimes a PDF does nothing more than present a full page image, but often there is additional content that would be lost.
+A PDF is similar to an HTML file, in that it contains document structure
+along with images. Sometimes a PDF does nothing more than present a full
+page image, but often there is additional content that would be lost.
 
 A manual process could work like either of these:
 
-1. Rasterize each page as an image, OCR the images, and combine the output into a PDF. This preserves the layout of each page, but resamples all images (possibly losing quality, increasing file size, introducing compression artifacts, etc.).
+1. Rasterize each page as an image, OCR the images, and combine the
+   output into a PDF. This preserves the layout of each page, but
+   resamples all images (possibly losing quality, increasing file size,
+   introducing compression artifacts, etc.).
+2. Extract each image, OCR, and combine the output into a PDF. This
+   loses the context in which images are used in the PDF, meaning that
+   cropping, rotation and scaling of pages may be lost. Some scanned
+   PDFs use multiple images segmented into black and white, grayscale
+   and color regions, with stencil masks to prevent overlap, as this can
+   enhance the appearance of a file while reducing file size. Clearly,
+   reassembling these images will be easy. This also loses and text or
+   vector art on any pages in a PDF with both scanned and pure digital
+   content.
 
-2. Extract each image, OCR, and combine the output into a PDF. This loses the context in which images are used in the PDF, meaning that cropping, rotation and scaling of pages may be lost. Some scanned PDFs use multiple images segmented into black and white, grayscale and color regions, with stencil masks to prevent overlap, as this can enhance the appearance of a file while reducing file size. Clearly, reassembling these images will be easy. This also loses and text or vector art on any pages in a PDF with both scanned and pure digital content.
+In the case of a PDF that is nothing other than a container of images
+(no rotation, scaling, cropping, one image per page), the second
+approach can be lossless.
 
-In the case of a PDF that is nothing other than a container of images (no rotation, scaling, cropping, one image per page), the second approach can be lossless.
-
-OCRmyPDF uses several strategies depending on input options and the input PDF itself, but generally speaking it rasterizes a page for OCR and then grafts the OCR back onto the original. As such it can handle complex PDFs and still preserve their contents as much as possible.
+OCRmyPDF uses several strategies depending on input options and the
+input PDF itself, but generally speaking it rasterizes a page for OCR
+and then grafts the OCR back onto the original. As such it can handle
+complex PDFs and still preserve their contents as much as possible.
 
+OCRmyPDF also supports a many, many edge cases that have cropped over
+several years of development. We support PDF features like images inside
+of Form XObjects, and pages with UserUnit scaling. We support rare image
+formats like non-monochrome 1-bit images. We warn about files you may
+not to OCR. Thanks to pikepdf and QPDF, we auto-repair PDFs that are
+damaged. (Not that you need to know what any of these are! You should be
+able to throw any PDF at it.)
 
 Limitations
------------
+===========
 
-OCRmyPDF is limited by the Tesseract OCR engine.  As such it experiences these limitations, as do any other programs that rely on Tesseract:
+OCRmyPDF is limited by the Tesseract OCR engine. As such it experiences
+these limitations, as do any other programs that rely on Tesseract:
 
-* The OCR is not as accurate as commercial solutions such as Abbyy.
-* It is not capable of recognizing handwriting.
-* It may find gibberish and report this as OCR output.
-* If a document contains languages outside of those given in the ``-l LANG`` arguments, results may be poor.
-* It is not always good at analyzing the natural reading order of documents. For example, it may fail to recognize that a document contains two columns and join text across the columns.
-* Poor quality scans may produce poor quality OCR. Garbage in, garbage out.
-* PDFs that use transparent layers are not currently checked in the test suite, so they may not work correctly.
-* It does not expose information about what font family text belongs to.
+-  The OCR is not as accurate as commercial solutions such as Abbyy.
+-  It is not capable of recognizing handwriting.
+-  It may find gibberish and report this as OCR output.
+-  If a document contains languages outside of those given in the
+   ``-l LANG`` arguments, results may be poor.
+-  It is not always good at analyzing the natural reading order of
+   documents. For example, it may fail to recognize that a document
+   contains two columns, and may try to join text across columns.
+-  Poor quality scans may produce poor quality OCR. Garbage in, garbage
+   out.
+-  It does not expose information about what font family text belongs
+   to.
 
 OCRmyPDF is also limited by the PDF specification:
 
-* PDF encodes the position of text glyphs but does not encode document structure.  There is no markup that divides a document in sections, paragraphs, sentences, or even words (since blank spaces are not represented). As such all elements of document structure including the spaces between words must be derived heuristically.  Some PDF viewers do a better job of this than others.
-* Because some popular open source PDF viewers have a particularly hard time with spaces betweem words, OCRmyPDF appends a space to each text element as a workaround. While this mixes document structure with graphical information that ideally should be left to the PDF viewer to interpret, it improves compatibility with some viewers and does not cause problems for better ones.
+-  PDF encodes the position of text glyphs but does not encode document
+   structure. There is no markup that divides a document in sections,
+   paragraphs, sentences, or even words (since blank spaces are not
+   represented). As such all elements of document structure including
+   the spaces between words must be derived heuristically. Some PDF
+   viewers do a better job of this than others.
+-  Because some popular open source PDF viewers have a particularly hard
+   time with spaces betweem words, OCRmyPDF appends a space to each text
+   element as a workaround (when using ``--pdf-renderer hocr``). While
+   this mixes document structure with graphical information that ideally
+   should be left to the PDF viewer to interpret, it improves
+   compatibility with some viewers and does not cause problems for
+   better ones.
 
 Ghostscript also imposes some limitations:
 
-* PDFs containing JBIG2-encoded content will be converted to CCITT Group4 encoding, which has lower compression ratios, if Ghostscript PDF/A is enabled.
-* PDFs containing JPEG 2000-encoded content will be converted to JPEG encoding, which may introduce compression artifacts, if Ghostscript PDF/A is enabled.
-* Ghostscript may transcode grayscale and color images, either lossy to lossless or lossless to lossy, based on an internal algorithm. This behavior can be suppressed by setting ``--pdfa-image-compression`` to ``jpeg`` or ``lossless`` to set all images to one type or the other. Ghostscript has no option to maintain the input image's format.
-  
-OCRmyPDF is currently not designed to be used as a Python API; it is designed to be run as a command line tool. ``import ocrmypf`` currently attempts to process the command line on ``sys.argv`` at import time so it has side effects that will interfere with its use as a package. The API it presents should not be considered stable.
+-  PDFs containing JBIG2-encoded content will be converted to CCITT
+   Group4 encoding, which has lower compression ratios, if Ghostscript
+   PDF/A is enabled.
+-  PDFs containing JPEG 2000-encoded content will be converted to JPEG
+   encoding, which may introduce compression artifacts, if Ghostscript
+   PDF/A is enabled.
+-  Ghostscript may transcode grayscale and color images, either lossy to
+   lossless or lossless to lossy, based on an internal algorithm. This
+   behavior can be suppressed by setting ``--pdfa-image-compression`` to
+   ``jpeg`` or ``lossless`` to set all images to one type or the other.
+   Ghostscript has no option to maintain the input image's format.
+   (Ghostscript 9.25+ can copy JPEG images without transcoding them;
+   earlier versions will transcode.)
+-  Ghostscript's PDF/A conversion removes any XMP metadata that is not
+   one of the standard XMP metadata namespaces for PDFs. In particular,
+   PRISM Metdata is removed.
 
+Regarding OCRmyPDF itself:
+
+-  PDFs that use transparency are not currently represented in the test
+   suite
 
 Similar programs
-----------------
+================
 
-To the author's knowledge, OCRmyPDF is the most feature-rich and thoroughly tested command line OCR PDF conversion tool. If it does not meet your needs, contributions and suggestions are welcome. If not, consider one of these similar open source programs:
+To the author's knowledge, OCRmyPDF is the most feature-rich and
+thoroughly tested command line OCR PDF conversion tool. If it does not
+meet your needs, contributions and suggestions are welcome. If not,
+consider one of these similar open source programs:
 
-* pdf2pdfocr
-* pdfsandwich
-* pypdfocr
-* pdfbeads
+-  pdf2pdfocr
+-  pdfsandwich
+-  pypdfocr
+-  pdfbeads
 
 Web front-ends
---------------
+==============
 
-* `Nextcloud OCR <https://github.com/janis91/ocr>`_ is a free software plugin for the Nextcloud private cloud software
-* `OCRmyPDF-web <https://github.com/sseemayer/OCRmyPDF-web>`_, a micro web-frontend for OCRmyPDF (third-party, not actively maintained)
+The Docker image ``ocrmypdf-alpine`` provides a web service front-end
+that allows files to submitted over HTTP and the results "downloaded".
+This is an HTTP server intended to simplify web services deployments; it
+is not intended to be deployed on the public internet and no real
+security measures to speak of.
 
-Bear in mind that OCRmyPDF is not designed to be secure against malware-bearing PDFs (see `Using OCRmyPDF online`_).
+In addition, the following third-party integrations are available:
 
+-  `Nextcloud OCR <https://github.com/janis91/ocr>`__ is a free software
+   plugin for the Nextcloud private cloud software
+
+OCRmyPDF is not designed to be secure against malware-bearing PDFs (see
+`Using OCRmyPDF online <ocr-service>`__). Users should ensure they
+comply with OCRmyPDF's licenses and the licenses of all dependencies. In
+particular, OCRmyPDF requires Ghostscript, which is licensed under
+AGPLv3.
+
+.. |image| image:: images/bitmap_vs_svg.svg

docs/jbig2.rst

@@ -0,0 +1,55 @@
+.. _jbig2:
+
+============================
+Installing the JBIG2 encoder
+============================
+
+Most Linux distributions do not include a JBIG2 encoder since JBIG2
+encoding was patented for a long time. All known JBIG2 US patents have
+expired as of 2017, but it is possible that unknown patents exist.
+
+JBIG2 encoding is recommended for OCRmyPDF and is used to losslessly
+create smaller PDFs. If JBIG2 encoding not available, lower quality
+encodings will be used.
+
+JBIG2 decoding is not patented and is performed automatically by most
+PDF viewers. It is widely supported has been part of the PDF
+specification since 2001.
+
+On macOS, Homebrew packages jbig2enc and OCRmyPDF includes it by
+default. The Docker image for OCRmyPDF also builds its own JBIG2 encoder
+from source.
+
+For all other Linux, you must build a JBIG2 encoder from source:
+
+.. code-block:: bash
+
+   git clone https://github.com/agl/jbig2enc
+   cd jbig2enc
+   ./autogen.sh
+   ./configure && make
+   [sudo] make install
+
+.. _jbig2-lossy:
+
+Lossy mode JBIG2
+================
+
+OCRmyPDF provides lossy mode JBIG2 as an advanced feature. Users should
+`review the technical concerns with JBIG2 in lossy
+mode <https://abbyy.technology/en:kb:tip:jbig2_compression_and_ocr>`__
+and decide if this feature is acceptable for their use case.
+
+JBIG2 lossy mode does achieve higher compression ratios than any other
+monochrome (bitonal) compression technology; for large text documents
+the savings are considerable. JBIG2 lossless still gives great
+compression ratios and is a major improvement over the older CCITT G4
+standard. As explained above, there is some risk of substitution errors.
+
+To turn on JBIG2 lossy mode, add the argument ``--jbig2-lossy``.
+``--optimize {1,2,3}`` are necessary for the argument to take effect
+also required. Also, a JBIG2 encoder must be installed as described in
+the previous section.
+
+*Due to an oversight, ocrmypdf v7.0 and v7.1 used lossy mode by
+default.*

docs/languages.rst

@@ -1,61 +1,59 @@
 .. _lang-packs:
 
+====================================
 Installing additional language packs
 ====================================
 
-OCRmyPDF uses Tesseract for OCR, and relies on its language packs for languages other than English. 
+OCRmyPDF uses Tesseract for OCR, and relies on its language packs for
+languages other than English.
 
-Tesseract supports `most languages <https://github.com/tesseract-ocr/tesseract/blob/master/doc/tesseract.1.asc#languages>`_.
+Tesseract supports `most
+languages <https://github.com/tesseract-ocr/tesseract/blob/master/doc/tesseract.1.asc#languages>`__.
 
-For Linux users, you can often find packages that provide language packs:
+For Linux users, you can often find packages that provide language
+packs:
 
 Debian and Ubuntu users
------------------------
+=======================
 
 .. code-block:: bash
 
    # Display a list of all Tesseract language packs
    apt-cache search tesseract-ocr
 
-   # Debian/Ubuntu users
-   apt-get install tesseract-ocr-chi-sim  # Example: Install Chinese Simplified language back
-   
-You can then pass the ``-l LANG`` argument to OCRmyPDF to give a hint as to what languages it should search for. Multiple
-languages can be requested using either ``-l eng+fre`` (English and French) or ``-l eng -l fre``.
+   # Install Chinese Simplified language pack
+   apt-get install tesseract-ocr-chi-sim
+
+You can then pass the ``-l LANG`` argument to OCRmyPDF to give a hint as
+to what languages it should search for. Multiple languages can be
+requested using either ``-l eng+fre`` (English and French) or
+``-l eng -l fre``.
+
+Fedora users
+============
+
+.. code-block:: bash
+
+   # Display a list of all Tesseract language packs
+   dnf search tesseract
+
+   # Install Chinese Simplified language pack
+   dnf install tesseract-langpack-chi_sim
+
+You can then pass the ``-l LANG`` argument to OCRmyPDF to give a hint as
+to what languages it should search for. Multiple languages can be
+requested using either ``-l eng+fre`` (English and French) or
+``-l eng -l fre``.
 
 macOS users
------------
+===========
 
-You can install additional language packs by :ref:`installing Tesseract using Homebrew with all language packs <macos-all-languages>`.
+You can install additional language packs by
+:ref:`installing Tesseract using Homebrew with all language packs <macos-all-languages>`.
 
 Docker users
-------------
+============
 
-Users of the Docker image may use the alternative :ref:`"polyglot" container <docker-polyglot>` which includes all languages.
-
-Adding individual language packs to a Docker image
-""""""""""""""""""""""""""""""""""""""""""""""""""
-
-If you wish to add a single language pack, you could do the following:
-
-* Download the desired ``.trainedata`` file from the `tessdata <https://github.com/tesseract-ocr/tessdata>`_ repository. Let's use Hebrew in this example (``heb.traineddata``)
-
-* Copy the file to ``/home/user/downloads/heb.traineddata``.
-
-* Create a new container based on the ocrmypdf-tess4 image and jump into it with a terminal:
-
-.. code-block:: bash
-
-	host$ docker run  -v /home/user/downloads:/home/docker -it --entrypoint /bin/bash ocrmypdf-tess4
-
-* Put the file where Tesseract expects it:
-
-.. code-block:: bash
-
-	docker$ cp /home/docker/heb.traineddata /usr/share/tesseract-ocr/tessdata
-
-* Note the container id, and save it as a new image (in this example, ``ocrmypdf-tess4-heb``)
-
-.. code-block:: bash
-
-    host$ docker commit <container_id> ocrmypdf-tess4-heb
+Users of the OCRmyPDF Docker image should install language packs into a
+derived Docker image as
+:ref:`described in that section <docker-lang-packs>`.

docs/plugins.rst

@@ -0,0 +1,24 @@
+=======
+Plugins
+=======
+
+You can use plugins to customize the behavior of OCRmyPDF at certain
+points of interest.
+
+Currently, it is possible to: - override the decision for whether or not
+to perform OCR on a particular file - modify the image is about to be
+sent for OCR
+
+How plugins are imported
+========================
+
+Plugins are imported on demand, by the OCRmyPDF worker process that
+needs to use them. As such, plugins cannot share state with each other,
+and will be imported many times, once for each worker process.
+
+Plugins currently cannot override the same hook.
+
+How plugins are invoked
+=======================
+
+Plugins may be called from the command line:

docs/security.rst

@@ -1,77 +1,162 @@
+===================
 PDF security issues
 ===================
 
-	OCRmyPDF should only be used on PDFs you trust. It is not designed to protect you against malware.
+   OCRmyPDF should only be used on PDFs you trust. It is not designed to
+   protect you against malware.
 
-Recognizing that many users have an interest in handling PDFs and applying OCR to PDFs they did not generate themselves, this article discusses the security implications of PDFs and how users can protect themselves.
+Recognizing that many users have an interest in handling PDFs and
+applying OCR to PDFs they did not generate themselves, this article
+discusses the security implications of PDFs and how users can protect
+themselves.
 
 The disclaimer applies: this software has no warranties of any kind.
 
 PDFs may contain malware
-------------------------
+========================
 
-PDF is a rich, complex file format. The official PDF 1.7 specification, ISO 32000:2008, is hundreds of pages long and references several annexes each of which are similar in length. PDFs can contain video, audio, XML, JavaScript and other programming, and forms. In some cases, they can open internet connections to pre-selected URLs. All of these possible attack vectors.
+PDF is a rich, complex file format. The official PDF 1.7 specification,
+ISO 32000:2008, is hundreds of pages long and references several annexes
+each of which are similar in length. PDFs can contain video, audio, XML,
+JavaScript and other programming, and forms. In some cases, they can
+open internet connections to pre-selected URLs. All of these possible
+attack vectors.
 
-In short, PDFs `may contain viruses <https://security.stackexchange.com/questions/64052/can-a-pdf-file-contain-a-virus>`_.
+In short, PDFs `may contain
+viruses <https://security.stackexchange.com/questions/64052/can-a-pdf-file-contain-a-virus>`__.
 
-This `article <https://theinvisiblethings.blogspot.ca/2013/02/converting-untrusted-pdfs-into-trusted.html>`_ describes a high-paranoia method which allows potentially hostile PDFs to be viewed and rasterized safely in a disposable virtual machine. A trusted PDF created in this manner is converted to images and loses all information making it searchable and losing all compression. OCRmyPDF could be used restore searchability.
+This
+`article <https://theinvisiblethings.blogspot.ca/2013/02/converting-untrusted-pdfs-into-trusted.html>`__
+describes a high-paranoia method which allows potentially hostile PDFs
+to be viewed and rasterized safely in a disposable virtual machine. A
+trusted PDF created in this manner is converted to images and loses all
+information making it searchable and losing all compression. OCRmyPDF
+could be used restore searchability.
 
 How OCRmyPDF processes PDFs
----------------------------
+===========================
 
-OCRmyPDF must open and interpret your PDF in order to insert an OCR layer. First, it runs all PDFs through `pikepdf <https://github.com/pikepdf/pikepdf>`_, a library based on `qpdf <https://github.com/qpdf/qpdf>`_, a program that repairs PDFs with syntax errors. This is done because, in the author's experience, a significant number of PDFs in the wild especially those created by scanners are not well-formed files. qpdf makes it more likely that OCRmyPDF will succeed, but offers no security guarantees. qpdf is also used to split the PDF into single page PDFs.
+OCRmyPDF must open and interpret your PDF in order to insert an OCR
+layer. First, it runs all PDFs through
+`pikepdf <https://github.com/pikepdf/pikepdf>`__, a library based on
+`qpdf <https://github.com/qpdf/qpdf>`__, a program that repairs PDFs
+with syntax errors. This is done because, in the author's experience, a
+significant number of PDFs in the wild especially those created by
+scanners are not well-formed files. qpdf makes it more likely that
+OCRmyPDF will succeed, but offers no security guarantees. qpdf is also
+used to split the PDF into single page PDFs.
 
-Finally, OCRmyPDF rasterizes each page of the PDF using `Ghostscript <http://ghostscript.com/>`_ in ``-dSAFER`` mode.
+Finally, OCRmyPDF rasterizes each page of the PDF using
+`Ghostscript <http://ghostscript.com/>`__ in ``-dSAFER`` mode.
 
-Depending on the options specified, OCRmyPDF may graft the OCR layer into the existing PDF or it may essentially reconstruct ("re-fry") a visually identical PDF that may be quite different at the binary level. That said, OCRmyPDF is not a tool designed for sanitizing PDFs.
+Depending on the options specified, OCRmyPDF may graft the OCR layer
+into the existing PDF or it may essentially reconstruct ("re-fry") a
+visually identical PDF that may be quite different at the binary level.
+That said, OCRmyPDF is not a tool designed for sanitizing PDFs.
+
+.. _ocr-service:
 
 Using OCRmyPDF online or as a service
--------------------------------------
+=====================================
 
-OCRmyPDF should not be deployed as a public-facing service, such as a website where a potential attacker could upload a PDF of their choice for OCR. OCRmyPDF is not designed to be secure against PDF malware. Another concern is PDFs specifically designed to be a denial of service attack: PDFs can contain recursive data structures that sometimes send parsers into infinite loops, and issue complex graphics drawing commands.
+OCRmyPDF is not designed for use as a public web service where a
+malicious user could upload a chosen PDF. In particular, it is not
+necessarily secure against PDF malware or PDFs that cause denial of
+service. OCRmyPDF relies on Ghostscript, and therefore, if deployed
+online one should be prepared to comply with Ghostscript's Affero GPL
+license, OCRmyPDF's GPL license, and any other licenses.
 
-Setting aside these concerns, a side effect of OCRmyPDF is it may incidentally sanitize PDFs that contain malware. It runs ``qpdf`` to repair the PDF, which could correct malformed PDF structures that are part of an attack. When PDF/A output is selected (the default), the input PDF is partially reconstructed by Ghostscript. When ``--force-ocr`` is used, all pages are rasterized and reconverted to PDF, which could remove malware in embedded images. No guarantees.
+Setting aside these concerns, a side effect of OCRmyPDF is it may
+incidentally sanitize PDFs that contain certain types of malware. It
+runs ``qpdf`` to repair the PDF, which could correct malformed PDF
+structures that are part of an attack. When PDF/A output is selected
+(the default), the input PDF is partially reconstructed by Ghostscript.
+When ``--force-ocr`` is used, all pages are rasterized and reconverted
+to PDF, which could remove malware in embedded images.
 
-OCRmyPDF should be relatively safe to use in a trusted intranet, with some considerations:
+OCRmyPDF should be relatively safe to use in a trusted intranet, with
+some considerations:
 
 Limiting CPU usage
-^^^^^^^^^^^^^^^^^^
+------------------
 
-OCRmyPDF will attempt to use all available CPUs and storage, so executing ``nice ocrmypdf`` or limiting the number of jobs with the ``-j`` argument may ensure the server remains available. Another option would be run OCRmyPDF jobs inside a Docker container, a virtual machine, or a cloud instance, which can impose its own limits on CPU usage and be terminated "from orbit" if it fails to complete.
+OCRmyPDF will attempt to use all available CPUs and storage, so
+executing ``nice ocrmypdf`` or limiting the number of jobs with the
+``-j`` argument may ensure the server remains available. Another option
+would be run OCRmyPDF jobs inside a Docker container, a virtual machine,
+or a cloud instance, which can impose its own limits on CPU usage and be
+terminated "from orbit" if it fails to complete.
 
 Temporary storage requirements
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+------------------------------
 
-OCRmyPDF will use a large amount of temporary storage for its work, proportional to the total number of pixels needed to rasterize the PDF. The raster image of a 8.5×11" color page at 300 DPI takes 25 MB uncompressed; OCRmyPDF saves its intermediates as PNG, but that still means it requires about 9 MB per intermediate based on average compression ratios. Multiple intermediates per page are also required, depending on the command line given. A rule of thumb would be to allow 100 MB of temporary storage per page in a file – meaning that a small cloud servers or small VM partitions should be provisioned with plenty of extra space, if say, a 500 page file might be sent.
+OCRmyPDF will use a large amount of temporary storage for its work,
+proportional to the total number of pixels needed to rasterize the PDF.
+The raster image of a 8.5×11" color page at 300 DPI takes 25 MB
+uncompressed; OCRmyPDF saves its intermediates as PNG, but that still
+means it requires about 9 MB per intermediate based on average
+compression ratios. Multiple intermediates per page are also required,
+depending on the command line given. A rule of thumb would be to allow
+100 MB of temporary storage per page in a file – meaning that a small
+cloud servers or small VM partitions should be provisioned with plenty
+of extra space, if say, a 500 page file might be sent.
 
-To check temporary storage usage on actual files, run ``ocrmypdf -k ...`` which will preserve and print the path to temporary storage when the job is done.
+To check temporary storage usage on actual files, run
+``ocrmypdf -k ...`` which will preserve and print the path to temporary
+storage when the job is done.
 
-To change where temporary files are stored, change the ``TMPDIR`` environment variable for ocrmypdf's environment. (Python's ``tempfile.gettempdir()`` returns the root directory in which temporary files will be stored.) For example, one could redirect ``TMPDIR`` to a large RAM disk to avoid wear on HDD/SSD and potentially improve performance. On Amazon Web Services, ``TMPDIR`` can be set to `empheral storage <https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/InstanceStorage.html>`_.
+To change where temporary files are stored, change the ``TMPDIR``
+environment variable for ocrmypdf's environment. (Python's
+``tempfile.gettempdir()`` returns the root directory in which temporary
+files will be stored.) For example, one could redirect ``TMPDIR`` to a
+large RAM disk to avoid wear on HDD/SSD and potentially improve
+performance. On Amazon Web Services, ``TMPDIR`` can be set to `empheral
+storage <https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/InstanceStorage.html>`__.
 
 Timeouts
-^^^^^^^^
+--------
 
-To prevent excessively long OCR jobs consider setting ``--tesseract-timeout`` and/or ``--skip-big`` arguments. ``--skip-big`` is particularly helpful if your PDFs include documents such as reports on standard page sizes with large images attached - often large images are not worth OCR'ing anyway.
+To prevent excessively long OCR jobs consider setting
+``--tesseract-timeout`` and/or ``--skip-big`` arguments. ``--skip-big``
+is particularly helpful if your PDFs include documents such as reports
+on standard page sizes with large images attached - often large images
+are not worth OCR'ing anyway.
 
 Commercial alternatives
-^^^^^^^^^^^^^^^^^^^^^^^
+-----------------------
 
-The author also provides professional services that include OCR and building databases around PDFs, and is happy to provide consultation.
-
-Abbyy Cloud OCR is a viable commercial alternative with a web services API.
+The author also provides professional services that include OCR and
+building databases around PDFs, and is happy to provide consultation.
 
+Abbyy Cloud OCR is a viable commercial alternative with a web services
+API.
 
 Password protection, digital signatures and certification
----------------------------------------------------------
+=========================================================
 
-Password protected PDFs usually have two passwords, and owner and user password. When the user password is set to empty, PDF readers will open the file automatically and marked it as "(SECURED)". While not as reliable as a digital signature, this indicates that whoever set the password approved of the file at that time. When the user password is set, the document cannot be viewed without the password.
+Password protected PDFs usually have two passwords, and owner and user
+password. When the user password is set to empty, PDF readers will open
+the file automatically and marked it as "(SECURED)". While not as
+reliable as a digital signature, this indicates that whoever set the
+password approved of the file at that time. When the user password is
+set, the document cannot be viewed without the password.
 
-Either way, OCRmyPDF does not remove passwords from PDFs and exits with an error on encountering them.
+Either way, OCRmyPDF does not remove passwords from PDFs and exits with
+an error on encountering them.
 
-``qpdf``, one of OCRmyPDF's dependencies, can remove passwords. If the owner and user password are set, a password is required for ``qpdf``. If only the owner password is set, then the password can be stripped, even if one does not have the owner password.
+``qpdf``, one of OCRmyPDF's dependencies, can remove passwords. If the
+owner and user password are set, a password is required for ``qpdf``. If
+only the owner password is set, then the password can be stripped, even
+if one does not have the owner password.
 
-After OCR is applied, password protection is not permitted on PDF/A documents but the file can be converted to regular PDF.
+After OCR is applied, password protection is not permitted on PDF/A
+documents but the file can be converted to regular PDF.
 
-Many programs exist which are capable of inserting an image of someone's signature. On its own, this offers no security guarantees. It is trivial to remove the signature image and apply it to other files. This practice offers no real security.
+Many programs exist which are capable of inserting an image of someone's
+signature. On its own, this offers no security guarantees. It is trivial
+to remove the signature image and apply it to other files. This practice
+offers no real security.
 
-Important documents can be digitally signed and certified to attest to their authorship. OCRmyPDF cannot do this. Open source tools such as pdfbox (Java) have this capability as does Adobe Acrobat.
+Important documents can be digitally signed and certified to attest to
+their authorship. OCRmyPDF cannot do this. Open source tools such as
+pdfbox (Java) have this capability as does Adobe Acrobat.

misc/completion/ocrmypdf.bash

@@ -0,0 +1,87 @@
+# ocrmypdf completion                                     -*- shell-script -*-
+
+_ocrmypdf()
+{
+    local cur prev cword words split
+    _init_completion -s || return
+
+    case $prev in
+        --version|-h|--help)
+            return
+            ;;
+        --user-words|--user-patterns|--tesseract-config)
+            _filedir
+            return
+            ;;
+        --output-type)
+            COMPREPLY=( $( compgen -W 'pdfa pdf pdfa-1 pdfa-2 pdfa-3' -- \
+                "$cur" ) )
+            return
+            ;;
+        --pdf-renderer)
+            COMPREPLY=( $( compgen -W 'auto hocr sandwich' -- "$cur" ) )
+            return
+            ;;
+        --pdfa-image-compression)
+            COMPREPLY=( $( compgen -W 'auto jpeg lossless' -- "$cur" ) )
+            return
+            ;;
+        -O|--optimize|--tesseract-oem)
+            COMPREPLY=( $( compgen -W '{0..3}' -- "$cur" ) )
+            return
+            ;;
+        --jpeg-quality|--png-quality)
+            COMPREPLY=( $( compgen -W '{0..100}' -- "$cur" ) )
+            return
+            ;;
+        -l|--language)
+            COMPREPLY=$( command tesseract --list-langs 2>/dev/null )
+            COMPREPLY=( $( compgen -W '${COMPREPLY[@]##*:}' -- "$cur" ) )
+            return
+            ;;
+        --image-dpi|--oversample|--skip-big|--max-image-mpixels|\
+        --tesseract-timeout|--rotate-pages-threshold)
+            COMPREPLY=( $( compgen -P "$cur" -W '{0..9}' ) )
+            return
+            ;;
+        -j|--jobs)
+            COMPREPLY=( $( compgen -W '{1..'$( _ncpus )'}' -- "$cur" ) )
+            return
+            ;;
+        -v|--verbose)
+            COMPREPLY=( $( compgen -W '{1..9}' -- "$cur" ) ) # max level ?
+            return
+            ;;
+        --tesseract-pagesegmode)
+            COMPREPLY=( $( compgen -W '{1..13}' -- "$cur" ) )
+            return
+            ;;
+        --sidecar|--title|--author|--subject|--keywords|--unpaper-args)
+            # argument required but no completions available
+            return
+            ;;
+    esac
+
+    $split && return
+
+    if [[ $cur == -* ]]; then
+        COMPREPLY=( $( compgen -W '--language --image-dpi --output-type
+            --sidecar --version --jobs --quiet --verbose --title --author
+            --subject --keywords --rotate-pages --remove-background --deskew
+            --clean --clean-final --unpaper-args --oversample --remove-vectors
+            --mask-barcodes --threshold --force-ocr --skip-text --redo-ocr
+            --skip-big --jpeg-quality --png-quality --jbig2-lossy
+            --max-image-mpixels --tesseract-config --tesseract-pagesegmode
+            --help --tesseract-oem --pdf-renderer --tesseract-timeout
+            --rotate-pages-threshold --pdfa-image-compression --user-words
+            --user-patterns --keep-temporary-files --flowchart --output-type' \
+            --  "$cur" ) )
+        return
+    else
+        _filedir
+        return
+    fi
+} &&
+complete -F _ocrmypdf ocrmypdf
+
+# ex: filetype=sh

misc/completion/ocrmypdf.fish

@@ -0,0 +1,70 @@
+complete -c ocrmypdf -l version
+complete -c ocrmypdf -l help
+
+complete -c ocrmypdf -l sidecar -r -d "write OCR to text file"
+complete -c ocrmypdf -s q -l quiet
+
+complete -c ocrmypdf -s r -l rotate-pages -d "rotate pages to correct orientation"
+complete -c ocrmypdf -s d -l deskew -d "fix small horizontal alignment skew"
+complete -c ocrmypdf -s c -l clean -d "clean document images before OCR"
+complete -c ocrmypdf -s i -l clean-final -d "clean document images and keep result"
+complete -c ocrmypdf -l remove-vectors -d "don't send vector objects to OCR"
+complete -c ocrmypdf -l mask-barcodes -d "mask barcodes from OCR"
+complete -c ocrmypdf -l threshold -d "threshold images before OCR"
+
+complete -c ocrmypdf -s f -l force-ocr -d "OCR documents that already have printable text"
+complete -c ocrmypdf -s s -l skip-ocr -d "skip OCR on pages that text, otherwise try OCR"
+complete -c ocrmypdf -l redo-ocr -d "redo OCR on any pages that seem to have OCR already"
+
+complete -c ocrmypdf -s k -l keep-temporary-files -d "keep temporary files (debug)"
+
+complete -c ocrmypdf -x -s l -l language -d 'language'
+complete -c ocrmypdf -x -s l -l language -a '(tesseract --list-langs)'
+
+complete -c ocrmypdf -x -l image-dpi -d "assume this DPI if input image DPI is unknown"
+
+function __fish_ocrmypdf_output_type
+    echo -e "pdfa\t"(_ "output a PDF/A (default)")
+    echo -e "pdf\t"(_ "output a standard PDF")
+    echo -e "pdfa-1\t"(_ "output a PDF/A-1b")
+    echo -e "pdfa-2\t"(_ "output a PDF/A-2b")
+    echo -e "pdfa-3\t"(_ "output a PDF/A-3b")
+end
+complete -c ocrmypdf -x -l output-type -a '(__fish_ocrmypdf_output_type)' -d "select PDF output options"
+
+function __fish_ocrmypdf_pdf_renderer
+    echo -e "auto\t"(_ "auto select PDF renderer")
+    echo -e "hocr\t"(_ "use hocr renderer")
+    echo -e "sandwich\t"(_ "use sandwich renderer")
+end
+complete -c ocrmypdf -x -l pdf-render -a '(__fish_ocrmypdf_pdf_renderer)' -d "select PDF renderer options"
+
+function __fish_ocrmypdf_optimize
+    echo -e "0\t"(_ "do not optimize")
+    echo -e "1\t"(_ "do safe, lossless optimizations (default)")
+    echo -e "2\t"(_ "do some lossy optimizations")
+    echo -e "3\t"(_ "do aggressive lossy optimizations (including lossy JBIG2)")
+end
+complete -c ocrmypdf -x -s O -l optimize -a '(__fish_ocrmypdf_optimize)' -d "select optimization level"
+
+complete -c ocrmypdf -x -s j -l jobs -d "how many worker processes to use"
+complete -c ocrmypdf -x -s v -a '(seq 1 9)'
+complete -c ocrmypdf -x -l title -d "set metadata"
+complete -c ocrmypdf -x -l author -d "set metadata"
+complete -c ocrmypdf -x -l subject -d "set metadata"
+complete -c ocrmypdf -x -l keywords -d "set metadata"
+complete -c ocrmypdf -x -l oversample -d "oversample images to this DPI"
+complete -c ocrmypdf -x -l skip-big -d "skip OCR on pages larger than this many MPixels"
+
+complete -c ocrmypdf -x -l jpeg-quality -d "JPEG quality [0..100]"
+complete -c ocrmypdf -x -l png-quality -d "PNG quality [0..100]"
+complete -c ocrmypdf -x -l jbig2-lossy -d "enable lossy JBIG2 (see docs)"
+complete -c ocrmypdf -x -l max-image-mpixels -d "image decompression bomb threshold"
+complete -c ocrmypdf -x -l tesseract-config -d "set custom tesseract config file"
+complete -c ocrmypdf -x -l tesseract-pagesegmode -d "set tesseract --psm"
+complete -c ocrmypdf -x -l tesseract-oem -d "set tesseract --oem"
+complete -c ocrmypdf -x -l tesseract-timeout -d "maximum number of seconds to wait for OCR"
+complete -c ocrmypdf -x -l rotate-pages-threshold -d "page rotation confidence"
+complete -c ocrmypdf -x -l pdfa-image-compression -a 'auto jpeg lossless' -d "set PDF/A image compression options"
+
+complete -c ocrmypdf -x -a "(__fish_complete_suffix .pdf)"

misc/webservice.py

@@ -0,0 +1,125 @@
+# webservice.py wrapper for OCRmyPDF
+# Copyright (C) 2019 James R. Barlow: github.com/jbarlow83
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Affero General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU Affero General Public License for more details.
+#
+# You should have received a copy of the GNU Affero General Public License
+# along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+"""This is a simple web service/HTTP wrapper for OCRmyPDF
+
+This may be more convenient than the command line tool for some Docker users.
+Note that OCRmyPDF uses Ghostscript, which is licensed under AGPLv3+. While
+OCRmyPDF is under GPLv3, this file is distributed under the Affero GPLv3+ license,
+to emphasize that SaaS deployments should make sure they comply with
+Ghostscript's license as well as OCRmyPDF's.
+"""
+
+from flask import (
+    Flask,
+    Response,
+    flash,
+    request,
+    redirect,
+    url_for,
+    abort,
+    send_from_directory,
+)
+from subprocess import run, PIPE
+from tempfile import TemporaryDirectory
+from werkzeug.utils import secure_filename
+import os
+import shlex
+
+app = Flask(__name__)
+app.secret_key = "secret"
+app.config['MAX_CONTENT_LENGTH'] = 50_000_000
+app.config.from_envvar("OCRMYPDF_WEBSERVICE_SETTINGS", silent=True)
+
+ALLOWED_EXTENSIONS = set(["pdf"])
+
+
+def allowed_file(filename):
+    return "." in filename and filename.rsplit(".", 1)[1].lower() in ALLOWED_EXTENSIONS
+
+
+def do_ocrmypdf(file):
+    uploaddir = TemporaryDirectory(prefix="ocrmypdf-upload")
+    downloaddir = TemporaryDirectory(prefix="ocrmypdf-download")
+
+    filename = secure_filename(file.filename)
+    up_file = os.path.join(uploaddir.name, filename)
+    file.save(up_file)
+
+    down_file = os.path.join(downloaddir.name, filename)
+
+    cmd_args = [arg for arg in shlex.split(request.form["params"])]
+    if "--sidecar" in cmd_args:
+        return Response("--sidecar not supported", 501, mimetype='text/plain')
+
+    ocrmypdf_args = ["ocrmypdf", *cmd_args, up_file, down_file]
+    proc = run(ocrmypdf_args, stdout=PIPE, stderr=PIPE, encoding="utf-8")
+    if proc.returncode != 0:
+        stderr = proc.stderr
+        return Response(stderr, 400, mimetype='text/plain')
+
+    return send_from_directory(downloaddir.name, filename)
+
+
+@app.route("/", methods=["GET", "POST"])
+def upload_file():
+    if request.method == "POST":
+        if "file" not in request.files:
+            return Response("No file in POST", 400, mimetype='text/plain')
+        file = request.files["file"]
+        if file.filename == "":
+            return Response("Empty filename", 400, mimetype='text/plain')
+        if not allowed_file(file.filename):
+            return Response("Invalid filename", 400, mimetype='text/plain')
+        if file and allowed_file(file.filename):
+            return do_ocrmypdf(file)
+        return Response("Some other problem", 400, mimetype='text/plain')
+
+    return """
+    <!doctype html>
+    <title>OCRmyPDF webservice</title>
+    <h1>Upload a PDF (debug UI)</h1>
+    <form method=post enctype=multipart/form-data>
+      <label for="args">Command line parameters</label>
+      <input type=textbox name=params>
+      <label for="file">File to upload</label>
+      <input type=file name=file>
+      <input type=submit value=Upload>
+    </form>
+    <h4>Notice</h2>
+    <div style="font-size: 70%; max-width: 34em;">
+    <p>This is a webservice wrapper for OCRmyPDF.</p>
+    <p>Copyright 2019 James R. Barlow</p>
+    <p>This program is free software: you can redistribute it and/or modify
+    it under the terms of the GNU Affero General Public License as published by
+    the Free Software Foundation, either version 3 of the License, or
+    (at your option) any later version.
+    </p>
+    <p>This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+    </p>
+    <p>
+    You should have received a copy of the GNU Affero General Public License
+    along with this program.  If not, see &lt;http://www.gnu.org/licenses/&gt;.
+    </p>
+    </div>
+    """
+
+
+if __name__ == "__main__":
+    app.run(host='0.0.0.0', port=5000)

pyproject.toml

@@ -0,0 +1,32 @@
+[build-system]
+requires = [
+  "setuptools >= 30.3.0",
+  "wheel",
+  "cffi",
+  "setuptools_scm",
+  "setuptools_scm_git_archive"
+]
+build-backend = "setuptools.build_meta"
+
+[tool.black]
+line-length = 88
+py36 = true
+skip-string-normalization = true
+include = '\.pyi?$'
+exclude = '''
+/(
+    \.eggs
+  | \.git
+  | \.hg
+  | \.mypy_cache
+  | \.tox
+  | \.venv
+  | _build
+  | buck-out
+  | build
+  | dist
+  | docs
+  | misc
+  | \.egg-info
+)/
+'''

requirements/dev.txt

@@ -1,4 +1,4 @@
 check-manifest >= 0.35
 twine >= 1.8.1
-coverage >= 4.4
+coverage >= 4.5
 GitPython == 2.1.3

requirements/main.txt

@@ -1,11 +1,13 @@
 # requirements.txt can be used to replicate the developer's build environment
 # setup.py lists a separate set of requirements that are looser to simplify
 # installation
-cffi == 1.11.5
-img2pdf == 0.2.4
-pikepdf == 0.2.2
+chardet == 3.0.4
+cffi == 1.12.2
+img2pdf == 0.3.3
+pdfminer.six == 20181108
+pikepdf == 1.3.0
 Pillow >= 5.0.0, != 5.1.0 ; sys_platform == "darwin"
-pycparser == 2.18
+pycparser == 2.19
 python-xmp-toolkit == 2.0.1
-reportlab == 3.4.0
-https://github.com/jbarlow83/ruffus/archive/py37.zip
+reportlab == 3.5.13
+tqdm == 4.32.1

requirements/test.txt

@@ -1,7 +1,7 @@
-pytest >= 3.2
-pytest-helpers-namespace
-pytest-xdist
-pytest-cov
+pytest >= 4.4.1, < 5
+pytest-helpers-namespace >= 2019.1.8
+pytest-xdist == 1.28.0
+pytest-cov >= 2.6.1
 python-xmp-toolkit   # requires apt-get install libexempi3
                      # or brew install exempi
 PyPDF2 >= 1.26.0

setup.cfg

@@ -11,6 +11,17 @@ ignore =
 [tool:pytest]
 norecursedirs = lib .pc .git output cache resources
 testpaths = tests
+filterwarnings =
+	ignore:.*XMLParser.*:DeprecationWarning
+markers =
+	slow
+
+[isort]
+multi_line_output=3
+include_trailing_comma=True
+force_grid_wrap=0
+use_parentheses=True
+line_length=88
 
 [metadata]
-license_file = LICENSE
+license_file = LICENSE

setup.py

@@ -20,198 +20,45 @@
 from __future__ import print_function, unicode_literals
 
 import sys
-if sys.version_info < (3, 5):
-    print("Python 3.5 or newer is required", file=sys.stderr)
+
+if sys.version_info < (3, 6):
+    print("Python 3.6 or newer is required", file=sys.stderr)
     sys.exit(1)
 
-from setuptools import setup, find_packages  # nopep8
-from subprocess import STDOUT, check_output, CalledProcessError  # nopep8
-from collections.abc import Mapping  # nopep8
-import re  # nopep8
-import os  # nopep8
+from setuptools import setup, find_packages
 
-
-missing_program = '''
-The program '{program}' could not be executed or was not found on your
-system PATH.
-'''
-
-unknown_version = '''
-OCRmyPDF requires '{program}' {need_version} or higher.  Your system has
-'{program}' but we cannot tell what version is installed.  Contact the
-package maintainer.
-'''
-
-old_version = '''
-OCRmyPDF requires '{program}' {need_version} or higher.  Your system appears
-to have {found_version}.  Please update this program.
-'''
-
-okay_its_optional = '''
-This program is OPTIONAL, so installation of OCRmyPDF can proceed, but
-some functionality may be missing.
-'''
-
-not_okay_its_required = '''
-This program is REQUIRED for OCRmyPDF to work.  Installation will abort.
-'''
-
-osx_install_advice = '''
-If you have homebrew installed, try these command to install the missing
-packages:
-    brew update
-    brew upgrade
-    brew install {package}
-'''
-
-linux_install_advice = '''
-On systems with the aptitude package manager (Debian, Ubuntu), try these
-commands:
-    sudo apt-get update
-    sudo apt-get install {package}
-
-On RPM-based systems (Red Hat, Fedora), search for instructions on
-installing the RPM for {program}.
-'''
-
-
-def get_platform():
-    if sys.platform.startswith('freebsd'):
-        return 'freebsd'
-    elif sys.platform.startswith('linux'):
-        return 'linux'
-    return sys.platform
-
-
-def _error_trailer(program, package, optional, **kwargs):
-    if optional:
-        print(okay_its_optional.format(**locals()), file=sys.stderr)
-    else:
-        print(not_okay_its_required.format(**locals()), file=sys.stderr)
-
-    if isinstance(package, Mapping):
-        package = package[get_platform()]
-
-    if get_platform() == 'darwin':
-        print(osx_install_advice.format(**locals()), file=sys.stderr)
-    elif get_platform() == 'linux':
-        print(linux_install_advice.format(**locals()), file=sys.stderr)
-
-
-def error_missing_program(
-        program,
-        package,
-        optional
-        ):
-    print(missing_program.format(**locals()), file=sys.stderr)
-    _error_trailer(**locals())
-
-
-def error_unknown_version(
-        program,
-        package,
-        optional,
-        need_version
-        ):
-    print(unknown_version.format(**locals()), file=sys.stderr)
-    _error_trailer(**locals())
-
-
-def error_old_version(
-        program,
-        package,
-        optional,
-        need_version,
-        found_version
-        ):
-    print(old_version.format(**locals()), file=sys.stderr)
-    _error_trailer(**locals())
-
-
-def check_external_program(
-        program,
-        need_version,
-        package,
-        version_check_args=['--version'],
-        version_scrape_regex=re.compile(r'(\d+\.\d+(?:\.\d+)?)'),
-        optional=False):
-
-    print('Checking for {program} >= {need_version}...'.format(
-            program=program, need_version=need_version))
-    try:
-        result = check_output(
-                [program] + version_check_args,
-                universal_newlines=True, stderr=STDOUT)
-    except (CalledProcessError, FileNotFoundError):
-        error_missing_program(program, package, optional)
-        if not optional:
-            sys.exit(1)
-        print('Continuing install without {program}'.format(program=program))
-        return
-
-    try:
-        found_version = version_scrape_regex.search(result).group(1)
-    except AttributeError:
-        error_unknown_version(program, package, optional, need_version)
-        sys.exit(1)
-
-    if found_version < need_version:
-        error_old_version(program, package, optional, need_version,
-                          found_version)
-
-    print('Found {program} {found_version}'.format(
-            program=program, found_version=found_version))
+# pylint: disable=w0613
 
 
 command = next((arg for arg in sys.argv[1:] if not arg.startswith('-')), '')
-forced = ('--force' in sys.argv)
+if command.startswith('install') or command in [
+    'check',
+    'test',
+    'nosetests',
+    'easy_install',
+]:
+    forced = '--force' in sys.argv
+    if forced:
+        print("The argument --force is deprecated. Please discontinue use.")
 
 
-if not forced and command.startswith('install') or \
-        command in ['check', 'test', 'nosetests', 'easy_install']:
-    check_external_program(
-        program='tesseract',
-        need_version='3.04',  # using backport for Travis CI
-        package={'darwin': 'tesseract', 'linux': 'tesseract-ocr'}
-    )
-    check_external_program(
-        program='gs',
-        need_version='9.15',  # limited by Travis CI / Ubuntu 14.04 backports
-        package='ghostscript'
-    )
-    check_external_program(
-        program='unpaper',
-        need_version='6.1',   # latest sane version
-        package='unpaper',
-        optional=True
-    )
-    check_external_program(
-        program='qpdf',
-        need_version='8.0.2', # test suite known to fail on 5.1.1
-        package='qpdf'
-    )
-    check_external_program(
-        program='pngquant',
-        need_version='2.0.0',
-        package='pngquant',
-        optional=True
-    )
-
 if 'upload' in sys.argv[1:]:
     print('Use twine to upload the package - setup.py upload is insecure')
     sys.exit(1)
 
-tests_require = open('test_requirements.txt').read().splitlines()
+tests_require = open('requirements/test.txt', encoding='utf-8').read().splitlines()
 
 
 def readme():
-    with open('README.rst') as f:
+    with open('README.md', encoding='utf-8') as f:
         return f.read()
 
+
 setup(
     name='ocrmypdf',
     description='OCRmyPDF adds an OCR text layer to scanned PDF files, allowing them to be searched',
     long_description=readme(),
+    long_description_content_type='text/markdown',
     url='https://github.com/jbarlow83/OCRmyPDF',
     author='James R. Barlow',
     author_email='jim@purplerock.ca',
@@ -219,7 +66,6 @@ setup(
     package_dir={'': 'src'},
     keywords=['PDF', 'OCR', 'optical character recognition', 'PDF/A', 'scanning'],
     classifiers=[
-        "Programming Language :: Python :: 3.5",
         "Programming Language :: Python :: 3.6",
         "Programming Language :: Python :: 3.7",
         "Development Status :: 5 - Production/Stable",
@@ -235,37 +81,37 @@ setup(
         "Topic :: Scientific/Engineering :: Image Recognition",
         "Topic :: Text Processing :: Indexing",
         "Topic :: Text Processing :: Linguistic",
-        ],
-    python_requires=' >= 3.5',
-    setup_requires=[
-        'cffi >= 1.9.1',        # to build the leptonica module
-        'pytest-runner',        # to enable python setup.py test
-        'setuptools_scm',       # so that version will work
+    ],
+    python_requires=' >= 3.6',
+    setup_requires=[  # can be removed whenever we can drop pip 9 support
+        'cffi >= 1.9.1',  # to build the leptonica module
+        'pytest-runner',  # to enable python setup.py test
+        'setuptools_scm',  # so that version will work
+        'setuptools_scm_git_archive',  # enable version from github tarballs
     ],
     use_scm_version={'version_scheme': 'post-release'},
-    cffi_modules=[
-        'src/ocrmypdf/lib/compile_leptonica.py:ffibuilder'
-    ],
+    cffi_modules=['src/ocrmypdf/lib/compile_leptonica.py:ffibuilder'],
     install_requires=[
-        'cffi >= 1.9.1',          # must be a setup and install requirement
-        'img2pdf >= 0.2.4',       # pure Python, so track HEAD closely
-        'pikepdf >= 0.2.2, < 0.3',
+        'chardet >= 3.0.4, < 4',  # unlisted requirement of pdfminer.six 20181108
+        'cffi >= 1.9.1',  # must be a setup and install requirement
+        'img2pdf >= 0.3.0, < 0.4',  # pure Python, so track HEAD closely
+        'pdfminer.six == 20181108 ; sys_platform != "darwin"',
+        'pikepdf >= 1.3.0, < 2',
         'Pillow >= 4.0.0, != 5.1.0 ; sys_platform == "darwin"',
-                                  # Pillow < 4 has BytesIO/TIFF bug w/img2pdf 0.2.3
-                                  # block 5.1.0, broken wheels
-        'python-xmp-toolkit >= 2, < 3',
-        'reportlab >= 3.3.0',     # oldest released version with sane image handling
-        'ruffus',        # pinned - ocrmypdf implements a 2.6.3 workaround
+        # Pillow < 4 has BytesIO/TIFF bug w/img2pdf 0.2.3
+        # block 5.1.0, broken wheels
+        'reportlab >= 3.3.0',  # oldest released version with sane image handling
+        'tqdm >= 4',
     ],
-    extras_require={
-        'fitz': [],  # Backward compatibility
-    },
+    extras_require={'pdfminer': ['pdfminer.six == 20181108']},
     tests_require=tests_require,
-    entry_points={
-        'console_scripts': [
-            'ocrmypdf = ocrmypdf.__main__:run_pipeline'
-        ],
-    },
+    entry_points={'console_scripts': ['ocrmypdf = ocrmypdf.__main__:run']},
     package_data={'ocrmypdf': ['data/sRGB.icc']},
     include_package_data=True,
-    zip_safe=False)
+    zip_safe=False,
+    project_urls={
+        'Documentation': 'https://ocrmypdf.readthedocs.io/',
+        'Source': 'https://github.com/jbarlow83/ocrmypdf',
+        'Tracker': 'https://github.com/jbarlow83/ocrmypdf/issues',
+    },
+)

src/ocrmypdf/__init__.py

@@ -23,3 +23,25 @@ PROGRAM_NAME = 'ocrmypdf'
 __version__ = pkg_resources.get_distribution('ocrmypdf').version
 
 VERSION = __version__
+
+from .exceptions import (
+    ExitCode,
+    BadArgsError,
+    PdfMergeFailedError,
+    MissingDependencyError,
+    UnsupportedImageFormatError,
+    DpiError,
+    OutputFileAccessError,
+    PriorOcrFoundError,
+    InputFileError,
+    SubprocessOutputError,
+    EncryptedPdfError,
+    TesseractConfigError,
+)
+
+from . import helpers
+from . import hocrtransform
+from . import leptonica
+from . import pdfa
+from . import pdfinfo
+from .api import run, configure_logging, Verbosity

src/ocrmypdf/__main__.py

@@ -1,5 +1,5 @@
 #!/usr/bin/env python3
-# © 2015-17 James R. Barlow: github.com/jbarlow83
+# © 2015-19 James R. Barlow: github.com/jbarlow83
 #
 # This file is part of OCRmyPDF.
 #
@@ -16,943 +16,54 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-from tempfile import mkdtemp
-from collections.abc import Sequence
-from pathlib import Path
-import sys
-import os
-import re
-import atexit
-import textwrap
 import logging
-import argparse
+import os
+import sys
 
-import PIL
+from . import __version__
+from .cli import parser
+from .api import configure_logging, Verbosity
+from ._jobcontext import make_logger
+from ._sync import run_pipeline
+from ._validation import check_closed_streams, check_options
+from .exceptions import ExitCode, BadArgsError, MissingDependencyError
 
-import ruffus.ruffus_exceptions as ruffus_exceptions
-import ruffus.cmdline as cmdline
-import ruffus.proxy_logger as proxy_logger
 
-from ._jobcontext import JobContext, JobContextManager, cleanup_working_files
-from .pipeline import build_pipeline
-from .pdfa import file_claims_pdfa
-from .helpers import is_iterable_notstr, re_symlink, is_file_writable, \
-    available_cpu_count
-from .exec import tesseract, qpdf, ghostscript
-from . import PROGRAM_NAME, VERSION
-
-from .exceptions import ExitCode, ExitCodeException, MissingDependencyError, \
-    InputFileError, BadArgsError, OutputFileAccessError
-from . import exceptions as ocrmypdf_exceptions
-from ._unicodefun import verify_python3_env
-
-
-# -------------
-# External dependencies
-
-MINIMUM_TESS_VERSION = '3.04'
-
-HOCR_OK_LANGS = frozenset([
-    'eng', 'deu', 'spa', 'ita', 'por'
-])
-
-def complain(message):
-    print(*textwrap.wrap(message), file=sys.stderr)
-
-
-# Hack to help debugger context find /usr/local/bin
-if 'IDE_PROJECT_ROOTS' in os.environ:
-    os.environ['PATH'] = '/usr/local/bin:' + os.environ['PATH']
-
-# --------
-# Critical environment tests
-
-verify_python3_env()
-
-if tesseract.version() < MINIMUM_TESS_VERSION:
-    complain(
-        "Please install tesseract {0} or newer "
-        "(currently installed version is {1})".format(
-            MINIMUM_TESS_VERSION, tesseract.version()))
-    sys.exit(ExitCode.missing_dependency)
-
-# -------------
-# Parser
-
-def numeric(basetype, min_=None, max_=None):
-    """Validator for numeric params"""
-    min_ = basetype(min_) if min_ is not None else None
-    max_ = basetype(max_) if max_ is not None else None
-    def _numeric(string):
-        value = basetype(string)
-        if (min_ is not None and value < min_
-                or max_ is not None and value > max_):
-            msg = "%r not in valid range %r" % (string, (min_, max_))
-            raise argparse.ArgumentTypeError(msg)
-        return value
-    _numeric.__name__ = basetype.__name__
-    return _numeric
-
-
-parser = argparse.ArgumentParser(
-    prog=PROGRAM_NAME,
-    fromfile_prefix_chars='@',
-    formatter_class=argparse.RawDescriptionHelpFormatter,
-    description="""\
-Generates a searchable PDF or PDF/A from a regular PDF.
-
-OCRmyPDF rasterizes each page of the input PDF, optionally corrects page
-rotation and performs image processing, runs the Tesseract OCR engine on the
-image, and then creates a PDF from the OCR information.
-""",
-    epilog="""\
-OCRmyPDF attempts to keep the output file at about the same size.  If a file
-contains losslessly compressed images, and output file will be losslessly
-compressed as well.
-
-PDF is a page description file that attempts to preserve a layout exactly.
-A PDF can contain vector objects (such as text or lines) and raster objects
-(images).  A page might have multiple images.  OCRmyPDF is prepared to deal
-with the wide variety of PDFs that exist in the wild.
-
-When a PDF page contains text, OCRmyPDF assumes that the page has already
-been OCRed or is a "born digital" page that should not be OCRed.  The default
-behavior is to exit in this case without producing a file.  You can use the
-option --skip-text to ignore pages with text, or --force-ocr to rasterize
-all objects on the page and produce an image-only PDF as output.
-
-    ocrmypdf --skip-text file_with_some_text_pages.pdf output.pdf
-
-    ocrmypdf --force-ocr word_document.pdf output.pdf
-
-If you are concerned about long-term archiving of PDFs, use the default option
---output-type pdfa which converts the PDF to a standardized PDF/A-2b.  This
-converts images to sRGB colorspace, removes some features from the PDF such
-as Javascript or forms. If you want to minimize the number of changes made to
-your PDF, use --output-type pdf.
-
-If OCRmyPDF is given an image file as input, it will attempt to convert the
-image to a PDF before processing.  For more control over the conversion of
-images to PDF, use the Python package img2pdf or other image to PDF software.
-
-For example, this command uses img2pdf to convert all .png files beginning
-with the 'page' prefix to a PDF, fitting each image on A4-sized paper, and
-sending the result to OCRmyPDF through a pipe.  img2pdf is a dependency of
-ocrmypdf so it is already installed.
-
-    img2pdf --pagesize A4 page*.png | ocrmypdf - myfile.pdf
-
-Online documentation is located at:
-    https://ocrmypdf.readthedocs.io/en/latest/introduction.html
-
-""")
-
-parser.add_argument(
-    'input_file', metavar="input_pdf_or_image",
-    help="PDF file containing the images to be OCRed (or '-' to read from "
-         "standard input)")
-parser.add_argument(
-    'output_file', metavar="output_pdf",
-    help="Output searchable PDF file (or '-' to write to standard output). "
-         "Existing files will be ovewritten. If same as input file, the "
-         "input file will be updated only if processing is successful.")
-parser.add_argument(
-    '-l', '--language', action='append',
-    help="Language(s) of the file to be OCRed (see tesseract --list-langs for "
-         "all language packs installed in your system). Use -l eng+deu for "
-         "multiple languages.")
-parser.add_argument(
-    '--image-dpi', metavar='DPI', type=int,
-    help="For input image instead of PDF, use this DPI instead of file's.")
-parser.add_argument(
-    '--output-type', choices=['pdfa', 'pdf', 'pdfa-1', 'pdfa-2', 'pdfa-3'],
-    default='pdfa',
-    help="Choose output type. 'pdfa' creates a PDF/A-2b compliant file for "
-         "long term archiving (default, recommended) but may not suitable "
-         "for users who want their file altered as little as possible. 'pdfa' "
-         "also has problems with full Unicode text. 'pdf' attempts to "
-         "preserve file contents as much as possible. 'pdf-a1' creates a "
-         "PDF/A1-b file. 'pdf-a2' is equivalent to 'pdfa'. 'pdf-a3' creates a "
-         "PDF/A3-b file."
-         )
-
-# Use null string '\0' as sentinel to indicate the user supplied no argument,
-# since that is the only invalid character for filepaths on all platforms
-# bool('\0') is True in Python
-parser.add_argument(
-    '--sidecar', nargs='?', const='\0', default=None, metavar='FILE',
-    help="Generate sidecar text files that contain the same text recognized "
-         "by Tesseract. This may be useful for building a OCR text database. "
-         "If FILE is omitted, the sidecar file be named {output_file}.txt "
-         "If FILE is set to '-', the sidecar is written to stdout (a "
-         "convenient way to preview OCR quality). The output file and sidecar "
-         "may not both use stdout at the same time.")
-
-parser.add_argument(
-    '--version', action='version', version=VERSION,
-    help="Print program version and exit")
-
-jobcontrol = parser.add_argument_group(
-    "Job control options")
-jobcontrol.add_argument(
-    '-j', '--jobs', metavar='N', type=numeric(int, 0, 256),
-    help="Use up to N CPU cores simultaneously (default: use all).")
-jobcontrol.add_argument(
-    '-q', '--quiet', action='store_true', help="Suppress INFO messages")
-jobcontrol.add_argument(
-    '-v', '--verbose', const="+", default=[], nargs='?', action="append",
-    help="Print more verbose messages for each additional verbose level")
-
-metadata = parser.add_argument_group(
-    "Metadata options",
-    "Set output PDF/A metadata (default: copy input document's metadata)")
-metadata.add_argument(
-    '--title', type=str,
-    help="Set document title (place multiple words in quotes)")
-metadata.add_argument(
-    '--author', type=str,
-    help="Set document author")
-metadata.add_argument(
-    '--subject', type=str,
-    help="Set document subject description")
-metadata.add_argument(
-    '--keywords', type=str,
-    help="Set document keywords")
-
-preprocessing = parser.add_argument_group(
-    "Image preprocessing options",
-    "Options to improve the quality of the final PDF and OCR")
-preprocessing.add_argument(
-    '-r', '--rotate-pages', action='store_true',
-    help="Automatically rotate pages based on detected text orientation")
-preprocessing.add_argument(
-    '--remove-background', action='store_true',
-    help="Attempt to remove background from gray or color pages, setting it "
-         "to white ")
-preprocessing.add_argument(
-    '-d', '--deskew', action='store_true',
-    help="Deskew each page before performing OCR")
-preprocessing.add_argument(
-    '-c', '--clean', action='store_true',
-    help="Clean pages from scanning artifacts before performing OCR, and send "
-         "the cleaned page to OCR, but do not include the cleaned page in "
-         "the output")
-preprocessing.add_argument(
-    '-i', '--clean-final', action='store_true',
-    help="Clean page as above, and incorporate the cleaned image in the final "
-         "PDF.  Might remove desired content.")
-preprocessing.add_argument(
-    '--oversample', metavar='DPI', type=numeric(int, 0, 5000), default=0,
-    help="Oversample images to at least the specified DPI, to improve OCR "
-         "results slightly")
-
-ocrsettings = parser.add_argument_group(
-    "OCR options",
-    "Control how OCR is applied")
-ocrsettings.add_argument(
-    '-f', '--force-ocr', action='store_true',
-    help="Rasterize any fonts or vector objects on each page, apply OCR, and "
-         "save the rastered output (this rewrites the PDF)")
-ocrsettings.add_argument(
-    '-s', '--skip-text', action='store_true',
-    help="Skip OCR on any pages that already contain text, but include the "
-         "page in final output; useful for PDFs that contain a mix of "
-         "images, text pages, and/or previously OCRed pages")
-ocrsettings.add_argument(
-    '--skip-big', type=numeric(float, 0, 5000), metavar='MPixels',
-    help="Skip OCR on pages larger than the specified amount of megapixels, "
-         "but include skipped pages in final output")
-
-optimizing = parser.add_argument_group(
-    "Optimization options",
-    "Control how the PDF is optimized after OCR"
-)
-optimizing.add_argument(
-    '-O', '--optimize', type=int, choices=range(0, 4), default=1,
-    help=("Control how PDF is optimized after processing:"
-        "0 - do not optimize;"
-        "1 - do safe, lossless optimizations (default);"
-        "2 - do lossy optimizations; "
-        "3 - do aggressive lossy optimizations"
-    )
-)
-optimizing.add_argument(
-    '--jpeg-quality', type=numeric(int, 0, 100), default=0, metavar='Q',
-    help=("Adjust JPEG quality level for JPEG optimization. "
-          "100 is best quality and largest output size; "
-          "1 is lowest quality and smallest output"
-          "0 uses the default."
-    )
-)
-optimizing.add_argument(
-    '--jpg-quality', type=numeric(int, 0, 100), default=0, metavar='Q',
-    dest='jpeg_quality',
-    help=argparse.SUPPRESS  # Alias for --jpeg-quality
-)
-optimizing.add_argument(
-    '--png-quality', type=numeric(int, 0, 100), default=0, metavar='Q',
-    help=("Adjust PNG quality level to use when quantizing PNGs. "
-          "Values have same meaning as with --jpeg-quality"
-    )
-)
-
-advanced = parser.add_argument_group(
-    "Advanced",
-    "Advanced options to control Tesseract's OCR behavior")
-advanced.add_argument(
-    '--max-image-mpixels', action='store', type=numeric(float, 0),
-    metavar='MPixels',
-    help="Set maximum number of pixels to unpack before treating an image as a "
-         "decompression bomb",
-    default=128.0)
-advanced.add_argument(
-    '--tesseract-config', action='append', metavar='CFG', default=[],
-    help="Additional Tesseract configuration files -- see documentation")
-advanced.add_argument(
-    '--tesseract-pagesegmode', action='store', type=int, metavar='PSM',
-    choices=range(0, 14),
-    help="Set Tesseract page segmentation mode (see tesseract --help)")
-advanced.add_argument(
-    '--tesseract-oem', action='store', type=int, metavar='MODE',
-    choices=range(0, 4),
-    help=("Set Tesseract 4.0 OCR engine mode: "
-         "0 - original Tesseract only; "
-         "1 - neural nets LSTM only; "
-         "2 - Tesseract + LSTM; "
-         "3 - default.")
-    )
-advanced.add_argument(
-    '--pdf-renderer',
-    choices=['auto', 'hocr', 'sandwich'], default='auto',
-    help="Choose OCR PDF renderer - the default option is to let OCRmyPDF "
-         "choose.  See documentation for discussion."
-    )
-advanced.add_argument(
-    '--tesseract-timeout', default=180.0, type=numeric(float, 0),
-    metavar='SECONDS',
-    help='Give up on OCR after the timeout, but copy the preprocessed page '
-         'into the final output')
-advanced.add_argument(
-    '--rotate-pages-threshold', default=14.0, type=numeric(float, 0, 1000),
-    metavar='CONFIDENCE',
-    help="Only rotate pages when confidence is above this value (arbitrary "
-         "units reported by tesseract)")
-advanced.add_argument(
-    '--pdfa-image-compression', choices=['auto', 'jpeg', 'lossless'],
-    default='auto',
-    help="Specify how to compress images in the output PDF/A. 'auto' lets "
-         "OCRmyPDF decide.  'jpeg' changes all grayscale and color images to "
-         "JPEG compression.  'lossless' uses PNG-style lossless compression "
-         "for all images.  Monochrome images are always compressed using a "
-         "lossless codec.  Compression settings "
-         "are applied to all pages, including those for which OCR was "
-         "skipped.  Not supported for --output-type=pdf ; that setting "
-         "preserves the original compression of all images.")
-advanced.add_argument(
-    '--user-words', metavar='FILE',
-    help="Specify the location of the Tesseract user words file. This is a "
-         "list of words Tesseract should consider while performing OCR in "
-         "addition to its standard language dictionaries. This can improve "
-         "OCR quality especially for specialized and technical documents.")
-advanced.add_argument(
-    '--user-patterns', metavar='FILE',
-    help="Specify the location of the Tesseract user patterns file.")
-
-debugging = parser.add_argument_group(
-    "Debugging",
-    "Arguments to help with troubleshooting and debugging")
-debugging.add_argument(
-    '-k', '--keep-temporary-files', action='store_true',
-    help="Keep temporary files (helpful for debugging)")
-debugging.add_argument(
-    '--flowchart', type=str,
-    help="Generate the pipeline execution flowchart")
-
-
-def check_options_languages(options, _log):
-    if not options.language:
-        options.language = ['eng']  # Enforce English hegemony
-
-    # Support v2.x "eng+deu" language syntax
-    if '+' in options.language[0]:
-        options.language = options.language[0].split('+')
-
-    languages = set(options.language)
-    if not languages.issubset(tesseract.languages()):
-        msg = (
-            "The installed version of tesseract does not have language "
-            "data for the following requested languages: \n")
-        for lang in (languages - tesseract.languages()):
-            msg += lang + '\n'
-        raise MissingDependencyError(msg)
-
-
-def check_options_output(options, log):
-    # We have these constraints to check for.
-    # 1. Ghostscript < 9.20 mangles multibyte Unicode
-    # 2. Tesseract < 3.05 embeds an older version of GlyphlessFont with which
-    #    no version of Ghostscript handles correctly.
-    # 3. hocr doesn't work on non-Latin languages (so don't select it)
-
-    languages = set(options.language)
-    is_latin = languages.issubset(HOCR_OK_LANGS)
-
-    if options.pdf_renderer == 'hocr' and not is_latin:
-        msg = (
-            "The 'hocr' PDF renderer is known to cause problems with one "
-            "or more of the languages in your document.  Use "
-            "--pdf-renderer auto (the default) to avoid this issue.")
-        log.warning(msg)
-
-    if ghostscript.version() < '9.20' \
-            and options.output_type != 'pdf' \
-            and not is_latin:
-        # https://bugs.ghostscript.com/show_bug.cgi?id=696874
-        # Ghostscript < 9.20 fails to encode multibyte characters properly
-        msg = (
-            "The installed version of Ghostscript does not work correctly "
-            "with the OCR languages you specified. Use --output-type pdf or "
-            "upgrade to Ghostscript 9.20 or later to avoid this issue.")
-        msg += "Found Ghostscript {}".format(ghostscript.version())
-        log.warning(msg)
-
-    # Decide on what renderer to use
-    if options.pdf_renderer == 'auto':
-        if tesseract.version() < '3.05' \
-                and options.output_type.startswith('pdfa') \
-                and is_latin:
-            options.pdf_renderer = 'hocr'
-        else:
-            options.pdf_renderer = 'sandwich'
-
-    if options.pdf_renderer == 'sandwich' \
-            and tesseract.version() < '3.05':
-        msg = (
-            "Ghostscript will corrupt the OCR text of PDFs produced by "
-            "Tesseract 3.04.xx and older.  For best results, upgrade to a "
-            "newer release of Tesseract. "
-        )
-
-        if options.output_type.startswith('pdfa'):
-            msg += (
-                "The argument --output-type=pdfa* requires Ghostscript, so "
-                "the PDF will be invalid.  If you cannot upgrade Tesseract, "
-                "use --output-type=pdf.")
-            raise MissingDependencyError(msg)
-        log.warning(msg)
-
-    if options.output_type == 'pdfa':
-        options.output_type = 'pdfa-2'
-
-    if options.output_type == 'pdfa-3' and ghostscript.version() < '9.19':
-        raise MissingDependencyError(
-            "--output-type pdfa-3 requires Ghostscript 9.19 or later"
-        )
-
-    lossless_reconstruction = False
-    if options.pdf_renderer in ('hocr', 'sandwich'):
-        if not any((options.deskew, options.clean_final, options.force_ocr,
-                   options.remove_background)):
-            lossless_reconstruction = True
-    options.lossless_reconstruction = lossless_reconstruction
-
-
-def check_options_sidecar(options, log):
-    if options.sidecar == '\0':
-        if options.output_file == '-':
-            raise argparse.ArgumentError(
-                None,
-                "--sidecar filename must be specified when output file is "
-                "stdout.")
-        options.sidecar = options.output_file + '.txt'
-
-
-def _optional_program_check(name, version_fn, min_version, for_argument):
-    try:
-        if version_fn() < min_version:
-            raise MissingDependencyError(
-                "The installed '{}' is not supported. "
-                "Install version {} or newer.".format(name, min_version))
-    except FileNotFoundError:
-        raise MissingDependencyError(
-            "Install the '{}' program to use {}.".format(name, for_argument))
-
-
-def check_options_preprocessing(options, log):
-    if any((options.clean, options.clean_final)):
-        from .exec import unpaper
-        _optional_program_check(
-            'unpaper', unpaper.version, '6.1', '--clean, --clean-final'
-        )
-
-
-def check_options_ocr_behavior(options, log):
-    if options.force_ocr and options.skip_text:
-        raise argparse.ArgumentError(
-            None,
-            "Error: --force-ocr and --skip-text are mutually exclusive.")
-
-
-def check_options_optimizing(options, log):
-    if options.optimize >= 2:
-        from .exec import pngquant, jbig2enc
-        _optional_program_check(
-            'pngquant', pngquant.version, '2.0.1', '--optimize {2,3}'
-        )
-        _optional_program_check(
-            'jbig2', jbig2enc.version, '0.28', '--optimize {2,3}'
-        )
-
-
-def check_options_advanced(options, log):
-    if options.tesseract_oem and not tesseract.v4():
-        log.warning(
-            "--tesseract-oem requires Tesseract 4.x -- argument ignored")
-    if options.pdfa_image_compression != 'auto' and \
-            options.output_type.startswith('pdfa'):
-        log.warning(
-            "--pdfa-image-compression argument has no effect when "
-            "--output-type is not 'pdfa', 'pdfa-1', or 'pdfa-2'"
-        )
-
-    if tesseract.v4() and (options.user_words or options.user_patterns):
-        log.warning(
-            'Tesseract 4.x ignores --user-words, so this has no effect')
-
-
-def check_options_metadata(options, log):
-    import unicodedata
-    docinfo = [options.title, options.author, options.keywords,
-               options.subject]
-    for s in (m for m in docinfo if m):
-        for c in s:
-            if unicodedata.category(c) == 'Co' or ord(c) >= 0x10000:
-                raise ValueError(
-                    "One of the metadata strings contains "
-                    "an unsupported Unicode character: '{}' (U+{})".format(
-                        c, hex(ord(c))[2:].upper()
-                ))
-
-
-def check_options_pillow(options, log):
-    PIL.Image.MAX_IMAGE_PIXELS = int(options.max_image_mpixels * 1000000)
-    if PIL.Image.MAX_IMAGE_PIXELS == 0:
-        PIL.Image.MAX_IMAGE_PIXELS = None
-
-
-def check_options(options, log):
-    try:
-        check_options_languages(options, log)
-        check_options_metadata(options, log)
-        check_options_output(options, log)
-        check_options_sidecar(options, log)
-        check_options_preprocessing(options, log)
-        check_options_ocr_behavior(options, log)
-        check_options_optimizing(options, log)
-        check_options_advanced(options, log)
-        check_options_pillow(options, log)
-    except ValueError as e:
-        log.error(e)
-        sys.exit(ExitCode.bad_args)
-    except argparse.ArgumentError as e:
-        log.error(e)
-        sys.exit(ExitCode.bad_args)
-    except MissingDependencyError as e:
-        log.error(e)
-        sys.exit(ExitCode.missing_dependency)
-
-
-# ----------
-# Logging
-
-
-def logging_factory(logger_name, logger_args):
-    verbose = logger_args['verbose']
-    quiet = logger_args['quiet']
-
-    root_logger = logging.getLogger(logger_name)
-    root_logger.setLevel(logging.DEBUG)
-
-    handler = logging.StreamHandler(sys.stderr)
-    formatter_ = logging.Formatter("%(levelname)7s - %(message)s")
-    handler.setFormatter(formatter_)
-    if verbose:
-        handler.setLevel(logging.DEBUG)
-    elif quiet:
-        handler.setLevel(logging.WARNING)
-    else:
-        handler.setLevel(logging.INFO)
-    root_logger.addHandler(handler)
-    return root_logger
-
-
-def cleanup_ruffus_error_message(msg):
-    msg = re.sub(r'\s+', r' ', msg)
-    msg = re.sub(r"\((.+?)\)", r'\1', msg)
-    msg = msg.strip()
-    return msg
-
-
-def do_ruffus_exception(ruffus_five_tuple, options, log):
-    """Replace the elaborate ruffus stack trace with a user friendly
-    description of the error message that occurred."""
-    exit_code = None
-
-    task_name, job_name, exc_name, exc_value, exc_stack = ruffus_five_tuple
-    task_name = task_name  # unused
-    job_name = job_name  # unused
-    if exc_name == 'builtins.SystemExit':
-        match = re.search(r"\.(.+?)\)", exc_value)
-        exit_code_name = match.groups()[0]
-        exit_code = getattr(ExitCode, exit_code_name, 'other_error')
-    elif exc_name == 'ruffus.ruffus_exceptions.MissingInputFileError':
-        log.error(cleanup_ruffus_error_message(exc_value))
-        exit_code = ExitCode.input_file
-    elif exc_name == 'builtins.KeyboardInterrupt':
-        log.error("Interrupted by user")
-        exit_code = ExitCode.ctrl_c
-    elif exc_name == 'subprocess.CalledProcessError':
-        # It's up to the subprocess handler to report something useful
-        msg = "Error occurred while running this command:"
-        log.error(msg + '\n' + exc_value)
-        exit_code = ExitCode.child_process_error
-    elif exc_name == 'ocrmypdf.exceptions.EncryptedPdfError':
-        log.error(textwrap.dedent("""\
-            Input PDF is encrypted. The encryption must be removed to
-            perform OCR.
-
-            For information about this PDF's security use
-                qpdf --show-encryption infilename
-
-            You can remove the encryption using
-                qpdf --decrypt [--password=[password]] infilename
-
-            """))
-        exit_code = ExitCode.encrypted_pdf
-    elif exc_name == 'ocrmypdf.exceptions.PdfMergeFailedError':
-        log.error(textwrap.dedent("""\
-            Failed to merge PDF image layer with OCR layer
-
-            Usually this happens because the input PDF file is mal-formed and
-            ocrmypdf cannot automatically correct the problem on its own.
-
-            Try using
-                ocrmypdf --pdf-renderer sandwich  [..other args..]
-            """))
-        exit_code = ExitCode.input_file
-    elif exc_name.startswith('ocrmypdf.exceptions.'):
-        base_exc_name = exc_name.replace('ocrmypdf.exceptions.', '')
-        exc_class = getattr(ocrmypdf_exceptions, base_exc_name)
-        exit_code = exc_class.exit_code
-    elif exc_name == 'PIL.Image.DecompressionBombError':
-        msg = cleanup_ruffus_error_message(exc_value)
-        msg += ("\nUse the --max-image-mpixels argument to set increase the "
-                "maximum number of megapixels to accept.")
-        log.error(msg)
-        exit_code = ExitCode.input_file
-
-    if exit_code is not None:
-        return exit_code
-
-    if not options.verbose:
-        log.error(exc_stack)
-    return ExitCode.other_error
-
-
-def traverse_ruffus_exception(e_args, options, log):
-    """Walk through a RethrownJobError and find the first exception.
-
-    Ruffus flattens exception to 5 element tuples. Because of a bug
-    in <= 2.6.3 it may present either the single:
-      (task, job, exc, value, stack)
-    or something like:
-      [[(task, job, exc, value, stack)]]
-
-    For Ruffus > 2.6.4 the RethrownJobException has a list of exceptions like
-    e.job_exceptions = [(5-tuple), (5-tuple), ...]
-
-    Generally cross-process exception marshalling doesn't work well
-    and ruffus doesn't support because BaseException has its own
-    implementation of __reduce__ that attempts to reconstruct the
-    exception based on e.__init__(e.args).
-
-    Attempting to log the exception directly marshalls it to the logger
-    which is probably in another process, so it's better to log only
-    data from the exception at this point.
-
-    The exit code will be based on this, even if multiple exceptions occurred
-    at the same time."""
-
-    if isinstance(e_args, Sequence) and isinstance(e_args[0], str) \
-            and len(e_args) == 5:
-        return do_ruffus_exception(e_args, options, log)
-    elif is_iterable_notstr(e_args):
-        for exc in e_args:
-            return traverse_ruffus_exception(exc, options, log)
-
-
-def check_closed_streams(options):
-    """Work around Python issue with multiprocessing forking on closed streams
-
-    https://bugs.python.org/issue28326
-
-    Attempting to a fork/exec a new Python process when any of std{in,out,err}
-    are closed or not flushable for some reason may raise an exception.
-    Fix this by opening devnull if the handle seems to be closed.  Do this
-    globally to avoid tracking places all places that fork.
-
-    Seems to be specific to multiprocessing.Process not all Python process
-    forkers.
-
-    The error actually occurs when the stream object is not flushable,
-    but replacing an open stream object that is not flushable with
-    /dev/null is a bad idea since it will create a silent failure.  Replacing
-    a closed handle with /dev/null seems safe.
-
-    """
-
-    if sys.version_info[0:3] >= (3, 6, 4):
-        return True  # Issued fixed in Python 3.6.4+
-
-    if sys.stderr is None:
-        sys.stderr = open(os.devnull, 'w')
-
-    if sys.stdin is None:
-        if options.input_file == '-':
-            print("Trying to read from stdin but stdin seems closed",
-                  file=sys.stderr)
-            return False
-        sys.stdin = open(os.devnull, 'r')
-
-    if sys.stdout is None:
-        if options.output_file == '-':
-            # Can't replace stdout if the user is piping
-            # If this case can even happen, it must be some kind of weird
-            # stream.
-            print(textwrap.dedent("""\
-                Output was set to stdout '-' but the stream attached to
-                stdout does not support the flush() system call.  This
-                will fail."""), file=sys.stderr)
-            return False
-        sys.stdout = open(os.devnull, 'w')
-
-    return True
-
-
-def log_page_orientations(pdfinfo, _log):
-    direction = {0: 'n', 90: 'e',
-                180: 's', 270: 'w'}
-    orientations = []
-    for n, page in enumerate(pdfinfo):
-        angle = page.rotation or 0
-        if angle != 0:
-            orientations.append('{0}{1}'.format(
-                n + 1,
-                direction.get(angle, '')))
-    if orientations:
-        _log.info('Page orientations detected: ' + ' '.join(orientations))
-
-
-def preamble(_log):
-    _log.debug('ocrmypdf ' + VERSION)
-    _log.debug('tesseract ' + tesseract.version())
-    _log.debug('qpdf ' + qpdf.version())
-
-
-def check_environ(options, _log):
-    old_envvars = (
-        'OCRMYPDF_TESSERACT',
-        'OCRMYPDF_QPDF',
-        'OCRMYPDF_GS',
-        'OCRMYPDF_UNPAPER')
-    for k in old_envvars:
-        if k in os.environ:
-            _log.warning(textwrap.dedent("""\
-                OCRmyPDF no longer uses the environment variable {}.
-                Change PATH to select alternate programs.""".format(k)))
-
-
-def check_input_file(options, _log, start_input_file):
-    if options.input_file == '-':
-        # stdin
-        _log.info('reading file from standard input')
-        with open(start_input_file, 'wb') as stream_buffer:
-            from shutil import copyfileobj
-            copyfileobj(sys.stdin.buffer, stream_buffer)
-    else:
-        try:
-            re_symlink(options.input_file, start_input_file, _log)
-        except FileNotFoundError:
-            _log.error("File not found - " + options.input_file)
-            raise InputFileError()
-
-
-def check_requested_output_file(options, _log):
-    if options.output_file == '-':
-        if sys.stdout.isatty():
-            _log.error(textwrap.dedent("""\
-                Output was set to stdout '-' but it looks like stdout
-                is connected to a terminal.  Please redirect stdout to a
-                file."""))
-            raise BadArgsError()
-    elif not is_file_writable(options.output_file):
-        _log.error(
-            "Output file location (" + options.output_file + ") " +
-            "is not a writable file.")
-        raise OutputFileAccessError()
-
-
-def report_output_file_size(options, _log, input_file, output_file):
-    try:
-        output_size = Path(output_file).stat().st_size
-        input_size = Path(input_file).stat().st_size
-    except FileNotFoundError:
-        return  # Outputting to stream or something
-    ratio = output_size / input_size
-    if ratio < 1.35 or input_size < 25000:
-        return  # Seems fine
-
-    reasons = []
-    image_preproc = {
-        'deskew',
-        'clean_final',
-        'remove_background',
-        'oversample',
-        'force_ocr'
-    }
-    for arg in image_preproc:
-        attr = getattr(options, arg, None)
-        if not attr:
-            continue
-        reasons.append(
-                "The argument --{} was issued, causing transcoding.".format(
-                        arg.replace('_', '-')))
-
-    if reasons:
-        explanation = (
-            "Possible reasons for this include:\n" + '\n'.join(reasons) + "\n")
-    else:
-        explanation = (
-            "No reason for this increase is known.  Please report this issue.")
-
-    _log.warning(textwrap.dedent("""\
-        The output file size is {:.2f}× larger than the input file.
-        {}
-        """.format(ratio, explanation)))
-
-
-def run_pipeline():
-    options = parser.parse_args()
-    options.verbose_abbreviated_path = 1
+def run(args=None):
+    options = parser.parse_args(args=args)
 
     if not check_closed_streams(options):
         return ExitCode.bad_args
 
-    logger_args = {'verbose': options.verbose, 'quiet': options.quiet}
-
-    _log, _log_mutex = proxy_logger.make_shared_logger_and_proxy(
-        logging_factory, __name__, logger_args)
-    preamble(_log)
-    check_options(options, _log)
-
-    # Complain about qpdf version < 7.0.0
-    # Suppress the warning if in the test suite, since there are no PPAs
-    # for qpdf 7.0.0 for Ubuntu trusty (i.e. Travis)
-    if qpdf.version() < '7.0.0' and not os.environ.get('PYTEST_CURRENT_TEST'):
-        complain(
-            "You are using qpdf version {0} which has known issues including "
-            "security vulnerabilities with certain malformed PDFs. Consider "
-            "upgrading to version 7.0.0 or newer.".format(qpdf.version()))
-
-    # Any changes to options will not take effect for options that are already
-    # bound to function parameters in the pipeline. (For example
-    # options.input_file, options.pdf_renderer are already bound.)
-    if not options.jobs:
-        options.jobs = available_cpu_count()
-
-    # Performance is improved by setting Tesseract to single threaded. In tests
-    # this gives better throughput than letting a smaller number of Tesseract
-    # jobs run multithreaded. Same story for pngquant. Tess <4 ignores this
-    # variable, but harmless to set if ignored.
-    os.environ.setdefault('OMP_THREAD_LIMIT', '1')
-
-    check_environ(options, _log)
-    if os.environ.get('PYTEST_CURRENT_TEST'):
-        os.environ['_OCRMYPDF_TEST_INFILE'] = options.input_file
+    if hasattr(os, 'nice'):
+        os.nice(5)
 
+    verbosity = options.verbose
+    if not os.isatty(sys.stderr.fileno()):
+        options.progress_bar = False
+    if options.quiet:
+        verbosity = Verbosity.quiet
+        options.progress_bar = False
+    configure_logging(
+        verbosity, progress_bar_friendly=options.progress_bar, manage_root_logger=True
+    )
+    log = make_logger('ocrmypdf')
+    log.debug('ocrmypdf ' + __version__)
     try:
-        work_folder = mkdtemp(prefix="com.github.ocrmypdf.")
-        options.history_file = os.path.join(
-            work_folder, 'ruffus_history.sqlite')
-        start_input_file = os.path.join(
-            work_folder, 'origin')
-
-        check_input_file(options, _log, start_input_file)
-        check_requested_output_file(options, _log)
-
-        manager = JobContextManager()
-        manager.register('JobContext', JobContext)  # pylint: disable=no-member
-        manager.start()
-
-        context = manager.JobContext()  # pylint: disable=no-member
-        context.set_options(options)
-        context.set_work_folder(work_folder)
-
-        build_pipeline(options, work_folder, _log, context)
-        atexit.register(cleanup_working_files, work_folder, options)
-        cmdline.run(options)
-    except ruffus_exceptions.RethrownJobError as e:
-        if options.verbose:
-            _log.debug(str(e))  # stringify exception so logger doesn't have to
-        if hasattr(e, 'job_exceptions'):
-            exceptions = e.job_exceptions
-        else:
-            exceptions = e.args  # ruffus 2.6.3
-        exitcode = traverse_ruffus_exception(exceptions, options, _log)
-        if exitcode is None:
-            _log.error("Unexpected ruffus exception: " + str(e))
-            _log.error(repr(e))
-            return ExitCode.other_error
-        return exitcode
-    except ExitCodeException as e:
+        check_options(options)
+    except ValueError as e:
+        log.error(e)
+        return ExitCode.bad_args
+    except BadArgsError as e:
+        log.error(e)
         return e.exit_code
-    except Exception as e:
-        _log.error(str(e))
-        return ExitCode.other_error
+    except MissingDependencyError as e:
+        log.error(e)
+        return ExitCode.missing_dependency
 
-    if options.flowchart:
-        _log.info("Flowchart saved to {}".format(options.flowchart))
-        return ExitCode.ok
-    elif options.output_file == '-':
-        _log.info("Output sent to stdout")
-    elif os.path.samefile(options.output_file, os.devnull):
-        pass  # Say nothing when sending to dev null
-    else:
-        if options.output_type.startswith('pdfa'):
-            pdfa_info = file_claims_pdfa(options.output_file)
-            if pdfa_info['pass']:
-                msg = 'Output file is a {} (as expected)'
-                _log.info(msg.format(pdfa_info['conformance']))
-            else:
-                msg = 'Output file is okay but is not PDF/A (seems to be {})'
-                _log.warning(msg.format(pdfa_info['conformance']))
-                return ExitCode.pdfa_conversion_failed
-        if not qpdf.check(options.output_file, _log):
-            _log.warning('Output file: The generated PDF is INVALID')
-            return ExitCode.invalid_output_pdf
-
-        report_output_file_size(options, _log, start_input_file,
-                                options.output_file)
-
-    pdfinfo = context.get_pdfinfo()
-    if options.verbose:
-        from pprint import pformat
-        _log.debug(pformat(pdfinfo))
-
-    log_page_orientations(pdfinfo, _log)
-
-    return ExitCode.ok
+    result = run_pipeline(options=options)
+    return result
 
 
 if __name__ == '__main__':
-    sys.exit(run_pipeline())
+    sys.exit(run())

src/ocrmypdf/_graft.py

@@ -0,0 +1,283 @@
+# © 2018 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+import os
+from contextlib import suppress
+from pathlib import Path
+
+import pikepdf
+
+from .exec import tesseract
+
+MAX_REPLACE_PAGES = int(os.environ.get('_OCRMYPDF_MAX_REPLACE_PAGES', 100))
+
+
+def _update_page_resources(*, page, font, font_key, procset):
+    """Update this page's fonts with a reference to the Glyphless font"""
+
+    if '/Resources' not in page:
+        page['/Resources'] = pikepdf.Dictionary({})
+    resources = page['/Resources']
+    try:
+        fonts = resources['/Font']
+    except KeyError:
+        fonts = pikepdf.Dictionary({})
+    if font_key is not None and font_key not in fonts:
+        fonts[font_key] = font
+    resources['/Font'] = fonts
+
+    # Reassign /ProcSet to one that just lists everything - ProcSet is
+    # obsolete and doesn't matter but recommended for old viewer support
+    resources['/ProcSet'] = procset
+
+
+def strip_invisible_text(pdf, page):
+    stream = []
+    in_text_obj = False
+    render_mode = 0
+    text_objects = []
+
+    page.page_contents_coalesce()
+    for operands, operator in pikepdf.parse_content_stream(page, ''):
+        if not in_text_obj:
+            if operator == pikepdf.Operator('BT'):
+                in_text_obj = True
+                render_mode = 0
+                text_objects.append((operands, operator))
+            else:
+                stream.append((operands, operator))
+        else:
+            if operator == pikepdf.Operator('Tr'):
+                render_mode = operands[0]
+            text_objects.append((operands, operator))
+            if operator == pikepdf.Operator('ET'):
+                in_text_obj = False
+                if render_mode != 3:
+                    stream.extend(text_objects)
+                text_objects.clear()
+
+    def convert(op):
+        try:
+            return op.unparse()
+        except AttributeError:
+            return str(op).encode('ascii')
+
+    lines = []
+
+    for operands, operator in stream:
+        if operator == pikepdf.Operator('INLINE IMAGE'):
+            iim = operands[0]
+            line = iim.unparse()
+        else:
+            line = b' '.join(convert(op) for op in operands) + b' ' + operator.unparse()
+        lines.append(line)
+
+    content_stream = b'\n'.join(lines)
+    page.Contents = pikepdf.Stream(pdf, content_stream)
+
+
+def _graft_text_layer(
+    *, pdf_base, page_num, text, font, font_key, procset, rotation, strip_old_text, log
+):
+    """Insert the text layer from text page 0 on to pdf_base at page_num"""
+
+    log.debug("Grafting")
+    if Path(text).stat().st_size == 0:
+        return
+
+    # This is a pointer indicating a specific page in the base file
+    pdf_text = pikepdf.open(text)
+    pdf_text_contents = pdf_text.pages[0].Contents.read_bytes()
+
+    base_page = pdf_base.pages.p(page_num)
+
+    # The text page always will be oriented up by this stage but the original
+    # content may have a rotation applied. Wrap the text stream with a rotation
+    # so it will be oriented the same way as the rest of the page content.
+    # (Previous versions OCRmyPDF rotated the content layer to match the text.)
+    mediabox = [float(pdf_text.pages[0].MediaBox[v]) for v in range(4)]
+    wt, ht = mediabox[2] - mediabox[0], mediabox[3] - mediabox[1]
+
+    mediabox = [float(base_page.MediaBox[v]) for v in range(4)]
+    wp, hp = mediabox[2] - mediabox[0], mediabox[3] - mediabox[1]
+
+    translate = pikepdf.PdfMatrix().translated(-wt / 2, -ht / 2)
+    untranslate = pikepdf.PdfMatrix().translated(wp / 2, hp / 2)
+    # -rotation because the input is a clockwise angle and this formula
+    # uses CCW
+    rotation = -rotation % 360
+    rotate = pikepdf.PdfMatrix().rotated(rotation)
+
+    # Because of rounding of DPI, we might get a text layer that is not
+    # identically sized to the target page. Scale to adjust. Normally this
+    # is within 0.998.
+    if rotation in (90, 270):
+        wt, ht = ht, wt
+    scale_x = wp / wt
+    scale_y = hp / ht
+
+    # log.debug('%r', scale_x, scale_y)
+    scale = pikepdf.PdfMatrix().scaled(scale_x, scale_y)
+
+    # Translate the text so it is centered at (0, 0), rotate it there, adjust
+    # for a size different between initial and text PDF, then untranslate
+    ctm = translate @ rotate @ scale @ untranslate
+
+    pdf_text_contents = b'q %s cm\n' % ctm.encode() + pdf_text_contents + b'\nQ\n'
+
+    new_text_layer = pikepdf.Stream(pdf_base, pdf_text_contents)
+
+    if strip_old_text:
+        strip_invisible_text(pdf_base, base_page)
+
+    base_page.page_contents_add(new_text_layer, prepend=True)
+
+    _update_page_resources(
+        page=base_page, font=font, font_key=font_key, procset=procset
+    )
+    pdf_text.close()
+
+
+def _find_font(text, pdf_base):
+    """Copy a font from the filename text into pdf_base"""
+
+    font, font_key = None, None
+    possible_font_names = ('/f-0-0', '/F1')
+    try:
+        with pikepdf.open(text) as pdf_text:
+            try:
+                pdf_text_fonts = pdf_text.pages[0].Resources.get('/Font', {})
+            except (AttributeError, IndexError, KeyError):
+                return None, None
+            for f in possible_font_names:
+                pdf_text_font = pdf_text_fonts.get(f, None)
+                if pdf_text_font is not None:
+                    font_key = f
+                    break
+            if pdf_text_font:
+                font = pdf_base.copy_foreign(pdf_text_font)
+            return font, font_key
+    except (FileNotFoundError, pikepdf.PdfError):
+        # PdfError occurs if a 0-length file is written e.g. due to OCR timeout
+        return None, None
+
+
+class OcrGrafter:
+    def __init__(self, context):
+        self.context = context
+        self.log = context.log
+        self.path_base = Path(context.origin).resolve()
+
+        self.pdf_base = pikepdf.open(self.path_base)
+        self.font, self.font_key = None, None
+
+        self.pdfinfo = context.pdfinfo
+        self.output_file = context.get_path('graft_layers.pdf')
+
+        self.procset = self.pdf_base.make_indirect(
+            pikepdf.Object.parse(b'[ /PDF /Text /ImageB /ImageC /ImageI ]')
+        )
+
+        self.emplacements = 1
+        self.interim_count = 0
+
+    def graft_page(self, page_result):
+        pageno, image, text, _sidecar, autorotate_correction = page_result
+        if text and not self.font:
+            self.font, self.font_key = _find_font(text, self.pdf_base)
+
+        emplaced_page = False
+        content_rotation = self.pdfinfo[pageno].rotation
+        path_image = Path(image).resolve() if image else None
+        if path_image is not None and path_image != self.path_base:
+            # We are updating the old page with a rasterized PDF of the new
+            # page (without changing objgen, to preserve references)
+            self.log.debug("Emplacement update")
+            with pikepdf.open(image) as pdf_image:
+                self.emplacements += 1
+                foreign_image_page = pdf_image.pages[0]
+                self.pdf_base.pages.append(foreign_image_page)
+                local_image_page = self.pdf_base.pages[-1]
+                self.pdf_base.pages[pageno].emplace(local_image_page)
+                del self.pdf_base.pages[-1]
+            emplaced_page = True
+
+        if emplaced_page:
+            content_rotation = autorotate_correction
+        text_rotation = autorotate_correction
+        text_misaligned = (text_rotation - content_rotation) % 360
+        self.log.debug(
+            '%r',
+            [text_rotation, autorotate_correction, text_misaligned, content_rotation],
+        )
+
+        if text and self.font:
+            # Graft the text layer onto this page, whether new or old
+            strip_old = self.context.options.redo_ocr
+            _graft_text_layer(
+                pdf_base=self.pdf_base,
+                page_num=pageno + 1,
+                text=text,
+                font=self.font,
+                font_key=self.font_key,
+                rotation=text_misaligned,
+                procset=self.procset,
+                strip_old_text=strip_old,
+                log=self.log,
+            )
+
+        # Correct the rotation if applicable
+        self.pdf_base.pages[pageno].Rotate = (
+            content_rotation - autorotate_correction
+        ) % 360
+
+        if self.emplacements % MAX_REPLACE_PAGES == 0:
+            self.save_and_reload()
+
+    def save_and_reload(self):
+        # Periodically save and reload the Pdf object. This will keep a
+        # lid on our memory usage for very large files. Attach the font to
+        # page 1 even if page 1 doesn't use it, so we have a way to get it
+        # back.
+        page0 = self.pdf_base.pages[0]
+        _update_page_resources(
+            page=page0, font=self.font, font_key=self.font_key, procset=self.procset
+        )
+
+        # We cannot read and write the same file, that will corrupt it
+        # but we don't to keep more copies than we need to. Delete intermediates.
+        # {interim_count} is the opened file we were updateing
+        # {interim_count - 1} can be deleted
+        # {interim_count + 1} is the new file will produce and open
+        old_file = self.output_file + f'_working{self.interim_count - 1}.pdf'
+        if not self.context.options.keep_temporary_files:
+            with suppress(FileNotFoundError):
+                os.unlink(old_file)
+
+        next_file = self.output_file + f'_working{self.interim_count + 1}.pdf'
+        self.pdf_base.save(next_file)
+        self.pdf_base.close()
+
+        self.pdf_base = pikepdf.open(next_file)
+        self.procset = self.pdf_base.pages[0].Resources.ProcSet
+        self.font, self.font_key = None, None  # Ensure we reacquire this information
+        self.interim_count += 1
+
+    def finalize(self):
+        self.pdf_base.save(self.output_file)
+        self.pdf_base.close()
+        return self.output_file

src/ocrmypdf/_jobcontext.py

@@ -15,70 +15,110 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-from contextlib import suppress
-from multiprocessing.managers import SyncManager
-import sys
+import logging
 import shutil
-
-from .pdfinfo import PdfInfo
+import sys
+import os
 
 
-class JobContext:
-    """Holds our context for a particular run of the pipeline
+class PicklableLoggerMixin:
+    def __init__(self):
+        self._log = None
 
-    A multiprocessing manager effectively creates a separate process
-    that keeps the master job context object.  Other threads access
-    job context via multiprocessing proxy objects.
+    @property
+    def log(self):
+        if not self._log:
+            self._log = self.get_logger()
+        return self._log
 
-    While this would naturally lend itself @property's it seems to make
-    a little more sense to use functions to make it explicitly that the
-    invocation requires marshalling data across a process boundary.
+    def __getstate__(self):
+        # Python 3.6 is incapable of pickling a logger and marshalling it to another
+        # process (threading._RLock error), so we disconnect it before pickling,
+        # and create a new logger in the worker process.
+        state = self.__dict__.copy()
+        state['_log'] = None
+        return state
 
+
+class PDFContext(PicklableLoggerMixin):
+    """Holds our context for a particular run of the pipeline"""
+
+    def __init__(self, options, work_folder, origin, pdfinfo):
+        PicklableLoggerMixin.__init__(self)
+        self.options = options
+        self.work_folder = work_folder
+        self.origin = origin
+        self.pdfinfo = pdfinfo
+        if options:
+            self.name = os.path.basename(options.input_file)
+        else:
+            self.name = 'origin.pdf'
+        if self.name == '-':
+            self.name = 'stdin'
+
+    def get_logger(self):
+        return make_logger(self.options, filename=self.name)
+
+    def get_path(self, name):
+        return os.path.join(self.work_folder, name)
+
+    def get_page_contexts(self):
+        npages = len(self.pdfinfo)
+        for n in range(npages):
+            yield PageContext(self, n)
+
+
+class PageContext(PicklableLoggerMixin):
+    """Holds our context for a page
+
+    Must be pickable, so only store intrinsic/simple data elements
     """
 
-    def __init__(self):
-        self.pdfinfo = None
-        self.options = None
-        self.work_folder = None
-        self.rotations = {}
+    def __init__(self, pdf_context, pageno):
+        PicklableLoggerMixin.__init__(self)
+        self.work_folder = pdf_context.work_folder
+        self.origin = pdf_context.origin
+        self.options = pdf_context.options
+        self.name = pdf_context.name
+        self.pageno = pageno
+        self.pageinfo = pdf_context.pdfinfo[pageno]
+        self._log = None
 
-    def generate_pdfinfo(self, infile):
-        self.pdfinfo = PdfInfo(infile)
+    def get_logger(self):
+        return make_logger(self.options, filename=self.name, page=self.pageno + 1)
 
-    def get_pdfinfo(self):
-        "What we know about the input PDF"
-        return self.pdfinfo
-
-    def set_pdfinfo(self, pdfinfo):
-        self.pdfinfo = pdfinfo
-
-    def get_options(self):
-        return self.options
-
-    def set_options(self, options):
-        self.options = options
-
-    def get_work_folder(self):
-        return self.work_folder
-
-    def set_work_folder(self, work_folder):
-        self.work_folder = work_folder
-
-    def get_rotation(self, pageno):
-        return self.rotations.get(pageno, 0)
-
-    def set_rotation(self, pageno, value):
-        self.rotations[pageno] = value
-
-
-class JobContextManager(SyncManager):
-    pass
+    def get_path(self, name):
+        return os.path.join(self.work_folder, "%06d_%s" % (self.pageno + 1, name))
 
 
 def cleanup_working_files(work_folder, options):
     if options.keep_temporary_files:
-        print("Temporary working files saved at:\n{0}".format(work_folder),
-              file=sys.stderr)
+        print(f"Temporary working files retained at:\n{work_folder}", file=sys.stderr)
     else:
-        with suppress(FileNotFoundError):
-            shutil.rmtree(work_folder)
+        shutil.rmtree(work_folder, ignore_errors=True)
+
+
+class LogNameAdapter(logging.LoggerAdapter):
+    def process(self, msg, kwargs):
+        # return '[%s] %s' % (self.extra['filename'], msg), kwargs
+        return '%s' % (msg,), kwargs
+
+
+class LogNamePageAdapter(logging.LoggerAdapter):
+    def process(self, msg, kwargs):
+        return (
+            #'[%s:%05u] %s' % (self.extra['filename'], self.extra['page'], msg),
+            '%4u: %s' % (self.extra['page'], msg),
+            kwargs,
+        )
+
+
+def make_logger(options=None, prefix='ocrmypdf', filename=None, page=None):
+    log = logging.getLogger(prefix)
+    if filename and page:
+        adapter = LogNamePageAdapter(log, dict(filename=filename, page=page))
+    elif filename:
+        adapter = LogNameAdapter(log, dict(filename=filename))
+    else:
+        adapter = log
+    return adapter

src/ocrmypdf/_pipeline.py

@@ -0,0 +1,811 @@
+# © 2016 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+import os
+import re
+import sys
+from datetime import datetime, timezone
+from shutil import copyfileobj
+
+import img2pdf
+from PIL import Image
+
+import pikepdf
+from pikepdf.models.metadata import encode_pdf_date
+
+from . import PROGRAM_NAME, VERSION, leptonica
+
+from ._plugins import load_plugin
+from .exceptions import (
+    DpiError,
+    EncryptedPdfError,
+    InputFileError,
+    UnsupportedImageFormatError,
+    PriorOcrFoundError,
+)
+from .exec import ghostscript, tesseract
+from .helpers import re_symlink
+from .hocrtransform import HocrTransform
+from .optimize import optimize
+from .pdfa import generate_pdfa_ps
+from .pdfinfo import Colorspace, PdfInfo
+
+VECTOR_PAGE_DPI = 400
+
+
+def triage_image_file(input_file, output_file, options, log):
+    try:
+        log.info("Input file is not a PDF, checking if it is an image...")
+        im = Image.open(input_file)
+    except EnvironmentError as e:
+        # Recover the original filename
+        log.error(str(e).replace(input_file, options.input_file))
+        raise UnsupportedImageFormatError() from e
+    else:
+        log.info("Input file is an image")
+
+        if 'dpi' in im.info:
+            if im.info['dpi'] <= (96, 96) and not options.image_dpi:
+                log.info("Image size: (%d, %d)" % im.size)
+                log.info("Image resolution: (%d, %d)" % im.info['dpi'])
+                log.error(
+                    "Input file is an image, but the resolution (DPI) is "
+                    "not credible.  Estimate the resolution at which the "
+                    "image was scanned and specify it using --image-dpi."
+                )
+                raise DpiError()
+        elif not options.image_dpi:
+            log.info("Image size: (%d, %d)" % im.size)
+            log.error(
+                "Input file is an image, but has no resolution (DPI) "
+                "in its metadata.  Estimate the resolution at which "
+                "image was scanned and specify it using --image-dpi."
+            )
+            raise DpiError()
+
+        if im.mode in ('RGBA', 'LA'):
+            log.error(
+                "The input image has an alpha channel. Remove the alpha "
+                "channel first."
+            )
+            raise UnsupportedImageFormatError()
+
+        if 'iccprofile' not in im.info:
+            if im.mode == 'RGB':
+                log.info('Input image has no ICC profile, assuming sRGB')
+            elif im.mode == 'CMYK':
+                log.info('Input CMYK image has no ICC profile, not usable')
+                raise UnsupportedImageFormatError()
+        im.close()
+
+    try:
+        log.info("Image seems valid. Try converting to PDF...")
+        layout_fun = img2pdf.default_layout_fun
+        if options.image_dpi:
+            layout_fun = img2pdf.get_fixed_dpi_layout_fun(
+                (options.image_dpi, options.image_dpi)
+            )
+        with open(output_file, 'wb') as outf:
+            img2pdf.convert(
+                input_file, layout_fun=layout_fun, with_pdfrw=False, outputstream=outf
+            )
+        log.info("Successfully converted to PDF, processing...")
+    except img2pdf.ImageOpenError as e:
+        log.error(e)
+        raise UnsupportedImageFormatError() from e
+
+
+def _pdf_guess_version(input_file, search_window=1024):
+    """Try to find version signature at start of file.
+
+    Not robust enough to deal with appended files.
+
+    Returns empty string if not found, indicating file is probably not PDF.
+    """
+
+    with open(input_file, 'rb') as f:
+        signature = f.read(search_window)
+    m = re.search(br'%PDF-(\d\.\d)', signature)
+    if m:
+        return m.group(1)
+    return ''
+
+
+def triage(input_file, output_file, options, log):
+    try:
+        if _pdf_guess_version(input_file):
+            if options.image_dpi:
+                log.warning(
+                    "Argument --image-dpi ignored because the "
+                    "input file is a PDF, not an image."
+                )
+            # Origin file is a pdf create a symlink with pdf extension
+            re_symlink(input_file, output_file)
+            return output_file
+    except EnvironmentError as e:
+        log.error(e)
+        raise InputFileError() from e
+
+    triage_image_file(input_file, output_file, options, log)
+    return output_file
+
+
+def get_pdfinfo(input_file, detailed_page_analysis=False, progbar=False):
+    try:
+        return PdfInfo(
+            input_file, detailed_page_analysis=detailed_page_analysis, progbar=progbar
+        )
+    except pikepdf.PasswordError:
+        raise EncryptedPdfError()
+    except pikepdf.PdfError:
+        raise InputFileError()
+
+
+def validate_pdfinfo_options(context):
+    log = context.log
+    pdfinfo = context.pdfinfo
+    options = context.options
+
+    if options.plugin_validation:
+        validate = load_plugin(options.plugin_validation)
+        result = validate(context)
+        if result is not None:
+            return result
+
+    if pdfinfo.needs_rendering:
+        log.error(
+            "This PDF contains dynamic XFA forms created by Adobe LiveCycle "
+            "Designer and can only be read by Adobe Acrobat or Adobe Reader."
+        )
+        raise InputFileError()
+    if pdfinfo.has_userunit and options.output_type.startswith('pdfa'):
+        log.error(
+            "This input file uses a PDF feature that is not supported "
+            "by Ghostscript, so you cannot use --output-type=pdfa for this "
+            "file. (Specifically, it uses the PDF-1.6 /UserUnit feature to "
+            "support very large or small page sizes, and Ghostscript cannot "
+            "output these files.)  Use --output-type=pdf instead."
+        )
+        raise InputFileError()
+    if pdfinfo.has_acroform:
+        if options.redo_ocr:
+            log.error(
+                "This PDF has a user fillable form. --redo-ocr is not "
+                "currently possible on such files."
+            )
+            raise InputFileError()
+        else:
+            log.warn(
+                "This PDF has a fillable form. "
+                "Chances are it is a pure digital "
+                "document that does not need OCR."
+            )
+            if not options.force_ocr:
+                log.info(
+                    "Use the option --force-ocr to produce an image of the "
+                    "form and all filled form fields. The output PDF will be "
+                    "'flattened' and will no longer be fillable."
+                )
+
+
+def get_page_dpi(pageinfo, options):
+    "Get the DPI when nonsquare DPI is tolerable"
+    xres = max(
+        pageinfo.xres or VECTOR_PAGE_DPI,
+        options.oversample or 0,
+        VECTOR_PAGE_DPI if pageinfo.has_vector else 0,
+    )
+    yres = max(
+        pageinfo.yres or VECTOR_PAGE_DPI,
+        options.oversample or 0,
+        VECTOR_PAGE_DPI if pageinfo.has_vector else 0,
+    )
+    return (float(xres), float(yres))
+
+
+def get_page_square_dpi(pageinfo, options):
+    "Get the DPI when we require xres == yres, scaled to physical units"
+    xres = pageinfo.xres or 0
+    yres = pageinfo.yres or 0
+    userunit = pageinfo.userunit or 1
+    return float(
+        max(
+            (xres * userunit) or VECTOR_PAGE_DPI,
+            (yres * userunit) or VECTOR_PAGE_DPI,
+            VECTOR_PAGE_DPI if pageinfo.has_vector else 0,
+            options.oversample or 0,
+        )
+    )
+
+
+def get_canvas_square_dpi(pageinfo, options):
+    """Get the DPI when we require xres == yres, in Postscript units"""
+    return float(
+        max(
+            (pageinfo.xres) or VECTOR_PAGE_DPI,
+            (pageinfo.yres) or VECTOR_PAGE_DPI,
+            VECTOR_PAGE_DPI if pageinfo.has_vector else 0,
+            options.oversample or 0,
+        )
+    )
+
+
+def is_ocr_required(page_context):
+    pageinfo = page_context.pageinfo
+    options = page_context.options
+    log = page_context.log
+
+    ocr_required = True
+
+    if options.pages and pageinfo.pageno not in options.pages:
+        log.debug(f"skipped {pageinfo.pageno} as requested by --pages {options.pages}")
+        ocr_required = False
+    elif pageinfo.has_text:
+        if not options.force_ocr and not (options.skip_text or options.redo_ocr):
+            raise PriorOcrFoundError(
+                "page already has text! - aborting (use --force-ocr to force OCR)"
+            )
+        elif options.force_ocr:
+            log.info("page already has text! - rasterizing text and running OCR anyway")
+            ocr_required = True
+        elif options.redo_ocr:
+            if pageinfo.has_corrupt_text:
+                log.warn(
+                    "some text on this page cannot be mapped to characters: "
+                    "consider using --force-ocr instead"
+                )
+            else:
+                log.info("redoing OCR")
+            ocr_required = True
+        elif options.skip_text:
+            log.info("skipping all processing on this page")
+            ocr_required = False
+    elif not pageinfo.images and not options.lossless_reconstruction:
+        # We found a page with no images and no text. That means it may
+        # have vector art that the user wants to OCR. If we determined
+        # lossless reconstruction is not possible then we have to rasterize
+        # the image. So if OCR is being forced, take that to mean YES, go
+        # ahead and rasterize. If not forced, then pretend there's no text
+        # on the page at all so we don't lose anything.
+        # This could be made smarter by explicitly searching for vector art.
+        if options.force_ocr and options.oversample:
+            # The user really wants to reprocess this file
+            log.info(
+                "page has no images - "
+                f"rasterizing at {options.oversample} DPI because "
+                "--force-ocr --oversample was specified"
+            )
+        elif options.force_ocr:
+            # Warn the user they might not want to do this
+            log.warn(
+                "page has no images - "
+                "all vector content will be "
+                f"rasterized at {VECTOR_PAGE_DPI} DPI, losing some resolution and likely "
+                "increasing file size. Use --oversample to adjust the "
+                "DPI."
+            )
+        else:
+            log.info(
+                "page has no images - "
+                "skipping all processing on this page to avoid losing detail. "
+                "Use --force-ocr if you wish to perform OCR on pages that "
+                "have vector content."
+            )
+            ocr_required = False
+
+    if ocr_required and options.skip_big and pageinfo.images:
+        pixel_count = pageinfo.width_pixels * pageinfo.height_pixels
+        if pixel_count > (options.skip_big * 1_000_000):
+            ocr_required = False
+            log.warn(
+                "page too big, skipping OCR "
+                f"({(pixel_count / 1_000_000):.1f} MPixels > {options.skip_big:.1f} MPixels --skip-big)"
+            )
+    return ocr_required
+
+
+def rasterize_preview(input_file, page_context):
+    output_file = page_context.get_path('rasterize_preview.jpg')
+    canvas_dpi = get_canvas_square_dpi(page_context.pageinfo, page_context.options)
+    page_dpi = get_page_square_dpi(page_context.pageinfo, page_context.options)
+    ghostscript.rasterize_pdf(
+        input_file,
+        output_file,
+        xres=canvas_dpi,
+        yres=canvas_dpi,
+        raster_device='jpeggray',
+        log=page_context.log,
+        page_dpi=(page_dpi, page_dpi),
+        pageno=page_context.pageinfo.pageno + 1,
+    )
+    return output_file
+
+
+def get_orientation_correction(preview, page_context):
+    """
+    Work out orientation correct for each page.
+
+    We ask Ghostscript to draw a preview page, which will rasterize with the
+    current /Rotate applied, and then ask Tesseract which way the page is
+    oriented. If the value of /Rotate is correct (e.g., a user already
+    manually fixed rotation), then Tesseract will say the page is pointing
+    up and the correction is zero. Otherwise, the orientation found by
+    Tesseract represents the clockwise rotation, or the counterclockwise
+    correction to rotation.
+
+    When we draw the real page for OCR, we rotate it by the CCW correction,
+    which points it (hopefully) upright. _graft.py takes care of the orienting
+    the image and text layers.
+
+    """
+
+    orient_conf = tesseract.get_orientation(
+        preview,
+        engine_mode=page_context.options.tesseract_oem,
+        timeout=page_context.options.tesseract_timeout,
+        log=page_context.log,
+        tesseract_env=page_context.options.tesseract_env,
+    )
+
+    direction = {0: '⇧', 90: '⇨', 180: '⇩', 270: '⇦'}
+
+    existing_rotation = page_context.pageinfo.rotation
+
+    correction = orient_conf.angle % 360
+
+    apply_correction = False
+    action = ''
+    if orient_conf.confidence >= page_context.options.rotate_pages_threshold:
+        if correction != 0:
+            apply_correction = True
+            action = ' - will rotate'
+        else:
+            action = ' - rotation appears correct'
+    else:
+        if correction != 0:
+            action = ' - confidence too low to rotate'
+        else:
+            action = ' - no change'
+
+    facing = ''
+    if existing_rotation != 0:
+        facing = 'with existing rotation {}, '.format(
+            direction.get(existing_rotation, '?')
+        )
+    facing += 'page is facing {}'.format(direction.get(orient_conf.angle, '?'))
+
+    page_context.log.debug(
+        '{pagenum:4d}: {facing}, confidence {conf:.2f}{action}'.format(
+            pagenum=page_context.pageinfo.pageno,
+            facing=facing,
+            conf=orient_conf.confidence,
+            action=action,
+        )
+    )
+
+    if apply_correction:
+        return correction
+    return 0
+
+
+def rasterize(
+    input_file, page_context, correction=0, output_tag='', remove_vectors=None
+):
+    colorspaces = ['pngmono', 'pnggray', 'png256', 'png16m']
+    device_idx = 0
+
+    if remove_vectors is None:
+        remove_vectors = page_context.options.remove_vectors
+
+    output_file = page_context.get_path(f'rasterize{output_tag}.png')
+    pageinfo = page_context.pageinfo
+
+    def at_least(cs):
+        return max(device_idx, colorspaces.index(cs))
+
+    for image in pageinfo.images:
+        if image.type_ != 'image':
+            continue  # ignore masks
+        if image.bpc > 1:
+            if image.color == Colorspace.index:
+                device_idx = at_least('png256')
+            elif image.color == Colorspace.gray:
+                device_idx = at_least('pnggray')
+            else:
+                device_idx = at_least('png16m')
+
+    if pageinfo.has_vector:
+        device_idx = at_least('png16m')
+
+    device = colorspaces[device_idx]
+
+    page_context.log.debug(f"Rasterize with {device}")
+
+    # Produce the page image with square resolution or else deskew and OCR
+    # will not work properly.
+    canvas_dpi = get_canvas_square_dpi(pageinfo, page_context.options)
+    page_dpi = get_page_square_dpi(pageinfo, page_context.options)
+
+    ghostscript.rasterize_pdf(
+        input_file,
+        output_file,
+        xres=canvas_dpi,
+        yres=canvas_dpi,
+        raster_device=device,
+        log=page_context.log,
+        page_dpi=(page_dpi, page_dpi),
+        pageno=pageinfo.pageno + 1,
+        rotation=correction,
+        filter_vector=remove_vectors,
+    )
+    return output_file
+
+
+def preprocess_remove_background(input_file, page_context):
+    if any(image.bpc > 1 for image in page_context.pageinfo.images):
+        output_file = page_context.get_path('pp_rm_bg.png')
+        leptonica.remove_background(input_file, output_file)
+        return output_file
+    else:
+        page_context.log.info("background removal skipped on mono page")
+        return input_file
+
+
+def preprocess_deskew(input_file, page_context):
+    output_file = page_context.get_path('pp_deskew.png')
+    dpi = get_page_square_dpi(page_context.pageinfo, page_context.options)
+    leptonica.deskew(input_file, output_file, dpi)
+    return output_file
+
+
+def preprocess_clean(input_file, page_context):
+    from .exec import unpaper
+
+    output_file = page_context.get_path('pp_clean.png')
+    dpi = get_page_square_dpi(page_context.pageinfo, page_context.options)
+    unpaper.clean(
+        input_file,
+        output_file,
+        dpi,
+        page_context.log,
+        page_context.options.unpaper_args,
+    )
+    return output_file
+
+
+def create_ocr_image(image, page_context):
+    """Create the image we send for OCR. May not be the same as the display
+    image depending on preprocessing. This image will never be shown to the
+    user."""
+
+    output_file = page_context.get_path('ocr.png')
+    options = page_context.options
+    with Image.open(image) as im:
+        from PIL import ImageColor
+        from PIL import ImageDraw
+
+        white = ImageColor.getcolor('#ffffff', im.mode)
+        # pink = ImageColor.getcolor('#ff0080', im.mode)
+        draw = ImageDraw.ImageDraw(im)
+
+        xres, yres = im.info['dpi']
+        page_context.log.debug('resolution %r %r' % (xres, yres))
+
+        if not options.force_ocr:
+            # Do not mask text areas when forcing OCR, because we need to OCR
+            # all text areas
+            mask = None  # Exclude both visible and invisible text from OCR
+            if options.redo_ocr:
+                mask = True  # Mask visible text, but not invisible text
+
+            for textarea in page_context.pageinfo.get_textareas(
+                visible=mask, corrupt=None
+            ):
+                # Calculate resolution based on the image size and page dimensions
+                # without regard whatever resolution is in pageinfo (may differ or
+                # be None)
+                bbox = [float(v) for v in textarea]
+                xscale, yscale = float(xres) / 72.0, float(yres) / 72.0
+                pixcoords = [
+                    bbox[0] * xscale,
+                    im.height - bbox[3] * yscale,
+                    bbox[2] * xscale,
+                    im.height - bbox[1] * yscale,
+                ]
+                pixcoords = [int(round(c)) for c in pixcoords]
+                page_context.log.debug('blanking %r', pixcoords)
+                draw.rectangle(pixcoords, fill=white)
+                # draw.rectangle(pixcoords, outline=pink)
+
+        if options.threshold:
+            pix = leptonica.Pix.frompil(im)
+            pix = pix.masked_threshold_on_background_norm()
+            im = pix.topil()
+
+        if options.filter_ocr_image:
+            filt = load_plugin(options.filter_ocr_image)
+            im = filt(im)
+
+        del draw
+        # Pillow requires integer DPI
+        dpi = round(xres), round(yres)
+        im.save(output_file, dpi=dpi)
+    return output_file
+
+
+def ocr_tesseract_hocr(input_file, page_context):
+    hocr_out = page_context.get_path('ocr_hocr.hocr')
+    hocr_text_out = page_context.get_path('ocr_hocr.txt')
+    options = page_context.options
+    tesseract.generate_hocr(
+        input_file=input_file,
+        output_files=[hocr_out, hocr_text_out],
+        language=options.language,
+        engine_mode=options.tesseract_oem,
+        tessconfig=options.tesseract_config,
+        timeout=options.tesseract_timeout,
+        pagesegmode=options.tesseract_pagesegmode,
+        user_words=options.user_words,
+        user_patterns=options.user_patterns,
+        tesseract_env=options.tesseract_env,
+        log=page_context.log,
+    )
+    return (hocr_out, hocr_text_out)
+
+
+def should_visible_page_image_use_jpg(pageinfo):
+    # If all images were JPEGs originally, produce a JPEG as output
+    return pageinfo.images and all(im.enc == 'jpeg' for im in pageinfo.images)
+
+
+def create_visible_page_jpg(image, page_context):
+    output_file = page_context.get_path('visible.jpg')
+    with Image.open(image) as im:
+        # At this point the image should be a .png, but deskew, unpaper
+        # might have removed the DPI information. In this case, fall back to
+        # square DPI used to rasterize. When the preview image was
+        # rasterized, it was also converted to square resolution, which is
+        # what we want to give tesseract, so keep it square.
+        fallback_dpi = get_page_square_dpi(page_context.pageinfo, page_context.options)
+        dpi = im.info.get('dpi', (fallback_dpi, fallback_dpi))
+
+        # Pillow requires integer DPI
+        dpi = round(dpi[0]), round(dpi[1])
+        im.save(output_file, format='JPEG', dpi=dpi)
+    return output_file
+
+
+def create_pdf_page_from_image(image, page_context):
+    # We rasterize a square DPI version of each page because most image
+    # processing tools don't support rectangular DPI. Use the square DPI as it
+    # accurately describes the image. It would be possible to resample the image
+    # at this stage back to non-square DPI to more closely resemble the input,
+    # except that the hocr renderer does not understand non-square DPI. The
+    # sandwich renderer would be fine.
+    output_file = page_context.get_path('visible.pdf')
+    dpi = get_page_square_dpi(page_context.pageinfo, page_context.options)
+    layout_fun = img2pdf.get_fixed_dpi_layout_fun((dpi, dpi))
+
+    # This create a single page PDF
+    with open(image, 'rb') as imfile, open(output_file, 'wb') as pdf:
+        page_context.log.debug('convert')
+        img2pdf.convert(
+            imfile, with_pdfrw=False, layout_fun=layout_fun, outputstream=pdf
+        )
+        page_context.log.debug('convert done')
+    return output_file
+
+
+def render_hocr_page(hocr, page_context):
+    output_file = page_context.get_path('ocr_hocr.pdf')
+    dpi = get_page_square_dpi(page_context.pageinfo, page_context.options)
+    hocrtransform = HocrTransform(hocr, dpi)
+    hocrtransform.to_pdf(
+        output_file,
+        imageFileName=None,
+        showBoundingboxes=False,
+        invisibleText=True,
+        interwordSpaces=True,
+    )
+    return output_file
+
+
+def ocr_tesseract_textonly_pdf(input_image, page_context):
+    output_pdf = page_context.get_path('ocr_tess.pdf')
+    output_text = page_context.get_path('ocr_tess.txt')
+    options = page_context.options
+    tesseract.generate_pdf(
+        input_image=input_image,
+        skip_pdf=None,
+        output_pdf=output_pdf,
+        output_text=output_text,
+        language=options.language,
+        engine_mode=options.tesseract_oem,
+        text_only=True,
+        tessconfig=options.tesseract_config,
+        timeout=options.tesseract_timeout,
+        pagesegmode=options.tesseract_pagesegmode,
+        user_words=options.user_words,
+        user_patterns=options.user_patterns,
+        tesseract_env=options.tesseract_env,
+        log=page_context.log,
+    )
+    return (output_pdf, output_text)
+
+
+def get_docinfo(base_pdf, options):
+    def from_document_info(key):
+        try:
+            s = base_pdf.docinfo[key]
+            return str(s)
+        except (KeyError, TypeError):
+            return ''
+
+    pdfmark = {
+        k: from_document_info(k)
+        for k in ('/Title', '/Author', '/Keywords', '/Subject', '/CreationDate')
+    }
+    renderer_tag = 'OCR'
+    if options is not None:
+        if options.title:
+            pdfmark['/Title'] = options.title
+        if options.author:
+            pdfmark['/Author'] = options.author
+        if options.keywords:
+            pdfmark['/Keywords'] = options.keywords
+        if options.subject:
+            pdfmark['/Subject'] = options.subject
+
+        if options.pdf_renderer == 'sandwich':
+            renderer_tag = 'OCR-PDF'
+
+    pdfmark['/Creator'] = (
+        f'{PROGRAM_NAME} {VERSION} / ' f'Tesseract {renderer_tag} {tesseract.version()}'
+    )
+    pdfmark['/Producer'] = f'pikepdf {pikepdf.__version__}'
+    if 'OCRMYPDF_CREATOR' in os.environ:
+        pdfmark['/Creator'] = os.environ['OCRMYPDF_CREATOR']
+    if 'OCRMYPDF_PRODUCER' in os.environ:
+        pdfmark['/Producer'] = os.environ['OCRMYPDF_PRODUCER']
+
+    pdfmark['/ModDate'] = encode_pdf_date(datetime.now(timezone.utc))
+    return pdfmark
+
+
+def generate_postscript_stub(context):
+    output_file = context.get_path('pdfa.ps')
+    generate_pdfa_ps(output_file)
+    return output_file
+
+
+def convert_to_pdfa(input_pdf, input_ps_stub, context):
+    options = context.options
+    input_pdfinfo = context.pdfinfo
+    output_file = context.get_path('pdfa.pdf')
+
+    # If the DocumentInfo record contains NUL characters, Ghostscript will
+    # produce XMP metadata which contains invalid XML entities (&#0;).
+    # NULs in DocumentInfo seem to be common since older Acrobats included them.
+    # pikepdf can deal with this, but we make the world a better place by
+    # stamping them out as soon as possible.
+    with pikepdf.open(input_pdf) as pdf_file:
+        if pdf_file.docinfo:
+            modified = False
+            for k, v in pdf_file.docinfo.items():
+                if b'\x00' in bytes(v):
+                    pdf_file.docinfo[k] = bytes(v).replace(b'\x00', b'')
+                    modified = True
+            if modified:
+                pdf_file.save(input_pdf)
+
+    ghostscript.generate_pdfa(
+        pdf_version=input_pdfinfo.min_version,
+        pdf_pages=[input_pdf, input_ps_stub],
+        output_file=output_file,
+        compression=options.pdfa_image_compression,
+        log=context.log,
+        pdfa_part=options.output_type[-1],  # is pdfa-1, pdfa-2, or pdfa-3
+    )
+
+    return output_file
+
+
+def metadata_fixup(working_file, context):
+    output_file = context.get_path('metafix.pdf')
+    options = context.options
+    original = pikepdf.open(context.origin)
+    docinfo = get_docinfo(original, options)
+    pdf = pikepdf.open(working_file)
+    with pdf.open_metadata() as meta:
+        meta.load_from_docinfo(docinfo, delete_missing=False)
+        # If xmp:CreateDate is missing, set it to the modify date to
+        # match Ghostscript, for consistency
+        if 'xmp:CreateDate' not in meta:
+            meta['xmp:CreateDate'] = meta.get('xmp:ModifyDate', '')
+
+        meta_original = original.open_metadata()
+        not_copied = set(meta_original.keys()) - set(meta.keys())
+        if not_copied:
+            if options.output_type.startswith('pdfa'):
+                context.log.warning(
+                    "Some input metadata could not be copied because it is not "
+                    "permitted in PDF/A. You may wish to examine the output "
+                    "PDF's XMP metadata."
+                )
+                context.log.debug(
+                    "The following metadata fields were not copied: %r", not_copied
+                )
+            else:
+                context.log.error(
+                    "Some input metadata could not be copied."
+                    "You may wish to examine the output PDF's XMP metadata."
+                )
+                context.log.info(
+                    "The following metadata fields were not copied: %r", not_copied
+                )
+
+    pdf.save(
+        output_file,
+        compress_streams=True,
+        object_stream_mode=pikepdf.ObjectStreamMode.generate,
+    )
+    original.close()
+    pdf.close()
+    return output_file
+
+
+def optimize_pdf(input_file, context):
+    output_file = context.get_path('optimize.pdf')
+    optimize(input_file, output_file, context)
+    return output_file
+
+
+def merge_sidecars(txt_files, context):
+    output_file = context.get_path('sidecar.txt')
+    with open(output_file, 'w', encoding="utf-8") as stream:
+        for page_num, txt_file in enumerate(txt_files):
+            if page_num != 0:
+                stream.write('\f')  # Form feed between pages
+            if txt_file:
+                with open(txt_file, 'r', encoding="utf-8") as in_:
+                    txt = in_.read()
+                    # Tesseract v4 alpha started adding form feeds in
+                    # commit aa6eb6b
+                    # No obvious way to detect what binaries will do this, so
+                    # for consistency just ignore its form feeds and insert our
+                    # own
+                    if txt.endswith('\f'):
+                        stream.write(txt[:-1])
+                    else:
+                        stream.write(txt)
+            else:
+                stream.write(f'[OCR skipped on page {(page_num + 1)}]')
+    return output_file
+
+
+def copy_final(input_file, output_file, context):
+    context.log.debug('%s -> %s', input_file, output_file)
+    with open(input_file, 'rb') as input_stream:
+        if output_file == '-':
+            copyfileobj(input_stream, sys.stdout.buffer)
+            sys.stdout.flush()
+        else:
+            # At this point we overwrite the output_file specified by the user
+            # use copyfileobj because then we use open() to create the file and
+            # get the appropriate umask, ownership, etc.
+            with open(output_file, 'wb') as output_stream:
+                copyfileobj(input_stream, output_stream)

src/ocrmypdf/_plugins.py

@@ -0,0 +1,79 @@
+# © 2019 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+import logging
+import importlib
+import os
+import sys
+from pathlib import Path
+
+log = logging.getLogger(__name__)
+
+
+def _load_object_from_module(location):
+    """Load a object given a module location
+
+    For location=a.b.c, will effectively run "from a.b import c"
+
+    Example:
+        _load_object_from_module("a.b.c")
+
+    """
+    module_parts = location.split('.')
+    module_name = '.'.join(module_parts[:-1])
+    object_name = module_parts[-1]
+    module = importlib.import_module(module_name)
+    obj = getattr(module, object_name)
+    log.debug(f"Loaded object: from {module_name} import {object_name}")
+    return obj
+
+
+def _load_object_from_pyfile(location):
+    """Load a object from a file
+
+    Example:
+        _load_object_from_pyfile("test.py::blur_filter")
+    """
+    filename, object_name = location.split('::', maxsplit=1)
+    log.debug(f"Loading object {object_name} from {filename}")
+
+    module_name = Path(filename).stem
+    spec = importlib.util.spec_from_file_location(module_name, filename)
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    obj = getattr(module, object_name)
+    return obj
+
+
+def load_plugin(plugin):
+    if callable(plugin):
+        return plugin
+
+    if not isinstance(plugin, str):
+        raise TypeError()
+
+    if '::' not in plugin:
+        plugin = _load_object_from_module(plugin)
+    else:
+        plugin = _load_object_from_pyfile(plugin)
+
+    return plugin
+
+
+def check_plugin_loadable(plugin):
+    load_plugin(plugin)
+    return plugin

src/ocrmypdf/_sync.py

@@ -0,0 +1,381 @@
+# © 2016 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+import logging
+import logging.handlers
+import multiprocessing
+import os
+import signal
+import sys
+import threading
+from collections import namedtuple
+from tempfile import mkdtemp
+
+from tqdm import tqdm
+
+from . import __version__
+from ._jobcontext import PDFContext, cleanup_working_files, make_logger
+from ._pipeline import (
+    convert_to_pdfa,
+    copy_final,
+    create_ocr_image,
+    create_pdf_page_from_image,
+    create_visible_page_jpg,
+    generate_postscript_stub,
+    get_orientation_correction,
+    get_pdfinfo,
+    is_ocr_required,
+    merge_sidecars,
+    metadata_fixup,
+    ocr_tesseract_hocr,
+    ocr_tesseract_textonly_pdf,
+    optimize_pdf,
+    preprocess_clean,
+    preprocess_deskew,
+    preprocess_remove_background,
+    rasterize,
+    rasterize_preview,
+    render_hocr_page,
+    should_visible_page_image_use_jpg,
+    triage,
+    validate_pdfinfo_options,
+)
+from ._validation import (
+    check_requested_output_file,
+    create_input_file,
+    report_output_file_size,
+)
+from ._graft import OcrGrafter
+from .exceptions import ExitCode, ExitCodeException
+from .exec import qpdf
+from .helpers import available_cpu_count
+from .pdfa import file_claims_pdfa
+
+PageResult = namedtuple(
+    'PageResult', 'pageno, pdf_page_from_image, ocr, text, orientation_correction'
+)
+
+
+def preprocess(page_context, image, remove_background, deskew, clean):
+    if remove_background:
+        image = preprocess_remove_background(image, page_context)
+    if deskew:
+        image = preprocess_deskew(image, page_context)
+    if clean:
+        image = preprocess_clean(image, page_context)
+    return image
+
+
+def exec_page_sync(page_context):
+    options = page_context.options
+    orientation_correction = 0
+    pdf_page_from_image_out = None
+    ocr_out = None
+    text_out = None
+    if is_ocr_required(page_context):
+        if options.rotate_pages:
+            # Rasterize
+            rasterize_preview_out = rasterize_preview(page_context.origin, page_context)
+            orientation_correction = get_orientation_correction(
+                rasterize_preview_out, page_context
+            )
+
+        rasterize_out = rasterize(
+            page_context.origin,
+            page_context,
+            correction=orientation_correction,
+            remove_vectors=False,
+        )
+
+        if not any([options.clean, options.clean_final, options.remove_vectors]):
+            ocr_image = preprocess_out = preprocess(
+                page_context,
+                rasterize_out,
+                options.remove_background,
+                options.deskew,
+                clean=False,
+            )
+        else:
+            if not options.lossless_reconstruction:
+                preprocess_out = preprocess(
+                    page_context,
+                    rasterize_out,
+                    options.remove_background,
+                    options.deskew,
+                    clean=options.clean_final,
+                )
+            if options.remove_vectors:
+                rasterize_ocr_out = rasterize(
+                    page_context.origin,
+                    page_context,
+                    correction=orientation_correction,
+                    remove_vectors=True,
+                    output_tag='_ocr',
+                )
+            else:
+                rasterize_ocr_out = rasterize_out
+            ocr_image = preprocess(
+                page_context,
+                rasterize_ocr_out,
+                options.remove_background,
+                options.deskew,
+                clean=options.clean,
+            )
+
+        ocr_image_out = create_ocr_image(ocr_image, page_context)
+
+        pdf_page_from_image_out = None
+        if not options.lossless_reconstruction:
+            visible_image_out = preprocess_out
+            if should_visible_page_image_use_jpg(page_context.pageinfo):
+                visible_image_out = create_visible_page_jpg(
+                    visible_image_out, page_context
+                )
+            pdf_page_from_image_out = create_pdf_page_from_image(
+                visible_image_out, page_context
+            )
+
+        if options.pdf_renderer == 'hocr':
+            (hocr_out, text_out) = ocr_tesseract_hocr(ocr_image_out, page_context)
+            ocr_out = render_hocr_page(hocr_out, page_context)
+
+        if options.pdf_renderer == 'sandwich':
+            (ocr_out, text_out) = ocr_tesseract_textonly_pdf(
+                ocr_image_out, page_context
+            )
+
+    return PageResult(
+        pageno=page_context.pageno,
+        pdf_page_from_image=pdf_page_from_image_out,
+        ocr=ocr_out,
+        text=text_out,
+        orientation_correction=orientation_correction,
+    )
+
+
+def post_process(pdf_file, context):
+    pdf_out = pdf_file
+    if context.options.output_type.startswith('pdfa'):
+        ps_stub_out = generate_postscript_stub(context)
+        pdf_out = convert_to_pdfa(pdf_out, ps_stub_out, context)
+
+    pdf_out = metadata_fixup(pdf_out, context)
+    return optimize_pdf(pdf_out, context)
+
+
+def worker_init(queue):
+    """Initialize a process pool worker"""
+
+    # Ignore SIGINT (our parent process will kill us gracefully)
+    signal.signal(signal.SIGINT, signal.SIG_IGN)
+
+    # Reconfigure the root logger for this process to send all messages to a queue
+    h = logging.handlers.QueueHandler(queue)
+    root = logging.getLogger()
+    root.handlers = []
+    root.addHandler(h)
+
+
+def worker_thread_init(queue):
+    pass
+
+
+def log_listener(queue):
+    """Listen to the worker processes and forward the messages to logging
+
+    For simplicity this is a thread rather than a process. Only one process
+    should actually write to sys.stderr or whatever we're using, so if this is
+    made into a process the main application needs to be directed to it.
+
+    See https://docs.python.org/3/howto/logging-cookbook.html#logging-to-a-single-file-from-multiple-processes
+    """
+
+    while True:
+        try:
+            record = queue.get()
+            if record is None:
+                break
+            logger = logging.getLogger(record.name)
+            logger.handle(record)
+        except Exception:
+            import traceback
+
+            print("Logging problem", file=sys.stderr)
+            traceback.print_exc(file=sys.stderr)
+
+
+def exec_concurrent(context):
+    """Execute the pipeline concurrently"""
+
+    # Run exec_page_sync on every page context
+    max_workers = min(len(context.pdfinfo), context.options.jobs)
+    if max_workers > 1:
+        context.log.info("Start processing %d pages concurrent", max_workers)
+
+    # Tesseract 4.0 is multithreaded, and we also run multiple workers. We want to
+    # avoid the situation where we end up trying to run NxN jobs on N CPU cores,
+    # as that gives poor performance. Performance testing shows we're better off
+    # parallelizing ocrmypdf and forcing Tesseract to be single threaded, which we
+    # get by setting the envvar OMP_THREAD_LIMIT to 1. But if the page count of the
+    # input file is small, then we allow Tesseract to use threads, subject to the
+    # constraint: (ocrmypdf workers) * (tesseract threads) <= max_workers and limiting
+    # Tesseract to 4 threads.
+    tess_threads = min(4, context.options.jobs // max_workers)
+    if context.options.tesseract_env is None:
+        context.options.tesseract_env = os.environ.copy()
+    context.options.tesseract_env.setdefault('OMP_THREAD_LIMIT', str(tess_threads))
+    if tess_threads > 1:
+        context.log.info("Using Tesseract OpenMP thread limit %d", tess_threads)
+
+    if context.options.use_threads:
+        from multiprocessing.dummy import Pool
+
+        initializer = worker_thread_init
+    else:
+        Pool = multiprocessing.Pool
+        initializer = worker_init
+
+    sidecars = [None] * len(context.pdfinfo)
+    ocrgraft = OcrGrafter(context)
+
+    log_queue = multiprocessing.Queue(-1)
+    listener = threading.Thread(target=log_listener, args=(log_queue,))
+    listener.start()
+    with tqdm(
+        total=(2 * len(context.pdfinfo)),
+        desc='OCR',
+        unit='page',
+        unit_scale=0.5,
+        disable=not context.options.progress_bar,
+    ) as pbar, Pool(
+        processes=max_workers, initializer=initializer, initargs=(log_queue,)
+    ) as pool:
+        results = pool.imap_unordered(exec_page_sync, context.get_page_contexts())
+        while True:
+            try:
+                page_result = results.next()
+                sidecars[page_result.pageno] = page_result.text
+                pbar.update()
+                ocrgraft.graft_page(page_result)
+                pbar.update()
+            except StopIteration:
+                break
+            except (Exception, KeyboardInterrupt):
+                pool.terminate()
+                log_queue.put_nowait(None)  # Terminate log listener
+                # Don't try listener.join() here, will deadlock
+                raise
+
+    log_queue.put_nowait(None)
+    listener.join()
+
+    # Output sidecar text
+    if context.options.sidecar:
+        text = merge_sidecars(sidecars, context)
+        # Copy text file to destination
+        copy_final(text, context.options.sidecar, context)
+
+    # Merge layers to one single pdf
+    pdf = ocrgraft.finalize()
+
+    # PDF/A and metadata
+    pdf = post_process(pdf, context)
+
+    # Copy PDF file to destination
+    copy_final(pdf, context.options.output_file, context)
+
+
+class NeverRaise(Exception):
+    """An exception that is never raised"""
+
+    pass
+
+
+def run_pipeline(options, api=False):
+    log = make_logger(options, __name__)
+
+    # Any changes to options will not take effect for options that are already
+    # bound to function parameters in the pipeline. (For example
+    # options.input_file, options.pdf_renderer are already bound.)
+    if not options.jobs:
+        options.jobs = available_cpu_count()
+
+    work_folder = mkdtemp(prefix="com.github.ocrmypdf.")
+    try:
+        check_requested_output_file(options)
+        start_input_file = create_input_file(options, work_folder)
+
+        # Triage image or pdf
+        origin_pdf = triage(
+            start_input_file, os.path.join(work_folder, 'origin.pdf'), options, log
+        )
+
+        # Gather pdfinfo and create context
+        pdfinfo = get_pdfinfo(
+            origin_pdf,
+            detailed_page_analysis=options.redo_ocr,
+            progbar=options.progress_bar,
+        )
+        context = PDFContext(options, work_folder, origin_pdf, pdfinfo)
+
+        # Validate options are okay for this pdf
+        validate_pdfinfo_options(context)
+
+        # Execute the pipeline
+        exec_concurrent(context)
+    except (KeyboardInterrupt if not api else NeverRaise) as e:
+        if options.verbose >= 1:
+            log.exception("KeyboardInterrupt")
+        else:
+            log.error("KeyboardInterrupt")
+        return ExitCode.ctrl_c
+    except (ExitCodeException if not api else NeverRaise) as e:
+        if str(e):
+            log.error("%s: %s", type(e).__name__, str(e))
+        else:
+            log.error(type(e).__name__)
+        return e.exit_code
+    except (Exception if not api else NeverRaise) as e:
+        log.exception("An exception occurred while executing the pipeline")
+        return ExitCode.other_error
+    finally:
+        cleanup_working_files(work_folder, options)
+
+    if options.output_file == '-':
+        log.info("Output sent to stdout")
+    elif os.path.samefile(options.output_file, os.devnull):
+        pass  # Say nothing when sending to dev null
+    else:
+        if options.output_type.startswith('pdfa'):
+            pdfa_info = file_claims_pdfa(options.output_file)
+            if pdfa_info['pass']:
+                msg = f"Output file is a {pdfa_info['conformance']} (as expected)"
+                log.info(msg)
+            else:
+                msg = (
+                    f"Output file is okay but is not PDF/A "
+                    f"(seems to be {pdfa_info['conformance']})"
+                )
+                log.warning(msg)
+                return ExitCode.pdfa_conversion_failed
+        if not qpdf.check(options.output_file, log):
+            log.warning('Output file: The generated PDF is INVALID')
+            return ExitCode.invalid_output_pdf
+
+        report_output_file_size(options, start_input_file, options.output_file)
+
+    return ExitCode.ok

src/ocrmypdf/_unicodefun.py

@@ -33,12 +33,12 @@
 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 
+import codecs
 import os
 import sys
-import codecs
 
 
-def verify_python3_env():
+def verify_python3_env():  # pragma: no cover
     """Ensures that the environment is good for unicode on Python 3."""
 
     # PEP 538 changes in Python 3.7 should make this wrangling unnecessary
@@ -47,6 +47,7 @@ def verify_python3_env():
 
     try:
         import locale
+
         fs_enc = codecs.lookup(locale.getpreferredencoding()).name
     except Exception:
         fs_enc = 'ascii'
@@ -56,8 +57,10 @@ def verify_python3_env():
     extra = ''
     if os.name == 'posix':
         import subprocess
-        rv = subprocess.Popen(['locale', '-a'], stdout=subprocess.PIPE,
-                              stderr=subprocess.PIPE).communicate()[0]
+
+        rv = subprocess.run(
+            ['locale', '-a'], stdout=subprocess.PIPE, stderr=subprocess.PIPE
+        ).stdout
         good_locales = set()
         has_c_utf8 = False
 
@@ -108,6 +111,8 @@ def verify_python3_env():
                 'is not supported'
             ) % bad_locale
 
-    raise RuntimeError('ocrmypdf will abort further execution because Python 3 '
-                       'was configured to use ASCII as encoding for the '
-                       'environment.' + extra)
+    raise RuntimeError(
+        'ocrmypdf will abort further execution because Python 3 '
+        'was configured to use ASCII as encoding for the '
+        'environment.' + extra
+    )

src/ocrmypdf/_validation.py

@@ -0,0 +1,445 @@
+#!/usr/bin/env python3
+# © 2015-17 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+
+import logging
+import os
+import sys
+from pathlib import Path
+from shutil import copyfileobj
+
+import PIL
+
+from ._unicodefun import verify_python3_env
+from .exceptions import (
+    BadArgsError,
+    InputFileError,
+    MissingDependencyError,
+    OutputFileAccessError,
+)
+from .exec import (
+    check_external_program,
+    ghostscript,
+    jbig2enc,
+    pngquant,
+    qpdf,
+    tesseract,
+    unpaper,
+)
+from .helpers import is_file_writable, re_symlink, is_iterable_notstr, monotonic
+
+# -------------
+# External dependencies
+
+HOCR_OK_LANGS = frozenset(['eng', 'deu', 'spa', 'ita', 'por'])
+
+log = logging.getLogger(__name__)
+
+
+# --------
+# Critical environment tests
+verify_python3_env()
+
+
+def check_options_languages(options):
+    if not options.language:
+        options.language = ['eng']  # Enforce English hegemony
+
+    # Support v2.x "eng+deu" language syntax
+    if '+' in options.language[0]:
+        options.language = options.language[0].split('+')
+
+    languages = set(options.language)
+    if not languages.issubset(tesseract.languages()):
+        msg = (
+            "The installed version of tesseract does not have language "
+            "data for the following requested languages: \n"
+        )
+        for lang in languages - tesseract.languages():
+            msg += lang + '\n'
+        raise MissingDependencyError(msg)
+
+
+def check_options_output(options):
+    # We have these constraints to check for.
+    # 1. Ghostscript < 9.20 mangles multibyte Unicode
+    # 2. hocr doesn't work on non-Latin languages (so don't select it)
+
+    languages = set(options.language)
+    is_latin = languages.issubset(HOCR_OK_LANGS)
+
+    if options.pdf_renderer == 'hocr' and not is_latin:
+        msg = (
+            "The 'hocr' PDF renderer is known to cause problems with one "
+            "or more of the languages in your document.  Use "
+            "--pdf-renderer auto (the default) to avoid this issue."
+        )
+        log.warning(msg)
+
+    if ghostscript.version() < '9.20' and options.output_type != 'pdf' and not is_latin:
+        # https://bugs.ghostscript.com/show_bug.cgi?id=696874
+        # Ghostscript < 9.20 fails to encode multibyte characters properly
+        msg = (
+            "The installed version of Ghostscript does not work correctly "
+            "with the OCR languages you specified. Use --output-type pdf or "
+            "upgrade to Ghostscript 9.20 or later to avoid this issue."
+        )
+        msg += f"Found Ghostscript {ghostscript.version()}"
+        log.warning(msg)
+
+    # Decide on what renderer to use
+    if options.pdf_renderer == 'auto':
+        options.pdf_renderer = 'sandwich'
+
+    if options.pdf_renderer == 'sandwich' and not tesseract.has_textonly_pdf(
+        options.tesseract_env
+    ):
+        raise MissingDependencyError(
+            "You are using an alpha version of Tesseract 4.0 that does not support "
+            "the textonly_pdf parameter. We don't support versions this old."
+        )
+
+    if options.output_type == 'pdfa':
+        options.output_type = 'pdfa-2'
+
+    if options.output_type == 'pdfa-3' and ghostscript.version() < '9.19':
+        raise MissingDependencyError(
+            "--output-type pdfa-3 requires Ghostscript 9.19 or later"
+        )
+
+    lossless_reconstruction = False
+    if not any(
+        (
+            options.deskew,
+            options.clean_final,
+            options.force_ocr,
+            options.remove_background,
+        )
+    ):
+        lossless_reconstruction = True
+    options.lossless_reconstruction = lossless_reconstruction
+
+    if not options.lossless_reconstruction and options.redo_ocr:
+        raise BadArgsError(
+            "--redo-ocr is not currently compatible with --deskew, "
+            "--clean-final, and --remove-background"
+        )
+
+
+def check_options_sidecar(options):
+    if options.sidecar == '\0':
+        if options.output_file == '-':
+            raise BadArgsError(
+                "--sidecar filename must be specified when output file is stdout."
+            )
+        options.sidecar = options.output_file + '.txt'
+
+
+def check_options_preprocessing(options):
+    if options.clean_final:
+        options.clean = True
+    if options.unpaper_args and not options.clean:
+        raise BadArgsError("--clean is required for --unpaper-args")
+    if options.clean:
+        check_external_program(
+            program='unpaper',
+            package='unpaper',
+            version_checker=unpaper.version,
+            need_version='6.1',
+            required_for=['--clean, --clean-final'],
+        )
+        try:
+            if options.unpaper_args:
+                options.unpaper_args = unpaper.validate_custom_args(
+                    options.unpaper_args
+                )
+        except Exception as e:
+            raise BadArgsError(str(e))
+
+
+def _pages_from_ranges(ranges):
+    if is_iterable_notstr(ranges):
+        return set(ranges)
+    pages = []
+    page_groups = ranges.replace(' ', '').split(',')
+    for g in page_groups:
+        if not g:
+            continue
+        try:
+            start, end = g.split('-')
+        except ValueError:
+            pages.append(int(g) - 1)
+        else:
+            pages.extend(range(int(start) - 1, int(end)))
+
+    if not monotonic(pages):
+        log.warning(
+            "List of pages to process contains duplicate pages, or pages that are "
+            "out of order"
+        )
+    if any(page < 0 for page in pages):
+        raise BadArgsError("pages refers to a page number less than 1")
+
+    log.debug("OCRing only these pages: %s", pages)
+    return set(pages)
+
+
+def check_options_ocr_behavior(options):
+    exclusive_options = sum(
+        [
+            (1 if opt else 0)
+            for opt in (options.force_ocr, options.skip_text, options.redo_ocr)
+        ]
+    )
+    if exclusive_options >= 2:
+        raise BadArgsError("Choose only one of --force-ocr, --skip-text, --redo-ocr.")
+    if options.pages and options.sidecar:
+        raise BadArgsError("--pages and --sidecar are mutually exclusive")
+    if options.pages:
+        options.pages = _pages_from_ranges(options.pages)
+
+
+def check_options_optimizing(options):
+    if options.optimize >= 2:
+        check_external_program(
+            program='pngquant',
+            package='pngquant',
+            version_checker=pngquant.version,
+            need_version='2.0.1',
+            required_for='--optimize {2,3}',
+        )
+
+    if options.optimize >= 2:
+        # Although we use JBIG2 for optimize=1, don't nag about it unless the
+        # user is asking for more optimization
+        check_external_program(
+            program='jbig2',
+            package='jbig2enc',
+            version_checker=jbig2enc.version,
+            need_version='0.28',
+            required_for='--optimize {2,3} | --jbig2-lossy',
+            recommended=True if not options.jbig2_lossy else False,
+        )
+
+    if options.optimize == 0 and any(
+        [options.jbig2_lossy, options.png_quality, options.jpeg_quality]
+    ):
+        log.warning(
+            "The arguments --jbig2-lossy, --png-quality, and --jpeg-quality "
+            "will be ignored because --optimize=0."
+        )
+
+
+def check_options_advanced(options):
+    if options.pdfa_image_compression != 'auto' and options.output_type.startswith(
+        'pdfa'
+    ):
+        log.warning(
+            "--pdfa-image-compression argument has no effect when "
+            "--output-type is not 'pdfa', 'pdfa-1', or 'pdfa-2'"
+        )
+    if tesseract.v4(options.tesseract_env) and (
+        options.user_words or options.user_patterns
+    ):
+        log.warning('Tesseract 4.x ignores --user-words, so this has no effect')
+
+
+def check_options_metadata(options):
+    import unicodedata
+
+    docinfo = [options.title, options.author, options.keywords, options.subject]
+    for s in (m for m in docinfo if m):
+        for c in s:
+            if unicodedata.category(c) == 'Co' or ord(c) >= 0x10000:
+                raise ValueError(
+                    "One of the metadata strings contains "
+                    "an unsupported Unicode character: '{}' (U+{})".format(
+                        c, hex(ord(c))[2:].upper()
+                    )
+                )
+
+
+def check_options_pillow(options):
+    PIL.Image.MAX_IMAGE_PIXELS = int(options.max_image_mpixels * 1_000_000)
+    if PIL.Image.MAX_IMAGE_PIXELS == 0:
+        PIL.Image.MAX_IMAGE_PIXELS = None
+
+
+def check_options(options):
+    check_options_languages(options)
+    check_options_metadata(options)
+    check_options_output(options)
+    check_options_sidecar(options)
+    check_options_preprocessing(options)
+    check_options_ocr_behavior(options)
+    check_options_optimizing(options)
+    check_options_advanced(options)
+    check_options_pillow(options)
+    check_dependency_versions(options)
+
+
+def check_closed_streams(options):
+    """Work around Python issue with multiprocessing forking on closed streams
+
+    https://bugs.python.org/issue28326
+
+    Attempting to a fork/exec a new Python process when any of std{in,out,err}
+    are closed or not flushable for some reason may raise an exception.
+    Fix this by opening devnull if the handle seems to be closed.  Do this
+    globally to avoid tracking places all places that fork.
+
+    Seems to be specific to multiprocessing.Process not all Python process
+    forkers.
+
+    The error actually occurs when the stream object is not flushable,
+    but replacing an open stream object that is not flushable with
+    /dev/null is a bad idea since it will create a silent failure.  Replacing
+    a closed handle with /dev/null seems safe.
+
+    """
+
+    if sys.version_info[0:3] >= (3, 6, 4):
+        return True  # Issued fixed in Python 3.6.4+
+
+    if sys.stderr is None:
+        sys.stderr = open(os.devnull, 'w')
+
+    if sys.stdin is None:
+        if options.input_file == '-':
+            log.error("Trying to read from stdin but stdin seems closed")
+            return False
+        sys.stdin = open(os.devnull, 'r')
+
+    if sys.stdout is None:
+        if options.output_file == '-':
+            # Can't replace stdout if the user is piping
+            # If this case can even happen, it must be some kind of weird
+            # stream.
+            log.error(
+                "Output was set to stdout '-' but the stream attached to "
+                "stdout does not support the flush() system call.  This "
+                "will fail."
+            )
+            return False
+        sys.stdout = open(os.devnull, 'w')
+
+    return True
+
+
+def log_page_orientations(pdfinfo):
+    direction = {0: 'n', 90: 'e', 180: 's', 270: 'w'}
+    orientations = []
+    for n, page in enumerate(pdfinfo):
+        angle = page.rotation or 0
+        if angle != 0:
+            orientations.append('{0}{1}'.format(n + 1, direction.get(angle, '')))
+    if orientations:
+        log.info('Page orientations detected: %s', ' '.join(orientations))
+
+
+def create_input_file(options, work_folder):
+    if options.input_file == '-':
+        # stdin
+        log.info('reading file from standard input')
+        target = os.path.join(work_folder, 'stdin')
+        with open(target, 'wb') as stream_buffer:
+            copyfileobj(sys.stdin.buffer, stream_buffer)
+        return target
+    else:
+        try:
+            target = os.path.join(work_folder, 'origin')
+            re_symlink(options.input_file, target)
+            return target
+        except FileNotFoundError:
+            raise InputFileError(f"File not found - {options.input_file}")
+
+
+def check_requested_output_file(options):
+    if options.output_file == '-':
+        if sys.stdout.isatty():
+            raise BadArgsError(
+                "Output was set to stdout '-' but it looks like stdout "
+                "is connected to a terminal.  Please redirect stdout to a "
+                "file."
+            )
+    elif not is_file_writable(options.output_file):
+        raise OutputFileAccessError(
+            f"Output file location ({options.output_file}) is not a writable file."
+        )
+
+
+def report_output_file_size(options, input_file, output_file):
+    try:
+        output_size = Path(output_file).stat().st_size
+        input_size = Path(input_file).stat().st_size
+    except FileNotFoundError:
+        return  # Outputting to stream or something
+    ratio = output_size / input_size
+    if ratio < 1.35 or input_size < 25000:
+        return  # Seems fine
+
+    reasons = []
+    image_preproc = {
+        'deskew',
+        'clean_final',
+        'remove_background',
+        'oversample',
+        'force_ocr',
+    }
+    for arg in image_preproc:
+        if getattr(options, arg, False):
+            reasons.append(
+                f"The argument --{arg.replace('_', '-')} was issued, causing transcoding."
+            )
+
+    if reasons:
+        explanation = "Possible reasons for this include:\n" + '\n'.join(reasons) + "\n"
+    else:
+        explanation = "No reason for this increase is known.  Please report this issue."
+
+    log.warning(
+        f"The output file size is {ratio:.2f}× larger than the input file.\n"
+        f"{explanation}"
+    )
+
+
+def check_dependency_versions(options):
+    check_external_program(
+        program='tesseract',
+        package={'darwin': 'tesseract', 'linux': 'tesseract-ocr'},
+        version_checker=tesseract.version,
+        need_version='4.0.0',  # using backport for Travis CI
+    )
+    check_external_program(
+        program='gs',
+        package='ghostscript',
+        version_checker=ghostscript.version,
+        need_version='9.15',  # limited by Travis CI / Ubuntu 14.04 backports
+    )
+    if ghostscript.version() == '9.24':
+        raise MissingDependencyError(
+            "Ghostscript 9.24 contains serious regressions and is not "
+            "supported. Please upgrade to Ghostscript 9.25 or use an older "
+            "version."
+        )
+    check_external_program(
+        program='qpdf',
+        package='qpdf',
+        version_checker=qpdf.version,
+        need_version='8.0.2',
+    )

src/ocrmypdf/_weave.py

@@ -1,332 +0,0 @@
-# © 2018 James R. Barlow: github.com/jbarlow83
-#
-# This file is part of OCRmyPDF.
-#
-# OCRmyPDF is free software: you can redistribute it and/or modify
-# it under the terms of the GNU General Public License as published by
-# the Free Software Foundation, either version 3 of the License, or
-# (at your option) any later version.
-#
-# OCRmyPDF is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-# GNU General Public License for more details.
-#
-# You should have received a copy of the GNU General Public License
-# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
-
-from pathlib import Path
-from itertools import groupby
-
-import pikepdf
-
-from .helpers import flatten_groups, page_number
-from .exec import tesseract
-
-
-def _update_page_resources(*, page, font, font_key, procset):
-    "Update this page's fonts with a reference to the Glyphless font"
-
-    if '/Resources' not in page:
-        page['/Resources'] = pikepdf.Dictionary({})
-    resources = page['/Resources']
-    try:
-        fonts = resources['/Font']
-    except KeyError:
-        fonts = pikepdf.Dictionary({})
-    if font_key not in fonts:
-        fonts[font_key] = font
-    resources['/Font'] = fonts
-
-    # Reassign /ProcSet to one that just lists everything - ProcSet is
-    # obsolete and doesn't matter but recommended for old viewer support
-    resources['/ProcSet'] = procset
-
-
-def _weave_layers_graft(
-        *, pdf_base, page_num, text, font, font_key, procset, rotation, log):
-    """Insert the text layer from text page 0 on to pdf_base at page_num"""
-
-    log.debug("Grafting")
-    if Path(text).stat().st_size == 0:
-        return
-
-    # This is a pointer indicating a specific page in the base file
-    pdf_text = pikepdf.open(text)
-    pdf_text_contents = pdf_text.pages[0].Contents.read_bytes()
-
-    if not tesseract.has_textonly_pdf():
-        # If we don't have textonly_pdf, edit the stream to delete the
-        # instruction to draw the image Tesseract generated, which we do not
-        # use.
-        stream = bytearray(pdf_text_contents)
-        pattern = b'/Im1 Do'
-        idx = stream.find(pattern)
-        stream[idx:(idx + len(pattern))] = b' ' * len(pattern)
-        pdf_text_contents = bytes(stream)
-
-    base_page = pdf_base.pages.p(page_num)
-
-    # The text page always will be oriented up by this stage but the original
-    # content may have a rotation applied. Wrap the text stream with a rotation
-    # so it will be oriented the same way as the rest of the page content.
-    # (Previous versions OCRmyPDF rotated the content layer to match the text.)
-    mediabox = [float(pdf_text.pages[0].MediaBox[v])
-                for v in range(4)]
-    wt, ht = mediabox[2] - mediabox[0], mediabox[3] - mediabox[1]
-
-    mediabox = [float(base_page.MediaBox[v])
-                for v in range(4)]
-    wp, hp = mediabox[2] - mediabox[0], mediabox[3] - mediabox[1]
-
-    translate = pikepdf.PdfMatrix().translated(-wt / 2, -ht / 2)
-    untranslate = pikepdf.PdfMatrix().translated(wp / 2, hp / 2)
-    # -rotation because the input is a clockwise angle and this formula
-    # uses CCW
-    rotation = -rotation % 360
-    rotate = pikepdf.PdfMatrix().rotated(rotation)
-
-    # Because of rounding of DPI, we might get a text layer that is not
-    # identically sized to the target page. Scale to adjust. Normally this
-    # is within 0.998.
-    if rotation in (90, 270):
-        wt, ht = ht, wt
-    scale_x = wp / wt
-    scale_y = hp / ht
-
-    log.debug('%r', (scale_x, scale_y))
-    scale = pikepdf.PdfMatrix().scaled(scale_x, scale_y)
-
-    # Translate the text so it is centered at (0, 0), rotate it there, adjust
-    # for a size different between initial and text PDF, then untranslate
-    ctm = translate @ rotate @ scale @ untranslate
-
-    pdf_text_contents = (
-        b'q %s cm\n' % ctm.encode() +
-        pdf_text_contents +
-        b'\nQ\n'
-    )
-
-    new_text_layer = pikepdf.Stream(pdf_base, pdf_text_contents)
-
-    base_page.page_contents_add(new_text_layer, prepend=True)
-
-    _update_page_resources(
-        page=base_page, font=font, font_key=font_key, procset=procset
-    )
-
-
-def _find_font(text, pdf_base):
-    "Copy a font from the filename text into pdf_base"
-
-    font, font_key = None, None
-    possible_font_names = ('/f-0-0', '/F1')
-    try:
-        pdf_text = pikepdf.open(text)
-        pdf_text_fonts = pdf_text.pages[0].Resources.get('/Font', {})
-    except Exception:
-        return None, None
-
-    for f in possible_font_names:
-        pdf_text_font = pdf_text_fonts.get(f, None)
-        if pdf_text_font is not None:
-            font_key = f
-            break
-    if pdf_text_font:
-        font = pdf_base.copy_foreign(pdf_text_font)
-    return font, font_key
-
-
-def _fix_toc(pdf_base, pageref_remap, log):
-    """Repair the table of contents
-
-    Whenever we replace a page wholesale, it gets assigned a new objgen number
-    and other references to it within the PDF become invalid, most notably in
-    the table of contents (/Outlines in PDF-speak).  In weave_layers we collect
-    pageref_remap, a mapping that describes the new objgen number given an old
-    one.  (objgen is a tuple, and the gen is almost always zero.)
-
-    The /Outlines data structure is a messy data structure, but rather than
-    navigating hierarchically we just track unique nodes.  Enqueue nodes when
-    we find them, and never visit them again.  set() is awesome.  We look for
-    the two types of object in the table of contents that can be page bookmarks
-    and update the page entry.
-
-    It may ultimately be better to find a way to rebuild a page in place.
-
-    """
-
-    visited = set()
-    queue = set()
-    link_keys = ('/Parent', '/First', '/Last', '/Prev', '/Next')
-
-    if not '/Outlines' in pdf_base.root:
-        return
-    if not pageref_remap:
-        return
-
-    def remap_dest(dest_node):
-        if not isinstance(dest_node, pikepdf.Array):
-            return
-        pageref = dest_node[0]
-        if pageref['/Type'] == '/Page' and \
-                pageref.objgen in pageref_remap:
-            new_objgen = pageref_remap[pageref.objgen]
-            dest_node[0] = pdf_base.get_object(new_objgen)
-
-    queue.add(pdf_base.root.Outlines.objgen)
-    while queue:
-        objgen = queue.pop()
-        visited.add(objgen)
-        node = pdf_base.get_object(objgen)
-        log.debug('fix toc: visiting %r', objgen)
-
-        # Enumerate other nodes we could visit from here
-        for key in link_keys:
-            if key not in node:
-                continue
-            item = node[key]
-            if not item.is_indirect:
-                continue
-            objgen = item.objgen
-            if objgen not in visited:
-                queue.add(objgen)
-
-        if '/Dest' in node:
-            remap_dest(node['/Dest'])
-        elif '/A' in node:
-            if '/S' in node['/A'] and node['/A']['/S'] == '/GoTo':
-                remap_dest(node['/A']['/D'])
-
-
-def weave_layers(
-        infiles,
-        output_file,
-        log,
-        context):
-    """Apply text layer and/or image layer changes to baseline file
-
-    This is where the magic happens. infiles will be the main PDF to modify,
-    and optional .text.pdf and .image-layer.pdf files, organized however ruffus
-    organizes them.
-
-    From .text.pdf, we copy the content stream (which contains the Tesseract
-    OCR results), and rotate it into place. The first time we do this, we also
-    copy the GlyphlessFont, and then reference that font again.
-
-    For .image-layer.pdf, we check if this is a "pointer" to the original file,
-    or a new file. If a new file, we replace the page and remember that we
-    replaced this page.
-
-    Every 100 open files, we save intermediate results, to avoid any resource
-    limits, since pikepdf/qpdf need to keep a lot of open file handles in the
-    background. When objects are copied from one file to another qpdf, qpdf
-    doesn't actually copy the data until asked to write, so all the resources
-    it may need to remain available.
-
-    For completeness, we set up a /ProcSet on every page, although it's
-    unlikely any PDF viewer cares about this anymore.
-
-    """
-
-    def input_sorter(key):
-        try:
-            return page_number(key)
-        except ValueError:
-            return -1
-    flat_inputs = sorted(flatten_groups(infiles), key=input_sorter)
-    groups = groupby(flat_inputs, key=input_sorter)
-
-    # Extract first item
-    _, basegroup = next(groups)
-    base = list(basegroup)[0]
-    path_base = Path(base).resolve()
-    pdf_base = pikepdf.open(path_base)
-    keep_open = []
-    font, font_key, procset = None, None, None
-    pdfinfo = context.get_pdfinfo()
-    pagerefs = {}
-
-    procset = pdf_base.make_indirect(
-        pikepdf.Object.parse(b'[ /PDF /Text /ImageB /ImageC /ImageI ]'))
-
-    # Iterate rest
-    for page_num, layers in groups:
-        layers = list(layers)
-        log.debug(page_num)
-        log.debug(layers)
-
-        text = next(
-            (ii for ii in layers if ii.endswith('.text.pdf')), None
-        )
-        image = next(
-            (ii for ii in layers if ii.endswith('.image-layer.pdf')), None
-        )
-
-        if text and not font:
-            font, font_key = _find_font(text, pdf_base)
-
-        replacing = False
-        content_rotation = pdfinfo[page_num - 1].rotation
-
-        path_image = Path(image).resolve() if image else None
-        if path_image is not None and path_image != path_base:
-            # We are replacing the old page with a rasterized PDF of the new
-            # page
-            log.debug("Replace")
-            old_objgen = pdf_base.pages[page_num - 1].objgen
-
-            pdf_image = pikepdf.open(image)
-            keep_open.append(pdf_image)
-            image_page = pdf_image.pages[0]
-            pdf_base.pages[page_num - 1] = image_page
-
-            # We're adding a new page, which will get a new objgen number pair,
-            # so we need to update any references to it.  qpdf did not like
-            # my attempt to update the old object in place, but that is an
-            # option to consider
-            pagerefs[old_objgen] = pdf_base.pages[page_num - 1].objgen
-            replacing = True
-
-        autorotate_correction = context.get_rotation(page_num - 1)
-        if replacing:
-            content_rotation = autorotate_correction
-        text_rotation = autorotate_correction
-        text_misaligned = (text_rotation - content_rotation) % 360
-        log.debug('%r', [
-            text_rotation, autorotate_correction, text_misaligned,
-            content_rotation]
-        )
-
-        if text and font:
-            # Graft the text layer onto this page, whether new or old
-            _weave_layers_graft(
-                pdf_base=pdf_base, page_num=page_num, text=text, font=font,
-                font_key=font_key, rotation=text_misaligned, procset=procset,
-                log=log
-            )
-
-        # Correct the rotation if applicable
-        pdf_base.pages[page_num - 1].Rotate = \
-            (content_rotation - autorotate_correction) % 360
-
-        if len(keep_open) > 100:
-            # qpdf limitations require us to keep files open when we intend
-            # to copy content from them before saving. However, we want to keep
-            # a lid on file handles and memory usage, so for big files we're
-            # going to stop and save periodically. Attach the font to page 1
-            # even if page 1 doesn't use it, so we have a way to get it back.
-            page0 = pdf_base.pages[0]
-            _update_page_resources(
-                page=page0, font=font, font_key=font_key, procset=procset)
-            interim = output_file + '_working{}.pdf'.format(page_num)
-            pdf_base.save(interim)
-            del pdf_base
-            keep_open = []
-
-            pdf_base = pikepdf.open(interim)
-            procset = pdf_base.pages[0].Resources.ProcSet
-            font = pdf_base.pages[0].Resources.Font.get(font_key)
-
-    _fix_toc(pdf_base, pagerefs, log)
-    pdf_base.save(output_file)

src/ocrmypdf/api.py

@@ -0,0 +1,246 @@
+# © 2019 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+import logging
+import os
+import sys
+from enum import IntEnum
+from pathlib import Path
+
+from tqdm import tqdm
+
+from .cli import parser
+from ._sync import run_pipeline
+from ._validation import check_options
+
+
+class TqdmConsole:
+    """Wrapper to log messages in a way that is compatible with tqdm progress bar"""
+
+    def __init__(self, file):
+        self.file = file
+        self.py36 = sys.version_info >= (3, 6)
+
+    def write(self, msg):
+        # When no progress bar is active, tqdm.write() routes to print()
+        if self.py36:
+            if msg.strip() != '':
+                tqdm.write(msg.rstrip(), end='\n', file=self.file)
+        else:
+            tqdm.write(msg.rstrip(), end='\n', file=self.file)
+
+    def flush(self):
+        if hasattr(self.file, "flush"):
+            self.file.flush()
+
+
+class Verbosity(IntEnum):
+    """Verbosity level for configure_logging."""
+
+    quiet = -1  #: Suppress most messages
+    default = 0  #: Default level of logging
+    debug = 1  #: Output ocrmypdf debug messages
+    debug_all = 2  #: More detailed debugging from ocrmypdf and dependent modules
+
+
+def configure_logging(verbosity, progress_bar_friendly=True, manage_root_logger=False):
+    """Set up logging.
+
+    Library users may wish to use this function if they want their log output to be
+    similar to ocrmypdf command line interface. If not used, the external application
+    should configure logging on its own.
+
+    ocrmypdf will perform all of its logging under the `"ocrmypdf"` logging namespace.
+    In addition, ocrmypdf imports pdfminer, which logs under `"pdfminer"`. A library
+    user may wish to configure both; note that pdfminer is extremely chatty at the log
+    level logging.INFO.
+
+    Library users may perform additional configuration afterwards.
+
+    Args:
+        verbosity (Verbosity): Verbosity level.
+        progress_bar_friendly (bool): Install the TqdmConsole log handler, which is
+            compatible with the tqdm progress bar; without this log messages will
+            overwrite the progress bar
+        manage_root_logger (bool): Configure the process's root logger, to ensure
+            all log output is sent through
+    """
+
+    prefix = '' if manage_root_logger else 'ocrmypdf'
+    log = logging.getLogger(prefix)
+    log.setLevel(logging.INFO)
+
+    if progress_bar_friendly:
+        console = logging.StreamHandler(stream=TqdmConsole(sys.stderr))
+    else:
+        console = logging.StreamHandler(stream=sys.stderr)
+
+    if verbosity < 0:
+        console.setLevel(logging.ERROR)
+    elif verbosity >= 1:
+        console.setLevel(logging.DEBUG)
+    else:
+        console.setLevel(logging.INFO)
+
+    formatter = logging.Formatter('%(levelname)7s - %(message)s')
+    if verbosity >= 1:
+        log.setLevel(logging.DEBUG)
+    if verbosity >= 2:
+        formatter = logging.Formatter('%(name)s - %(levelname)7s - %(message)s')
+
+    console.setFormatter(formatter)
+    log.addHandler(console)
+
+    if verbosity <= 1:
+        pdfminer_log = logging.getLogger('pdfminer')
+        pdfminer_log.setLevel(logging.ERROR)
+        pil_log = logging.getLogger('PIL')
+        pil_log.setLevel(logging.INFO)
+
+    if manage_root_logger:
+        logging.captureWarnings(True)
+
+
+def create_options(*, input_file, output_file, **kwargs):
+    cmdline = []
+    deferred = []
+
+    for arg, val in kwargs.items():
+        if val is None:
+            continue
+        if (arg.startswith('plugin') or arg.startswith('filter')) and (
+            callable(val) or isinstance(val, str)
+        ):
+            deferred.append((arg, val))
+            continue
+        elif arg == 'tesseract_env':
+            deferred.append((arg, val))
+            continue
+        cmd_style_arg = arg.replace('_', '-')
+        cmdline.append(f"--{cmd_style_arg}")
+        if isinstance(val, bool):
+            continue
+        if isinstance(val, (int, float)):
+            cmdline.append(str(val))
+        elif isinstance(val, str):
+            cmdline.append(val)
+        elif isinstance(val, Path):
+            cmdline.append(str(val))
+        else:
+            raise TypeError(f"{arg}: {val} ({type(val)})")
+
+    cmdline.append(str(input_file))
+    cmdline.append(str(output_file))
+
+    parser.api_mode = True
+    options = parser.parse_args(cmdline)
+    for keyword, val in deferred:
+        setattr(options, keyword, val)
+
+    # If we are running a Tesseract spoof, ensure it knows what the input file is
+    if os.environ.get('PYTEST_CURRENT_TEST') and options.tesseract_env:
+        options.tesseract_env['_OCRMYPDF_TEST_INFILE'] = input_file
+
+    return options
+
+
+def run(  # pylint: disable=unused-argument
+    input_file,
+    output_file,
+    *,
+    language=None,
+    image_dpi=None,
+    output_type=None,
+    sidecar=None,
+    jobs=None,
+    use_threads=None,
+    title=None,
+    author=None,
+    subject=None,
+    keywords=None,
+    rotate_pages=None,
+    remove_background=None,
+    deskew=None,
+    clean=None,
+    clean_final=None,
+    unpaper_args=None,
+    oversample=None,
+    remove_vectors=None,
+    threshold=None,
+    force_ocr=None,
+    skip_text=None,
+    redo_ocr=None,
+    skip_big=None,
+    optimize=None,
+    jpg_quality=None,
+    png_quality=None,
+    jbig2_lossy=None,
+    jbig2_page_group_size=None,
+    pages=None,
+    max_image_mpixels=None,
+    tesseract_config=None,
+    tesseract_pagesegmode=None,
+    tesseract_oem=None,
+    pdf_renderer=None,
+    tesseract_timeout=None,
+    rotate_pages_threshold=None,
+    pdfa_image_compression=None,
+    user_words=None,
+    user_patterns=None,
+    keep_temporary_files=None,
+    progress_bar=None,
+    filter_ocr_image=None,
+    plugin_validation=None,
+    tesseract_env=None,
+):
+    """Run OCRmyPDF on one PDF or image.
+
+    For most arguments, see documentation for the equivalent command line parameter.
+    A few specific arguments are discussed here:
+
+    Args:
+        use_threads (bool): Use worker threads instead of processes. This reduces
+            performance but may make debugging easier since it is easier to set
+            breakpoints.
+        tesseract_env (dict): Override environment variables for Tesseract
+    Raises:
+        ocrmypdf.PdfMergeFailedError: If the input PDF is malformed, preventing merging
+            with the OCR layer.
+        ocrmypdf.MissingDependencyError: If a required dependency program is missing or
+            was not found on PATH.
+        ocrmypdf.UnsupportedImageFormatError: If the input file type was an image that
+            could not be read, or some other file type that is not a PDF.
+        ocrmypdf.DpiError: If the input file is an image, but the resolution of the
+            image is not credible (allowing it to proceed would cause poor OCR).
+        ocrmypdf.OutputFileAccessError: If an attempt to write to the intended output
+            file failed.
+        ocrmypdf.PriorOcrFoundError: If the input PDF seems to have OCR or digital
+            text already, and settings did not tell us to proceed.
+        ocrmypdf.InputFileError: Any other problem with the input file.
+        ocrmypdf.SubprocessOutputError: Any error related to executing a subprocess.
+        ocrmypdf.EncryptedPdfERror: If the input PDF is encrypted (password protected).
+            OCRmyPDF does not remove passwords.
+        ocrmypdf.TesseractConfigError: If Tesseract reported its configuration was not
+            valid.
+
+    Returns:
+        :class:`ocrmypdf.ExitCode`
+    """
+
+    options = create_options(**locals())
+    check_options(options)
+    return run_pipeline(options, api=True)

src/ocrmypdf/cli.py

@@ -0,0 +1,486 @@
+# © 2015-19 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+import argparse
+
+from . import PROGRAM_NAME, VERSION
+from ._plugins import check_plugin_loadable
+
+
+def numeric(basetype, min_=None, max_=None):
+    """Validator for numeric params"""
+    min_ = basetype(min_) if min_ is not None else None
+    max_ = basetype(max_) if max_ is not None else None
+
+    def _numeric(string):
+        value = basetype(string)
+        if (min_ is not None and value < min_) or (max_ is not None and value > max_):
+            msg = "%r not in valid range %r" % (string, (min_, max_))
+            raise argparse.ArgumentTypeError(msg)
+        return value
+
+    _numeric.__name__ = basetype.__name__
+    return _numeric
+
+
+class ArgumentParser(argparse.ArgumentParser):
+    """Override parser's default behavior of calling sys.exit()
+
+    https://stackoverflow.com/questions/5943249/python-argparse-and-controlling-overriding-the-exit-status-code
+    """
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.api_mode = False
+
+    def error(self, message):
+        if not self.api_mode:
+            super().error(message)
+            return
+        raise ValueError(message)
+
+
+parser = ArgumentParser(
+    prog=PROGRAM_NAME,
+    fromfile_prefix_chars='@',
+    formatter_class=argparse.RawDescriptionHelpFormatter,
+    description="""\
+Generates a searchable PDF or PDF/A from a regular PDF.
+
+OCRmyPDF rasterizes each page of the input PDF, optionally corrects page
+rotation and performs image processing, runs the Tesseract OCR engine on the
+image, and then creates a PDF from the OCR information.
+""",
+    epilog="""\
+OCRmyPDF attempts to keep the output file at about the same size.  If a file
+contains losslessly compressed images, and output file will be losslessly
+compressed as well.
+
+PDF is a page description file that attempts to preserve a layout exactly.
+A PDF can contain vector objects (such as text or lines) and raster objects
+(images).  A page might have multiple images.  OCRmyPDF is prepared to deal
+with the wide variety of PDFs that exist in the wild.
+
+When a PDF page contains text, OCRmyPDF assumes that the page has already
+been OCRed or is a "born digital" page that should not be OCRed.  The default
+behavior is to exit in this case without producing a file.  You can use the
+option --skip-text to ignore pages with text, or --force-ocr to rasterize
+all objects on the page and produce an image-only PDF as output.
+
+    ocrmypdf --skip-text file_with_some_text_pages.pdf output.pdf
+
+    ocrmypdf --force-ocr word_document.pdf output.pdf
+
+If you are concerned about long-term archiving of PDFs, use the default option
+--output-type pdfa which converts the PDF to a standardized PDF/A-2b.  This
+converts images to sRGB colorspace, removes some features from the PDF such
+as Javascript or forms. If you want to minimize the number of changes made to
+your PDF, use --output-type pdf.
+
+If OCRmyPDF is given an image file as input, it will attempt to convert the
+image to a PDF before processing.  For more control over the conversion of
+images to PDF, use the Python package img2pdf or other image to PDF software.
+
+For example, this command uses img2pdf to convert all .png files beginning
+with the 'page' prefix to a PDF, fitting each image on A4-sized paper, and
+sending the result to OCRmyPDF through a pipe.  img2pdf is a dependency of
+ocrmypdf so it is already installed.
+
+    img2pdf --pagesize A4 page*.png | ocrmypdf - myfile.pdf
+
+Online documentation is located at:
+    https://ocrmypdf.readthedocs.io/en/latest/introduction.html
+
+""",
+)
+
+parser.add_argument(
+    'input_file',
+    metavar="input_pdf_or_image",
+    help="PDF file containing the images to be OCRed (or '-' to read from "
+    "standard input)",
+)
+parser.add_argument(
+    'output_file',
+    metavar="output_pdf",
+    help="Output searchable PDF file (or '-' to write to standard output). "
+    "Existing files will be ovewritten. If same as input file, the "
+    "input file will be updated only if processing is successful.",
+)
+parser.add_argument(
+    '-l',
+    '--language',
+    action='append',
+    help="Language(s) of the file to be OCRed (see tesseract --list-langs for "
+    "all language packs installed in your system). Use -l eng+deu for "
+    "multiple languages.",
+)
+parser.add_argument(
+    '--image-dpi',
+    metavar='DPI',
+    type=int,
+    help="For input image instead of PDF, use this DPI instead of file's.",
+)
+parser.add_argument(
+    '--output-type',
+    choices=['pdfa', 'pdf', 'pdfa-1', 'pdfa-2', 'pdfa-3'],
+    default='pdfa',
+    help="Choose output type. 'pdfa' creates a PDF/A-2b compliant file for "
+    "long term archiving (default, recommended) but may not suitable "
+    "for users who want their file altered as little as possible. 'pdfa' "
+    "also has problems with full Unicode text. 'pdf' attempts to "
+    "preserve file contents as much as possible. 'pdf-a1' creates a "
+    "PDF/A1-b file. 'pdf-a2' is equivalent to 'pdfa'. 'pdf-a3' creates a "
+    "PDF/A3-b file.",
+)
+
+# Use null string '\0' as sentinel to indicate the user supplied no argument,
+# since that is the only invalid character for filepaths on all platforms
+# bool('\0') is True in Python
+parser.add_argument(
+    '--sidecar',
+    nargs='?',
+    const='\0',
+    default=None,
+    metavar='FILE',
+    help="Generate sidecar text files that contain the same text recognized "
+    "by Tesseract. This may be useful for building a OCR text database. "
+    "If FILE is omitted, the sidecar file be named {output_file}.txt "
+    "If FILE is set to '-', the sidecar is written to stdout (a "
+    "convenient way to preview OCR quality). The output file and sidecar "
+    "may not both use stdout at the same time.",
+)
+
+parser.add_argument(
+    '--version',
+    action='version',
+    version=VERSION,
+    help="Print program version and exit",
+)
+
+jobcontrol = parser.add_argument_group("Job control options")
+jobcontrol.add_argument(
+    '-j',
+    '--jobs',
+    metavar='N',
+    type=numeric(int, 0, 256),
+    help="Use up to N CPU cores simultaneously (default: use all).",
+)
+jobcontrol.add_argument(
+    '-q', '--quiet', action='store_true', help="Suppress INFO messages"
+)
+jobcontrol.add_argument(
+    '-v',
+    '--verbose',
+    type=int,
+    default=0,
+    nargs='?',
+    help="Print more verbose messages for each additional verbose level. Use "
+    "`-v 1` typically for much more detailed logging. Higher numbers "
+    "are probably only useful in debugging.",
+)
+jobcontrol.add_argument(
+    '--no-progress-bar',
+    action='store_false',
+    dest='progress_bar',
+    help=argparse.SUPPRESS,
+)
+jobcontrol.add_argument('--use-threads', action='store_true', help=argparse.SUPPRESS)
+
+metadata = parser.add_argument_group(
+    "Metadata options",
+    "Set output PDF/A metadata (default: copy input document's metadata)",
+)
+metadata.add_argument(
+    '--title', type=str, help="Set document title (place multiple words in quotes)"
+)
+metadata.add_argument('--author', type=str, help="Set document author")
+metadata.add_argument('--subject', type=str, help="Set document subject description")
+metadata.add_argument('--keywords', type=str, help="Set document keywords")
+
+preprocessing = parser.add_argument_group(
+    "Image preprocessing options",
+    "Options to improve the quality of the final PDF and OCR",
+)
+preprocessing.add_argument(
+    '-r',
+    '--rotate-pages',
+    action='store_true',
+    help="Automatically rotate pages based on detected text orientation",
+)
+preprocessing.add_argument(
+    '--remove-background',
+    action='store_true',
+    help="Attempt to remove background from gray or color pages, setting it "
+    "to white ",
+)
+preprocessing.add_argument(
+    '-d', '--deskew', action='store_true', help="Deskew each page before performing OCR"
+)
+preprocessing.add_argument(
+    '-c',
+    '--clean',
+    action='store_true',
+    help="Clean pages from scanning artifacts before performing OCR, and send "
+    "the cleaned page to OCR, but do not include the cleaned page in "
+    "the output",
+)
+preprocessing.add_argument(
+    '-i',
+    '--clean-final',
+    action='store_true',
+    help="Clean page as above, and incorporate the cleaned image in the final "
+    "PDF.  Might remove desired content.",
+)
+preprocessing.add_argument(
+    '--unpaper-args',
+    type=str,
+    default=None,
+    help="A quoted string of arguments to pass to unpaper. Requires --clean. "
+    "Example: --unpaper-args '--layout double'.",
+)
+preprocessing.add_argument(
+    '--oversample',
+    metavar='DPI',
+    type=numeric(int, 0, 5000),
+    default=0,
+    help="Oversample images to at least the specified DPI, to improve OCR "
+    "results slightly",
+)
+preprocessing.add_argument(
+    '--remove-vectors',
+    action='store_true',
+    help="EXPERIMENTAL. Mask out any vector objects in the PDF so that they "
+    "will not be included in OCR. This can eliminate false characters.",
+)
+preprocessing.add_argument(
+    '--threshold',
+    action='store_true',
+    help="EXPERIMENTAL. Threshold image to 1bpp before sending it to Tesseract for OCR. Can "
+    "improve OCR quality compared to Tesseract's thresholder.",
+)
+
+ocrsettings = parser.add_argument_group("OCR options", "Control how OCR is applied")
+ocrsettings.add_argument(
+    '-f',
+    '--force-ocr',
+    action='store_true',
+    help="Rasterize any text or vector objects on each page, apply OCR, and "
+    "save the rastered output (this rewrites the PDF)",
+)
+ocrsettings.add_argument(
+    '-s',
+    '--skip-text',
+    action='store_true',
+    help="Skip OCR on any pages that already contain text, but include the "
+    "page in final output; useful for PDFs that contain a mix of "
+    "images, text pages, and/or previously OCRed pages",
+)
+ocrsettings.add_argument(
+    '--redo-ocr',
+    action='store_true',
+    help="Attempt to detect and remove the hidden OCR layer from files that "
+    "were previously OCRed with OCRmyPDF or another program. Apply OCR "
+    "to text found in raster images. Existing visible text objects will "
+    "not be changed. If there is no existing OCR, OCR will be added.",
+)
+ocrsettings.add_argument(
+    '--skip-big',
+    type=numeric(float, 0, 5000),
+    metavar='MPixels',
+    help="Skip OCR on pages larger than the specified amount of megapixels, "
+    "but include skipped pages in final output",
+)
+
+optimizing = parser.add_argument_group(
+    "Optimization options", "Control how the PDF is optimized after OCR"
+)
+optimizing.add_argument(
+    '-O',
+    '--optimize',
+    type=int,
+    choices=range(0, 4),
+    default=1,
+    help=(
+        "Control how PDF is optimized after processing:"
+        "0 - do not optimize; "
+        "1 - do safe, lossless optimizations (default); "
+        "2 - do some lossy optimizations; "
+        "3 - do aggressive lossy optimizations (including lossy JBIG2)"
+    ),
+)
+optimizing.add_argument(
+    '--jpeg-quality',
+    type=numeric(int, 0, 100),
+    default=0,
+    metavar='Q',
+    help=(
+        "Adjust JPEG quality level for JPEG optimization. "
+        "100 is best quality and largest output size; "
+        "1 is lowest quality and smallest output; "
+        "0 uses the default."
+    ),
+)
+optimizing.add_argument(
+    '--jpg-quality',
+    type=numeric(int, 0, 100),
+    default=0,
+    metavar='Q',
+    dest='jpeg_quality',
+    help=argparse.SUPPRESS,  # Alias for --jpeg-quality
+)
+optimizing.add_argument(
+    '--png-quality',
+    type=numeric(int, 0, 100),
+    default=0,
+    metavar='Q',
+    help=(
+        "Adjust PNG quality level to use when quantizing PNGs. "
+        "Values have same meaning as with --jpeg-quality"
+    ),
+)
+optimizing.add_argument(
+    '--jbig2-lossy',
+    action='store_true',
+    help=(
+        "Enable JBIG2 lossy mode (better compression, not suitable for some "
+        "use cases - see documentation)."
+    ),
+)
+optimizing.add_argument(
+    '--jbig2-page-group-size',
+    type=numeric(int, 1, 10000),
+    default=0,
+    metavar='N',
+    # Adjust number of pages to consider at once for JBIG2 compression
+    help=argparse.SUPPRESS,
+)
+
+advanced = parser.add_argument_group(
+    "Advanced", "Advanced options to control Tesseract's OCR behavior"
+)
+advanced.add_argument(
+    '--pages',
+    type=str,
+    help="Limit OCR to the specified pages (ranges or comma separated), skipping others",
+)
+advanced.add_argument(
+    '--max-image-mpixels',
+    action='store',
+    type=numeric(float, 0),
+    metavar='MPixels',
+    help="Set maximum number of pixels to unpack before treating an image as a "
+    "decompression bomb",
+    default=128.0,
+)
+advanced.add_argument(
+    '--tesseract-config',
+    action='append',
+    metavar='CFG',
+    default=[],
+    help="Additional Tesseract configuration files -- see documentation",
+)
+advanced.add_argument(
+    '--tesseract-pagesegmode',
+    action='store',
+    type=int,
+    metavar='PSM',
+    choices=range(0, 14),
+    help="Set Tesseract page segmentation mode (see tesseract --help)",
+)
+advanced.add_argument(
+    '--tesseract-oem',
+    action='store',
+    type=int,
+    metavar='MODE',
+    choices=range(0, 4),
+    help=(
+        "Set Tesseract 4.0 OCR engine mode: "
+        "0 - original Tesseract only; "
+        "1 - neural nets LSTM only; "
+        "2 - Tesseract + LSTM; "
+        "3 - default."
+    ),
+)
+advanced.add_argument(
+    '--pdf-renderer',
+    choices=['auto', 'hocr', 'sandwich'],
+    default='auto',
+    help="Choose OCR PDF renderer - the default option is to let OCRmyPDF "
+    "choose.  See documentation for discussion.",
+)
+advanced.add_argument(
+    '--tesseract-timeout',
+    default=180.0,
+    type=numeric(float, 0),
+    metavar='SECONDS',
+    help='Give up on OCR after the timeout, but copy the preprocessed page '
+    'into the final output',
+)
+advanced.add_argument(
+    '--rotate-pages-threshold',
+    default=14.0,
+    type=numeric(float, 0, 1000),
+    metavar='CONFIDENCE',
+    help="Only rotate pages when confidence is above this value (arbitrary "
+    "units reported by tesseract)",
+)
+advanced.add_argument(
+    '--pdfa-image-compression',
+    choices=['auto', 'jpeg', 'lossless'],
+    default='auto',
+    help="Specify how to compress images in the output PDF/A. 'auto' lets "
+    "OCRmyPDF decide.  'jpeg' changes all grayscale and color images to "
+    "JPEG compression.  'lossless' uses PNG-style lossless compression "
+    "for all images.  Monochrome images are always compressed using a "
+    "lossless codec.  Compression settings "
+    "are applied to all pages, including those for which OCR was "
+    "skipped.  Not supported for --output-type=pdf ; that setting "
+    "preserves the original compression of all images.",
+)
+advanced.add_argument(
+    '--user-words',
+    metavar='FILE',
+    help="Specify the location of the Tesseract user words file. This is a "
+    "list of words Tesseract should consider while performing OCR in "
+    "addition to its standard language dictionaries. This can improve "
+    "OCR quality especially for specialized and technical documents.",
+)
+advanced.add_argument(
+    '--user-patterns',
+    metavar='FILE',
+    help="Specify the location of the Tesseract user patterns file.",
+)
+
+plugins = parser.add_argument_group("Filters and Plugins", argparse.SUPPRESS)
+plugins.add_argument(
+    '--filter-ocr-image', help=argparse.SUPPRESS, type=check_plugin_loadable
+)
+plugins.add_argument(
+    '--plugin-validation', help=argparse.SUPPRESS, type=check_plugin_loadable
+)
+
+debugging = parser.add_argument_group(
+    "Debugging", "Arguments to help with troubleshooting and debugging"
+)
+debugging.add_argument(
+    '-k',
+    '--keep-temporary-files',
+    action='store_true',
+    help="Keep temporary files (helpful for debugging)",
+)
+debugging.add_argument('--tesseract-env', type=str, help=argparse.SUPPRESS)

src/ocrmypdf/exceptions.py

@@ -17,6 +17,8 @@
 
 
 from enum import IntEnum
+from textwrap import dedent
+
 
 class ExitCode(IntEnum):
     ok = 0
@@ -36,6 +38,13 @@ class ExitCode(IntEnum):
 
 class ExitCodeException(Exception):
     exit_code = ExitCode.other_error
+    message = ""
+
+    def __str__(self):
+        super_msg = super().__str__()  # Don't do str(super())
+        if self.message:
+            return self.message.format(super_msg)
+        return super_msg
 
 
 class BadArgsError(ExitCodeException):
@@ -44,6 +53,17 @@ class BadArgsError(ExitCodeException):
 
 class PdfMergeFailedError(ExitCodeException):
     exit_code = ExitCode.input_file
+    message = dedent(
+        '''\
+        Failed to merge PDF image layer with OCR layer
+
+        Usually this happens because the input PDF file is malformed and
+        ocrmypdf cannot automatically correct the problem on its own.
+
+        Try using
+            ocrmypdf --pdf-renderer sandwich  [..other args..]
+        '''
+    )
 
 
 class MissingDependencyError(ExitCodeException):
@@ -76,7 +96,20 @@ class SubprocessOutputError(ExitCodeException):
 
 class EncryptedPdfError(ExitCodeException):
     exit_code = ExitCode.encrypted_pdf
+    message = dedent(
+        '''\
+        Input PDF is encrypted. The encryption must be removed to
+        perform OCR.
+
+        For information about this PDF's security use
+            qpdf --show-encryption infilename
+
+        You can remove the encryption using
+            qpdf --decrypt [--password=[password]] infilename
+        '''
+    )
 
 
 class TesseractConfigError(ExitCodeException):
     exit_code = ExitCode.invalid_config
+    message = "Error occurred while parsing a Tesseract configuration file"

src/ocrmypdf/exec/__init__.py

@@ -17,44 +17,160 @@
 
 """Wrappers to manage subprocess calls"""
 
+import logging
 import os
 import re
 import sys
 from subprocess import run, STDOUT, PIPE, CalledProcessError
-from ..exceptions import MissingDependencyError
+from ..exceptions import MissingDependencyError, ExitCode
+from collections.abc import Mapping
+
+log = logging.Logger(__name__)
 
 
-
-def get_version(program, *,
-        version_arg='--version', regex=r'(\d+(\.\d+)*)'):
+def get_version(program, *, version_arg='--version', regex=r'(\d+(\.\d+)*)', env=None):
     "Get the version of the specified program"
-    args_prog = [
-        program,
-        version_arg
-    ]
+    args_prog = [program, version_arg]
     try:
         proc = run(
-            args_prog, close_fds=True, universal_newlines=True,
-            stdout=PIPE, stderr=STDOUT, check=True)
+            args_prog,
+            close_fds=True,
+            universal_newlines=True,
+            stdout=PIPE,
+            stderr=STDOUT,
+            check=True,
+            env=env,
+        )
         output = proc.stdout
     except FileNotFoundError as e:
         raise MissingDependencyError(
-            "Could not find program '{}' on the PATH".format(
-                program)) from e
+            f"Could not find program '{program}' on the PATH"
+        ) from e
     except CalledProcessError as e:
-        if e.returncode < 0:
+        if e.returncode != 0:
             raise MissingDependencyError(
-                "Ran program '{}' but it exited with an error:\n{}".format(
-                    program, e.output)) from e
+                f"Ran program '{program}' but it exited with an error:\n{e.output}"
+            ) from e
         raise MissingDependencyError(
-            "Could not find program '{}' on the PATH".format(
-                program)) from e
+            f"Could not find program '{program}' on the PATH"
+        ) from e
     try:
         version = re.match(regex, output.strip()).group(1)
     except AttributeError as e:
         raise MissingDependencyError(
-            ("The program '{}' did not report its version. "
-            "Message was:\n{}").format(program, output)
+            f"The program '{program}' did not report its version. "
+            f"Message was:\n{output}"
         )
 
     return version
+
+
+missing_program = '''
+The program '{program}' could not be executed or was not found on your
+system PATH.
+'''
+
+missing_optional_program = '''
+The program '{program}' could not be executed or was not found on your
+system PATH.  This program is required when you use the
+{required_for} arguments.  You could try omitting these arguments, or install
+the package.
+'''
+
+missing_recommend_program = '''
+The program '{program}' could not be executed or was not found on your
+system PATH.  This program is recommended when using the {required_for} arguments,
+but not required, so we will proceed.  For best results, install the program.
+'''
+
+old_version = '''
+OCRmyPDF requires '{program}' {need_version} or higher.  Your system appears
+to have {found_version}.  Please update this program.
+'''
+
+old_version_required_for = '''
+OCRmyPDF requires '{program}' {need_version} or higher when run with the
+{required_for} arguments.  If you omit these arguments, OCRmyPDF may be able to
+proceed.  For best results, install the program.
+'''
+
+osx_install_advice = '''
+If you have homebrew installed, try these command to install the missing
+package:
+    brew install {package}
+'''
+
+linux_install_advice = '''
+On systems with the aptitude package manager (Debian, Ubuntu), try these
+commands:
+    sudo apt-get update
+    sudo apt-get install {package}
+
+On RPM-based systems (Red Hat, Fedora), search for instructions on
+installing the RPM for {program}.
+'''
+
+
+def _get_platform():
+    if sys.platform.startswith('freebsd'):
+        return 'freebsd'
+    elif sys.platform.startswith('linux'):
+        return 'linux'
+    return sys.platform
+
+
+def _error_trailer(program, package, **kwargs):
+    if isinstance(package, Mapping):
+        package = package[_get_platform()]
+
+    if _get_platform() == 'darwin':
+        log.info(osx_install_advice.format(**locals()))
+    elif _get_platform() == 'linux':
+        log.info(linux_install_advice.format(**locals()))
+
+
+def _error_missing_program(program, package, required_for, recommended):
+    if required_for:
+        log.error(missing_optional_program.format(**locals()))
+    elif recommended:
+        log.info(missing_recommend_program.format(**locals()))
+    else:
+        log.error(missing_program.format(**locals()))
+    _error_trailer(**locals())
+
+
+def _error_old_version(program, package, need_version, found_version, required_for):
+    if required_for:
+        log.error(old_version_required_for.format(**locals()))
+    else:
+        log.error(old_version.format(**locals()))
+    _error_trailer(**locals())
+
+
+def check_external_program(
+    *,
+    program,
+    package,
+    version_checker,
+    need_version,
+    required_for=None,
+    recommended=False,
+    **kwargs,  # To consume log parameter
+):
+    if kwargs:
+        if not 'log' in kwargs:
+            log.warning('check_external_program(log=...) is deprecated')
+    try:
+        found_version = version_checker()
+    except (CalledProcessError, FileNotFoundError, MissingDependencyError):
+        _error_missing_program(program, package, required_for, recommended)
+        if not recommended:
+            raise MissingDependencyError()
+        return
+
+    if found_version < need_version:
+        _error_old_version(program, package, need_version, found_version, required_for)
+        if not recommended:
+            raise MissingDependencyError()
+
+    log.debug('Found %s %s', program, found_version)

src/ocrmypdf/exec/ghostscript.py

@@ -15,15 +15,21 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-from tempfile import NamedTemporaryFile
-from subprocess import run, PIPE, STDOUT
-from shutil import copy
-from functools import lru_cache
+import logging
 import re
+from functools import lru_cache
+from os import fspath
+from shutil import copy
+from subprocess import PIPE, STDOUT, run
+from tempfile import NamedTemporaryFile
+
 from PIL import Image
-from . import get_version
+
 from ..exceptions import SubprocessOutputError
-from ..helpers import fspath
+from . import get_version
+
+
+gslog = logging.getLogger()
 
 
 @lru_cache(maxsize=1)
@@ -32,15 +38,20 @@ def version():
 
 
 def jpeg_passthrough_available():
-    """
-    Ghostscript 9.23 introduced JPEG passthrough but it seems to corrupt the
-    last two bytes of certain images, for now we disable it for 9.23 and
-    do not mention it for < 9.23.
+    """Returns True if the installed version of Ghostscript supports JPEG passthru
 
+    Prior to 9.23, Ghostscript decode and re-encoded JPEGs internally. In 9.23
+    it gained the ability to keep JPEGs unmodified. However, the 9.23
+    implementation was buggy and would deletes the last two bytes of images in
+    some cases, as reported here.
     https://bugs.ghostscript.com/show_bug.cgi?id=699216
 
+    The issue was fixed for 9.24, hence that is the first version we consider
+    the feature available. (However, we don't use 9.24 at all, so the first
+    version that allows JPEG passthrough is 9.25.
+
     """
-    return False
+    return version() >= '9.24'
 
 
 def _gs_error_reported(stream):
@@ -48,8 +59,7 @@ def _gs_error_reported(stream):
 
 
 def extract_text(input_file, pageno=1):
-    """
-    Use the txtwrite device to get text layout information out
+    """Use the txtwrite device to get text layout information out
 
     For details on options of -dTextFormat see
     https://www.ghostscript.com/doc/current/VectorDevices.htm#TXT
@@ -60,39 +70,52 @@ def extract_text(input_file, pageno=1):
     <span bbox="left top right bottom" font="..." size="...">
     <char bbox="...." c="X"/>
 
+    :param pageno: number of page to extract, or all pages if None
     :return: XML-ish text representation in bytes
-
     """
 
-    args_gs = [
-        'gs',
-        '-dQUIET',
-        '-dSAFER',
-        '-dBATCH',
-        '-dNOPAUSE',
-        '-sDEVICE=txtwrite',
-        '-dTextFormat=0',
-        '-dFirstPage=%i' % pageno,
-        '-dLastPage=%i' % pageno,
-        '-o', '-',
-        input_file
-    ]
+    if pageno is not None:
+        pages = ['-dFirstPage=%i' % pageno, '-dLastPage=%i' % pageno]
+    else:
+        pages = []
+
+    args_gs = (
+        [
+            'gs',
+            '-dQUIET',
+            '-dSAFER',
+            '-dBATCH',
+            '-dNOPAUSE',
+            '-sDEVICE=txtwrite',
+            '-dTextFormat=0',
+        ]
+        + pages
+        + ['-o', '-', fspath(input_file)]
+    )
 
     p = run(args_gs, stdout=PIPE, stderr=PIPE)
     if p.returncode != 0:
         raise SubprocessOutputError(
-            'Ghostscript text extraction failed\n%s\n%s\n%s' % (
-                input_file, p.stdout.decode(), p.stderr.decode()
-            )
+            'Ghostscript text extraction failed\n%s\n%s\n%s'
+            % (input_file, p.stdout.decode(), p.stderr.decode())
         )
 
     return p.stdout
 
 
-def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
-                  pageno=1, page_dpi=None, rotation=None):
-    """
-    Rasterize one page of a PDF at resolution (xres, yres) in canvas units.
+def rasterize_pdf(
+    input_file,
+    output_file,
+    xres,
+    yres,
+    raster_device,
+    log,
+    pageno=1,
+    page_dpi=None,
+    rotation=None,
+    filter_vector=False,
+):
+    """Rasterize one page of a PDF at resolution (xres, yres) in canvas units.
 
     The image is sized to match the integer pixels dimensions implied by
     (xres, yres) even if those numbers are noninteger. The image's DPI will
@@ -106,59 +129,51 @@ def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
     :param log:
     :param pageno: page number to rasterize (beginning at page 1)
     :param page_dpi: resolution tuple (x, y) overriding output image DPI
+    :param rotation: 0, 90, 180, 270: clockwise angle to rotate page
+    :param filter_vector: if True, remove vector graphics objects
     :return:
     """
-    res = xres, yres
-    int_res = round(xres), round(yres)
+    res = round(xres, 6), round(yres, 6)
     if not page_dpi:
         page_dpi = res
-
-    autorotate = '/PageByPage' if rotation is None else '/None'
+    if not log:
+        log = gslog
 
     with NamedTemporaryFile(delete=True) as tmp:
-        args_gs = [
-            'gs',
-            '-dQUIET',
-            '-dSAFER',
-            '-dBATCH',
-            '-dNOPAUSE',
-            '-sDEVICE=%s' % raster_device,
-            '-dFirstPage=%i' % pageno,
-            '-dLastPage=%i' % pageno,
-            '-r{0}x{1}'.format(str(int_res[0]), str(int_res[1])),
-            '-o', tmp.name,
-            '-dAutoRotatePages=%s' % autorotate,
-            '-f',
-            fspath(input_file)
-        ]
+        args_gs = (
+            [
+                'gs',
+                '-dQUIET',
+                '-dSAFER',
+                '-dBATCH',
+                '-dNOPAUSE',
+                f'-sDEVICE={raster_device}',
+                f'-dFirstPage={pageno}',
+                f'-dLastPage={pageno}',
+                f'-r{res[0]:f}x{res[1]:f}',
+            ]
+            + (['-dFILTERVECTOR'] if filter_vector else [])
+            + [
+                '-o',
+                tmp.name,
+                '-dAutoRotatePages=/None',  # Probably has no effect on raster
+                '-f',
+                fspath(input_file),
+            ]
+        )
 
         log.debug(args_gs)
-        p = run(args_gs, stdout=PIPE, stderr=STDOUT,
-                universal_newlines=True)
+        p = run(args_gs, stdout=PIPE, stderr=STDOUT, universal_newlines=True)
         if _gs_error_reported(p.stdout):
             log.error(p.stdout)
         else:
             log.debug(p.stdout)
 
         if p.returncode != 0:
-            log.error('Ghostscript rasterizing failed')
-            raise SubprocessOutputError()
-
-        # Ghostscript only accepts integers for output resolution
-        # if the resolution happens to be fractional, then the discrepancy
-        # would change the size of the output page, especially if the DPI
-        # is quite low. Resize the image to the expected size
+            raise SubprocessOutputError('Ghostscript rasterizing failed')
 
         tmp.seek(0)
         with Image.open(tmp) as im:
-            expected_size = round(im.size[0] / int_res[0] * res[0]), \
-                            round(im.size[1] / int_res[1] * res[1])
-            if expected_size != im.size or page_dpi != (xres, yres):
-                log.debug(
-                    "Ghostscript: resize output image {} -> {}".format(
-                        im.size, expected_size))
-                im = im.resize(expected_size)
-
             if rotation is not None:
                 log.debug("Rotating output by %i", rotation)
                 # rotation is a clockwise angle and Image.ROTATE_* is
@@ -174,8 +189,35 @@ def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
             im.save(fspath(output_file), dpi=page_dpi)
 
 
-def generate_pdfa(pdf_pages, output_file, compression, log,
-                  threads=1, pdf_version='1.5', pdfa_part='2'):
+def generate_pdfa(
+    pdf_pages,
+    output_file,
+    compression,
+    log,
+    threads=1,
+    pdf_version='1.5',
+    pdfa_part='2',
+):
+    """Generate a PDF/A.
+
+    The pdf_pages, a list files, will be merged into output_file. One or more
+    PDF files may be merged. One of the files in this list must be a pdfmark
+    file that provides Ghostscript with details on how to perform the PDF/A
+    conversion. By default with we pick PDF/A-2b, but this works for 1 or 3.
+
+    compression can be 'jpeg', 'lossless', or an empty string. In 'jpeg',
+    Ghostscript is instructed to convert color and grayscale images to DCT
+    (JPEG encoding). In 'lossless' Ghostscript is told to convert images to
+    Flate (lossless/PNG). If the parameter is omitted Ghostscript is left to
+    make its own decisions about how to encode images; it appears to use a
+    heuristic to decide how to encode images. As of Ghostscript 9.25, we
+    support passthrough JPEG which allows Ghostscript to avoid transcoding
+    images entirely. (The feature was added in 9.23 but broken, and the 9.24
+    release of Ghostscript had regressions, so we don't support it until 9.25.)
+    """
+    if not log:
+        log = gslog
+
     compression_args = []
     if compression == 'jpeg':
         compression_args = [
@@ -209,27 +251,31 @@ def generate_pdfa(pdf_pages, output_file, compression, log,
         compression_args.append('-dPassThroughJPEGImages=false')
 
     with NamedTemporaryFile(delete=True) as gs_pdf:
-        args_gs = [
-            "gs",
-            "-dQUIET",
-            "-dBATCH",
-            "-dNOPAUSE",
-            "-dCompatibilityLevel=" + str(pdf_version),
-            "-dNumRenderingThreads=" + str(threads),
-            "-sDEVICE=pdfwrite",
-            "-dAutoRotatePages=/None",
-            "-sColorConversionStrategy=" + strategy,
-            "-sProcessColorModel=DeviceRGB"
-        ] + compression_args + [
-            "-dJPEGQ=95",
-            "-dPDFA=" + pdfa_part,
-            "-dPDFACompatibilityPolicy=1",
-            "-sOutputFile=" + gs_pdf.name,
-        ]
+        # nb no need to specify ProcessColorModel when ColorConversionStrategy
+        # is set; see:
+        # https://bugs.ghostscript.com/show_bug.cgi?id=699392
+        args_gs = (
+            [
+                "gs",
+                "-dQUIET",
+                "-dBATCH",
+                "-dNOPAUSE",
+                "-dCompatibilityLevel=" + str(pdf_version),
+                "-sDEVICE=pdfwrite",
+                "-dAutoRotatePages=/None",
+                "-sColorConversionStrategy=" + strategy,
+            ]
+            + compression_args
+            + [
+                "-dJPEGQ=95",
+                "-dPDFA=" + pdfa_part,
+                "-dPDFACompatibilityPolicy=1",
+                "-sOutputFile=" + gs_pdf.name,
+            ]
+        )
         args_gs.extend(fspath(s) for s in pdf_pages)  # Stringify Path objs
         log.debug(args_gs)
-        p = run(args_gs, stdout=PIPE, stderr=STDOUT,
-                universal_newlines=True)
+        p = run(args_gs, stdout=PIPE, stderr=STDOUT, universal_newlines=True)
 
         if _gs_error_reported(p.stdout):
             log.error(p.stdout)
@@ -240,7 +286,7 @@ def generate_pdfa(pdf_pages, output_file, compression, log,
             log.debug(
                 "Ghostscript had to remove PDF 'overprinting' from the "
                 "input file to complete PDF/A conversion. "
-                )
+            )
         else:
             log.debug(p.stdout)
 
@@ -249,5 +295,4 @@ def generate_pdfa(pdf_pages, output_file, compression, log,
             # PDF/A - check PDF/A status elsewhere
             copy(gs_pdf.name, fspath(output_file))
         else:
-            log.error('Ghostscript PDF/A rendering failed')
-            raise SubprocessOutputError()
+            raise SubprocessOutputError('Ghostscript PDF/A rendering failed')

src/ocrmypdf/exec/jbig2enc.py

@@ -15,14 +15,11 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-from subprocess import CalledProcessError, run, PIPE
 from functools import lru_cache
-import sys
-import os
-import shutil
+from subprocess import PIPE, run
 
 from . import get_version
-from ..exceptions import ExitCode, MissingDependencyError
+from ..exceptions import MissingDependencyError
 
 
 @lru_cache(maxsize=1)
@@ -30,7 +27,6 @@ def version():
     return get_version('jbig2', regex=r'jbig2enc (\d+(\.\d+)*).*')
 
 
-@lru_cache(maxsize=1)
 def available():
     try:
         version()
@@ -44,10 +40,20 @@ def convert_group(*, cwd, infiles, out_prefix):
         'jbig2',
         '-b',
         out_prefix,
-        '-s',
+        '-s',  # symbol mode (lossy)
+        # '-r', # refinement mode (lossless symbol mode, currently disabled in
+        # jbig2)
         '-p',
     ]
     args.extend(infiles)
     proc = run(args, cwd=cwd, stdout=PIPE, stderr=PIPE)
     proc.check_returncode()
-    return proc
+    return proc
+
+
+def convert_single(*, cwd, infile, outfile):
+    args = ['jbig2', '-p', infile]
+    with open(outfile, 'wb') as fstdout:
+        proc = run(args, cwd=cwd, stdout=fstdout, stderr=PIPE)
+    proc.check_returncode()
+    return proc

src/ocrmypdf/exec/pngquant.py

@@ -15,15 +15,14 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-from subprocess import CalledProcessError, run
-from tempfile import TemporaryFile
 from functools import lru_cache
-import sys
-import os
-import shutil
+from subprocess import run
+from tempfile import NamedTemporaryFile
+
+from PIL import Image
 
 from . import get_version
-from ..exceptions import ExitCode
+from ..exceptions import MissingDependencyError
 
 
 @lru_cache(maxsize=1)
@@ -31,15 +30,41 @@ def version():
     return get_version('pngquant', regex=r'(\d+(\.\d+)*).*')
 
 
+def available():
+    try:
+        version()
+    except MissingDependencyError:
+        return False
+    return True
+
+
 def quantize(input_file, output_file, quality_min, quality_max):
-    args = [
-        'pngquant',
-        '--force',
-        '--skip-if-larger',
-        '--output', output_file,
-        '--quality', '{}-{}'.format(quality_min, quality_max),
-        '--',
-        input_file
-    ]
-    proc = run(args)
-    proc.check_returncode()
+    if input_file.endswith('.jpg'):
+        im = Image.open(input_file)
+        with NamedTemporaryFile(suffix='.png') as tmp:
+            im.save(tmp)
+            args = [
+                'pngquant',
+                '--force',
+                '--skip-if-larger',
+                '--output',
+                output_file,
+                '--quality',
+                f'{quality_min}-{quality_max}',
+                '--',
+                tmp.name,
+            ]
+            run(args)
+    else:
+        args = [
+            'pngquant',
+            '--force',
+            '--skip-if-larger',
+            '--output',
+            output_file,
+            '--quality',
+            f'{quality_min}-{quality_max}',
+            '--',
+            input_file,
+        ]
+        run(args)

src/ocrmypdf/exec/qpdf.py

@@ -15,12 +15,11 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-from subprocess import CalledProcessError, STDOUT, PIPE, run
 from functools import lru_cache
+from os import fspath
+from subprocess import PIPE, STDOUT, CalledProcessError, run
 
-from ..exceptions import InputFileError, SubprocessOutputError, \
-    EncryptedPdfError
-from . import  get_version
+from . import get_version
 
 
 @lru_cache(maxsize=1)
@@ -29,22 +28,16 @@ def version():
 
 
 def check(input_file, log=None):
-    args_qpdf = [
-        'qpdf',
-        '--check',
-        input_file
-    ]
+    args_qpdf = ['qpdf', '--check', fspath(input_file)]
 
     if log is None:
         import logging as log
 
     try:
-        run(args_qpdf, stderr=STDOUT, stdout=PIPE, universal_newlines=True,
-            check=True)
+        run(args_qpdf, stderr=STDOUT, stdout=PIPE, universal_newlines=True, check=True)
     except CalledProcessError as e:
         if e.returncode == 2:
-            log.error("%s: not a valid PDF, and could not repair it.",
-                input_file)
+            log.error("%s: not a valid PDF, and could not repair it.", input_file)
             log.error("Details:")
             log.error(e.output)
         elif e.returncode == 3:

src/ocrmypdf/exec/tesseract.py

@@ -15,22 +15,24 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-import sys
 import os
 import shutil
-from functools import lru_cache
+import sys
 from collections import namedtuple
-from textwrap import dedent
-from subprocess import CalledProcessError, TimeoutExpired, check_output, STDOUT
 from contextlib import suppress
+from functools import lru_cache
+from os import fspath
+from subprocess import PIPE, STDOUT, CalledProcessError, TimeoutExpired, run
 
-from ..exceptions import MissingDependencyError, TesseractConfigError
-from ..helpers import page_number
 from . import get_version
+from ..exceptions import (
+    MissingDependencyError,
+    TesseractConfigError,
+    SubprocessOutputError,
+)
+from ..helpers import page_number
 
-OrientationConfidence = namedtuple(
-    'OrientationConfidence',
-    ('angle', 'confidence'))
+OrientationConfidence = namedtuple('OrientationConfidence', ('angle', 'confidence'))
 
 HOCR_TEMPLATE = """<?xml version="1.0" encoding="UTF-8"?>
 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
@@ -39,8 +41,8 @@ HOCR_TEMPLATE = """<?xml version="1.0" encoding="UTF-8"?>
  <head>
   <title></title>
 <meta http-equiv="Content-Type" content="text/html;charset=utf-8" />
-  <meta name='ocr-system' content='tesseract 3.05.00' />
-  <meta name='ocr-capabilities' content='ocr_page ocr_carea ocr_par ocr_line ocrx_word'/>
+  <meta name='ocr-system' content='tesseract 4.0.0' />
+  <meta name='ocr-capabilities' content='ocr_page ocr_carea ocr_par ocr_line ocrx_word ocrp_wconf'/>
 </head>
 <body>
   <div class='ocr_page' id='page_1' title='image "_blank.png"; bbox 0 0 {0} {1}; ppageno 0'>
@@ -50,70 +52,74 @@ HOCR_TEMPLATE = """<?xml version="1.0" encoding="UTF-8"?>
 """
 
 
-@lru_cache(maxsize=1)
-def version():
-    return get_version('tesseract', regex=r'tesseract\s(.+)')
+def version(tesseract_env=None):
+    return get_version('tesseract', regex=r'tesseract\s(.+)', env=tesseract_env)
 
 
-def v4():
+def v4(tesseract_env=None):
     "Is this Tesseract v4.0?"
-    return version() >= '4'
+    return version(tesseract_env) >= '4'
 
 
-@lru_cache(maxsize=1)
-def has_textonly_pdf():
+def has_textonly_pdf(tesseract_env=None):
     """Does Tesseract have textonly_pdf capability?
 
-    Available in 3.05.01, and v4.00.00alpha since January 2017. Best to
+    Available in v4.00.00alpha since January 2017. Best to
     parse the parameter list
     """
-    args_tess = [
-        'tesseract',
-        '--print-parameters',
-        'pdf'
-    ]
+    args_tess = ['tesseract', '--print-parameters', 'pdf']
     params = ''
     try:
-        params = check_output(
-                args_tess, universal_newlines=True, stderr=STDOUT)
+        proc = run(
+            args_tess,
+            check=True,
+            universal_newlines=True,
+            stdout=PIPE,
+            stderr=STDOUT,
+            env=tesseract_env,
+        )
+        params = proc.stdout
     except CalledProcessError as e:
-        print("Could not --print-parameters from tesseract",
-              file=sys.stderr)
-        raise MissingDependencyError from e
+        raise MissingDependencyError(
+            "Could not --print-parameters from tesseract"
+        ) from e
     if 'textonly_pdf' in params:
         return True
     return False
 
 
-def psm():
-    "If Tesseract 4.0, use argument --psm instead of -psm"
-    return '--psm' if v4() else '-psm'
+def languages(tesseract_env=None):
+    def lang_error(output):
+        msg = (
+            "Tesseract failed to report available languages.\n"
+            "Output from Tesseract:\n"
+            "-----------\n"
+        )
+        msg += output
+        return msg
 
-
-@lru_cache(maxsize=1)
-def languages():
-    args_tess = [
-        'tesseract',
-        '--list-langs'
-    ]
+    args_tess = ['tesseract', '--list-langs']
     try:
-        langs = check_output(
-                args_tess, universal_newlines=True, stderr=STDOUT)
+        proc = run(
+            args_tess,
+            universal_newlines=True,
+            stdout=PIPE,
+            stderr=STDOUT,
+            check=True,
+            env=tesseract_env,
+        )
+        output = proc.stdout
     except CalledProcessError as e:
-        msg = dedent("""Tesseract failed to report available languages.
-        Output from Tesseract:
-        -----------
-        """)
-        msg += e.output
-        print(msg, file=sys.stderr)
-        raise MissingDependencyError from e
-    return set(lang.strip() for lang in langs.splitlines()[1:])
+        raise MissingDependencyError(lang_error(e.output)) from e
+
+    header, *rest = output.splitlines()
+    if not header.startswith('List of available languages'):
+        raise MissingDependencyError(lang_error(output))
+    return set(lang.strip() for lang in rest)
 
 
 def tess_base_args(langs, engine_mode):
-    args = [
-        'tesseract',
-    ]
+    args = ['tesseract']
     if langs:
         args.extend(['-l', '+'.join(langs)])
     if engine_mode is not None and v4():
@@ -121,24 +127,34 @@ def tess_base_args(langs, engine_mode):
     return args
 
 
-def get_orientation(input_file, engine_mode, timeout: float, log):
+def get_orientation(input_file, engine_mode, timeout: float, log, tesseract_env=None):
     args_tesseract = tess_base_args(['osd'], engine_mode) + [
-        psm(), '0',
-        input_file,
-        'stdout'
+        '--psm',
+        '0',
+        fspath(input_file),
+        'stdout',
     ]
 
     try:
-        stdout = check_output(
-            args_tesseract, stderr=STDOUT, timeout=timeout)
+        p = run(
+            args_tesseract,
+            stdout=PIPE,
+            stderr=STDOUT,
+            timeout=timeout,
+            check=True,
+            env=tesseract_env,
+        )
+        stdout = p.stdout
     except TimeoutExpired:
         return OrientationConfidence(angle=0, confidence=0.0)
     except CalledProcessError as e:
         tesseract_log_output(log, e.output, input_file)
-        if (b'Too few characters. Skipping this page' in e.output or
-                b'Image too large' in e.output):
+        if (
+            b'Too few characters. Skipping this page' in e.output
+            or b'Image too large' in e.output
+        ):
             return OrientationConfidence(0, 0)
-        raise e from e
+        raise SubprocessOutputError() from e
     else:
         osd = {}
         for line in stdout.decode().splitlines():
@@ -148,32 +164,24 @@ def get_orientation(input_file, engine_mode, timeout: float, log):
                 osd[parts[0].strip()] = parts[1].strip()
 
         angle = int(osd.get('Orientation in degrees', 0))
-        if 'Orientation' in osd:
-            # Tesseract < 3.04.01
-            # reports "Orientation in degrees" as a counterclockwise angle
-            # We keep it clockwise
-            assert 'Rotate' not in osd
-            angle = -angle % 360
-        else:
-            # Tesseract >= 3.04.01
-            # reports "Orientation in degrees" as a clockwise angle
-            assert 'Rotate' in osd
-
         oc = OrientationConfidence(
-            angle=angle,
-            confidence=float(osd.get('Orientation confidence', 0)))
+            angle=angle, confidence=float(osd.get('Orientation confidence', 0))
+        )
         return oc
 
 
 def tesseract_log_output(log, stdout, input_file):
-    prefix = "{0:4d}: [tesseract] ".format(page_number(input_file))
+    prefix = "[tesseract] "
 
     try:
         text = stdout.decode()
     except UnicodeDecodeError:
-        log.error(prefix + "command line output was not utf-8. " +
-            "This usually means Tesseract's language packs do not match "
-            "the installed version of Tesseract.")
+        log.error(
+            prefix
+            + "command line output was not utf-8. "
+            + "This usually means Tesseract's language packs do not match "
+            "the installed version of Tesseract."
+        )
         text = stdout.decode('utf-8', 'backslashreplace')
 
     lines = text.splitlines()
@@ -190,6 +198,10 @@ def tesseract_log_output(log, stdout, input_file):
             pass  # Appears to be spurious/problem with nonwhite borders
         elif 'Error in boxClipToRectangle' in line:
             pass  # Always appears with pixScanForForeground message
+        elif 'parameter not found: ' in line.lower():
+            log.error(prefix + line.strip())
+            problem = line.split('found: ')[1]
+            raise TesseractConfigError(problem)
         elif 'error' in line.lower() or 'exception' in line.lower():
             log.error(prefix + line.strip())
         elif 'warning' in line.lower():
@@ -200,8 +212,10 @@ def tesseract_log_output(log, stdout, input_file):
             log.info(prefix + line.strip())
 
 
-def page_timedout(log, input_file):
-    prefix = "{0:4d}: [tesseract] ".format(page_number(input_file))
+def page_timedout(log, input_file, timeout):
+    if timeout == 0:
+        return
+    prefix = f"{(page_number(input_file)):4d}: [tesseract] "
     log.warning(prefix + " took too long to OCR - skipping")
 
 
@@ -219,10 +233,19 @@ def _generate_null_hocr(output_hocr, output_sidecar, image):
         f.write('[skipped page]')
 
 
-def generate_hocr(input_file, output_files, language: list, engine_mode,
-                  tessconfig: list,
-                  timeout: float, pagesegmode: int, user_words, user_patterns,
-                  log):
+def generate_hocr(
+    input_file,
+    output_files,
+    language: list,
+    engine_mode,
+    tessconfig: list,
+    timeout: float,
+    pagesegmode: int,
+    user_words,
+    user_patterns,
+    tesseract_env,
+    log,
+):
 
     output_hocr = next(o for o in output_files if o.endswith('.hocr'))
     output_sidecar = next(o for o in output_files if o.endswith('.txt'))
@@ -231,7 +254,7 @@ def generate_hocr(input_file, output_files, language: list, engine_mode,
     args_tesseract = tess_base_args(language, engine_mode)
 
     if pagesegmode is not None:
-        args_tesseract.extend([psm(), str(pagesegmode)])
+        args_tesseract.extend(['--psm', str(pagesegmode)])
 
     if user_words:
         args_tesseract.extend(['--user-words', user_words])
@@ -241,35 +264,31 @@ def generate_hocr(input_file, output_files, language: list, engine_mode,
 
     # Reminder: test suite tesseract spoofers will break after any changes
     # to the number of order parameters here
-    # Tesseract 3.04 requires the order here to be "hocr txt" and will fail
-    # on "txt hocr"
-
-    args_tesseract.extend([
-        input_file,
-        prefix,
-        'hocr',
-        'txt'
-    ] + tessconfig)
+    args_tesseract.extend([input_file, prefix, 'hocr', 'txt'] + tessconfig)
     try:
         log.debug(args_tesseract)
-        stdout = check_output(
-            args_tesseract, stderr=STDOUT,
-            timeout=timeout)
+        p = run(
+            args_tesseract,
+            stdout=PIPE,
+            stderr=STDOUT,
+            timeout=timeout,
+            check=True,
+            env=tesseract_env,
+        )
+        stdout = p.stdout
     except TimeoutExpired:
         # Generate a HOCR file with no recognized text if tesseract times out
         # Temporary workaround to hocrTransform not being able to function if
         # it does not have a valid hOCR file.
-        page_timedout(log, input_file)
+        page_timedout(log, input_file, timeout)
         _generate_null_hocr(output_hocr, output_sidecar, input_file)
     except CalledProcessError as e:
         tesseract_log_output(log, e.output, input_file)
-        if b'read_params_file: parameter not found' in e.output:
-            raise TesseractConfigError() from e
         if b'Image too large' in e.output:
             _generate_null_hocr(output_hocr, output_sidecar, input_file)
             return
 
-        raise e from e
+        raise SubprocessOutputError() from e
     else:
         tesseract_log_output(log, stdout, input_file)
         # The sidecar text file will get the suffix .txt; rename it to
@@ -294,11 +313,24 @@ def use_skip_page(text_only, skip_pdf, output_pdf, output_text):
         out.write(b'')
 
 
-def generate_pdf(*, input_image, skip_pdf=None, output_pdf, output_text,
-                 language: list, engine_mode, text_only: bool,
-                 tessconfig: list, timeout: float, pagesegmode: int,
-                 user_words, user_patterns, log):
-    '''Use Tesseract to render a PDF.
+def generate_pdf(
+    *,
+    input_image,
+    skip_pdf=None,
+    output_pdf,
+    output_text,
+    language: list,
+    engine_mode,
+    text_only: bool,
+    tessconfig: list,
+    timeout: float,
+    pagesegmode: int,
+    user_words,
+    user_patterns,
+    tesseract_env,
+    log,
+):
+    """Use Tesseract to render a PDF.
 
     input_image -- image to analyze
     skip_pdf -- if we time out, use this file as output
@@ -310,14 +342,14 @@ def generate_pdf(*, input_image, skip_pdf=None, output_pdf, output_text,
     tessconfig -- tesseract configuration
     timeout -- timeout (seconds)
     log -- logger object
-    '''
+    """
 
     args_tesseract = tess_base_args(language, engine_mode)
 
     if pagesegmode is not None:
-        args_tesseract.extend([psm(), str(pagesegmode)])
+        args_tesseract.extend(['--psm', str(pagesegmode)])
 
-    if text_only and has_textonly_pdf():
+    if text_only and has_textonly_pdf(tesseract_env):
         args_tesseract.extend(['-c', 'textonly_pdf=1'])
 
     if user_words:
@@ -331,31 +363,28 @@ def generate_pdf(*, input_image, skip_pdf=None, output_pdf, output_text,
     # Reminder: test suite tesseract spoofers might break after any changes
     # to the number of order parameters here
 
-    args_tesseract.extend([
-        input_image,
-        prefix,
-        'pdf',
-        'txt'
-    ] + tessconfig)
-
+    args_tesseract.extend([input_image, prefix, 'pdf', 'txt'] + tessconfig)
     try:
         log.debug(args_tesseract)
-        stdout = check_output(
-            args_tesseract, stderr=STDOUT,
-            timeout=timeout)
+        p = run(
+            args_tesseract,
+            stdout=PIPE,
+            stderr=STDOUT,
+            timeout=timeout,
+            check=True,
+            env=tesseract_env,
+        )
+        stdout = p.stdout
         if os.path.exists(prefix + '.txt'):
             shutil.move(prefix + '.txt', output_text)
     except TimeoutExpired:
-        page_timedout(log, input_image)
+        page_timedout(log, input_image, timeout)
         use_skip_page(text_only, skip_pdf, output_pdf, output_text)
     except CalledProcessError as e:
         tesseract_log_output(log, e.output, input_image)
-        if b'read_params_file: parameter not found' in e.output:
-            raise TesseractConfigError() from e
-
         if b'Image too large' in e.output:
             use_skip_page(text_only, skip_pdf, output_pdf, output_text)
             return
-        raise e from e
+        raise SubprocessOutputError() from e
     else:
         tesseract_log_output(log, stdout, input_image)

src/ocrmypdf/exec/unpaper.py

@@ -18,20 +18,18 @@
 # unpaper documentation:
 # https://github.com/Flameeyes/unpaper/blob/master/doc/basic-concepts.md
 
-from subprocess import CalledProcessError, STDOUT, check_output
-from tempfile import NamedTemporaryFile
-import sys
 import os
+import shlex
+import subprocess
+import sys
 from functools import lru_cache
-from ..exceptions import MissingDependencyError
+from subprocess import PIPE, STDOUT, CalledProcessError
+from tempfile import TemporaryDirectory
+
 from . import get_version
+from ..exceptions import MissingDependencyError, SubprocessOutputError
 
-
-try:
-    from PIL import Image
-except ImportError:
-    print("Could not find Python3 imaging library", file=sys.stderr)
-    raise
+from PIL import Image
 
 
 @lru_cache(maxsize=1)
@@ -40,11 +38,7 @@ def version():
 
 
 def run(input_file, output_file, dpi, log, mode_args):
-    args_unpaper = [
-        'unpaper',
-        '-v',
-        '--dpi', str(dpi)
-    ] + mode_args
+    args_unpaper = ['unpaper', '-v', '--dpi', str(dpi)] + mode_args
 
     SUFFIXES = {'1': '.pbm', 'L': '.pgm', 'RGB': '.ppm'}
 
@@ -57,46 +51,77 @@ def run(input_file, output_file, dpi, log, mode_args):
             else:
                 im = im.convert(mode='RGB')
         except IOError as e:
-            log.error("Could not convert image with type " + im.mode)
             im.close()
-            raise MissingDependencyError() from e
+            raise MissingDependencyError(
+                "Could not convert image with type " + im.mode
+            ) from e
 
     try:
         suffix = SUFFIXES[im.mode]
     except KeyError:
-        log.error("Failed to convert image to a supported format.")
         im.close()
-        raise MissingDependencyError() from e
+        raise MissingDependencyError(
+            "Failed to convert image to a supported format."
+        ) from e
 
-    with NamedTemporaryFile(suffix=suffix) as input_pnm, \
-            NamedTemporaryFile(suffix=suffix, mode="r+b") as output_pnm:
+    with TemporaryDirectory() as tmpdir:
+        input_pnm = os.path.join(tmpdir, f'input{suffix}')
+        output_pnm = os.path.join(tmpdir, f'output{suffix}')
         im.save(input_pnm, format='PPM')
         im.close()
 
-        os.unlink(output_pnm.name)
-
-        args_unpaper.extend([input_pnm.name, output_pnm.name])
+        # To prevent any shenanigans from accepting arbitrary parameters in
+        # --unpaper-args, we:
+        # 1) run with cwd set to a tmpdir with only unpaper's files
+        # 2) forbid the use of '/' in arguments, to prevent changing paths
+        # 3) append absolute paths for the input and output file
+        # This should ensure that a user cannot clobber some other file with
+        # their unpaper arguments (whether intentionally or otherwise)
+        args_unpaper.extend([input_pnm, output_pnm])
         try:
-            stdout = check_output(
-                args_unpaper, close_fds=True,
-                universal_newlines=True, stderr=STDOUT,
-                )
+            proc = subprocess.run(
+                args_unpaper,
+                check=True,
+                close_fds=True,
+                universal_newlines=True,
+                stderr=STDOUT,
+                cwd=tmpdir,
+                stdout=PIPE,
+            )
         except CalledProcessError as e:
             log.debug(e.output)
             raise e from e
         else:
-            log.debug(stdout)
-            # unpaper sets dpi to 72
-            Image.open(output_pnm.name).save(output_file, dpi=(dpi, dpi))
+            log.debug(proc.stdout)
+            # unpaper sets dpi to 72; fix this
+            try:
+                Image.open(output_pnm).save(output_file, dpi=(dpi, dpi))
+            except (FileNotFoundError, OSError):
+                raise SubprocessOutputError(
+                    "unpaper: failed to produce the expected output file. Called with: "
+                    + str(args_unpaper)
+                ) from None
 
 
-def clean(input_file, output_file, dpi, log):
-    run(input_file, output_file, dpi, log, [
-        '--layout', 'none',
-        '--mask-scan-size', '100',  # don't blank out narrow columns
+def validate_custom_args(args: str):
+    unpaper_args = shlex.split(args)
+    if any('/' in arg for arg in unpaper_args):
+        raise ValueError('No filenames allowed in --unpaper-args')
+    return unpaper_args
+
+
+def clean(input_file, output_file, dpi, log, unpaper_args=None):
+    default_args = [
+        '--layout',
+        'none',
+        '--mask-scan-size',
+        '100',  # don't blank out narrow columns
         '--no-border-align',  # don't align visible content to borders
-        '--no-mask-center',   # don't center visible content within page
-        '--no-grayfilter',    # don't remove light gray areas
-        '--no-blackfilter',   # don't remove solid black areas
-        '--no-deskew',        # don't deskew
-    ])
+        '--no-mask-center',  # don't center visible content within page
+        '--no-grayfilter',  # don't remove light gray areas
+        '--no-blackfilter',  # don't remove solid black areas
+        '--no-deskew',  # don't deskew
+    ]
+    if not unpaper_args:
+        unpaper_args = default_args
+    run(input_file, output_file, dpi, log, unpaper_args)

src/ocrmypdf/filters.py

@@ -0,0 +1,10 @@
+from PIL import Image
+import PIL.ImageOps
+
+
+def invert(im):
+    return PIL.ImageOps.invert(im.convert('L'))
+
+
+def whiteout(im):
+    return Image.new(im.mode, im.size)

src/ocrmypdf/helpers.py

@@ -15,64 +15,69 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-from functools import partial
+import logging
+import multiprocessing
+import os
+import warnings
 from collections.abc import Iterable
 from contextlib import suppress
+from functools import wraps
 from pathlib import Path
-import sys
-import os
-import multiprocessing
-import warnings
+
+log = logging.getLogger(__name__)
 
 
-def re_symlink(input_file, soft_link_name, log=None):
+def re_symlink(input_file, soft_link_name, *args, **kwargs):
     """
     Helper function: relinks soft symbolic link if necessary
     """
-    input_file = fspath(input_file)  # For Py3.5
-    soft_link_name = fspath(soft_link_name)
-    if log is None:
-        prdebug = partial(print, file=sys.stderr)
-    else:
-        prdebug = log.debug
+    if len(args) == 1 and isinstance(args[0], logging.Logger):
+        log.warning("Deprecated: re_symlink(,log)")
+    if 'log' in kwargs:
+        log.warning('Deprecated: re_symlink(...log=)')
+
+    input_file = os.fspath(input_file)
+    soft_link_name = os.fspath(soft_link_name)
 
     # Guard against soft linking to oneself
     if input_file == soft_link_name:
-        prdebug("Warning: No symbolic link made. You are using " +
-                "the original data directory as the working directory.")
+        log.warning(
+            "No symbolic link made. You are using "
+            "the original data directory as the working directory."
+        )
         return
 
     # Soft link already exists: delete for relink?
     if os.path.lexists(soft_link_name):
         # do not delete or overwrite real (non-soft link) file
         if not os.path.islink(soft_link_name):
-            raise FileExistsError(
-                "%s exists and is not a link" % soft_link_name)
+            raise FileExistsError(f"{soft_link_name} exists and is not a link")
         try:
             os.unlink(soft_link_name)
         except OSError:
-            prdebug("Can't unlink %s" % (soft_link_name))
+            log.debug("Can't unlink %s", soft_link_name)
 
     if not os.path.exists(input_file):
-        raise FileNotFoundError(
-            "trying to create a broken symlink to %s" % input_file)
+        raise FileNotFoundError(f"trying to create a broken symlink to {input_file}")
 
-    prdebug("os.symlink(%s, %s)" % (input_file, soft_link_name))
+    log.debug("os.symlink(%s, %s)", input_file, soft_link_name)
 
     # Create symbolic link using absolute path
-    os.symlink(
-        os.path.abspath(input_file),
-        soft_link_name
-    )
+    os.symlink(os.path.abspath(input_file), soft_link_name)
 
 
 def is_iterable_notstr(thing):
     return isinstance(thing, Iterable) and not isinstance(thing, str)
 
 
+def monotonic(L):
+    """Does list increase monotonically?"""
+    return all(b > a for a, b in zip(L, L[1:]))
+
+
 def page_number(input_file):
     """Get one-based page number implied by filename (000002.pdf -> 2)"""
-    return int(os.path.basename(fspath(input_file))[0:6])
+    return int(os.path.basename(os.fspath(input_file))[0:6])
 
 
 def available_cpu_count():
@@ -80,16 +85,9 @@ def available_cpu_count():
         return multiprocessing.cpu_count()
     except NotImplementedError:
         pass
-
-    try:
-        import psutil
-        return psutil.cpu_count()
-    except (ImportError, AttributeError):
-        pass
-
     warnings.warn(
-        "Could not get CPU count.  Assuming one (1) CPU."
-        "Use -j N to set manually.")
+        "Could not get CPU count.  Assuming one (1) CPU." "Use -j N to set manually."
+    )
     return 1
 
 
@@ -103,20 +101,15 @@ def is_file_writable(test_file):
     p = Path(test_file)
 
     if p.is_symlink():
-        # Python 3.5 does not accept parameters for Path.resolve() and behaves
-        # as if strict=True (throws an exception on failure). Python 3.6
-        # defaults to strict=False. This implements strict=False like behavior
-        # for Python 3.5.
-        if sys.version_info[0:2] <= (3, 5):
-            p = Path(os.path.realpath(fspath(p)))
-        else:
-            p = p.resolve(strict=False)
+        p = p.resolve(strict=False)
 
     # p.is_file() throws an exception in some cases
     if p.exists() and p.is_file():
         return os.access(
-            fspath(p), os.W_OK,
-            effective_ids=(os.access in os.supports_effective_ids))
+            os.fspath(p),
+            os.W_OK,
+            effective_ids=(os.access in os.supports_effective_ids),
+        )
     else:
         try:
             fp = p.open('wb')
@@ -129,42 +122,18 @@ def is_file_writable(test_file):
         return True
 
 
-if sys.version_info[0:2] <= (3, 5):
-    def fspath(path):
-        """https://www.python.org/dev/peps/pep-0519/#os"""
-        import pathlib
-        if isinstance(path, (str, bytes)):
-            return path
+def deprecated(func):
+    """Warn that function is deprecated"""
 
-        # Work from the object's type to match method resolution of other magic
-        # methods.
-        path_type = type(path)
-        try:
-            path = path_type.__fspath__(path)
-        except AttributeError:
-            # Added for Python 3.5 support.
-            if isinstance(path, pathlib.Path):
-                return str(path)
-            elif hasattr(path_type, '__fspath__'):
-                raise
-        else:
-            if isinstance(path, (str, bytes)):
-                return path
-            else:
-                raise TypeError("expected __fspath__() to return str or bytes, "
-                                "not " + type(path).__name__)
+    @wraps(func)
+    def new_func(*args, **kwargs):
+        warnings.simplefilter('always', DeprecationWarning)  # turn off filter
+        warnings.warn(
+            "Call to deprecated function {}.".format(func.__name__),
+            category=DeprecationWarning,
+            stacklevel=2,
+        )
+        warnings.simplefilter('default', DeprecationWarning)  # reset filter
+        return func(*args, **kwargs)
 
-        raise TypeError(
-            "expected str, bytes, pathlib.Path or os.PathLike object, not "
-            + path_type.__name__)
-
-else:
-    fspath = os.fspath
-
-
-def flatten_groups(groups):
-    for obj in groups:
-        if is_iterable_notstr(obj):
-            yield from obj
-        else:
-            yield obj
+    return new_func

src/ocrmypdf/hocrtransform.py

@@ -28,14 +28,14 @@
 # TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
 # SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
-from reportlab.pdfgen.canvas import Canvas
-from reportlab.lib.units import inch
-from xml.etree import ElementTree
-from collections import namedtuple
-from math import atan, sin, cos
-import re
 import argparse
+import re
+from collections import namedtuple
+from math import atan, cos, sin
+from xml.etree import ElementTree
 
+from reportlab.lib.units import inch
+from reportlab.pdfgen.canvas import Canvas
 
 Rect = namedtuple('Rect', ['x1', 'y1', 'x2', 'y2'])
 
@@ -44,7 +44,7 @@ class HocrTransformError(Exception):
     pass
 
 
-class HocrTransform():
+class HocrTransform:
 
     """
     A class for converting documents from the hOCR format.
@@ -53,17 +53,16 @@ class HocrTransform():
     """
 
     box_pattern = re.compile(r'bbox((\s+\d+){4})')
-    baseline_pattern = re.compile(r'''
+    baseline_pattern = re.compile(
+        r'''
         baseline \s+
         ([\-\+]?\d*\.?\d*) \s+  # +/- decimal float
-        ([\-\+]?\d+)            # +/- int''', re.VERBOSE)
-    ligatures = str.maketrans({
-        'ff': 'ff',
-        'ffi': 'f‌f‌i',
-        'ffl': 'f‌f‌l',
-        'fi': 'fi',
-        'fl': 'fl',
-    })
+        ([\-\+]?\d+)            # +/- int''',
+        re.VERBOSE,
+    )
+    ligatures = str.maketrans(
+        {'ff': 'ff', 'ffi': 'f‌f‌i', 'ffl': 'f‌f‌l', 'fi': 'fi', 'fl': 'fl'}
+    )
 
     def __init__(self, hocrFileName, dpi):
         self.dpi = dpi
@@ -78,8 +77,7 @@ class HocrTransform():
 
         # get dimension in pt (not pixel!!!!) of the OCRed image
         self.width, self.height = None, None
-        for div in self.hocr.findall(
-                ".//%sdiv[@class='ocr_page']" % (self.xmlns)):
+        for div in self.hocr.findall(".//%sdiv[@class='ocr_page']" % (self.xmlns)):
             coords = self.element_coordinates(div)
             pt_coords = self.pt_from_pixel(coords)
             self.width = pt_coords.x2 - pt_coords.x1
@@ -144,8 +142,7 @@ class HocrTransform():
         """
         Returns the quantity in PDF units (pt) given quantity in pixels
         """
-        return Rect._make(
-            (c / self.dpi * inch) for c in pxl)
+        return Rect._make((c / self.dpi * inch) for c in pxl)
 
     @classmethod
     def replace_unsupported_chars(cls, s):
@@ -156,8 +153,15 @@ class HocrTransform():
         """
         return s.translate(cls.ligatures)
 
-    def to_pdf(self, outFileName, imageFileName=None, showBoundingboxes=False,
-               fontname="Helvetica", invisibleText=False, interwordSpaces=False):
+    def to_pdf(
+        self,
+        outFileName,
+        imageFileName=None,
+        showBoundingboxes=False,
+        fontname="Helvetica",
+        invisibleText=False,
+        interwordSpaces=False,
+    ):
         """
         Creates a PDF file with an image superimposed on top of the text.
         Text is positioned according to the bounding box of the lines in
@@ -168,17 +172,15 @@ class HocrTransform():
         """
         # create the PDF file
         # page size in points (1/72 in.)
-        pdf = Canvas(
-            outFileName, pagesize=(self.width, self.height), pageCompression=1)
+        pdf = Canvas(outFileName, pagesize=(self.width, self.height), pageCompression=1)
 
         # draw bounding box for each paragraph
         # light blue for bounding box of paragraph
         pdf.setStrokeColorRGB(0, 1, 1)
         # light blue for bounding box of paragraph
         pdf.setFillColorRGB(0, 1, 1)
-        pdf.setLineWidth(0)		# no line for bounding box
-        for elem in self.hocr.findall(
-                ".//%sp[@class='%s']" % (self.xmlns, "ocr_par")):
+        pdf.setLineWidth(0)  # no line for bounding box
+        for elem in self.hocr.findall(".//%sp[@class='%s']" % (self.xmlns, "ocr_par")):
 
             elemtxt = self._get_element_text(elem).rstrip()
             if len(elemtxt) == 0:
@@ -190,38 +192,58 @@ class HocrTransform():
             # draw the bbox border
             if showBoundingboxes:
                 pdf.rect(
-                    pt.x1, self.height - pt.y2, pt.x2 - pt.x1, pt.y2 - pt.y1,
-                    fill=1)
+                    pt.x1, self.height - pt.y2, pt.x2 - pt.x1, pt.y2 - pt.y1, fill=1
+                )
 
         found_lines = False
         for line in self.hocr.findall(
-                ".//%sspan[@class='%s']" % (self.xmlns, "ocr_line")):
+            ".//%sspan[@class='%s']" % (self.xmlns, "ocr_line")
+        ):
             found_lines = True
-            self._do_line(pdf, line, "ocrx_word", fontname, invisibleText,
-                          interwordSpaces, showBoundingboxes)
+            self._do_line(
+                pdf,
+                line,
+                "ocrx_word",
+                fontname,
+                invisibleText,
+                interwordSpaces,
+                showBoundingboxes,
+            )
 
         if not found_lines:
             # Tesseract did not report any lines (just words)
             root = self.hocr.find(".//%sdiv[@class='%s']" % (self.xmlns, "ocr_page"))
-            self._do_line(pdf, root, "ocrx_word", fontname, invisibleText,
-                          interwordSpaces, showBoundingboxes)
+            self._do_line(
+                pdf,
+                root,
+                "ocrx_word",
+                fontname,
+                invisibleText,
+                interwordSpaces,
+                showBoundingboxes,
+            )
         # put the image on the page, scaled to fill the page
         if imageFileName is not None:
-            pdf.drawImage(imageFileName, 0, 0,
-                          width=self.width, height=self.height)
+            pdf.drawImage(imageFileName, 0, 0, width=self.width, height=self.height)
 
         # finish up the page and save it
         pdf.showPage()
         pdf.save()
 
-
     @classmethod
     def polyval(cls, poly, x):
         return x * poly[0] + poly[1]
 
-
-    def _do_line(self, pdf, line, elemclass, fontname, invisibleText,
-                 interwordSpaces, showBoundingboxes):
+    def _do_line(
+        self,
+        pdf,
+        line,
+        elemclass,
+        fontname,
+        invisibleText,
+        interwordSpaces,
+        showBoundingboxes,
+    ):
         pxl_line_coords = self.element_coordinates(line)
         line_box = self.pt_from_pixel(pxl_line_coords)
         line_height = line_box.y2 - line_box.y1
@@ -254,23 +276,20 @@ class HocrTransform():
             pdf.setLineWidth(0.5)
             # negate slope because it is defined as a rise/run in pixel
             # coordinates and page coordinates have the y axis flipped
-            pdf.line(line_box.x1,
-                     baseline_y2,
-                     line_box.x2,
-                     self.polyval((-slope, baseline_y2),
-                                  line_box.x2 - line_box.x1))
+            pdf.line(
+                line_box.x1,
+                baseline_y2,
+                line_box.x2,
+                self.polyval((-slope, baseline_y2), line_box.x2 - line_box.x1),
+            )
             # light green for bounding box of word/line
             pdf.setDash(6, 3)
             pdf.setStrokeColorRGB(1, 0, 0)
 
-        text.setTextTransform(
-            cos_a, -sin_a, sin_a, cos_a,
-            line_box.x1, baseline_y2
-        )
+        text.setTextTransform(cos_a, -sin_a, sin_a, cos_a, line_box.x1, baseline_y2)
         pdf.setFillColorRGB(0, 0, 0)  # text in black
 
-        elements = line.findall(
-                ".//%sspan[@class='%s']" % (self.xmlns, elemclass))
+        elements = line.findall(".//%sspan[@class='%s']" % (self.xmlns, elemclass))
         for elem in elements:
             elemtxt = self._get_element_text(elem).strip()
             elemtxt = self.replace_unsupported_chars(elemtxt)
@@ -287,22 +306,22 @@ class HocrTransform():
                 # though it would look better, because it will interfere with
                 # naive text extraction. \n does not work either.
                 elemtxt += ' '
-                box = Rect._make((
-                    box.x1,
-                    line_box.y1,
-                    box.x2 + pdf.stringWidth(' ', fontname, line_height),
-                    line_box.y2))
+                box = Rect._make(
+                    (
+                        box.x1,
+                        line_box.y1,
+                        box.x2 + pdf.stringWidth(' ', fontname, line_height),
+                        line_box.y2,
+                    )
+                )
             box_width = box.x2 - box.x1
             font_width = pdf.stringWidth(elemtxt, fontname, fontsize)
 
             # draw the bbox border
             if showBoundingboxes:
                 pdf.rect(
-                    box.x1,
-                    self.height - line_box.y2,
-                    box_width,
-                    line_height,
-                    fill=0)
+                    box.x1, self.height - line_box.y2, box_width, line_height, fill=0
+                )
 
             # Adjust relative position of cursor
             # This is equivalent to:
@@ -331,19 +350,40 @@ class HocrTransform():
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description='Convert hocr file to PDF')
-    parser.add_argument('-b', '--boundingboxes', action="store_true",
-                        default=False, help='Show bounding boxes borders')
-    parser.add_argument('-r', '--resolution', type=int,
-                        default=300,
-                        help='Resolution of the image that was OCRed')
-    parser.add_argument('-i', '--image', default=None,
-                        help='Path to the image to be placed above the text')
-    parser.add_argument('--interword-spaces', action='store_true',
-                         default=False, help='Add spaces between words')
-    parser.add_argument('hocrfile', help='Path to the hocr file to be parsed')
     parser.add_argument(
-        'outputfile', help='Path to the PDF file to be generated')
+        '-b',
+        '--boundingboxes',
+        action="store_true",
+        default=False,
+        help='Show bounding boxes borders',
+    )
+    parser.add_argument(
+        '-r',
+        '--resolution',
+        type=int,
+        default=300,
+        help='Resolution of the image that was OCRed',
+    )
+    parser.add_argument(
+        '-i',
+        '--image',
+        default=None,
+        help='Path to the image to be placed above the text',
+    )
+    parser.add_argument(
+        '--interword-spaces',
+        action='store_true',
+        default=False,
+        help='Add spaces between words',
+    )
+    parser.add_argument('hocrfile', help='Path to the hocr file to be parsed')
+    parser.add_argument('outputfile', help='Path to the PDF file to be generated')
     args = parser.parse_args()
 
     hocr = HocrTransform(args.hocrfile, args.resolution)
-    hocr.to_pdf(args.outputfile, args.image, args.boundingboxes, interwordSpaces=args.interword_spaces)
+    hocr.to_pdf(
+        args.outputfile,
+        args.image,
+        args.boundingboxes,
+        interwordSpaces=args.interword_spaces,
+    )

src/ocrmypdf/leptonica.py

@@ -21,36 +21,36 @@
 # Python FFI wrapper for Leptonica library
 
 import argparse
-import sys
-import os
 import logging
+import os
+import sys
 import warnings
-from tempfile import TemporaryFile
+from collections.abc import Sequence
+from contextlib import suppress
 from ctypes.util import find_library
 from functools import lru_cache
+from io import BytesIO
+from os import fspath
+from tempfile import TemporaryFile
 
 from .lib._leptonica import ffi
-from .helpers import fspath
 
-# pylint: disable=w0212
-
-lept = ffi.dlopen(find_library('lept'))
+# pylint: disable=protected-access
 
 logger = logging.getLogger(__name__)
 
-
-def stderr(*objs):
-    """Shorthand print to stderr."""
-    print("leptonica.py:", *objs, file=sys.stderr)
+lept = ffi.dlopen(find_library('lept'))
+lept.setMsgSeverity(lept.L_SEVERITY_WARNING)
 
 
 class _LeptonicaErrorTrap:
-    """Context manager to trap errors reported by Leptonica.
+    """
+    Context manager to trap errors reported by Leptonica.
 
-    Leptonica's error return codes are unreliable to the point of being
-    almost useless.  It does, however, write errors to stderr provided that is
-    not disabled at its compile time.  Fortunately this is done using error
-    macros so it is very self-consistent.
+    Leptonica's error return codes don't provide much informatino about what
+    went wrong. Leptonica does, however, write more detailed errors to stderr
+    (provided this is not disabled at compile time). The Leptonica source
+    code is very consistent in its use of macros to generate errors.
 
     This context manager redirects stderr to a temporary file which is then
     read and parsed for error messages.  As a side benefit, debug messages
@@ -61,40 +61,48 @@ class _LeptonicaErrorTrap:
     def __init__(self):
         self.tmpfile = None
         self.copy_of_stderr = -1
+        self.no_stderr = False
 
     def __enter__(self):
         from io import UnsupportedOperation
+
         self.tmpfile = TemporaryFile()
 
         # Save the old stderr, and redirect stderr to temporary file
-        sys.stderr.flush()
+        with suppress(AttributeError):
+            sys.stderr.flush()
         try:
             self.copy_of_stderr = os.dup(sys.stderr.fileno())
-            os.dup2(self.tmpfile.fileno(), sys.stderr.fileno(),
-                    inheritable=False)
+            os.dup2(self.tmpfile.fileno(), sys.stderr.fileno(), inheritable=False)
+        except AttributeError:
+            # We are in some unusual context where our Python process does not
+            # have a sys.stderr. Leptonica still expects to write to file
+            # descriptor 2, so we are going to ensure it is redirected.
+            self.copy_of_stderr = None
+            self.no_stderr = True
+            os.dup2(self.tmpfile.fileno(), 2, inheritable=False)
         except UnsupportedOperation:
             self.copy_of_stderr = None
         return
 
     def __exit__(self, exc_type, exc_value, traceback):
         # Restore old stderr
-        sys.stderr.flush()
+        with suppress(AttributeError):
+            sys.stderr.flush()
+
         if self.copy_of_stderr is not None:
             os.dup2(self.copy_of_stderr, sys.stderr.fileno())
             os.close(self.copy_of_stderr)
+        if self.no_stderr:
+            os.close(2)
 
-        # Get data from tmpfile (in with block to ensure it is closed)
-        with self.tmpfile as tmpfile:
-            tmpfile.seek(0)  # Cursor will be at end, so move back to beginning
-            leptonica_output = tmpfile.read().decode(errors='replace')
-
-        assert self.tmpfile.closed
-        assert not sys.stderr.closed
-
-        # If there are Python errors, let them bubble up
+        # Get data from tmpfile
+        self.tmpfile.seek(0)  # Cursor will be at end, so move back to beginning
+        leptonica_output = self.tmpfile.read().decode(errors='replace')
+        self.tmpfile.close()
+        # If there are Python errors, record them
         if exc_type:
             logger.warning(leptonica_output)
-            return False
 
         # If there are Leptonica errors, wrap them in Python excpetions
         if 'Error' in leptonica_output:
@@ -102,6 +110,8 @@ class _LeptonicaErrorTrap:
                 raise FileNotFoundError()
             if 'pixWrite: stream not opened' in leptonica_output:
                 raise LeptonicaIOError()
+            if 'index not valid' in leptonica_output:
+                raise IndexError()
             raise LeptonicaError(leptonica_output)
 
         return False
@@ -115,8 +125,45 @@ class LeptonicaIOError(LeptonicaError):
     pass
 
 
-class Pix:
-    """Wrapper around leptonica's PIX object.
+class LeptonicaObject:
+    """General wrapper for Leptonica objects
+
+    When Leptonica returns an object, we bundled it in a wrapper class, which
+    manages its memory. The wrapper class assumes that it will be calling some
+    sort of lept.thingDestroy() function when the instance is deleted. Most
+    Leptonica objects are reference counted, and destroy decrements the
+    refcount.
+
+    Most of the time, when Leptonica returns something, we wrap and it the job
+    is done. When wrapping objects that came from a Leptonica container, like
+    a PIXA returning PIX, the subclass must clone the object before passing it
+    here, to maintain the reference count.
+
+    CFFI ensures that the destroy function is called at garbage collection time
+    so we do not need to mess with __del__.
+    """
+
+    cdata_destroy = lambda cdata: None
+    LEPTONICA_TYPENAME = ''
+
+    def __init__(self, cdata):
+        if not cdata:
+            raise ValueError('Tried to wrap a NULL ' + self.LEPTONICA_TYPENAME)
+        self._cdata = ffi.gc(cdata, self._destroy)
+
+    @classmethod
+    def _destroy(cls, cdata):
+        """Destroy some cdata"""
+        # Leptonica API uses double-pointers for its destroy APIs to prevent
+        # dangling pointers. This means we need to put our single pointer,
+        # cdata, in a temporary CDATA**.
+        pp = ffi.new('{} **'.format(cls.LEPTONICA_TYPENAME), cdata)
+        cls.cdata_destroy(pp)
+
+
+class Pix(LeptonicaObject):
+    """
+    Wrapper around leptonica's PIX object.
 
     Leptonica uses referencing counting on PIX objects. Also, many Leptonica
     functions return the original object with an increased reference count
@@ -133,18 +180,21 @@ class Pix:
     modified objects.  This allows convenient chaining:
 
     >>>   Pix.open('filename.jpg').scale((0.5, 0.5)).deskew().show()
-
     """
 
-    def __init__(self, pix):
-        self._pix = ffi.gc(pix, Pix._pix_destroy)
+    LEPTONICA_TYPENAME = "PIX"
+    cdata_destroy = lept.pixDestroy
 
     def __repr__(self):
-        if self._pix:
+        if self._cdata:
             s = "<leptonica.Pix image size={0}x{1} depth={2}{4} at 0x{3:x}>"
-            return s.format(self._pix.w, self._pix.h, self._pix.d,
-                            int(ffi.cast('intptr_t', self._pix)),
-                            '(colormapped)' if self._pix.colormap else '')
+            return s.format(
+                self._cdata.w,
+                self._cdata.h,
+                self._cdata.d,
+                int(ffi.cast('intptr_t', self._cdata)),
+                '(colormapped)' if self._cdata.colormap else '',
+            )
         else:
             return "<leptonica.Pix image NULL>"
 
@@ -157,7 +207,7 @@ class Pix:
         data = ffi.new('l_uint8 **')
         size = ffi.new('size_t *')
 
-        err = lept.pixWriteMemPng(data, size, self._pix, 0)
+        err = lept.pixWriteMemPng(data, size, self._cdata, 0)
         if err != 0:
             raise LeptonicaIOError("pixWriteMemPng")
 
@@ -168,7 +218,7 @@ class Pix:
         data = ffi.new('l_uint32 **')
         size = ffi.new('size_t *')
 
-        err = lept.pixSerializeToMemory(self._pix, data, size)
+        err = lept.pixSerializeToMemory(self._cdata, data, size)
         if err != 0:
             raise LeptonicaIOError("pixSerializeToMemory")
 
@@ -185,32 +235,38 @@ class Pix:
         cdata_bytes = ffi.new('char[]', state['data'])
         cdata_uint32 = ffi.cast('l_uint32 *', cdata_bytes)
 
-        pix = lept.pixDeserializeFromMemory(
-            cdata_uint32, len(state['data']))
+        pix = lept.pixDeserializeFromMemory(cdata_uint32, len(state['data']))
         Pix.__init__(self, pix)
 
     def __eq__(self, other):
-        return self.__getstate__() == other.__getstate__()
+        if not isinstance(other, Pix):
+            return NotImplemented
+        same = ffi.new('l_int32 *', 0)
+        with _LeptonicaErrorTrap():
+            err = lept.pixEqual(self._cdata, other._cdata, same)
+            if err:
+                raise TypeError()
+        return bool(same[0])
 
     @property
     def width(self):
-        return self._pix.w
+        return self._cdata.w
 
     @property
     def height(self):
-        return self._pix.h
+        return self._cdata.h
 
     @property
     def depth(self):
-        return self._pix.d
+        return self._cdata.d
 
     @property
     def size(self):
-        return (self._pix.w, self._pix.h)
+        return (self._cdata.w, self._cdata.h)
 
     @property
     def info(self):
-        return {'dpi': (self._pix.xres, self._pix.yres)}
+        return {'dpi': (self._cdata.xres, self._cdata.yres)}
 
     @property
     def mode(self):
@@ -219,7 +275,7 @@ class Pix:
             return '1'
         elif self.depth >= 16:
             return 'RGB'
-        elif not self._pix.colormap:
+        elif not self._cdata.colormap:
             return 'L'
         else:
             return 'P'
@@ -240,8 +296,7 @@ class Pix:
         with _LeptonicaErrorTrap():
             return cls(lept.pixRead(os.fsencode(filename)))
 
-    def write_implied_format(
-            self, path, jpeg_quality=0, jpeg_progressive=0):
+    def write_implied_format(self, path, jpeg_quality=0, jpeg_progressive=0):
         """Write pix to the filename, with the extension indicating format.
 
         jpeg_quality -- quality (iff JPEG; 1 - 100, 0 for default)
@@ -250,11 +305,22 @@ class Pix:
         filename = fspath(path)
         with _LeptonicaErrorTrap():
             lept.pixWriteImpliedFormat(
-                os.fsencode(filename),
-                self._pix, jpeg_quality, jpeg_progressive)
+                os.fsencode(filename), self._cdata, jpeg_quality, jpeg_progressive
+            )
+
+    @classmethod
+    def frompil(self, pillow_image):
+        """Create a copy of a PIL.Image from this Pix"""
+        bio = BytesIO()
+        pillow_image.save(bio, format='png', compress_level=1)
+        py_buffer = bio.getbuffer()
+        c_buffer = ffi.from_buffer(py_buffer)
+        with _LeptonicaErrorTrap():
+            pix = Pix(lept.pixReadMem(c_buffer, len(c_buffer)))
+            return pix
 
     def topil(self):
-        "Returns a PIL.Image version of this Pix"
+        """Returns a PIL.Image version of this Pix"""
         from PIL import Image
 
         # Leptonica manages data in words, so it implicitly does an endian
@@ -267,17 +333,17 @@ class Pix:
                 raw_mode = 'ABGR'
             elif self.mode == '1':
                 raw_mode = '1;I'
-                pix = Pix(lept.pixEndianByteSwapNew(pix._pix))
+                pix = Pix(lept.pixEndianByteSwapNew(pix._cdata))
             else:
                 raw_mode = self.mode
-                pix = Pix(lept.pixEndianByteSwapNew(pix._pix))
+                pix = Pix(lept.pixEndianByteSwapNew(pix._cdata))
         else:
             raw_mode = self.mode  # no endian swap needed
 
-        size = (pix._pix.w, pix._pix.h)
-        bytecount = pix._pix.wpl * 4 * pix._pix.h
-        buf = ffi.buffer(pix._pix.data, bytecount)
-        stride = pix._pix.wpl * 4
+        size = (pix._cdata.w, pix._cdata.h)
+        bytecount = pix._cdata.wpl * 4 * pix._cdata.h
+        buf = ffi.buffer(pix._cdata.data, bytecount)
+        stride = pix._cdata.wpl * 4
 
         im = Image.frombytes(self.mode, size, buf, 'raw', raw_mode, stride)
 
@@ -296,21 +362,21 @@ class Pix:
             for skew angle
         """
         with _LeptonicaErrorTrap():
-            return Pix(lept.pixDeskew(self._pix, reduction_factor))
+            return Pix(lept.pixDeskew(self._cdata, reduction_factor))
 
     def scale(self, scale_xy):
         "Returns the pix object rescaled according to the proportions given."
         with _LeptonicaErrorTrap():
-            return Pix(lept.pixScale(self._pix, scale_xy[0], scale_xy[1]))
+            return Pix(lept.pixScale(self._cdata, scale_xy[0], scale_xy[1]))
 
     def rotate180(self):
         with _LeptonicaErrorTrap():
-            return Pix(lept.pixRotate180(ffi.NULL, self._pix))
+            return Pix(lept.pixRotate180(ffi.NULL, self._cdata))
 
     def rotate_orth(self, quads):
         "Orthographic rotation, quads: 0-3, number of clockwise rotations"
         with _LeptonicaErrorTrap():
-            return Pix(lept.pixRotateOrth(self._pix, quads))
+            return Pix(lept.pixRotateOrth(self._cdata, quads))
 
     def find_skew(self):
         """Returns a tuple (deskew angle in degrees, confidence value).
@@ -320,7 +386,7 @@ class Pix:
         with _LeptonicaErrorTrap():
             angle = ffi.new('float *', 0.0)
             confidence = ffi.new('float *', 0.0)
-            result = lept.pixFindSkew(self._pix, angle, confidence)
+            result = lept.pixFindSkew(self._cdata, angle, confidence)
             if result == 0:
                 return (angle[0], confidence[0])
             else:
@@ -328,7 +394,7 @@ class Pix:
 
     def convert_rgb_to_luminance(self):
         with _LeptonicaErrorTrap():
-            gray_pix = lept.pixConvertRGBToLuminance(self._pix)
+            gray_pix = lept.pixConvertRGBToLuminance(self._cdata)
             if gray_pix:
                 return Pix(gray_pix)
             return None
@@ -341,114 +407,165 @@ class Pix:
 
         """
         with _LeptonicaErrorTrap():
-            return Pix(lept.pixRemoveColormapGeneral(
-                    self._pix, removal_type, lept.L_COPY))
+            return Pix(
+                lept.pixRemoveColormapGeneral(self._cdata, removal_type, lept.L_COPY)
+            )
 
     def otsu_adaptive_threshold(
-            self, tile_size=(300, 300), kernel_size=(4, 4), scorefract=0.1):
+        self, tile_size=(300, 300), kernel_size=(4, 4), scorefract=0.1
+    ):
         with _LeptonicaErrorTrap():
             sx, sy = tile_size
             smoothx, smoothy = kernel_size
             p_pix = ffi.new('PIX **')
 
+            pix = Pix(lept.pixConvertTo8(self._cdata, 0))
             result = lept.pixOtsuAdaptiveThreshold(
-                self._pix,
-                sx, sy,
-                smoothx, smoothy,
-                scorefract,
-                ffi.NULL,
-                p_pix)
+                pix._cdata, sx, sy, smoothx, smoothy, scorefract, ffi.NULL, p_pix
+            )
             if result == 0:
                 return Pix(p_pix[0])
             else:
                 return None
 
     def otsu_threshold_on_background_norm(
-            self, mask=None, tile_size=(10, 15), thresh=100, mincount=50,
-            bgval=255, kernel_size=(2, 2), scorefract=0.1):
+        self,
+        mask=None,
+        tile_size=(10, 15),
+        thresh=100,
+        mincount=50,
+        bgval=255,
+        kernel_size=(2, 2),
+        scorefract=0.1,
+    ):
         with _LeptonicaErrorTrap():
             sx, sy = tile_size
             smoothx, smoothy = kernel_size
-            if mask is None:
-                mask = ffi.NULL
+            mask = ffi.NULL
             if isinstance(mask, Pix):
-                mask = mask._pix
+                mask = mask._cdata
 
+            pix = Pix(lept.pixConvertTo8(self._cdata, 0))
             thresh_pix = lept.pixOtsuThreshOnBackgroundNorm(
-                self._pix,
+                pix._cdata,
                 mask,
-                sx, sy,
-                thresh, mincount, bgval,
-                smoothx, smoothy,
+                sx,
+                sy,
+                thresh,
+                mincount,
+                bgval,
+                smoothx,
+                smoothy,
                 scorefract,
-                ffi.NULL
-                )
-            if thresh_pix == ffi.NULL:
-                return None
+                ffi.NULL,
+            )
+            return Pix(thresh_pix)
+
+    def masked_threshold_on_background_norm(
+        self,
+        mask=None,
+        tile_size=(10, 15),
+        thresh=100,
+        mincount=50,
+        kernel_size=(2, 2),
+        scorefract=0.1,
+    ):
+        with _LeptonicaErrorTrap():
+            sx, sy = tile_size
+            smoothx, smoothy = kernel_size
+            mask = ffi.NULL
+            if isinstance(mask, Pix):
+                mask = mask._cdata
+
+            pix = Pix(lept.pixConvertTo8(self._cdata, 0))
+            thresh_pix = lept.pixMaskedThreshOnBackgroundNorm(
+                pix._cdata,
+                mask,
+                sx,
+                sy,
+                thresh,
+                mincount,
+                smoothx,
+                smoothy,
+                scorefract,
+                ffi.NULL,
+            )
             return Pix(thresh_pix)
 
     def crop_to_foreground(
-            self, threshold=128, mindist=70, erasedist=30, pagenum=0,
-            showmorph=0, display=0, pdfdir=ffi.NULL):
+        self,
+        threshold=128,
+        mindist=70,
+        erasedist=30,
+        pagenum=0,
+        showmorph=0,
+        display=0,
+        pdfdir=ffi.NULL,
+    ):
         with _LeptonicaErrorTrap():
-            cropbox = Box(lept.pixFindPageForeground(
-                self._pix,
-                threshold,
-                mindist,
-                erasedist,
-                pagenum,
-                showmorph,
-                display,
-                pdfdir))
+            cropbox = Box(
+                lept.pixFindPageForeground(
+                    self._cdata,
+                    threshold,
+                    mindist,
+                    erasedist,
+                    pagenum,
+                    showmorph,
+                    display,
+                    pdfdir,
+                )
+            )
 
-            print(repr(cropbox))
-
-            cropped_pix = lept.pixClipRectangle(
-                self._pix,
-                cropbox._box,
-                ffi.NULL)
+            cropped_pix = lept.pixClipRectangle(self._cdata, cropbox._cdata, ffi.NULL)
 
             return Pix(cropped_pix)
 
     def clean_background_to_white(
-            self, mask=None, grayscale=None, gamma=1.0, black=0, white=255):
+        self, mask=None, grayscale=None, gamma=1.0, black=0, white=255
+    ):
         with _LeptonicaErrorTrap():
-            return Pix(lept.pixCleanBackgroundToWhite(
-                self._pix,
-                mask or ffi.NULL,
-                grayscale or ffi.NULL,
-                gamma,
-                black,
-                white))
+            return Pix(
+                lept.pixCleanBackgroundToWhite(
+                    self._cdata,
+                    mask or ffi.NULL,
+                    grayscale or ffi.NULL,
+                    gamma,
+                    black,
+                    white,
+                )
+            )
 
     def gamma_trc(self, gamma=1.0, minval=0, maxval=255):
         with _LeptonicaErrorTrap():
-            return Pix(lept.pixGammaTRC(
-                ffi.NULL,
-                self._pix,
-                gamma,
-                minval,
-                maxval
-                ))
+            return Pix(lept.pixGammaTRC(ffi.NULL, self._cdata, gamma, minval, maxval))
 
     def background_norm(
-            self, mask=None, grayscale=None, tile_size=(10, 15), fg_threshold=60,
-            min_count=40, bg_val=200, smooth_kernel=(2, 1)):
+        self,
+        mask=None,
+        grayscale=None,
+        tile_size=(10, 15),
+        fg_threshold=60,
+        min_count=40,
+        bg_val=200,
+        smooth_kernel=(2, 1),
+    ):
         # Background norm doesn't work on color mapped Pix, so remove colormap
         target_pix = self.remove_colormap(lept.REMOVE_CMAP_BASED_ON_SRC)
         with _LeptonicaErrorTrap():
-            return Pix(lept.pixBackgroundNorm(
-                target_pix._pix,
-                mask or ffi.NULL,
-                grayscale or ffi.NULL,
-                tile_size[0],
-                tile_size[1],
-                fg_threshold,
-                min_count,
-                bg_val,
-                smooth_kernel[0],
-                smooth_kernel[1]
-            ))
+            return Pix(
+                lept.pixBackgroundNorm(
+                    target_pix._cdata,
+                    mask or ffi.NULL,
+                    grayscale or ffi.NULL,
+                    tile_size[0],
+                    tile_size[1],
+                    fg_threshold,
+                    min_count,
+                    bg_val,
+                    smooth_kernel[0],
+                    smooth_kernel[1],
+                )
+            )
 
     @staticmethod
     @lru_cache(maxsize=1)
@@ -462,53 +579,93 @@ class Pix:
             # implementation of pixCorrelationBinary that overflows on larger
             # images.  Ubuntu 14.04/trusty has 1.70. Ubuntu PPA
             # ppa:alex-p/tesseract-ocr has leptonlib 1.75.
-            pix1_count = ffi.new('l_int32 *')
-            pix2_count = ffi.new('l_int32 *')
-            pixn_count = ffi.new('l_int32 *')
-            tab8 = Pix.make_pixel_sum_tab8()
+            raise LeptonicaError("Leptonica version is too old")
 
-            lept.pixCountPixels(pix1._pix, pix1_count, tab8)
-            lept.pixCountPixels(pix2._pix, pix2_count, tab8)
-            pixn = Pix(lept.pixAnd(ffi.NULL, pix1._pix, pix2._pix))
-            lept.pixCountPixels(pixn._pix, pixn_count, tab8)
-
-            # Python converts these int32s to larger units as needed
-            # to avoid overflow. Overflow happens easily here.
-            correlation = (
-                    (pixn_count[0] * pixn_count[0]) /
-                    (pix1_count[0] * pix2_count[0])
-                    )
-            return correlation
-        else:
-            correlation = ffi.new('float *', 0.0)
-            result = lept.pixCorrelationBinary(pix1._pix, pix2._pix,
-                                               correlation)
-            if result != 0:
-                raise LeptonicaError("Correlation failed")
-            return correlation[0]
+        correlation = ffi.new('float *', 0.0)
+        result = lept.pixCorrelationBinary(pix1._cdata, pix2._cdata, correlation)
+        if result != 0:
+            raise LeptonicaError("Correlation failed")
+        return correlation[0]
 
     def generate_pdf_ci_data(self, type_, quality):
         "Convert to PDF data, with transcoding"
         p_compdata = ffi.new('L_COMP_DATA **')
-        result = lept.pixGenerateCIData(self._pix, type_, quality, 0,
-                                        p_compdata)
+        result = lept.pixGenerateCIData(self._cdata, type_, quality, 0, p_compdata)
         if result != 0:
             raise LeptonicaError("Generate PDF data failed")
         return CompressedData(p_compdata[0])
 
     def invert(self):
-        return Pix(lept.pixInvert(ffi.NULL, self._pix))
+        return Pix(lept.pixInvert(ffi.NULL, self._cdata))
 
-    @staticmethod
-    def _pix_destroy(pix):
-        p_pix = ffi.new('PIX **', pix)
-        lept.pixDestroy(p_pix)
-        # print('pix destroy ' + repr(pix))
+    def locate_barcodes(self):
+        try:
+            with _LeptonicaErrorTrap():
+                pix = Pix(lept.pixConvertTo8(self._cdata, 0))
+                pixa_candidates = PixArray(lept.pixExtractBarcodes(pix._cdata, 0))
+                if not pixa_candidates:
+                    return
+                sarray = StringArray(
+                    lept.pixReadBarcodes(
+                        pixa_candidates._cdata,
+                        lept.L_BF_ANY,
+                        lept.L_USE_WIDTHS,
+                        ffi.NULL,
+                        0,
+                    )
+                )
+        except (LeptonicaError, ValueError, IndexError):
+            return
+        finally:
+            leptonica_junk = ('junkpixt.png', 'junkpixt')
+            for junk in leptonica_junk:
+                with suppress(FileNotFoundError):
+                    os.unlink(junk)  # leptonica may produce this
+
+        for n, s in enumerate(sarray):
+            decoded = s.decode()
+            if decoded.strip() == '':
+                continue
+            box = pixa_candidates.get_box(n)
+            left, top = box.x, box.y
+            right, bottom = box.x + box.w, box.y + box.h
+            yield (decoded, (left, top, right, bottom))
+
+    def despeckle(self, size):
+        if size == 2:
+            speckle2 = """
+                oooo
+                oC o
+                o  o
+                oooo
+                """
+            sel1 = Sel.from_selstr(speckle2, 'speckle2')
+            sel2 = Sel.create_brick(2, 2, 0, 0, lept.SEL_HIT)
+        elif size == 3:
+            speckle3 = """
+                ooooo
+                oC  o
+                o   o
+                o   o
+                ooooo
+                """
+            sel1 = Sel.from_selstr(speckle3, 'speckle3')
+            sel2 = Sel.create_brick(3, 3, 0, 0, lept.SEL_HIT)
+        else:
+            raise ValueError(size)
+
+        pixhmt = Pix(lept.pixHMT(ffi.NULL, self._cdata, sel1._cdata))
+        pixdilated = Pix(lept.pixDilate(ffi.NULL, pixhmt._cdata, sel2._cdata))
+
+        pixsub = Pix(lept.pixSubtract(ffi.NULL, self._cdata, pixdilated._cdata))
+        return pixsub
 
 
-class CompressedData:
-    def __init__(self, compdata):
-        self._compdata = ffi.gc(compdata, CompressedData._destroy)
+class CompressedData(LeptonicaObject):
+    """Wrapper for L_COMP_DATA - abstract compressed image data"""
+
+    LEPTONICA_TYPENAME = 'L_COMP_DATA'
+    cdata_destroy = lept.l_CIDataDestroy
 
     @classmethod
     def open(cls, path, jpeg_quality=75):
@@ -517,70 +674,150 @@ class CompressedData:
 
         p_compdata = ffi.new('L_COMP_DATA **')
         result = lept.l_generateCIDataForPdf(
-            os.fsencode(filename), ffi.NULL, jpeg_quality, p_compdata)
+            os.fsencode(filename), ffi.NULL, jpeg_quality, p_compdata
+        )
         if result != 0:
             raise LeptonicaError("CompressedData.open")
         return CompressedData(p_compdata[0])
 
     def __len__(self):
-        return self._compdata.nbytescomp
+        return self._cdata.nbytescomp
 
     def read(self):
-        buf = ffi.buffer(self._compdata.datacomp, self._compdata.nbytescomp)
+        buf = ffi.buffer(self._cdata.datacomp, self._cdata.nbytescomp)
         return bytes(buf)
 
     def __getattr__(self, name):
-        if hasattr(self._compdata, name):
-            return getattr(self._compdata, name)
+        if hasattr(self._cdata, name):
+            return getattr(self._cdata, name)
         raise AttributeError(name)
 
     def get_palette_pdf_string(self):
         "Returns palette pre-formatted for use in PDF"
-        buflen = len('< ') + len(' rrggbb') * self._compdata.ncolors + len('>')
-        buf = ffi.buffer(self._compdata.cmapdatahex, buflen)
+        buflen = len('< ') + len(' rrggbb') * self._cdata.ncolors + len('>')
+        buf = ffi.buffer(self._cdata.cmapdatahex, buflen)
         return bytes(buf)
 
-    @staticmethod
-    def _destroy(compdata):
-        pp = ffi.new('L_COMP_DATA **', compdata)
-        lept.l_CIDataDestroy(pp)
+
+class PixArray(LeptonicaObject, Sequence):
+    """Wrapper around PIXA (array of PIX)"""
+
+    LEPTONICA_TYPENAME = 'PIXA'
+    cdata_destroy = lept.pixaDestroy
+
+    def __len__(self):
+        return self._cdata[0].n
+
+    def __getitem__(self, n):
+        with _LeptonicaErrorTrap():
+            return Pix(lept.pixaGetPix(self._cdata, n, lept.L_CLONE))
+
+    def get_box(self, n):
+        with _LeptonicaErrorTrap():
+            return Box(lept.pixaGetBox(self._cdata, n, lept.L_CLONE))
 
 
-class Box:
-    """Wrapper around Leptonica's BOX objects.
+class Box(LeptonicaObject):
+    """Wrapper around Leptonica's BOX objects (a pixel rectangle)
 
-    See class Pix for notes about reference counting.
+    Uses x, y, w, h coordinates.
     """
 
-    def __init__(self, box):
-        self._box = ffi.gc(box, Box._box_destroy)
+    LEPTONICA_TYPENAME = 'BOX'
+    cdata_destroy = lept.boxDestroy
 
     def __repr__(self):
-        if self._box:
+        if self._cdata:
             return '<leptonica.Box x={0} y={1} w={2} h={3}>'.format(
-                self.x, self.y, self.w, self.h)
+                self.x, self.y, self.w, self.h
+            )
         return '<leptonica.Box NULL>'
 
     @property
     def x(self):
-        return self._box.x
+        return self._cdata.x
 
     @property
     def y(self):
-        return self._box.y
+        return self._cdata.y
 
     @property
     def w(self):
-        return self._box.w
+        return self._cdata.w
 
     @property
     def h(self):
-        return self._box.h
+        return self._cdata.h
 
-    @staticmethod
-    def _box_destroy(box):
-        p_box = ffi.new('BOX **', box)
-        lept.boxDestroy(p_box)
+
+class BoxArray(LeptonicaObject, Sequence):
+    """Wrapper around Leptonica's BOXA (Array of BOX) objects."""
+
+    LEPTONICA_TYPENAME = 'BOXA'
+    cdata_destroy = lept.boxaDestroy
+
+    def __repr__(self):
+        if not self._cdata:
+            return '<BoxArray>'
+        boxes = (repr(box) for box in self)
+        return '<BoxArray [' + ', '.join(boxes) + ']>'
+
+    def __len__(self):
+        return self._cdata.n
+
+    def __getitem__(self, n):
+        if not isinstance(n, int):
+            raise TypeError('list indices must be integers')
+        if 0 <= n < len(self):
+            return Box(lept.boxaGetBox(self._cdata, n, lept.L_CLONE))
+        raise IndexError(n)
+
+
+class StringArray(LeptonicaObject, Sequence):
+    """Leptonica SARRAY/string array"""
+
+    LEPTONICA_TYPENAME = 'SARRAY'
+    cdata_destroy = lept.sarrayDestroy
+
+    def __len__(self):
+        return self._cdata.n
+
+    def __getitem__(self, n):
+        if 0 <= n < len(self):
+            return ffi.string(self._cdata.array[n])
+        raise IndexError(n)
+
+
+class Sel(LeptonicaObject):
+    """Leptonica 'sel'/selection element for hit-miss transform"""
+
+    LEPTONICA_TYPENAME = 'SEL'
+    cdata_destroy = lept.selDestroy
+
+    @classmethod
+    def from_selstr(cls, selstr, name):
+        # TODO this will strip a horizontal line of don't care's
+        lines = [line.strip() for line in selstr.split('\n') if line.strip()]
+        h = len(lines)
+        w = len(lines[0])
+        lengths = set(len(line) for line in lines)
+        if len(lengths) != 1:
+            raise ValueError("All lines in selstr must be same length")
+
+        repacked = ''.join(line.strip() for line in lines)
+        buf_selstr = ffi.from_buffer(repacked.encode('ascii'))
+        buf_name = ffi.from_buffer(name.encode('ascii'))
+        sel = lept.selCreateFromString(buf_selstr, h, w, buf_name)
+        return cls(sel)
+
+    @classmethod
+    def create_brick(cls, h, w, cy, cx, type_):
+        sel = lept.selCreateBrick(h, w, cy, cx, type_)
+        return cls(sel)
+
+    def __repr__(self):
+        selstr = ffi.gc(lept.selPrintToString(self._cdata), lept.lept_free)
+        return '<Sel \n' + ffi.string(selstr).decode('ascii') + '\n>'
 
 
 @lru_cache(maxsize=1)
@@ -612,15 +849,22 @@ def deskew(infile, outfile, dpi):
         raise LeptonicaIOError("Failed to open destination file: %s" % outfile)
 
 
-def remove_background(infile, outfile, tile_size=(40, 60), gamma=1.0,
-                      black_threshold=70, white_threshold=190):
+def remove_background(
+    infile,
+    outfile,
+    tile_size=(40, 60),
+    gamma=1.0,
+    black_threshold=70,
+    white_threshold=190,
+):
     try:
         pix = Pix.open(infile)
     except LeptonicaIOError:
         raise LeptonicaIOError("Failed to open file: %s" % infile)
 
     pix = pix.background_norm(tile_size=tile_size).gamma_trc(
-            gamma, black_threshold, white_threshold)
+        gamma, black_threshold, white_threshold
+    )
 
     try:
         pix.write_implied_format(outfile)
@@ -629,15 +873,22 @@ def remove_background(infile, outfile, tile_size=(40, 60), gamma=1.0,
 
 
 if __name__ == '__main__':
-    parser = argparse.ArgumentParser(
-        description="Python wrapper to access Leptonica")
+    parser = argparse.ArgumentParser(description="Python wrapper to access Leptonica")
 
-    subparsers = parser.add_subparsers(title='commands',
-                                       description='supported operations')
+    subparsers = parser.add_subparsers(
+        title='commands', description='supported operations'
+    )
 
     parser_deskew = subparsers.add_parser('deskew')
-    parser_deskew.add_argument('-r', '--dpi', dest='dpi', action='store',
-                               type=int, default=300, help='input resolution')
+    parser_deskew.add_argument(
+        '-r',
+        '--dpi',
+        dest='dpi',
+        action='store',
+        type=int,
+        default=300,
+        help='input resolution',
+    )
     parser_deskew.add_argument('infile', help='image to deskew')
     parser_deskew.add_argument('outfile', help='deskewed output image')
     parser_deskew.set_defaults(func=deskew)

src/ocrmypdf/lib/compile_leptonica.py

@@ -19,8 +19,8 @@
 from cffi import FFI
 
 ffibuilder = FFI()
-ffibuilder.set_source("ocrmypdf.lib._leptonica", None)
-ffibuilder.cdef("""
+ffibuilder.cdef(
+    """
 typedef signed char             l_int8;
 typedef unsigned char           l_uint8;
 typedef short                   l_int16;
@@ -32,6 +32,8 @@ typedef double                  l_float64;
 typedef long long               l_int64;
 typedef unsigned long long      l_uint64;
 
+typedef int l_ok; /*!< return type 0 if OK, 1 on error */
+
 struct Pix
 {
     l_uint32             w;           /* width in pixels                   */
@@ -61,6 +63,17 @@ struct PixColormap
 };
 typedef struct PixColormap  PIXCMAP;
 
+/*! Array of pix */
+struct Pixa
+{
+    l_int32             n;          /*!< number of Pix in ptr array        */
+    l_int32             nalloc;     /*!< number of Pix ptrs allocated      */
+    l_uint32            refcount;   /*!< reference count (1 if no clones)  */
+    struct Pix        **pix;        /*!< the array of ptrs to pix          */
+    struct Boxa        *boxa;       /*!< array of boxes                    */
+};
+typedef struct Pixa PIXA;
+
 struct Box
 {
     l_int32            x;
@@ -72,6 +85,25 @@ struct Box
 };
 typedef struct Box    BOX;
 
+/*! Array of Box */
+struct Boxa
+{
+    l_int32            n;           /*!< number of box in ptr array        */
+    l_int32            nalloc;      /*!< number of box ptrs allocated      */
+    l_uint32           refcount;    /*!< reference count (1 if no clones)  */
+    struct Box       **box;         /*!< box ptr array                     */
+};
+typedef struct Boxa BOXA;
+
+/*! String array: an array of C strings */
+struct Sarray
+{
+    l_int32          nalloc;    /*!< size of allocated ptr array         */
+    l_int32          n;         /*!< number of strings allocated         */
+    l_int32          refcount;  /*!< reference count (1 if no clones)    */
+    char           **array;     /*!< string array                        */
+};
+typedef struct Sarray SARRAY;
 
 /*! Pdf formatted encoding types */
 enum {
@@ -104,6 +136,18 @@ struct L_Compressed_Data
 };
 typedef struct L_Compressed_Data L_COMP_DATA;
 
+/*! Selection */
+struct Sel
+{
+    l_int32       sy;        /*!< sel height                               */
+    l_int32       sx;        /*!< sel width                                */
+    l_int32       cy;        /*!< y location of sel origin                 */
+    l_int32       cx;        /*!< x location of sel origin                 */
+    l_int32     **data;      /*!< {0,1,2}; data[i][j] in [row][col] order  */
+    char         *name;      /*!< used to find sel by name                 */
+};
+typedef struct Sel SEL;
+
 enum {
     REMOVE_CMAP_TO_BINARY = 0,     /*!< remove colormap for conv to 1 bpp  */
     REMOVE_CMAP_TO_GRAYSCALE = 1,  /*!< remove colormap for conv to 8 bpp  */
@@ -122,10 +166,50 @@ enum {
                       /*!< the array with clones (e.g., pix)              */
 };
 
-""")
+/*! Flags for method of extracting barcode widths */
+enum {
+    L_USE_WIDTHS = 1,     /*!< use histogram of barcode widths           */
+    L_USE_WINDOWS = 2     /*!< find best window for decoding transitions */
+};
 
-ffibuilder.cdef("""
+/*! Flags for barcode formats */
+enum {
+    L_BF_UNKNOWN = 0,     /*!< unknown format                            */
+    L_BF_ANY = 1,         /*!< try decoding with all known formats       */
+    L_BF_CODE128 = 2,     /*!< decode with Code128 format                */
+    L_BF_EAN8 = 3,        /*!< decode with EAN8 format                   */
+    L_BF_EAN13 = 4,       /*!< decode with EAN13 format                  */
+    L_BF_CODE2OF5 = 5,    /*!< decode with Code 2 of 5 format            */
+    L_BF_CODEI2OF5 = 6,   /*!< decode with Interleaved 2 of 5 format     */
+    L_BF_CODE39 = 7,      /*!< decode with Code39 format                 */
+    L_BF_CODE93 = 8,      /*!< decode with Code93 format                 */
+    L_BF_CODABAR = 9,     /*!< decode with Code93 format                 */
+    L_BF_UPCA = 10        /*!< decode with UPC A format                  */
+};
+
+enum {
+    L_SEVERITY_EXTERNAL = 0,   /* Get the severity from the environment   */
+    L_SEVERITY_ALL      = 1,   /* Lowest severity: print all messages     */
+    L_SEVERITY_DEBUG    = 2,   /* Print debugging and higher messages     */
+    L_SEVERITY_INFO     = 3,   /* Print informational and higher messages */
+    L_SEVERITY_WARNING  = 4,   /* Print warning and higher messages       */
+    L_SEVERITY_ERROR    = 5,   /* Print error and higher messages         */
+    L_SEVERITY_NONE     = 6    /* Highest severity: print no messages     */
+};
+
+enum {
+    SEL_DONT_CARE  = 0,
+    SEL_HIT        = 1,
+    SEL_MISS       = 2
+};
+
+"""
+)
+
+ffibuilder.cdef(
+    """
 PIX * pixRead ( const char *filename );
+PIX * pixReadMem ( const l_uint8 *data, size_t size );
 PIX * pixScale ( PIX *pixs, l_float32 scalex, l_float32 scaley );
 l_int32 pixFindSkew ( PIX *pixs, l_float32 *pangle, l_float32 *pconf );
 l_int32 pixWriteImpliedFormat ( const char *filename, PIX *pix, l_int32 quality, l_int32 progressive );
@@ -137,6 +221,11 @@ pixWriteMemPng(l_uint8  **pdata,
 
 void pixDestroy ( PIX **ppix );
 
+l_ok
+pixEqual(PIX      *pix1,
+         PIX      *pix2,
+         l_int32 *psame);
+
 PIX *
 pixEndianByteSwapNew(PIX  *pixs);
 
@@ -157,6 +246,8 @@ l_int32 pixSerializeToMemory ( PIX *pixs, l_uint32 **pdata, size_t *pnbytes );
 
 PIX * pixConvertRGBToLuminance(PIX *pixs);
 
+PIX * pixConvertTo8(PIX     *pixs, l_int32  cmapflag);
+
 PIX * pixRemoveColormap(PIX *pixs, l_int32  type);
 
 l_int32
@@ -182,6 +273,18 @@ pixOtsuThreshOnBackgroundNorm(PIX       *pixs,
                               l_float32  scorefract,
                               l_int32   *pthresh);
 
+PIX *
+pixMaskedThreshOnBackgroundNorm(PIX       *pixs,
+                                PIX       *pixim,
+                                l_int32    sx,
+                                l_int32    sy,
+                                l_int32    thresh,
+                                l_int32    mincount,
+                                l_int32    smoothx,
+                                l_int32    smoothy,
+                                l_float32  scorefract,
+                                l_int32   *pthresh);
+
 PIX *
 pixCleanBackgroundToWhite(PIX       *pixs,
                           PIX       *pixim,
@@ -265,13 +368,13 @@ pixGetAverageMaskedRGB(PIX        *pixs,
                        l_float32  *pgval,
                        l_float32  *pbval);
 
-PIX * 
+PIX *
 pixGlobalNormRGB(PIX * 	pixd,
                  PIX * 	pixs,
                  l_int32 	rval,
                  l_int32 	gval,
                  l_int32 	bval,
-                 l_int32 	mapval); 
+                 l_int32 	mapval);
 
 PIX *
 pixInvert(PIX * pixd,
@@ -289,22 +392,102 @@ pixGenerateCIData(PIX           *pixs,
                   l_int32        ascii85,
                   L_COMP_DATA **pcid);
 
-l_int32 
-l_generateCIDataForPdf(const char *fname, 
-                       PIX *pix, 
-                       l_int32 quality, 
+SARRAY *
+pixProcessBarcodes(PIX      *pixs,
+                   l_int32   format,
+                   l_int32   method,
+                   SARRAY  **psaw,
+                   l_int32 debugflag);
+
+PIX *
+pixaGetPix(PIXA    *pixa,
+           l_int32  index,
+           l_int32 accesstype);
+
+BOX*
+pixaGetBox 	(PIXA *  	pixa,
+		    l_int32  	index,
+		    l_int32  	accesstype );
+
+PIXA *
+pixExtractBarcodes(PIX     *pixs,
+                   l_int32 debugflag);
+
+BOXA *
+pixLocateBarcodes ( PIX *pixs, l_int32 thresh, PIX **ppixb, PIX **ppixm );
+
+SARRAY *
+pixReadBarcodes(PIXA     *pixa,
+                l_int32   format,
+                l_int32   method,
+                SARRAY  **psaw,
+                l_int32 debugflag);
+
+l_int32
+l_generateCIDataForPdf(const char *fname,
+                       PIX *pix,
+                       l_int32 quality,
                        L_COMP_DATA **pcid);
 
-void                 
+
+BOX *
+boxClone ( BOX *box );
+
+BOX *
+boxaGetBox ( BOXA *boxa, l_int32 index, l_int32 accessflag );
+
+SEL *
+selCreateFromString ( const char *text, l_int32 h, l_int32 w, const char *name );
+
+SEL *
+selCreateBrick ( l_int32 h, l_int32 w, l_int32 cy, l_int32 cx, l_int32 type );
+
+char *
+selPrintToString(SEL  *sel);
+
+PIX *
+pixDilate ( PIX *pixd, PIX *pixs, SEL *sel );
+
+PIX *
+pixErode ( PIX *pixd, PIX *pixs, SEL *sel );
+
+PIX *
+pixHMT ( PIX *pixd, PIX *pixs, SEL *sel );
+
+PIX *
+pixSubtract ( PIX *pixd, PIX *pixs1, PIX *pixs2 );
+
+void
 boxDestroy(BOX  **pbox);
 
+void
+boxaDestroy ( BOXA **pboxa );
+
+void
+pixaDestroy(PIXA **ppixa);
+
+l_ok
+pixRenderBoxa ( PIX *pix, BOXA *boxa, l_int32 width, l_int32 op );
+
 void
 l_CIDataDestroy(L_COMP_DATA **pcid);
 
 void
-lept_free(void *ptr);  
-""")
+sarrayDestroy(SARRAY  **psa);
 
+void
+lept_free(void *ptr);
+
+void selDestroy ( SEL **psel );
+
+l_int32
+setMsgSeverity(l_int32 newsev);
+
+"""
+)
+
+
+ffibuilder.set_source("ocrmypdf.lib._leptonica", None)
 
 if __name__ == '__main__':
     ffibuilder.compile(verbose=True)

src/ocrmypdf/optimize.py

@@ -15,28 +15,30 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-from pathlib import Path
 import concurrent.futures
-from collections import defaultdict, namedtuple
-import logging
 import sys
+import tempfile
+from collections import defaultdict
+from os import fspath
+from pathlib import Path
 
 from PIL import Image
-
+from tqdm import tqdm
 import pikepdf
+from pikepdf import Name, Dictionary
 
-from ._jobcontext import JobContext
 from . import leptonica
-from .helpers import re_symlink, fspath
-from .exec import pngquant, jbig2enc
+from ._jobcontext import PDFContext
+from .exec import jbig2enc, pngquant
+from .exceptions import OutputFileAccessError
+from .helpers import re_symlink
 
-PAGE_GROUP_SIZE = 10
 DEFAULT_JPEG_QUALITY = 75
-DEFAULT_PNG_QUALITY = (65, 75)
+DEFAULT_PNG_QUALITY = 70
 
 
 def img_name(root, xref, ext):
-    return fspath(root / '{:08d}{}'.format(xref, ext))
+    return fspath(root / f'{xref:08d}{ext}')
 
 
 def png_name(root, xref):
@@ -51,40 +53,58 @@ def tif_name(root, xref):
     return img_name(root, xref, '.tif')
 
 
-def extract_image(*, pike, root, log, image, xref, jbig2s,
-                  pngs, jpegs, options):
-    if image.Subtype != '/Image':
-        return False
+def extract_image_filter(pike, root, log, image, xref):
+    if image.Subtype != Name.Image:
+        return None
     if image.Length < 100:
-        log.debug("Skipping small image, xref {}".format(xref))
-        return False
+        log.debug("Skipping small image, xref %s", xref)
+        return None
 
     pim = pikepdf.PdfImage(image)
 
     if len(pim.filter_decodeparms) > 1:
-        log.debug("Skipping multiply filtered, xref {}".format(xref))
-        return False
+        log.debug("Skipping multiply filtered, xref %s", xref)
+        return None
     filtdp = pim.filter_decodeparms[0]
 
     if pim.bits_per_component > 8:
-        return False  # Don't mess with wide gamut images
+        return None  # Don't mess with wide gamut images
 
-    if filtdp[0] == '/JPXDecode':
-        return False  # Don't do JPEG2000
+    if filtdp[0] == Name.JPXDecode:
+        return None  # Don't do JPEG2000
 
-    if pim.bits_per_component == 1 \
-            and filtdp != '/JBIG2Decode' \
-            and jbig2enc.available():
+    return pim, filtdp
+
+
+def extract_image_jbig2(*, pike, root, log, image, xref, options):
+    result = extract_image_filter(pike, root, log, image, xref)
+    if result is None:
+        return None
+    pim, filtdp = result
+
+    if (
+        pim.bits_per_component == 1
+        and filtdp != Name.JBIG2Decode
+        and jbig2enc.available()
+    ):
         try:
-            imgname = Path(root / '{:08d}'.format(xref))
+            imgname = Path(root / f'{xref:08d}')
             with imgname.open('wb') as f:
                 ext = pim.extract_to(stream=f)
             imgname.rename(imgname.with_suffix(ext))
         except pikepdf.UnsupportedImageTypeError:
-            return False
-        jbig2s.append((xref, ext))
-    elif filtdp[0] == '/DCTDecode' \
-            and options.optimize >= 2:
+            return None
+        return xref, ext
+    return None
+
+
+def extract_image_generic(*, pike, root, log, image, xref, options):
+    result = extract_image_filter(pike, root, log, image, xref)
+    if result is None:
+        return None
+    pim, filtdp = result
+
+    if filtdp[0] == Name.DCTDecode and options.optimize >= 2:
         # This is a simple heuristic derived from some training data, that has
         # about a 70% chance of guessing whether the JPEG is high quality,
         # and possibly recompressible, or not. The number itself doesn't mean
@@ -92,7 +112,7 @@ def extract_image(*, pike, root, log, image, xref, jbig2s,
         # bytes_per_pixel = int(raw_jpeg.Length) / (w * h)
         # jpeg_quality_estimate = 117.0 * (bytes_per_pixel ** 0.213)
         # if jpeg_quality_estimate < 65:
-        #     return False
+        #     return None
 
         # We could get the ICC profile here, but there's no need to look at it
         # for quality transcoding
@@ -102,125 +122,206 @@ def extract_image(*, pike, root, log, image, xref, jbig2s,
         #     with Image.open(stream) as im:
         #         im.save(jpg_name(root, xref), icc_profile=iccbytes)
         try:
-            imgname = Path(root / '{:08d}'.format(xref))
+            imgname = Path(root / f'{xref:08d}')
             with imgname.open('wb') as f:
                 ext = pim.extract_to(stream=f)
             imgname.rename(imgname.with_suffix(ext))
         except pikepdf.UnsupportedImageTypeError:
-            return False
-        jpegs.append(xref)
-    elif pim.indexed \
-            and pim.colorspace in pim.SIMPLE_COLORSPACES \
-            and options.optimize >= 3:
+            return None
+        return xref, ext
+    elif (
+        pim.indexed
+        and pim.colorspace in pim.SIMPLE_COLORSPACES
+        and options.optimize >= 3
+    ):
         # Try to improve on indexed images - these are far from low hanging
         # fruit in most cases
         pim.as_pil_image().save(png_name(root, xref))
-        pngs.append(xref)
+        return xref, '.png'
     elif not pim.indexed and pim.colorspace in pim.SIMPLE_COLORSPACES:
         # An optimization opportunity here, not currently taken, is directly
         # generating a PNG from compressed data
         pim.as_pil_image().save(png_name(root, xref))
-        pngs.append(xref)
-    else:
-        return False
+        return xref, '.png'
 
-    return True
+    return None
 
 
-def extract_images(pike, root, log, options):
-    # Extract images we can improve
-    changed_xrefs = set()
-    jbig2_groups = defaultdict(lambda: [])
-    jpegs = []
-    pngs = []
+def extract_images(pike, root, log, options, extract_fn):
+    """Extract image using extract_fn
+
+    Enumerate images on each page, lookup their xref/ID number in the PDF.
+    Exclude images that are soft masks (i.e. alpha transparency related).
+    Record the page number on which an image is first used, since images may be
+    used on multiple pages (or multiple times on the same page).
+
+    Current we do not check Form XObjects or other objects that may contain
+    images, and we don't evaluate alternate images or thumbnails.
+
+    extract_fn must decide if wants to extract the image in this context. If
+    it does a tuple should be returned: (xref, ext) where .ext is the file
+    extension. extract_fn must also extract the file it finds interesting.
+    """
+
+    include_xrefs = set()
+    exclude_xrefs = set()
+    pageno_for_xref = {}
     errors = 0
     for pageno, page in enumerate(pike.pages):
-        group, _ = divmod(pageno, PAGE_GROUP_SIZE)
         try:
             xobjs = page.Resources.XObject
         except AttributeError:
             continue
-        for imname, image in dict(xobjs).items():
+        for _imname, image in dict(xobjs).items():
             if image.objgen[1] != 0:
                 continue  # Ignore images in an incremental PDF
             xref = image.objgen[0]
-            if xref in changed_xrefs:
-                continue  # Don't improve same image twice
-            try:
-                result = extract_image(
-                    pike=pike, root=root, log=log, image=image,
-                    xref=xref, jbig2s=jbig2_groups[group], pngs=pngs,
-                    jpegs=jpegs, options=options
-                )
-                if result:
-                    changed_xrefs.add(xref)
-            except Exception as e:
-                log.debug("Image {} xref {}".format(imname, xref))
-                log.debug(repr(e))
-                errors += 1
+            if hasattr(image, 'SMask'):
+                # Ignore soft masks
+                smask_xref = image.SMask.objgen[0]
+                exclude_xrefs.add(smask_xref)
+            include_xrefs.add(xref)
+            if xref not in pageno_for_xref:
+                pageno_for_xref[xref] = pageno
+
+    working_xrefs = include_xrefs - exclude_xrefs
+    for xref in working_xrefs:
+        image = pike.get_object((xref, 0))
+        try:
+            result = extract_fn(
+                pike=pike, root=root, log=log, image=image, xref=xref, options=options
+            )
+        except Exception as e:
+            log.debug("Image xref %s, error %s", xref, repr(e))
+            errors += 1
+        else:
+            if result:
+                _, ext = result
+                yield pageno_for_xref[xref], xref, ext
+
+
+def extract_images_generic(pike, root, log, options):
+    """Extract any >=2bpp image we think we can improve"""
+
+    jpegs = []
+    pngs = []
+    for _, xref, ext in extract_images(pike, root, log, options, extract_image_generic):
+        log.debug('xref = %s ext = %s', xref, ext)
+        if ext == '.png':
+            pngs.append(xref)
+        elif ext == '.jpg':
+            jpegs.append(xref)
+    log.debug("Optimizable images: JPEGs: %s PNGs: %s", len(jpegs), len(pngs))
+    return jpegs, pngs
+
+
+def extract_images_jbig2(pike, root, log, options):
+    """Extract any bitonal image that we think we can improve as JBIG2"""
+
+    jbig2_groups = defaultdict(list)
+    for pageno, xref, ext in extract_images(
+        pike, root, log, options, extract_image_jbig2
+    ):
+        group = pageno // options.jbig2_page_group_size
+        jbig2_groups[group].append((xref, ext))
 
     # Elide empty groups
-    jbig2_groups = {group: xrefs for group, xrefs in jbig2_groups.items()
-                    if len(xrefs) > 0}
-    log.debug(
-        "Optimizable images: "
-        "JBIG2 groups: {} JPEGs: {} PNGs: {} Errors: {}".format(
-            len(jbig2_groups), len(jpegs), len(pngs), errors
-    ))
-
-    return jbig2_groups, jpegs, pngs
+    jbig2_groups = {
+        group: xrefs for group, xrefs in jbig2_groups.items() if len(xrefs) > 0
+    }
+    log.debug("Optimizable images: JBIG2 groups: %s", (len(jbig2_groups),))
+    return jbig2_groups
 
 
-def convert_to_jbig2(pike, jbig2_groups, root, log, options):
-    """
-    Convert a group of JBIG2 images and insert into PDF.
+def _produce_jbig2_images(jbig2_groups, root, log, options):
+    """Produce JBIG2 images from their groups"""
 
-    We use a group because JBIG2 works best with a symbol dictionary that spans
-    multiple pages. When inserted back into the PDF, each JBIG2 must reference
-    the symbol dictionary it is associated with. So convert a group at a time,
-    and replace their streams with a parameter set that points to the
-    appropriate dictionary.
-
-    If too many pages shared the same dictionary JBIG2 encoding becomes more
-    expensive and less efficient.
-
-    """
-    with concurrent.futures.ThreadPoolExecutor(
-            max_workers=options.jobs) as executor:
-        futures = []
-        for group, xref_exts in jbig2_groups.items():
-            prefix = 'group{:08d}'.format(group)
+    def jbig2_group_futures(executor, root, groups):
+        for group, xref_exts in groups.items():
+            prefix = f'group{group:08d}'
             future = executor.submit(
                 jbig2enc.convert_group,
                 cwd=fspath(root),
                 infiles=(img_name(root, xref, ext) for xref, ext in xref_exts),
-                out_prefix=prefix
+                out_prefix=prefix,
             )
-            futures.append(future)
-        for future in concurrent.futures.as_completed(futures):
-            proc = future.result()
-            log.debug(proc.stderr.decode())
+            yield future
+
+    def jbig2_single_futures(executor, root, groups):
+        for group, xref_exts in groups.items():
+            prefix = f'group{group:08d}'
+            # Second loop is to ensure multiple images per page are unpacked
+            for n, xref_ext in enumerate(xref_exts):
+                xref, ext = xref_ext
+                future = executor.submit(
+                    jbig2enc.convert_single,
+                    cwd=fspath(root),
+                    infile=img_name(root, xref, ext),
+                    outfile=root / f'{prefix}.{n:04d}',
+                )
+                yield future
+
+    if options.jbig2_page_group_size > 1:
+        jbig2_futures = jbig2_group_futures
+    else:
+        jbig2_futures = jbig2_single_futures
+
+    with concurrent.futures.ThreadPoolExecutor(max_workers=options.jobs) as executor:
+        futures = jbig2_futures(executor, root, jbig2_groups)
+        with tqdm(
+            total=len(jbig2_groups),
+            desc="JBIG2",
+            unit='item',
+            disable=not options.progress_bar,
+        ) as pbar:
+            for future in concurrent.futures.as_completed(futures):
+                proc = future.result()
+                log.debug(proc.stderr.decode())
+                pbar.update()
+
+
+def convert_to_jbig2(pike, jbig2_groups, root, log, options):
+    """Convert images to JBIG2 and insert into PDF.
+
+    When the JBIG2 page group size is > 1 we do several JBIG2 images at once
+    and build a symbol dictionary that will span several pages. Each JBIG2
+    image must reference to its symbol dictionary. If too many pages shared the
+    same dictionary JBIG2 encoding becomes more expensive and less efficient.
+    The default value of 10 was determined through testing. Currently this
+    must be lossy encoding since jbig2enc does not support refinement coding.
+
+    When the JBIG2 symbolic coder is not used, each JBIG2 stands on its own
+    and needs no dictionary. Currently this must be lossless JBIG2.
+    """
+
+    _produce_jbig2_images(jbig2_groups, root, log, options)
 
     for group, xref_exts in jbig2_groups.items():
-        prefix = 'group{:08d}'.format(group)
-        jbig2_globals_data = (root / (prefix + '.sym')).read_bytes()
-        jbig2_globals = pikepdf.Stream(pike, jbig2_globals_data)
+        prefix = f'group{group:08d}'
+        jbig2_symfile = root / (prefix + '.sym')
+        if jbig2_symfile.exists():
+            jbig2_globals_data = jbig2_symfile.read_bytes()
+            jbig2_globals = pikepdf.Stream(pike, jbig2_globals_data)
+            jbig2_globals_dict = Dictionary(JBIG2Globals=jbig2_globals)
+        elif options.jbig2_page_group_size == 1:
+            jbig2_globals_dict = None
+        else:
+            raise FileNotFoundError(jbig2_symfile)
 
         for n, xref_ext in enumerate(xref_exts):
             xref, _ = xref_ext
-            jbig2_im_file = root / (prefix + '.{:04d}'.format(n))
+            jbig2_im_file = root / (prefix + f'.{n:04d}')
             jbig2_im_data = jbig2_im_file.read_bytes()
             im_obj = pike.get_object(xref, 0)
             im_obj.write(
-                jbig2_im_data, pikepdf.Name('/JBIG2Decode'),
-                pikepdf.Dictionary({
-                    '/JBIG2Globals': jbig2_globals
-                })
+                jbig2_im_data, filter=Name.JBIG2Decode, decode_parms=jbig2_globals_dict
             )
 
 
 def transcode_jpegs(pike, jpegs, root, log, options):
-    for xref in jpegs:
+    for xref in tqdm(
+        jpegs, desc="JPEGs", unit='image', disable=not options.progress_bar
+    ):
         in_jpg = Path(jpg_name(root, xref))
         opt_jpg = in_jpg.with_suffix('.opt.jpg')
 
@@ -229,129 +330,176 @@ def transcode_jpegs(pike, jpegs, root, log, options):
         # 'close'.  Seems to be mostly harmless
         # https://github.com/python-pillow/Pillow/issues/1144
         with Image.open(fspath(in_jpg)) as im:
-            im.save(fspath(opt_jpg),
-                    optimize=True,
-                    quality=options.jpeg_quality)
-        # pylint: disable=E1101
+            im.save(fspath(opt_jpg), optimize=True, quality=options.jpeg_quality)
+
         if opt_jpg.stat().st_size > in_jpg.stat().st_size:
-            log.debug("xref {}, jpeg, made larger - skip".format(xref))
+            log.debug("xref %s, jpeg, made larger - skip", xref)
             continue
 
         compdata = leptonica.CompressedData.open(opt_jpg)
         im_obj = pike.get_object(xref, 0)
-        im_obj.write(
-            compdata.read(), pikepdf.Name('/DCTDecode'),
-            pikepdf.Null()
-        )
+        im_obj.write(compdata.read(), filter=Name.DCTDecode)
 
 
-def transcode_pngs(pike, pngs, root, log, options):
+def transcode_pngs(pike, images, image_name_fn, root, log, options):
     if options.optimize >= 2:
         png_quality = (
             max(10, options.png_quality - 10),
-            min(100, options.png_quality + 10)
+            min(100, options.png_quality + 10),
         )
         with concurrent.futures.ThreadPoolExecutor(
-                max_workers=options.jobs) as executor:
-            for xref in pngs:
-                executor.submit(
-                    pngquant.quantize,
-                    png_name(root, xref), png_name(root, xref),
-                    png_quality[0], png_quality[1])
+            max_workers=options.jobs
+        ) as executor:
+            futures = []
+            for xref in images:
+                log.debug(image_name_fn(root, xref))
+                futures.append(
+                    executor.submit(
+                        pngquant.quantize,
+                        image_name_fn(root, xref),
+                        png_name(root, xref),
+                        png_quality[0],
+                        png_quality[1],
+                    )
+                )
+            with tqdm(
+                desc="PNGs",
+                total=len(futures),
+                unit='image',
+                disable=not options.progress_bar,
+            ) as pbar:
+                for _future in concurrent.futures.as_completed(futures):
+                    pbar.update()
 
-    for xref in pngs:
+    for xref in images:
         im_obj = pike.get_object(xref, 0)
-
-        # Open, transcode (!), package for PDF
         try:
-            pix = leptonica.Pix.open(png_name(root, xref))
-            if pix.depth == 1:
-                pix = pix.invert()  # PDF assumes 1 is black for monochrome
-            compdata = pix.generate_pdf_ci_data(
-                leptonica.lept.L_FLATE_ENCODE, 0
-            )
+            compdata = leptonica.CompressedData.open(png_name(root, xref))
         except leptonica.LeptonicaError as e:
+            # Most likely this means file not found, i.e. quantize did not
+            # produce an improved version
             log.error(e)
             continue
 
-        # This is what we should be doing: open the compressed data without
-        # transcoding. However this shifts each pixel row by one for some
-        # reason.
-        #compdata = leptonica.CompressedData.open(png_name(root, xref))
+        # If re-coded image is larger don't use it - we test here because
+        # pngquant knows the size of the temporary output file but not the actual
+        # object in the PDF
         if len(compdata) > int(im_obj.stream_dict.Length):
-            continue  # If we produced a larger image, don't use
+            log.debug(
+                f"pngquant: pngquant did not improve over original image "
+                f"{len(compdata)} > {int(im_obj.stream_dict.Length)}"
+            )
+            continue
 
-        predictor = pikepdf.Null()
-        if compdata.predictor > 0:
-            predictor = pikepdf.Dictionary({'/Predictor': compdata.predictor})
+        # When a PNG is inserted into a PDF, we more or less copy the IDAT section from
+        # the PDF and transfer the rest of the PNG headers to PDF image metadata.
+        # One thing we have to do is tell the PDF reader whether a predictor was used
+        # on the image before Flate encoding. (Typically one is.)
+        # According to Leptonica source, PDF readers don't actually need us
+        # to specify the correct predictor, they just need a value of either:
+        #   1 - no predictor
+        #   10-14 - there is a predictor
+        # Leptonica's compdata->predictor only tells TRUE or FALSE
+        # From there the PNG decoder can infer the rest from the file.
+        # In practice the predictor should be Paeth, 14, so we'll use that.
+        # See:
+        #   - PDF RM 7.4.4.4 Table 10
+        #   - https://github.com/DanBloomberg/leptonica/blob/master/src/pdfio2.c#L757
+        predictor = 14 if compdata.predictor > 0 else 1
+        dparms = Dictionary(Predictor=predictor)
+        if predictor > 1:
+            dparms.BitsPerComponent = compdata.bps  # Yes, this is redundant
+            dparms.Colors = compdata.spp
+            dparms.Columns = compdata.w
 
         im_obj.BitsPerComponent = compdata.bps
         im_obj.Width = compdata.w
         im_obj.Height = compdata.h
 
         if compdata.ncolors > 0:
+            # .ncolors is the number of colors in the palette, not the number of
+            # colors used in a true color image
             palette_pdf_string = compdata.get_palette_pdf_string()
             palette_data = pikepdf.Object.parse(palette_pdf_string)
             palette_stream = pikepdf.Stream(pike, bytes(palette_data))
-            palette = [pikepdf.Name('/Indexed'), pikepdf.Name('/DeviceRGB'),
-                       compdata.ncolors - 1, palette_stream]
+            palette = [
+                Name.Indexed,
+                Name.DeviceRGB,
+                compdata.ncolors - 1,
+                palette_stream,
+            ]
             cs = palette
         else:
             if compdata.spp == 1:
-                cs = pikepdf.Name('/DeviceGray')
+                # PDF interprets binary-1 as black in 1bpp, but PNG sets
+                # black to 0 for 1bpp. Create a palette that informs the PDF
+                # of the mapping - seems cleaner to go this way but pikepdf
+                # needs to be patched to support it.
+                # palette = [Name.Indexed, Name.DeviceGray, 1, b"\xff\x00"]
+                # cs = palette
+                cs = Name.DeviceGray
             elif compdata.spp == 3:
-                cs = pikepdf.Name('/DeviceRGB')
+                cs = Name.DeviceRGB
             elif compdata.spp == 4:
-                cs = pikepdf.Name('/DeviceCMYK')
+                cs = Name.DeviceCMYK
+        if compdata.bps == 1:
+            im_obj.Decode = [1, 0]  # Bit of a kludge but this inverts photometric too
         im_obj.ColorSpace = cs
-        im_obj.write(compdata.read(), pikepdf.Name('/FlateDecode'), predictor)
+        im_obj.write(compdata.read(), filter=Name.FlateDecode, decode_parms=dparms)
 
 
-def optimize(
-    input_file,
-    output_file,
-    log,
-    context):
-
-    options = context.get_options()
+def optimize(input_file, output_file, context):
+    log = context.log
+    options = context.options
     if options.optimize == 0:
-        re_symlink(input_file, output_file, log)
+        re_symlink(input_file, output_file)
         return
 
     if options.jpeg_quality == 0:
-        options.jpeg_quality = \
-                DEFAULT_JPEG_QUALITY if options.optimize < 3 else 40
+        options.jpeg_quality = DEFAULT_JPEG_QUALITY if options.optimize < 3 else 40
     if options.png_quality == 0:
-        options.png_quality = \
-                DEFAULT_PNG_QUALITY if options.optimize < 3 else 30
+        options.png_quality = DEFAULT_PNG_QUALITY if options.optimize < 3 else 30
+    if options.jbig2_page_group_size == 0:
+        options.jbig2_page_group_size = 10 if options.jbig2_lossy else 1
 
-    pike = pikepdf.Pdf.open(input_file)
+    with pikepdf.Pdf.open(input_file) as pike:
+        root = Path(output_file).parent / 'images'
+        root.mkdir(exist_ok=True)
 
-    root = Path(output_file).parent / 'images'
-    root.mkdir(exist_ok=True)  # pylint: disable=E1101
-    jbig2_groups, jpegs, pngs = extract_images(
-        pike, root, log, options)
+        jpegs, pngs = extract_images_generic(pike, root, log, options)
+        transcode_jpegs(pike, jpegs, root, log, options)
+        # if options.optimize >= 2:
+        # Try pngifying the jpegs
+        #    transcode_pngs(pike, jpegs, jpg_name, root, log, options)
+        transcode_pngs(pike, pngs, png_name, root, log, options)
 
-    convert_to_jbig2(pike, jbig2_groups, root, log, options)
-    transcode_jpegs(pike, jpegs, root, log, options)
-    transcode_pngs(pike, pngs, root, log, options)
+        jbig2_groups = extract_images_jbig2(pike, root, log, options)
+        convert_to_jbig2(pike, jbig2_groups, root, log, options)
 
-    # Not object_stream_mode + preserve_pdfa generates noncompliant PDFs
-    target_file = Path(output_file).with_suffix('.opt.pdf')
-    pike.save(target_file, preserve_pdfa=True)
+        target_file = Path(output_file).with_suffix('.opt.pdf')
+        pike.remove_unreferenced_resources()
+        pike.save(
+            target_file,
+            preserve_pdfa=True,
+            object_stream_mode=pikepdf.ObjectStreamMode.generate,
+        )
 
     input_size = Path(input_file).stat().st_size
     output_size = Path(target_file).stat().st_size
+    if output_size == 0:
+        raise OutputFileAccessError(
+            f"Output file not created after optimizing. We probably ran "
+            f"out of disk space in the temporary folder: {tempfile.gettempdir()}."
+        )
     ratio = input_size / output_size
     savings = 1 - output_size / input_size
-    log.info("Optimize ratio: {:.2f} savings: {:.1f}%".format(
-        ratio, 100 * savings))
+    log.info(f"Optimize ratio: {ratio:.2f} savings: {(100 * savings):.1f}%")
 
     if savings < 0:
         log.info("Optimize did not improve the file - discarded")
-        re_symlink(input_file, output_file, log)
+        re_symlink(input_file, output_file)
     else:
-        re_symlink(target_file, output_file, log)
+        re_symlink(target_file, output_file)
 
 
 def main(infile, outfile, level, jobs=1):
@@ -361,27 +509,32 @@ def main(infile, outfile, level, jobs=1):
     class OptimizeOptions:
         """Emulate ocrmypdf's options"""
 
-        def __init__(self, jobs, optimize, jpeg_quality, png_quality):
+        def __init__(
+            self, input_file, jobs, optimize, jpeg_quality, png_quality, jb2lossy
+        ):
+            self.input_file = input_file
             self.jobs = jobs
             self.optimize = optimize
             self.jpeg_quality = jpeg_quality
             self.png_quality = png_quality
+            self.jbig2_page_group_size = 0
+            self.jbig2_lossy = jb2lossy
+            self.quiet = True
+            self.progress_bar = False
 
-    logging.basicConfig(level=logging.DEBUG)
-    log = logging.getLogger()
-
-    ctx = JobContext()
     options = OptimizeOptions(
+        input_file=infile,
         jobs=jobs,
         optimize=int(level),
         jpeg_quality=0,  # Use default
-        png_quality=0
+        png_quality=0,
+        jb2lossy=False,
     )
-    ctx.set_options(options)
 
     with TemporaryDirectory() as td:
+        context = PDFContext(options, td, infile, None)
         tmpout = Path(td) / 'out.pdf'
-        optimize(infile, tmpout, log, ctx)
+        optimize(infile, tmpout, context)
         copy(fspath(tmpout), fspath(outfile))
 
 

src/ocrmypdf/pdfa.py

@@ -15,38 +15,44 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-# Generate a PDFA_def.ps file for Ghostscript >= 9.14
+"""
+Generate a PDFMARK file for Ghostscript >= 9.14, for PDF/A conversion
 
-from string import Template
+pdfmark is an extension to the Postscript language that describes some PDF
+features like bookmarks and annotations. It was originally specified Adobe
+Distiller, for Postscript to PDF conversion:
+https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/pdfmark_reference.pdf
+
+Ghostscript uses pdfmark for PDF to PDF/A conversion as well. To use Ghostscript
+to create a PDF/A, we need to create a pdfmark file with the necessary metadata.
+
+This takes care of the many version-specific bugs and pecularities in
+Ghostscript's handling of pdfmark.
+
+"""
+
+import os
 from binascii import hexlify
-from datetime import datetime
-import pkg_resources
-from libxmp.utils import file_to_dict
-from libxmp import consts
 from pathlib import Path
+from string import Template
 
+import pkg_resources
+
+import pikepdf
 
 ICC_PROFILE_RELPATH = 'data/sRGB.icc'
 
-SRGB_ICC_PROFILE = pkg_resources.resource_filename(
-    'ocrmypdf', ICC_PROFILE_RELPATH)
+SRGB_ICC_PROFILE = pkg_resources.resource_filename('ocrmypdf', ICC_PROFILE_RELPATH)
 
 
 # This is a template written in PostScript which is needed to create PDF/A
 # files, from the Ghostscript documentation. Lines beginning with % are
 # comments. Python substitution variables have a '$' prefix.
 pdfa_def_template = u"""%!
-% This is derived from Ghostscript's template for creating a PDF/A document.
-% This is a small PostScript program that includes some necessary information
-% to create a PDF/A compliant file.
-
 % Define entries in the document Info dictionary :
-/ICCProfile ($icc_profile)
+/ICCProfile $icc_profile
 def
 
-[$docinfo
-  /DOCINFO pdfmark
-
 % Define an ICC profile :
 
 [/_objdef {icc_PDFA} /type /stream /OBJ pdfmark
@@ -77,142 +83,41 @@ def
 """
 
 
-def encode_text_string(s: str) -> str:
-    '''Encode text string to hex string for use in a PDF
+def generate_pdfa_ps(target_filename, icc='sRGB'):
+    """Create a Postscript pdfmark file for Ghostscript PDF/A conversion
 
-    From PDF 32000-1:2008 a string object may be included in hexademical form
-    if it is enclosed in angle brackets.  For general Unicode the string should
-    be UTF-16 (big endian) with byte order marks.  Many strings including all
-    ASCII strings could be encoded as PdfDocEncoding literals provided
-    that certain Postscript sequences are escaped.  But it's far simpler to
-    encode everything as UTF-16.
-    '''
+    A pdfmark file is a small Postscript program that provides some information
+    Ghostscript needs to perform PDF/A conversion. The only information we put
+    in specifies that we want the file to be a PDF/A, and we want to Ghostscript
+    to convert objects to the sRGB colorspace if it runs into any object that
+    it decides must be converted.
 
-    # Sometimes lazy C programmers leave their NULs at the end of strings they
-    # insert into PDFs
-    # tests/resources/aspect.pdf is one example (created by ImageMagick)
-    s = s.replace('\x00', '')
+    See the Adobe pdfmark Reference for details:
+    https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/pdfmark_reference.pdf
 
-    if s == '':
-        return ''
+    :param target_filename: filename to save
+    :param icc: ICC identifier such as 'sRGB'
 
-    utf16_bytes = s.encode('utf-16be')
-    ascii_hex_bytes = hexlify(b'\xfe\xff' + utf16_bytes)
-    ascii_hex_str = ascii_hex_bytes.decode('ascii').lower()
-    return ascii_hex_str
-
-
-def encode_pdf_date(d: datetime) -> str:
-    """Encode Python datetime object as PDF date string
-
-    From Adobe pdfmark manual:
-    (D:YYYYMMDDHHmmSSOHH'mm')
-    D: is an optional prefix. YYYY is the year. All fields after the year are
-    optional. MM is the month (01-12), DD is the day (01-31), HH is the
-    hour (00-23), mm are the minutes (00-59), and SS are the seconds
-    (00-59). The remainder of the string defines the relation of local
-    time to GMT. O is either + for a positive difference (local time is
-    later than GMT) or - (minus) for a negative difference. HH' is the
-    absolute value of the offset from GMT in hours, and mm' is the
-    absolute value of the offset in minutes. If no GMT information is
-    specified, the relation between the specified time and GMT is
-    considered unknown. Regardless of whether or not GMT
-    information is specified, the remainder of the string should specify
-    the local time.
+    :returns: a string containing the entire pdfmark
     """
-
-    pdfmark_date_fmt = r'%Y%m%d%H%M%S'
-    s = d.strftime(pdfmark_date_fmt)
-
-    tz = d.strftime('%z')
-    if tz == 'Z' or tz == '':
-        # Ghostscript <= 9.23 handles missing timezones incorrectly, so if
-        # timezone is missing, move it into GMT.
-        # https://bugs.ghostscript.com/show_bug.cgi?id=699182
-        s += "+00'00'"
-    else:
-        sign, tz_hours, tz_mins = tz[0], tz[1:3], tz[3:5]
-        s += "{}{}'{}'".format(sign, tz_hours, tz_mins)
-    return s
-
-
-def decode_pdf_date(s: str) -> datetime:
-    if s.startswith('D:'):
-        s = s[2:]
-
-    # Literal Z00'00', is incorrect but found in the wild,
-    # probably made by OS X Quartz -- standardize
-    if s.endswith("Z00'00'"):
-        s = s.replace("Z00'00'", '+0000')
-    elif s.endswith('Z'):
-        s = s.replace('Z', '+0000')
-
-    s = s.replace("'", "")  # Remove apos from PDF time strings
-
-    return datetime.strptime(s, r'%Y%m%d%H%M%S%z')
-
-
-def _get_pdfmark_dates(pdfmark):
-    """Encode dates for pdfmark Postscript.  The best way to deal with a
-    missing date entry is set it to null, because if the key is omitted
-    Ghostscript will set it to now - we do not want to erase the fact that
-    the value was unknown.  Setting to an empty string breaks Ghostscript
-    9.22 as reported here:
-    https://bugs.ghostscript.com/show_bug.cgi?id=699182
-    """
-
-    for key in ('/CreationDate', '/ModDate'):
-        if key not in pdfmark:
-            continue
-        if pdfmark[key].strip() == '':
-            yield '  {} null'.format(key)
-            continue
-        date_str = pdfmark[key]
-        if date_str.startswith('D:'):
-            date_str = date_str[2:]
-        try:
-            yield '  {} (D:{})'.format(
-                    key,
-                    encode_pdf_date(decode_pdf_date(date_str)))
-        except ValueError:
-            yield '  {} null'.format(key)
-
-
-def _get_pdfa_def(icc_profile, icc_identifier, pdfmark):
-    """Create a Postscript file for Ghostscript.  pdfmark contains the various
-    objects as strings; these must be encoded in ASCII, and dates have a
-    special format."""
-
-    # Ghostscript <= 9.21 has a bug where null entries in DOCINFO might produce
-    # ERROR: VMerror (-25) on closing pdfwrite device.
-    # https://bugs.ghostscript.com/show_bug.cgi?id=697684
-    # Work around this by only adding keys that have a nontrivial value
-    docinfo_keys = ('/Title', '/Author', '/Subject', '/Creator', '/Keywords')
-    docinfo_line_template = '  {key} <{value}>'
-
-    def docinfo_gen():
-        yield from _get_pdfmark_dates(pdfmark)
-        for key in docinfo_keys:
-            if key in pdfmark and pdfmark[key].strip() != '':
-                line = docinfo_line_template.format(
-                    key=key, value=encode_text_string(pdfmark[key]))
-                yield line
-    docinfo = '\n'.join(docinfo_gen())
-
-    t = Template(pdfa_def_template)
-    result = t.substitute(icc_profile=icc_profile,
-                          icc_identifier=icc_identifier,
-                          docinfo=docinfo)
-    return result
-
-
-def generate_pdfa_ps(target_filename, pdfmark, icc='sRGB'):
     if icc == 'sRGB':
         icc_profile = SRGB_ICC_PROFILE
     else:
         raise NotImplementedError("Only supporting sRGB")
 
-    ps = _get_pdfa_def(icc_profile, icc, pdfmark)
+    # pdfmark must contain the full path to the ICC profile, and pdfmark must be
+    # also encoded in ASCII. ocrmypdf can be installed anywhere, including to
+    # paths that have a non-ASCII character in the filename. Ghostscript
+    # accepts hex-encoded strings and converts them to byte strings, so
+    # we encode the path with fsencode() and use the hex representation.
+    # UTF-16 not accepted here. (Even though ASCII encodable is the usual case,
+    # do this always to avoid making it a rare conditional.)
+    bytes_icc_profile = os.fsencode(icc_profile)
+    hex_icc_profile = hexlify(bytes_icc_profile)
+    icc_profile = '<' + hex_icc_profile.decode('ascii') + '>'
+
+    t = Template(pdfa_def_template)
+    ps = t.substitute(icc_profile=icc_profile, icc_identifier=icc)
 
     # We should have encoded everything to pure ASCII by this point, and
     # to be safe, only allow ASCII in PostScript
@@ -222,41 +127,23 @@ def generate_pdfa_ps(target_filename, pdfmark, icc='sRGB'):
 def file_claims_pdfa(filename):
     """Determines if the file claims to be PDF/A compliant
 
-    Checking if a file is a truly compliant PDF/A is a massive undertaking
-    that no open source tool does properly.  Some commercial tools are
-    generally reliable (Acrobat).
-
-    This checks if the XMP metadata contains a PDF/A marker.
+    This only checks if the XMP metadata contains a PDF/A marker. It does not
+    do full PDF/A validation.
     """
 
-    xmp = file_to_dict(filename)
-    if not xmp:
-        return {'pass': False, 'output': 'pdf',
-                'conformance': 'No XMP metadata'}
-
-    if not consts.XMP_NS_PDFA_ID in xmp:
-        return {'pass': False, 'output': 'pdf',
-                'conformance': 'No PDF/A metadata in XMP'}
-
-    pdfa_node = xmp[consts.XMP_NS_PDFA_ID]
-    def read_node(node, key):
-        return next(
-            (v for k, v, meta in node if k == key), ''
-        )
-
-    part = read_node(pdfa_node, 'pdfaid:part')
-    conformance = read_node(pdfa_node, 'pdfaid:conformance')
-
-    part_conformance = part + conformance
-    valid_part_conforms = {'1A', '1B', '2A', '2B', '2U', '3A', '3B', '3U'}
-
-    conformance = 'PDF/A-{}'.format(
-        part_conformance)
-
-    pdfa_dict = {}
-    if part_conformance in valid_part_conforms:
-        pdfa_dict['pass'] = True
-        pdfa_dict['output'] = 'pdfa'
-    pdfa_dict['conformance'] = conformance
-
+    with pikepdf.open(filename) as pdf:
+        pdfmeta = pdf.open_metadata()
+        if not pdfmeta.pdfa_status:
+            return {
+                'pass': False,
+                'output': 'pdf',
+                'conformance': 'No PDF/A metadata in XMP',
+            }
+        valid_part_conforms = {'1A', '1B', '2A', '2B', '2U', '3A', '3B', '3U'}
+        conformance = f'PDF/A-{pdfmeta.pdfa_status}'
+        pdfa_dict = {}
+        if pdfmeta.pdfa_status in valid_part_conforms:
+            pdfa_dict['pass'] = True
+            pdfa_dict['output'] = 'pdfa'
+        pdfa_dict['conformance'] = conformance
     return pdfa_dict

src/ocrmypdf/pdfinfo/__init__.py

@@ -0,0 +1,19 @@
+#!/usr/bin/env python3
+# © 2015 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+from .info import PdfInfo, Colorspace, Encoding

src/ocrmypdf/pdfinfo/ghosttext.py

@@ -0,0 +1,101 @@
+# © 2018 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+import logging
+import re
+import xml.etree.ElementTree as ET
+
+from ..exec import ghostscript
+
+gslog = logging.getLogger()
+
+# Forgive me for I have sinned
+# I am using regular expressions to parse XML. However the XML in this case,
+# generated by Ghostscript, is self-consistent enough to be parseable.
+regex_remove_char_tags = re.compile(
+    br"""
+    <char\b
+    (?:   [^>]   # anything single character but >
+        | \">\"  # special case: trap ">"
+    )*
+    />           # terminate with '/>'
+""",
+    re.VERBOSE,
+)
+
+
+def page_get_textblocks(infile, pageno, xmltext, height):
+    """Get text boxes out of Ghostscript txtwrite xml"""
+
+    root = xmltext
+    if not hasattr(xmltext, 'findall'):
+        return []
+
+    def blocks():
+        for span in root.findall('.//span'):
+            bbox_str = span.attrib['bbox']
+            font_size = span.attrib['size']
+            pts = [int(pt) for pt in bbox_str.split()]
+            pts[1] = pts[1] - int(float(font_size) + 0.5)
+            bbox_topdown = tuple(pts)
+            bb = bbox_topdown
+            bbox_bottomup = (bb[0], height - bb[3], bb[2], height - bb[1])
+            yield bbox_bottomup
+
+    def joined_blocks():
+        prev = None
+        for bbox in blocks():
+            if prev is None:
+                prev = bbox
+            if bbox[1] == prev[1] and bbox[3] == prev[3]:
+                gap = prev[2] - bbox[0]
+                height = abs(bbox[3] - bbox[1])
+                if gap < height:
+                    # Join boxes
+                    prev = (prev[0], prev[1], bbox[2], bbox[3])
+                    continue
+            # yield previously joined bboxes and start anew
+            yield prev
+            prev = bbox
+        if prev is not None:
+            yield prev
+
+    return [block for block in joined_blocks()]
+
+
+def extract_text_xml(infile, pdf, pageno=None, log=gslog):
+    existing_text = ghostscript.extract_text(infile, pageno=None)
+    existing_text = regex_remove_char_tags.sub(b' ', existing_text)
+
+    try:
+        root = ET.fromstringlist([b'<document>\n', existing_text, b'</document>\n'])
+        page_xml = root.findall('page')
+    except ET.ParseError as e:
+        log.error(
+            "An error occurred while attempting to retrieve existing text in "
+            "the input file. Will attempt to continue assuming that there is "
+            "no existing text in the file. The error was:"
+        )
+        log.error(e)
+        page_xml = [None] * len(pdf.pages)
+
+    page_count_difference = len(pdf.pages) - len(page_xml)
+    if page_count_difference != 0:
+        log.error("The number of pages in the input file is inconsistent.")
+        if page_count_difference > 0:
+            page_xml.extend([None] * page_count_difference)
+    return page_xml

src/ocrmypdf/pdfinfo/info.py

@@ -16,36 +16,30 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-from decimal import Decimal
-from math import hypot, isclose
-import re
 from collections import namedtuple
-from pathlib import Path
+from decimal import Decimal
 from enum import Enum
-
-from .exec import ghostscript
-from .helpers import fspath
+import logging
+from math import hypot, isclose
+from os import fspath
+from pathlib import Path
+from warnings import warn
+import re
 
 from pikepdf import PdfMatrix
 import pikepdf
+from tqdm import tqdm
 
-Colorspace = Enum('Colorspace',
-                  'gray rgb cmyk lab icc index sep devn pattern jpeg2000')
+from . import ghosttext
+from ocrmypdf.exceptions import EncryptedPdfError, MissingDependencyError
 
-Encoding = Enum('Encoding',
-                'ccitt jpeg jpeg2000 jbig2 asciihex ascii85 lzw flate ' + \
-                'runlength')
+logger = logging.getLogger()
 
-# Forgive me for I have sinned
-# I am using regular expressions to parse XML. However the XML in this case,
-# generated by Ghostscript, is self-consistent enough to be parseable.
-regex_remove_char_tags = re.compile(br"""
-    <char\b
-    (?:   [^>]   # anything single character but >
-        | \">\"  # special case: trap ">"
-    )*
-    />           # terminate with '/>'
-""", re.VERBOSE)
+Colorspace = Enum('Colorspace', 'gray rgb cmyk lab icc index sep devn pattern jpeg2000')
+
+Encoding = Enum(
+    'Encoding', 'ccitt jpeg jpeg2000 jbig2 asciihex ascii85 lzw flate ' + 'runlength'
+)
 
 FRIENDLY_COLORSPACE = {
     '/DeviceGray': Colorspace.gray,
@@ -76,7 +70,7 @@ FRIENDLY_ENCODING = {
     '/A85': Encoding.ascii85,
     '/LZW': Encoding.lzw,
     '/Fl': Encoding.flate,
-    '/RL': Encoding.runlength
+    '/RL': Encoding.runlength,
 }
 
 FRIENDLY_COMP = {
@@ -84,63 +78,48 @@ FRIENDLY_COMP = {
     Colorspace.rgb: 3,
     Colorspace.cmyk: 4,
     Colorspace.lab: 3,
-    Colorspace.index: 1
+    Colorspace.index: 1,
 }
 
 
 UNIT_SQUARE = (1.0, 0.0, 0.0, 1.0, 0.0, 0.0)
 
 
-def _matrix_from_shorthand(shorthand):
-    """Convert from PDF matrix shorthand to full matrix
-
-    PDF 1.7 spec defines a shorthand for describing the entries of a matrix
-    since the last column is always (0, 0, 1).
-    """
-
-    a, b, c, d, e, f = map(float, shorthand)
-    return ((a, b, 0),
-            (c, d, 0),
-            (e, f, 1))
-
-
-def _shorthand_from_matrix(matrix):
-    """Convert from transformation matrix to PDF shorthand."""
-    a, b = matrix[0][0], matrix[0][1]
-    c, d = matrix[1][0], matrix[1][1]
-    e, f = matrix[2][0], matrix[2][1]
-    return tuple(map(float, (a, b, c, d, e, f)))
-
-
 def _is_unit_square(shorthand):
     values = map(float, shorthand)
     pairwise = zip(values, UNIT_SQUARE)
     return all([isclose(a, b, rel_tol=1e-3) for a, b in pairwise])
 
-XobjectSettings = namedtuple('XobjectSettings',
-    ['name', 'shorthand', 'stack_depth'])
 
-InlineSettings = namedtuple('InlineSettings',
-    ['iimage', 'shorthand', 'stack_depth'])
+XobjectSettings = namedtuple('XobjectSettings', ['name', 'shorthand', 'stack_depth'])
 
-ContentsInfo = namedtuple('ContentsInfo',
-    ['xobject_settings', 'inline_images', 'found_text'])
+InlineSettings = namedtuple('InlineSettings', ['iimage', 'shorthand', 'stack_depth'])
+
+ContentsInfo = namedtuple(
+    'ContentsInfo', ['xobject_settings', 'inline_images', 'found_vector']
+)
+
+TextboxInfo = namedtuple('TextboxInfo', ['bbox', 'is_visible', 'is_corrupt'])
 
 
+class VectorInfo:
+    def __init__(self):
+        pass
 
-def _normalize_stack(operations):
-    """Convert runs of qQ's in the stack into single operations"""
-    for operands, command in operations:
-        command = str(command)
-        if re.match(r'Q*q+$', command):   # Zero or more Q, one or more q
-            for char in command:          # Split into individual
-                yield ([], char)          # Yield individual
+
+def _normalize_stack(graphobjs):
+    """Convert runs of qQ's in the stack into single graphobjs"""
+    for operands, operator in graphobjs:
+        operator = str(operator)
+        if re.match(r'Q*q+$', operator):  # Zero or more Q, one or more q
+            for char in operator:  # Split into individual
+                yield ([], char)  # Yield individual
         else:
-            yield (operands, command)
+            yield (operands, operator)
 
 
 def _interpret_contents(contentstream, initial_shorthand=UNIT_SQUARE):
-    """Interpret the PDF content stream
+    """Interpret the PDF content stream.
 
     The stack represents the state of the PDF graphics stack.  We are only
     interested in the current transformation matrix (CTM) so we only track
@@ -158,53 +137,66 @@ def _interpret_contents(contentstream, initial_shorthand=UNIT_SQUARE):
 
     PDF units suit our needs so we initialize ctm to the identity matrix.
 
+    According to the PDF specification, the maximum stack depth is 32. Other
+    viewers tolerate some amount beyond this.  We issue a warning if the
+    stack depth exceeds the spec limit and set a hard limit beyond this to
+    bound our memory requirements.  If the stack underflows behavior is
+    undefined in the spec, but we just pretend nothing happened and leave the
+    CTM unchanged.
     """
 
     stack = []
     ctm = PdfMatrix(initial_shorthand)
     xobject_settings = []
     inline_images = []
-    found_text = False
-    text_operators = set(['Tj', 'TJ', '"', "'"])
-    operator_whitelist = """q Q Do cm TJ Tj " ' BI ID EI"""
+    found_vector = False
+    vector_ops = set('S s f F f* B B* b b*'.split())
+    image_ops = set('BI ID EI q Q Do cm'.split())
+    operator_whitelist = ' '.join(vector_ops | image_ops)
 
-    for n, op in enumerate(_normalize_stack(
-            pikepdf.parse_content_stream(contentstream, operator_whitelist))):
-        operands, command = op
-
-        if command == 'q':
+    for n, graphobj in enumerate(
+        _normalize_stack(
+            pikepdf.parse_content_stream(contentstream, operator_whitelist)
+        )
+    ):
+        operands, operator = graphobj
+        if operator == 'q':
             stack.append(ctm)
-            if len(stack) > 32:
-                raise RuntimeError(
-                    "PDF graphics stack overflow, command %i" % n)
-        elif command == 'Q':
+            if len(stack) > 32:  # See docstring
+                if len(stack) > 128:
+                    raise RuntimeError(
+                        "PDF graphics stack overflowed hard limit, operator %i" % n
+                    )
+                warn("PDF graphics stack overflowed spec limit")
+        elif operator == 'Q':
             try:
                 ctm = stack.pop()
             except IndexError:
-                raise RuntimeError(
-                    "PDF graphics stack underflow, command %i" % n)
-        elif command == 'cm':
+                # Keeping the ctm the same seems to be the only sensible thing
+                # to do. Just pretend nothing happened, keep calm and carry on.
+                warn("PDF graphics stack underflowed - PDF may be malformed")
+        elif operator == 'cm':
             ctm = PdfMatrix(operands) @ ctm
-        elif command == 'Do':
+        elif operator == 'Do':
             image_name = operands[0]
             settings = XobjectSettings(
-                name=image_name, shorthand=ctm.shorthand,
-                stack_depth=len(stack))
+                name=image_name, shorthand=ctm.shorthand, stack_depth=len(stack)
+            )
             xobject_settings.append(settings)
-        elif command == 'INLINE IMAGE':
+        elif operator == 'INLINE IMAGE':  # BI/ID/EI are grouped into this
             iimage = operands[0]
             inline = InlineSettings(
-                iimage=iimage, shorthand=ctm.shorthand,
-                stack_depth=len(stack))
+                iimage=iimage, shorthand=ctm.shorthand, stack_depth=len(stack)
+            )
             inline_images.append(inline)
-        elif command in text_operators:
-            found_text = True
-
+        elif operator in vector_ops:
+            found_vector = True
 
     return ContentsInfo(
         xobject_settings=xobject_settings,
         inline_images=inline_images,
-        found_text=found_text)
+        found_vector=found_vector,
+    )
 
 
 def _get_dpi(ctm_shorthand, image_size):
@@ -274,8 +266,7 @@ def _get_dpi(ctm_shorthand, image_size):
 class ImageInfo:
     DPI_PREC = Decimal('1.000')
 
-    def __init__(self, *, name='', pdfimage=None, inline=None,
-                 shorthand=None):
+    def __init__(self, *, name='', pdfimage=None, inline=None, shorthand=None):
 
         self._name = str(name)
         self._shorthand = shorthand
@@ -360,19 +351,24 @@ class ImageInfo:
         return _get_dpi(self._shorthand, (self._width, self._height))[1]
 
     def __repr__(self):
-        class_locals = {attr: getattr(self, attr, None) for attr in dir(self)
-                        if not attr.startswith('_')}
+        class_locals = {
+            attr: getattr(self, attr, None)
+            for attr in dir(self)
+            if not attr.startswith('_')
+        }
         return (
             "<ImageInfo '{name}' {type_} {width}x{height} {color} "
-            "{comp} {bpc} {enc} {xres}x{yres}>").format(**class_locals)
+            "{comp} {bpc} {enc} {xres}x{yres}>"
+        ).format(**class_locals)
 
 
 def _find_inline_images(contentsinfo):
     "Find inline images in the contentstream"
 
     for n, inline in enumerate(contentsinfo.inline_images):
-        yield ImageInfo(name='inline-%02d' % n, shorthand=inline.shorthand,
-                        inline=inline)
+        yield ImageInfo(
+            name='inline-%02d' % n, shorthand=inline.shorthand, inline=inline
+        )
 
 
 def _image_xobjects(container):
@@ -396,6 +392,8 @@ def _image_xobjects(container):
     xobjs = resources['/XObject'].as_dict()
     for xobj in xobjs:
         candidate = xobjs[xobj]
+        if not '/Subtype' in candidate:
+            continue
         if candidate['/Subtype'] == '/Image':
             pdfimage = candidate
             yield (pdfimage, xobj)
@@ -408,7 +406,6 @@ def _find_regular_images(container, contentsinfo):
     that contains images.
 
     Generates images with their DPI at time of drawing.
-
     """
 
     for pdfimage, xobj in _image_xobjects(container):
@@ -427,8 +424,7 @@ def _find_regular_images(container, contentsinfo):
                 # these from our DPI calculation for the page.
                 continue
 
-            yield ImageInfo(name=draw.name, pdfimage=pdfimage, shorthand=
-                            draw.shorthand)
+            yield ImageInfo(name=draw.name, pdfimage=pdfimage, shorthand=draw.shorthand)
 
 
 def _find_form_xobject_images(pdf, container, contentsinfo):
@@ -458,11 +454,12 @@ def _find_form_xobject_images(pdf, container, contentsinfo):
             # but in practice both Form XObjects and multiple drawing of the
             # same object are both very rare.
             ctm_shorthand = settings.shorthand
-            yield from _find_images(
-                pdf=pdf, container=form_xobject, shorthand=ctm_shorthand)
+            yield from _process_content_streams(
+                pdf=pdf, container=form_xobject, shorthand=ctm_shorthand
+            )
 
 
-def _find_images(*, pdf, container, shorthand=None):
+def _process_content_streams(*, pdf, container, shorthand=None):
     """Find all individual instances of images drawn in the container
 
     Usually the container is a page, but it may also be a Form XObject.
@@ -471,7 +468,7 @@ def _find_images(*, pdf, container, shorthand=None):
     in an XObject.
 
     Form XObjects may include inline images, XObject images,
-    and recursively, other Form XObjects; and also vector drawing commands.
+    and recursively, other Form XObjects; and also vector graphic objects.
 
     Every instance of an image being drawn somewhere is flattened and
     treated as a unique image, since if the same image is drawn multiple times
@@ -483,8 +480,7 @@ def _find_images(*, pdf, container, shorthand=None):
 
     if container.get('/Type') == '/Page' and '/Contents' in container:
         initial_shorthand = shorthand or UNIT_SQUARE
-    elif container.get('/Type') == '/XObject' and \
-            container['/Subtype'] == '/Form':
+    elif container.get('/Type') == '/XObject' and container['/Subtype'] == '/Form':
         # Set the CTM to the state it was when the "Do" operator was
         # encountered that is drawing this instance of the Form XObject
         ctm = PdfMatrix(shorthand) if shorthand else PdfMatrix.identity()
@@ -503,62 +499,13 @@ def _find_images(*, pdf, container, shorthand=None):
 
     contentsinfo = _interpret_contents(container, initial_shorthand)
 
+    if contentsinfo.found_vector:
+        yield VectorInfo()
     yield from _find_inline_images(contentsinfo)
     yield from _find_regular_images(container, contentsinfo)
     yield from _find_form_xobject_images(pdf, container, contentsinfo)
 
 
-def _page_get_textblocks(infile, pageno):
-    """Smarter text detection"""
-    import xml.etree.ElementTree as ET
-
-    gstext = ghostscript.extract_text(infile, pageno+1)
-
-    # Remove all <char /> tags, because they might contain invalid XML entities
-    # like <char bbox="348 596 348 596" c="&#x1;"/> which chokes on the
-    # inclusion of U+0001. Understandably.
-    # Just remove the whole <char /> tag since we don't use it at all, and they
-    # are only generated as innermost self-closing tags.
-    gstext = regex_remove_char_tags.sub(b' ', gstext)
-
-    if gstext.strip() == '':
-        return []
-
-    try:
-        root = ET.fromstring(gstext)
-    except ET.ParseError as e:
-        return []  # If we can't parse, assume none...
-
-    def blocks():
-        for span in root.findall('.//span'):
-            bbox_str = span.attrib['bbox']
-            font_size = span.attrib['size']
-            pts = [int(pt) for pt in bbox_str.split()]
-            pts[1] = pts[1] - int(float(font_size) + 0.5)
-            bbox = tuple(pts)
-            yield bbox
-
-    def joined_blocks():
-        prev = None
-        for bbox in blocks():
-            if prev is None:
-                prev = bbox
-            if bbox[1] == prev[1] and bbox[3] == prev[3]:
-                gap = prev[2] - bbox[0]
-                height = bbox[3] - bbox[1]
-                if gap < height:
-                    # Join boxes
-                    prev = (prev[0], prev[1], bbox[2], bbox[3])
-                    continue
-            # yield previously joined bboxes and start anew
-            yield prev
-            prev = bbox
-        if prev is not None:
-            yield prev
-
-    return [block for block in joined_blocks()]
-
-
 def _page_has_text(text_blocks, page_width, page_height):
     """Smarter text detection that ignores text in margins"""
 
@@ -566,18 +513,19 @@ def _page_has_text(text_blocks, page_width, page_height):
 
     margin_ratio = 0.125
     interior_bbox = (
-        margin_ratio * pw, margin_ratio * ph,
-        (1 - margin_ratio) * pw, (1 - margin_ratio) * ph
+        margin_ratio * pw,  # left
+        (1 - margin_ratio) * ph,  # top
+        (1 - margin_ratio) * pw,  # right
+        margin_ratio * ph,  # bottom  (first quadrant: bottom < top)
     )
 
     def rects_intersect(a, b):
         """
         Where (a,b) are 4-tuple rects (left-0, top-1, right-2, bottom-3)
         https://stackoverflow.com/questions/306316/determine-if-two-rectangles-overlap-each-other
-        Negative signs to account for our coordinates being in the fourth quadrant
-        and the formula assuming the first
+        Formula assumes all boxes are in first quadrant
         """
-        return a[0] < b[2] and a[2] > b[0] and -a[1] > -b[3] and -a[3] < -b[1]
+        return a[0] < b[2] and a[2] > b[0] and a[1] > b[3] and a[3] < b[1]
 
     has_text = False
     for bbox in text_blocks:
@@ -587,21 +535,50 @@ def _page_has_text(text_blocks, page_width, page_height):
     return has_text
 
 
-def _pdf_get_pageinfo(pdf, pageno: int, infile):
+def simplify_textboxes(miner, textbox_getter):
+    """Extract only limited content from text boxes
+
+    We do this to save memory and ensure that our objects are pickleable.
+    """
+    for box in textbox_getter(miner):
+        first_line = box._objs[0]
+        first_char = first_line._objs[0]
+
+        visible = first_char.rendermode != 3
+        corrupt = first_char.get_text() == '\ufffd'
+        yield TextboxInfo(box.bbox, visible, corrupt)
+
+
+def _pdf_get_pageinfo(pdf, pageno: int, infile, xmltext):
     pageinfo = {}
     pageinfo['pageno'] = pageno
     pageinfo['images'] = []
 
     page = pdf.pages[pageno]
-
-    pageinfo['textinfo'] = _page_get_textblocks(fspath(infile), pageno)
-
     mediabox = [Decimal(d) for d in page.MediaBox.as_list()]
     width_pt = mediabox[2] - mediabox[0]
     height_pt = mediabox[3] - mediabox[1]
 
-    pageinfo['has_text'] = _page_has_text(
-        pageinfo['textinfo'], width_pt, height_pt)
+    if xmltext is not None:
+        bboxes = ghosttext.page_get_textblocks(
+            fspath(infile), pageno, xmltext=xmltext, height=height_pt
+        )
+        pageinfo['bboxes'] = bboxes
+    else:
+        # pdfminer required for this section
+        try:
+            from .layout import get_page_analysis, get_text_boxes
+        except ImportError:
+            raise MissingDependencyError(
+                "pdfminer is required for this feature. Your distribution "
+                "may not have installed it."
+            )
+        pscript5_mode = str(pdf.docinfo.get('/Creator')).startswith('PScript5')
+        miner = get_page_analysis(infile, pageno, pscript5_mode)
+        pageinfo['textboxes'] = list(simplify_textboxes(miner, get_text_boxes))
+        bboxes = (box.bbox for box in pageinfo['textboxes'])
+
+    pageinfo['has_text'] = _page_has_text(bboxes, width_pt, height_pt)
 
     userunit = page.get('/UserUnit', Decimal(1.0))
     if not isinstance(userunit, Decimal):
@@ -616,31 +593,59 @@ def _pdf_get_pageinfo(pdf, pageno: int, infile):
         pageinfo['rotate'] = 0
 
     userunit_shorthand = (userunit, 0, 0, userunit, 0, 0)
-    pageinfo['images'] = [im for im in
-                          _find_images(pdf=pdf, container=page,
-                                       shorthand=userunit_shorthand)]
+    contentsinfo = [
+        ci
+        for ci in _process_content_streams(
+            pdf=pdf, container=page, shorthand=userunit_shorthand
+        )
+    ]
+
+    pageinfo['has_vector'] = False
+    if any(isinstance(ci, VectorInfo) for ci in contentsinfo):
+        pageinfo['has_vector'] = True
+
+    pageinfo['images'] = [im for im in contentsinfo if isinstance(im, ImageInfo)]
     if pageinfo['images']:
         xres = Decimal(max(image.xres for image in pageinfo['images']))
         yres = Decimal(max(image.yres for image in pageinfo['images']))
         pageinfo['xres'], pageinfo['yres'] = xres, yres
-        pageinfo['width_pixels'] = \
-            int(round(xres * pageinfo['width_inches']))
-        pageinfo['height_pixels'] = \
-            int(round(yres * pageinfo['height_inches']))
+        pageinfo['width_pixels'] = int(round(xres * pageinfo['width_inches']))
+        pageinfo['height_pixels'] = int(round(yres * pageinfo['height_inches']))
 
     return pageinfo
 
 
-def _pdf_get_all_pageinfo(infile):
-    pdf = pikepdf.open(infile)
-    return [PageInfo(pdf, n, infile) for n in range(len(pdf.pages))]
+def _pdf_get_all_pageinfo(infile, detailed_analysis=False, log=None, progbar=False):
+    pdf = pikepdf.open(infile)  # Do not close in this function
+    if pdf.is_encrypted:
+        pdf.close()
+        raise EncryptedPdfError()  # Triggered by encryption with empty passwd
+    if detailed_analysis:
+        pages_xml = None
+    else:
+        pages_xml = ghosttext.extract_text_xml(infile, pdf, pageno=None, log=log)
+
+    pages = []
+    for n, _ in tqdm(
+        enumerate(pdf.pages),
+        total=len(pdf.pages),
+        desc="Scan",
+        unit='page',
+        disable=not progbar,
+    ):
+        page_xml = pages_xml[n] if pages_xml else None
+        page = PageInfo(pdf, n, infile, page_xml, detailed_analysis)
+        pages.append(page)
+
+    return pages, pdf
 
 
 class PageInfo:
-    def __init__(self, pdf, pageno, infile):
+    def __init__(self, pdf, pageno, infile, xmltext, detailed_analysis=False):
         self._pageno = pageno
         self._infile = infile
-        self._pageinfo = _pdf_get_pageinfo(pdf, pageno, infile)
+        self._pageinfo = _pdf_get_pageinfo(pdf, pageno, infile, xmltext)
+        self._detailed_analysis = detailed_analysis
 
     @property
     def pageno(self):
@@ -650,6 +655,16 @@ class PageInfo:
     def has_text(self):
         return self._pageinfo['has_text']
 
+    @property
+    def has_corrupt_text(self):
+        if not self._detailed_analysis:
+            raise NotImplementedError('Did not do detailed analysis')
+        return any(tbox.is_corrupt for tbox in self._pageinfo['textboxes'])
+
+    @property
+    def has_vector(self):
+        return self._pageinfo['has_vector']
+
     @property
     def width_inches(self):
         return self._pageinfo['width_inches']
@@ -681,8 +696,27 @@ class PageInfo:
     def images(self):
         return self._pageinfo['images']
 
-    def get_textareas(self):
-        yield from self._pageinfo['textinfo']
+    def get_textareas(self, visible=None, corrupt=None):
+        def predicate(obj, want_visible, want_corrupt):
+            result = True
+            if want_visible is not None:
+                if obj.is_visible != want_visible:
+                    result = False
+            if want_corrupt is not None:
+                if obj.is_corrupt != want_corrupt:
+                    result = False
+            return result
+
+        if 'textboxes' not in self._pageinfo:
+            if visible is not None and corrupt is not None:
+                raise NotImplementedError('Ghostscript textboxes cannot be classified')
+            return self._pageinfo['bboxes']
+
+        return (
+            obj.bbox
+            for obj in self._pageinfo['textboxes']
+            if predicate(obj, visible, corrupt)
+        )
 
     @property
     def xres(self):
@@ -705,21 +739,34 @@ class PageInfo:
 
     def __repr__(self):
         return (
-            '<PageInfo '
-            'pageno={} {}"x{}" rotation={} res={}x{} has_text={}>').format(
-            self.pageno, self.width_inches, self.height_inches,
+            '<PageInfo ' 'pageno={} {}"x{}" rotation={} res={}x{} has_text={}>'
+        ).format(
+            self.pageno,
+            self.width_inches,
+            self.height_inches,
             self.rotation,
-            self.xres, self.yres, self.has_text
+            self.xres,
+            self.yres,
+            self.has_text,
         )
 
 
 class PdfInfo:
-    """Get summary information about a PDF
+    """Get summary information about a PDF"""
 
-    """
-    def __init__(self, infile):
+    def __init__(self, infile, detailed_page_analysis=False, log=logger, progbar=False):
         self._infile = infile
-        self._pages = _pdf_get_all_pageinfo(infile)
+        self._pages, pdf = _pdf_get_all_pageinfo(
+            infile, detailed_page_analysis, log=log, progbar=progbar
+        )
+        self._needs_rendering = pdf.root.get('/NeedsRendering', False)
+        self._has_acroform = False
+        if '/AcroForm' in pdf.root:
+            if len(pdf.root.AcroForm.get('/Fields', [])) > 0:
+                self._has_acroform = True
+            elif '/XFA' in pdf.root.AcroForm:
+                self._has_acroform = True
+        pdf.close()
 
     @property
     def pages(self):
@@ -734,12 +781,20 @@ class PdfInfo:
     def has_userunit(self):
         return any(page.userunit != 1.0 for page in self.pages)
 
+    @property
+    def has_acroform(self):
+        return self._has_acroform
+
     @property
     def filename(self):
         if not isinstance(self._infile, (str, Path)):
             raise NotImplementedError("can't get filename from stream")
         return self._infile
 
+    @property
+    def needs_rendering(self):
+        return self._needs_rendering
+
     def __getitem__(self, item):
         return self._pages[item]
 
@@ -747,24 +802,18 @@ class PdfInfo:
         return len(self._pages)
 
     def __repr__(self):
-        return "<PdfInfo('...'), page count={}>".format(len(self))
-
-    # def __getstate__(self):
-    #     state = {'_infile': self._infile}
-    #     return state
-    #
-    # def __setstate__(self, state):
-    #     self._infile = state['_infile']
-    #     self._pages = _pdf_get_all_pageinfo(self._infile)
+        return f"<PdfInfo('...'), page count={len(self)}>"
 
 
 def main():
     import argparse
+
     parser = argparse.ArgumentParser()
     parser.add_argument('infile')
     args = parser.parse_args()
     info = _pdf_get_all_pageinfo(args.infile)
     from pprint import pprint
+
     pprint(info)
 
 

src/ocrmypdf/pdfinfo/layout.py

@@ -0,0 +1,300 @@
+# © 2018 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+import re
+from math import copysign
+from pathlib import Path
+from unittest.mock import patch
+
+import pdfminer.encodingdb
+import pdfminer.pdfdevice
+import pdfminer.pdfinterp
+from pdfminer.converter import PDFLayoutAnalyzer
+from pdfminer.glyphlist import glyphname2unicode
+from pdfminer.layout import LAParams, LTChar, LTPage, LTTextBox
+from pdfminer.pdfdocument import PDFTextExtractionNotAllowed
+from pdfminer.pdffont import PDFFont, PDFSimpleFont, PDFUnicodeNotDefined
+from pdfminer.pdfpage import PDFPage
+from pdfminer.utils import bbox2str, matrix2str
+
+from ..exceptions import EncryptedPdfError
+
+STRIP_NAME = re.compile(r'[0-9]+')
+
+#
+# Unconditional pdfminer patches
+#
+
+
+def name2unicode(name):
+    """Fix pdfminer's name2unicode function
+
+    Font cids that are mapped to names of the form /g123 seem to be, by convention
+    characters with no corresponding Unicode entry. These can be subsetted fonts
+    or symbolic fonts. There seems to be no way to map /g123 fonts to Unicode,
+    barring a ToUnicode data structure.
+    """
+    if name in glyphname2unicode:
+        return glyphname2unicode[name]
+    if name.startswith('g') or name.startswith('a'):
+        raise KeyError(name)
+    if name.startswith('uni'):
+        try:
+            return chr(int(name[3:], 16))
+        except ValueError:  # Not hexadecimal
+            raise KeyError(name)
+    m = STRIP_NAME.search(name)
+    if not m:
+        raise KeyError(name)
+    return chr(int(m.group(0)))
+
+
+pdfminer.encodingdb.name2unicode = name2unicode
+
+original_PDFFont_init = PDFFont.__init__
+
+
+def PDFFont__init__(self, descriptor, widths, default_width=None):
+    original_PDFFont_init(self, descriptor, widths, default_width)
+    # PDF spec says descent should be negative
+    # A font with a positive descent implies it floats entirely above the
+    # baseline, i.e. it's not really a baseline anymore. I have fonts that
+    # claim a positive descent, but treating descent as positive always seems
+    # to misposition text.
+    if self.descent > 0:
+        self.descent = -self.descent
+
+
+PDFFont.__init__ = PDFFont__init__
+
+original_PDFSimpleFont_init = PDFSimpleFont.__init__
+
+
+def PDFSimpleFont__init__(self, descriptor, widths, spec):
+    # Font encoding is specified either by a name of
+    # built-in encoding or a dictionary that describes
+    # the differences.
+    original_PDFSimpleFont_init(self, descriptor, widths, spec)
+    # pdfminer is incorrect. If there is no ToUnicode and no Encoding, do not
+    # assume Unicode conversion is possible. RM 9.10.2
+    if not self.unicode_map and 'Encoding' not in spec:
+        self.cid2unicode = {}
+    return
+
+
+PDFSimpleFont.__init__ = PDFSimpleFont__init__
+#
+# pdfminer patches when creator is PScript5.dll
+#
+
+
+def PDFType3Font__PScript5_get_height(self):
+    h = self.bbox[3] - self.bbox[1]
+    if h == 0:
+        h = self.ascent - self.descent
+    return h * copysign(1.0, self.vscale)
+
+
+def PDFType3Font__PScript5_get_descent(self):
+    return self.descent * copysign(1.0, self.vscale)
+
+
+def PDFType3Font__PScript5_get_ascent(self):
+    return self.ascent * copysign(1.0, self.vscale)
+
+
+class LTStateAwareChar(LTChar):
+    """A subclass of LTChar that tracks text render mode at time of drawing"""
+
+    __slots__ = (
+        'rendermode',
+        '_text',
+        'matrix',
+        'fontname',
+        'adv',
+        'upright',
+        'size',
+        'width',
+        'height',
+        'bbox',
+        'x0',
+        'x1',
+        'y0',
+        'y1',
+    )
+
+    def __init__(
+        self,
+        matrix,
+        font,
+        fontsize,
+        scaling,
+        rise,
+        text,
+        textwidth,
+        textdisp,
+        ncs,
+        graphicstate,
+        textstate,
+    ):
+        super().__init__(
+            matrix,
+            font,
+            fontsize,
+            scaling,
+            rise,
+            text,
+            textwidth,
+            textdisp,
+            ncs,
+            graphicstate,
+        )
+        self.rendermode = textstate.render
+
+    def is_compatible(self, obj):
+        """Check if characters can be combined into a textline
+
+        We consider characters compatible if:
+            - the Unicode mapping is known, and both have the same render mode
+            - the Unicode mapping is unknown but both are part of the same font
+        """
+        # pylint: disable=protected-access
+        both_unicode_mapped = isinstance(self._text, str) and isinstance(obj._text, str)
+        try:
+            if both_unicode_mapped:
+                return self.rendermode == obj.rendermode
+            font0, _ = self._text
+            font1, _ = obj._text
+            return font0 == font1 and self.rendermode == obj.rendermode
+        except (ValueError, AttributeError):
+            return False
+
+    def get_text(self):
+        if isinstance(self._text, tuple):
+            return '\ufffd'  # standard 'Unknown symbol'
+        return self._text
+
+    def __repr__(self):
+        return '<%s %s matrix=%s rendermode=%r font=%r adv=%s text=%r>' % (
+            self.__class__.__name__,
+            bbox2str(self.bbox),
+            matrix2str(self.matrix),
+            self.rendermode,
+            self.fontname,
+            self.adv,
+            self.get_text(),
+        )
+
+
+class TextPositionTracker(PDFLayoutAnalyzer):
+    """A page layout analyzer that pays attention to text visibility"""
+
+    def __init__(self, rsrcmgr, pageno=1, laparams=None):
+        super().__init__(rsrcmgr, pageno, laparams)
+        self.textstate = None
+        self.result = None
+
+    def begin_page(self, page, ctm):
+        super().begin_page(page, ctm)
+        self.cur_item = LTPage(self.pageno, page.mediabox)
+
+    def end_page(self, page):
+        assert not self._stack, str(len(self._stack))
+        assert isinstance(self.cur_item, LTPage), str(type(self.cur_item))
+        if self.laparams is not None:
+            self.cur_item.analyze(self.laparams)
+        self.pageno += 1
+        self.receive_layout(self.cur_item)
+
+    def render_string(self, textstate, seq, ncs, graphicstate):
+        self.textstate = textstate.copy()
+        super().render_string(self.textstate, seq, ncs, graphicstate)
+
+    def render_char(
+        self, matrix, font, fontsize, scaling, rise, cid, ncs, graphicstate
+    ):
+        try:
+            text = font.to_unichr(cid)
+            assert isinstance(text, str), str(type(text))
+        except PDFUnicodeNotDefined:
+            text = self.handle_undefined_char(font, cid)
+        textwidth = font.char_width(cid)
+        textdisp = font.char_disp(cid)
+        item = LTStateAwareChar(
+            matrix,
+            font,
+            fontsize,
+            scaling,
+            rise,
+            text,
+            textwidth,
+            textdisp,
+            ncs,
+            graphicstate,
+            self.textstate,
+        )
+        self.cur_item.add(item)
+        return item.adv
+
+    def handle_undefined_char(self, font, cid):
+        # log.info('undefined: %r, %r', font, cid)
+        return (font.fontname, cid)
+
+    def receive_layout(self, ltpage):
+        self.result = ltpage
+
+    def get_result(self):
+        return self.result
+
+
+def get_page_analysis(infile, pageno, pscript5_mode):
+    rman = pdfminer.pdfinterp.PDFResourceManager(caching=True)
+    dev = TextPositionTracker(rman, laparams=LAParams())
+    interp = pdfminer.pdfinterp.PDFPageInterpreter(rman, dev)
+
+    if pscript5_mode:
+        patcher = patch.multiple(
+            'pdfminer.pdffont.PDFType3Font',
+            spec=True,
+            get_ascent=PDFType3Font__PScript5_get_ascent,
+            get_descent=PDFType3Font__PScript5_get_descent,
+            get_height=PDFType3Font__PScript5_get_height,
+        )
+        patcher.start()
+
+    try:
+        with Path(infile).open('rb') as f:
+            page = PDFPage.get_pages(f, pagenos=[pageno], maxpages=0)
+            interp.process_page(next(page))
+    except PDFTextExtractionNotAllowed:
+        raise EncryptedPdfError()
+    finally:
+        if pscript5_mode:
+            patcher.stop()
+
+    return dev.get_result()
+
+
+def get_text_boxes(obj):
+    for child in obj:
+        if isinstance(child, (LTTextBox)):
+            yield child
+        else:
+            try:
+                yield from get_text_boxes(child)
+            except TypeError:
+                continue

tests/cache/2400dpi/__-l__eng__000001.ocr.png__000001.text__pdf__txt/stderr.bin

@@ -1 +0,0 @@
-Tesseract Open Source OCR Engine v3.05.01 with Leptonica

tests/cache/2400dpi/__-l__eng__000001.ocr.png__000001.text__pdf__txt/txt.bin

@@ -1,25 +0,0 @@
-é la Waterman
-
-h ens linzen
-
-3 liter water
-
-3 mien
-
-bloem} hater
-
-2 kopjes melk
-
-laurier, kruidnagel, kerrie, gout
-
-Be linzen wassen en in¢l liter kokend wa-
-ter 1 dag laten weken. 2 liter water bi'
-de linzen voegen, gender hat water wagrin
-ze geweekt zijn af te gieten. De helft Van
-de mien bakken met laurier en kruiénagal.
-Alla mien, kerrie en zout bij de linzen
-voegen. A1135 aan de Rook brengem. Van de
-bloem mat boter-en melk een papje maken en
-verder afmaken met de soap. A13 de linzen
-gaar zijn is de seep klaar.
-

tests/cache/2400dpi/__-l__eng__000001_ocr.png__000001_ocr_tess__pdf__txt/stderr.bin

@@ -0,0 +1 @@
+Tesseract Open Source OCR Engine v4.0.0 with Leptonica

tests/cache/2400dpi/__-l__eng__000001_ocr.png__000001_ocr_tess__pdf__txt/txt.bin

@@ -0,0 +1,25 @@
+a la Waterman
+
+4 ons linzen
+
+3 liter water
+
+3 uien
+
+bloem, boter
+
+2 kopjes melk
+
+laurier, kruidnagel, kerrie, zout
+
+De linzgen wassen en in -l liter kokend wa-
+ter 1 dag laten weken, 2 liter water bij
+de linzen voegen, zonder het water waarin
+ze geweekt zijn af te gieten, De helft van
+de uien bakken met laurier en Kruicdnagel.
+Alle uien, kerrie en gout bij de linzen
+voegen, Alles aan de kook brengen,. Van de
+bloem met boter en melk een papje maken en
+verder afmaken met de soep, Als de linzen
+gaar Zijn is de soep klaar.
+

tests/cache/aspect/__-l__eng__000001.ocr.png__000001.text__pdf__txt/stderr.bin

@@ -1,3 +0,0 @@
-Tesseract Open Source OCR Engine v3.05.01 with Leptonica
-Too few characters. Skipping this page
-OSD: Weak margin (0.00) for 24 blob text block, but using orientation anyway: 0

tests/cache/aspect/__-l__eng__000001.ocr.png__000001.text__pdf__txt/txt.bin

@@ -1,4 +0,0 @@
- 
-
-This shnuld be a perfect circle-
-

tests/cache/aspect/__-l__eng__000001.ocr.png__000001__hocr__txt/stderr.bin

@@ -1,3 +0,0 @@
-Tesseract Open Source OCR Engine v3.05.01 with Leptonica
-Too few characters. Skipping this page
-OSD: Weak margin (0.00) for 24 blob text block, but using orientation anyway: 0

tests/cache/aspect/__-l__eng__000001.ocr.png__000001__hocr__txt/txt.bin

@@ -1,4 +0,0 @@
- 
-
-This shnuld be a perfect circle-
-

tests/cache/aspect/__-l__eng__000001_ocr.png__000001_ocr_hocr__hocr__txt/hocr.bin

@@ -5,20 +5,27 @@
  <head>
   <title></title>
 <meta http-equiv="Content-Type" content="text/html;charset=utf-8" />
-  <meta name='ocr-system' content='tesseract 3.05.01' />
-  <meta name='ocr-capabilities' content='ocr_page ocr_carea ocr_par ocr_line ocrx_word'/>
+  <meta name='ocr-system' content='tesseract 4.0.0' />
+  <meta name='ocr-capabilities' content='ocr_page ocr_carea ocr_par ocr_line ocrx_word ocrp_wconf'/>
 </head>
 <body>
-  <div class='ocr_page' id='page_1' title='image "/var/folders/37/78_114p552q16vv6vmgm5kr00000gn/T/com.github.ocrmypdf.7vjmeixk/000001.ocr.png"; bbox 0 0 1000 800; ppageno 0'>
+  <div class='ocr_page' id='page_1' title='image "/var/folders/2s/7t022mgj0h5cprbq0dtb1ksm0000gn/T/com.github.ocrmypdf.xam82ph5/000001_ocr.png"; bbox 0 0 1000 800; ppageno 0'>
    <div class='ocr_carea' id='block_1_1' title="bbox 296 96 704 504">
     <p class='ocr_par' id='par_1_1' lang='eng' title="bbox 296 96 704 504">
-     <span class='ocr_line' id='line_1_1' title="bbox 296 96 704 504; baseline 0 296; x_size 169.33333; x_descenders 42.333332; x_ascenders 42.333336"><span class='ocrx_word' id='word_1_1' title='bbox 296 96 704 504; x_wconf 95'><strong> </strong></span> 
+     <span class='ocr_line' id='line_1_1' title="bbox 296 96 704 504; baseline 0 296; x_size 169.33333; x_descenders 42.333332; x_ascenders 42.333336">
+      <span class='ocrx_word' id='word_1_1' title='bbox 296 96 704 504; x_wconf 95'> </span>
      </span>
     </p>
    </div>
    <div class='ocr_carea' id='block_1_2' title="bbox 150 592 841 622">
     <p class='ocr_par' id='par_1_2' lang='eng' title="bbox 150 592 841 622">
-     <span class='ocr_line' id='line_1_2' title="bbox 150 592 841 622; baseline 0 -6; x_size 30; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_2' title='bbox 150 592 230 616; x_wconf 85'><strong>This</strong></span> <span class='ocrx_word' id='word_1_3' title='bbox 260 592 384 616; x_wconf 81'>shnuld</span> <span class='ocrx_word' id='word_1_4' title='bbox 413 592 449 616; x_wconf 81'><strong>be</strong></span> <span class='ocrx_word' id='word_1_5' title='bbox 479 600 493 616; x_wconf 85'><em>a</em></span> <span class='ocrx_word' id='word_1_6' title='bbox 523 592 668 622; x_wconf 76'>perfect</span> <span class='ocrx_word' id='word_1_7' title='bbox 698 592 841 616; x_wconf 76'><strong>circle-</strong></span> 
+     <span class='ocr_line' id='line_1_2' title="bbox 150 592 841 622; baseline 0 -6; x_size 30; x_descenders 6; x_ascenders 8">
+      <span class='ocrx_word' id='word_1_2' title='bbox 150 592 230 616; x_wconf 96'>This</span>
+      <span class='ocrx_word' id='word_1_3' title='bbox 260 592 384 616; x_wconf 95'>should</span>
+      <span class='ocrx_word' id='word_1_4' title='bbox 413 592 449 616; x_wconf 95'>be</span>
+      <span class='ocrx_word' id='word_1_5' title='bbox 479 600 493 616; x_wconf 95'>a</span>
+      <span class='ocrx_word' id='word_1_6' title='bbox 523 592 668 622; x_wconf 95'>perfect</span>
+      <span class='ocrx_word' id='word_1_7' title='bbox 698 592 841 616; x_wconf 55'>circle:</span>
      </span>
     </p>
    </div>

tests/cache/aspect/__-l__eng__000001_ocr.png__000001_ocr_hocr__hocr__txt/stderr.bin

@@ -0,0 +1 @@
+Tesseract Open Source OCR Engine v4.0.0 with Leptonica

tests/cache/aspect/__-l__eng__000001_ocr.png__000001_ocr_hocr__hocr__txt/txt.bin

@@ -0,0 +1,4 @@
+ 
+
+This should be a perfect circle:
+

tests/cache/aspect/__-l__eng__000001_ocr.png__000001_ocr_tess__pdf__txt/stderr.bin

@@ -0,0 +1 @@
+Tesseract Open Source OCR Engine v4.0.0 with Leptonica

tests/cache/aspect/__-l__eng__000001_ocr.png__000001_ocr_tess__pdf__txt/txt.bin

@@ -0,0 +1,4 @@
+ 
+
+This should be a perfect circle:
+

tests/cache/cardinal/__-l__eng__000001.ocr.png__000001.text__pdf__txt/stderr.bin

@@ -1 +0,0 @@
-Tesseract Open Source OCR Engine v3.05.01 with Leptonica

tests/cache/cardinal/__-l__eng__000001.ocr.png__000001.text__pdf__txt/txt.bin

@@ -1,123 +0,0 @@
-The LinnSequencer
-32 Track MIDI Sequence Recorder
-
-The LinnSequencer is a state—of—the-art composition and performance tool for the professional musician. It is
-
-extremely powerful, yet amazingly simple to learn and use. It’s many remarkable features include:
-
-0 Operation is similar to multi—track tape recorder with PLAY, STOP, RECORD, FAST
-FORWARD, REWIND, and LOCATE controls.
-
-0 Each of the 100 sequences contains 32 simultaneous, polyphonic tracks. Each track may
-be assigned to one of 16 MIDI channels. Simultaneously plays up to 16 polyphonic
-
-synthesizers !
-
-0 Ultra—fast 3 1/2 ” disk drive stores complex songs in seconds and holds over 110,000 notes
-
-per disk!
-
-0 One or all tracks may be TRANSPOSED at the touch of a key.
-0 Exclusive real—time ERASE function makes editing FAST.
-0 Exclusive REPEAT function automatically repeats any held notes at a pre-selected
-
-rhythmic value.
-
-0 TIMING CORRECTION works during playback and operates without ‘chopping’ notes.
-
-0 Optional SMPTE time code synchronization.
-
-0 Optional remote control.
-
-Recording 3 Sequence
-
-To record a sequence, simply press RECORD and PLAY,
-then play your MIDI keyboard in time to the Sequencer’s
-click track. When the sequence loops back around to bar 1,
-you’ll hear what you played—only all timing errors will be
-
-corrected! (Timing correction may be adjusted or defeated).
-
-Any additional notes played will be added into the track
-—existing notes are not erased while recording!
-
-FAST FORWARD, REWIND, and LOCATE controls
-may be used at any time to quickly access any location in
-your sequence for spot—recording. To overdub a new part,
-select a different track and start recording~while you
-record, the first‘track will play in perfect sync (unless you
-MUTE it, or SOLO another track). In this way, up to 32
-tracks may be overdubbed! All MIDI effects are recorded
-including pitch bend, modulation, velocity, aftertouch,
-sustain pedal, and program changes!
-
-Editing
-
-To erase a wrong note, simply hold ERASE and press
-the note to be erased just before it plays in the sequence—
-when played back, it will be gone. Notes may also be
-
-added, erased, or changed using the SINGLE STEP func-
-tion. To overdub notes at specific points within a sequence,
-
-Additional Features
-
-simply use LOCATE, FAST FORWARD, or REWIND to
-find the desired bar number, then start recording.
-
-The INSERT/COPY function allows you to move bars
-from one location to another—in the same sequence or a
-different one. For example, you might insert a copy of the
-first verse between the second chorus and the bridge.
-DELETE BARS operates the same way to remove
-unwanted sections.
-
-Creating a Song
-
-One way to create a song is to record each track all the
-way through (up to 999 bars). Another way is to record
-each basic section (verse, chorus, etc.) in individual
-sequences, then use the CREATE SONG function to “chain”
-them together. CREATE SONG will then automatically
-copy all the parts into a new sequence. If desired, you can
-even set the last few bars to repeat infinitely, for a fadeout.
-
-Composition Without Compromise
-
-The technology you use should never be so complex that
-it interferes with the creative process. That’s precisely why
-the LinnSequencer is designed to let you compose, record
-and edit while devoting your undivided attention to your
-music. See your Linn dealer today for a demonstration!
-
-0 Simple, easy to learn operation—the 32 character LCD display clearly guides you through all operations. If needed, the
-
-HELP button displays additional explanations.
-
-0 Non-destructive recording—existing notes are not erased while recording.
-0 Two FOOTSWITCH INPUTS may be assigned to remotely control many of the commonly used functions, including
-
-ERASE, REPEAT, PLAY/STOP, or LOCATE.
-
-0 Two TRIGGER OUTPUTS may be programmed to output pulses at any selected note value.
-
-0 Will sync to standard LinnDrum or Linn 9000 sync tone.
-
-' Utilizes ultra high—speed, 8 MHz 80186 16 bit computer internally for FAST operation.
-0 TEMPO may be specified in BEATS-PER—MINUTE or FRAMES-PER—BEAT at 24, 25, or 30 frames per second,
-
-(even drop frame!)
-
-0 TEMPO may be entered numerically, adjustable in tenths of a Beat-Per-Minute increments, or by tapping quarter notes
-
-on the TAP TEMPO button.
-
-0 TEMPO CHANGES may be programmed into a sequence, with smooth transitions if desired.
-° Any TIME SIGNATURE may be used, and may be changed within a song.
-
-E1111
-Linn Electronics, Inc.
-
-18720 Oxnard Street, Tarzana, CA 91356
-(818) 7088131 T ELEX #298949 LINN UR
-

tests/cache/cardinal/__-l__eng__000001.ocr.png__000001__hocr__txt/hocr.bin

@@ -1,335 +0,0 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
-    "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
-<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
- <head>
-  <title></title>
-<meta http-equiv="Content-Type" content="text/html;charset=utf-8" />
-  <meta name='ocr-system' content='tesseract 3.05.01' />
-  <meta name='ocr-capabilities' content='ocr_page ocr_carea ocr_par ocr_line ocrx_word'/>
-</head>
-<body>
-  <div class='ocr_page' id='page_1' title='image "/var/folders/37/78_114p552q16vv6vmgm5kr00000gn/T/com.github.ocrmypdf.wahjk6fa/000001.ocr.png"; bbox 0 0 2550 3300; ppageno 0'>
-   <div class='ocr_carea' id='block_1_1' title="bbox 582 131 1968 303">
-    <p class='ocr_par' id='par_1_1' lang='eng' title="bbox 582 131 1968 303">
-     <span class='ocr_line' id='line_1_1' title="bbox 882 131 1657 217; baseline 0.001 -17; x_size 85; x_descenders 16; x_ascenders 19"><span class='ocrx_word' id='word_1_1' title='bbox 882 132 1036 202; x_wconf 86'>The</span> <span class='ocrx_word' id='word_1_2' title='bbox 1061 131 1657 217; x_wconf 84'>LinnSequencer</span> 
-     </span>
-     <span class='ocr_line' id='line_1_2' title="bbox 582 215 1968 303; baseline 0 -17; x_size 87; x_descenders 16; x_ascenders 21"><span class='ocrx_word' id='word_1_3' title='bbox 582 215 674 286; x_wconf 90'>32</span> <span class='ocrx_word' id='word_1_4' title='bbox 697 218 923 288; x_wconf 84'>Track</span> <span class='ocrx_word' id='word_1_5' title='bbox 948 218 1181 287; x_wconf 94'><strong>MIDI</strong></span> <span class='ocrx_word' id='word_1_6' title='bbox 1208 217 1575 303; x_wconf 89'>Sequence</span> <span class='ocrx_word' id='word_1_7' title='bbox 1600 218 1968 288; x_wconf 83'>Recorder</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_2' title="bbox 347 380 2188 574">
-    <p class='ocr_par' id='par_1_2' lang='eng' title="bbox 347 380 2188 423">
-     <span class='ocr_line' id='line_1_3' title="bbox 347 380 2188 423; baseline -0.001 -12; x_size 38; x_descenders 8; x_ascenders 10"><span class='ocrx_word' id='word_1_8' title='bbox 347 380 412 410; x_wconf 86'>The</span> <span class='ocrx_word' id='word_1_9' title='bbox 424 380 676 417; x_wconf 85'>LinnSequencer</span> <span class='ocrx_word' id='word_1_10' title='bbox 688 380 712 411; x_wconf 87'>is</span> <span class='ocrx_word' id='word_1_11' title='bbox 724 390 743 411; x_wconf 90'>a</span> <span class='ocrx_word' id='word_1_12' title='bbox 754 381 1005 423; x_wconf 79'>state—of—the-art</span> <span class='ocrx_word' id='word_1_13' title='bbox 1017 380 1226 418; x_wconf 86'>composition</span> <span class='ocrx_word' id='word_1_14' title='bbox 1238 381 1299 411; x_wconf 88'>and</span> <span class='ocrx_word' id='word_1_15' title='bbox 1311 380 1525 418; x_wconf 85'>performance</span> <span class='ocrx_word' id='word_1_16' title='bbox 1536 380 1602 411; x_wconf 85'>tool</span> <span class='ocrx_word' id='word_1_17' title='bbox 1615 380 1663 411; x_wconf 82'>for</span> <span class='ocrx_word' id='word_1_18' title='bbox 1674 381 1725 410; x_wconf 94'>the</span> <span class='ocrx_word' id='word_1_19' title='bbox 1737 380 1940 417; x_wconf 85'>professional</span> <span class='ocrx_word' id='word_1_20' title='bbox 1952 380 2112 411; x_wconf 87'>musician.</span> <span class='ocrx_word' id='word_1_21' title='bbox 2127 381 2152 410; x_wconf 90'>It</span> <span class='ocrx_word' id='word_1_22' title='bbox 2164 380 2188 410; x_wconf 90'>is</span> 
-     </span>
-    </p>
-
-    <p class='ocr_par' id='par_1_3' lang='eng' title="bbox 347 430 1988 468">
-     <span class='ocr_line' id='line_1_4' title="bbox 347 430 1988 468; baseline 0 -8; x_size 37; x_descenders 7; x_ascenders 9"><span class='ocrx_word' id='word_1_23' title='bbox 347 430 507 467; x_wconf 86'>extremely</span> <span class='ocrx_word' id='word_1_24' title='bbox 518 430 677 467; x_wconf 84'>powerful,</span> <span class='ocrx_word' id='word_1_25' title='bbox 691 435 739 467; x_wconf 91'>yet</span> <span class='ocrx_word' id='word_1_26' title='bbox 752 430 924 468; x_wconf 87'>amazingly</span> <span class='ocrx_word' id='word_1_27' title='bbox 936 430 1044 468; x_wconf 85'>simple</span> <span class='ocrx_word' id='word_1_28' title='bbox 1055 435 1087 461; x_wconf 87'>to</span> <span class='ocrx_word' id='word_1_29' title='bbox 1099 431 1183 461; x_wconf 88'>learn</span> <span class='ocrx_word' id='word_1_30' title='bbox 1195 431 1257 461; x_wconf 90'>and</span> <span class='ocrx_word' id='word_1_31' title='bbox 1269 440 1329 461; x_wconf 89'>use.</span> <span class='ocrx_word' id='word_1_32' title='bbox 1344 431 1393 461; x_wconf 87'>It’s</span> <span class='ocrx_word' id='word_1_33' title='bbox 1406 440 1499 467; x_wconf 91'>many</span> <span class='ocrx_word' id='word_1_34' title='bbox 1510 430 1701 461; x_wconf 88'>remarkable</span> <span class='ocrx_word' id='word_1_35' title='bbox 1714 430 1843 460; x_wconf 85'>features</span> <span class='ocrx_word' id='word_1_36' title='bbox 1856 430 1988 461; x_wconf 87'>include:</span> 
-     </span>
-    </p>
-
-    <p class='ocr_par' id='par_1_4' lang='eng' title="bbox 350 482 2093 574">
-     <span class='ocr_line' id='line_1_5' title="bbox 350 482 2093 527; baseline 0 -9; x_size 43; x_descenders 7; x_ascenders 11"><span class='ocrx_word' id='word_1_37' title='bbox 350 490 368 508; x_wconf 68'><strong>0</strong></span> <span class='ocrx_word' id='word_1_38' title='bbox 383 482 585 526; x_wconf 88'>Operation</span> <span class='ocrx_word' id='word_1_39' title='bbox 598 482 627 518; x_wconf 89'>is</span> <span class='ocrx_word' id='word_1_40' title='bbox 641 482 776 518; x_wconf 88'>similar</span> <span class='ocrx_word' id='word_1_41' title='bbox 789 488 829 518; x_wconf 88'>to</span> <span class='ocrx_word' id='word_1_42' title='bbox 843 482 1062 519; x_wconf 86'>multi—track</span> <span class='ocrx_word' id='word_1_43' title='bbox 1076 488 1160 527; x_wconf 92'>tape</span> <span class='ocrx_word' id='word_1_44' title='bbox 1173 483 1336 519; x_wconf 88'>recorder</span> <span class='ocrx_word' id='word_1_45' title='bbox 1350 482 1436 518; x_wconf 87'>with</span> <span class='ocrx_word' id='word_1_46' title='bbox 1451 483 1580 525; x_wconf 90'>PLAY,</span> <span class='ocrx_word' id='word_1_47' title='bbox 1598 483 1724 525; x_wconf 90'>STOP,</span> <span class='ocrx_word' id='word_1_48' title='bbox 1741 483 1957 525; x_wconf 89'>RECORD,</span> <span class='ocrx_word' id='word_1_49' title='bbox 1974 483 2093 518; x_wconf 88'>FAST</span> 
-     </span>
-     <span class='ocr_line' id='line_1_6' title="bbox 383 532 1345 574; baseline 0.001 -7; x_size 43; x_descenders 7; x_ascenders 11"><span class='ocrx_word' id='word_1_50' title='bbox 383 532 635 574; x_wconf 88'>FORWARD,</span> <span class='ocrx_word' id='word_1_51' title='bbox 652 532 865 574; x_wconf 89'>REWIND,</span> <span class='ocrx_word' id='word_1_52' title='bbox 882 532 956 568; x_wconf 91'>and</span> <span class='ocrx_word' id='word_1_53' title='bbox 971 532 1163 568; x_wconf 90'>LOCATE</span> <span class='ocrx_word' id='word_1_54' title='bbox 1177 532 1345 568; x_wconf 87'>controls.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_3' title="bbox 349 589 2136 685">
-    <p class='ocr_par' id='par_1_5' lang='eng' title="bbox 349 589 2136 685">
-     <span class='ocr_line' id='line_1_7' title="bbox 349 589 2136 634; baseline 0.001 -9; x_size 44; x_descenders 8; x_ascenders 11"><span class='ocrx_word' id='word_1_55' title='bbox 349 597 368 615; x_wconf 67'><strong>0</strong></span> <span class='ocrx_word' id='word_1_56' title='bbox 383 590 482 625; x_wconf 90'>Each</span> <span class='ocrx_word' id='word_1_57' title='bbox 496 589 539 625; x_wconf 87'>of</span> <span class='ocrx_word' id='word_1_58' title='bbox 549 590 611 626; x_wconf 92'>the</span> <span class='ocrx_word' id='word_1_59' title='bbox 625 590 686 625; x_wconf 91'>100</span> <span class='ocrx_word' id='word_1_60' title='bbox 700 601 895 634; x_wconf 89'>sequences</span> <span class='ocrx_word' id='word_1_61' title='bbox 909 590 1075 626; x_wconf 88'>contains</span> <span class='ocrx_word' id='word_1_62' title='bbox 1090 591 1135 626; x_wconf 89'>32</span> <span class='ocrx_word' id='word_1_63' title='bbox 1149 590 1418 633; x_wconf 87'>simultaneous,</span> <span class='ocrx_word' id='word_1_64' title='bbox 1435 590 1661 634; x_wconf 89'>polyphonic</span> <span class='ocrx_word' id='word_1_65' title='bbox 1674 590 1802 626; x_wconf 85'>tracks.</span> <span class='ocrx_word' id='word_1_66' title='bbox 1819 590 1921 626; x_wconf 89'>Each</span> <span class='ocrx_word' id='word_1_67' title='bbox 1934 590 2035 626; x_wconf 88'>track</span> <span class='ocrx_word' id='word_1_68' title='bbox 2050 600 2136 634; x_wconf 89'>may</span> 
-     </span>
-     <span class='ocr_line' id='line_1_8' title="bbox 383 639 2022 685; baseline 0.001 -10; x_size 44; x_descenders 8; x_ascenders 11"><span class='ocrx_word' id='word_1_69' title='bbox 383 639 428 675; x_wconf 95'>be</span> <span class='ocrx_word' id='word_1_70' title='bbox 442 639 607 684; x_wconf 88'>assigned</span> <span class='ocrx_word' id='word_1_71' title='bbox 621 645 659 676; x_wconf 92'>to</span> <span class='ocrx_word' id='word_1_72' title='bbox 674 650 745 676; x_wconf 91'>one</span> <span class='ocrx_word' id='word_1_73' title='bbox 759 640 801 676; x_wconf 84'>of</span> <span class='ocrx_word' id='word_1_74' title='bbox 813 641 851 676; x_wconf 92'>16</span> <span class='ocrx_word' id='word_1_75' title='bbox 864 641 980 675; x_wconf 90'>MIDI</span> <span class='ocrx_word' id='word_1_76' title='bbox 996 640 1176 676; x_wconf 88'>channels.</span> <span class='ocrx_word' id='word_1_77' title='bbox 1194 640 1498 684; x_wconf 90'>Simultaneously</span> <span class='ocrx_word' id='word_1_78' title='bbox 1510 640 1609 685; x_wconf 90'>plays</span> <span class='ocrx_word' id='word_1_79' title='bbox 1624 651 1674 684; x_wconf 91'>up</span> <span class='ocrx_word' id='word_1_80' title='bbox 1688 645 1727 676; x_wconf 90'>to</span> <span class='ocrx_word' id='word_1_81' title='bbox 1742 641 1780 676; x_wconf 94'>16</span> <span class='ocrx_word' id='word_1_82' title='bbox 1794 640 2022 684; x_wconf 87'>polyphonic</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_4' title="bbox 384 689 632 734">
-    <p class='ocr_par' id='par_1_6' lang='eng' title="bbox 384 689 632 734">
-     <span class='ocr_line' id='line_1_9' title="bbox 384 689 632 734; baseline 0.004 -9; x_size 46; x_descenders 9; x_ascenders 12"><span class='ocrx_word' id='word_1_83' title='bbox 384 689 618 734; x_wconf 84'>synthesizers</span> <span class='ocrx_word' id='word_1_84' title='bbox 625 690 632 726; x_wconf 85'><strong>!</strong></span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_5' title="bbox 349 748 2117 793">
-    <p class='ocr_par' id='par_1_7' lang='eng' title="bbox 349 748 2117 793">
-     <span class='ocr_line' id='line_1_10' title="bbox 349 748 2117 793; baseline 0 -9; x_size 45; x_descenders 9; x_ascenders 11"><span class='ocrx_word' id='word_1_85' title='bbox 349 755 367 774; x_wconf 66'><strong>0</strong></span> <span class='ocrx_word' id='word_1_86' title='bbox 383 748 573 784; x_wconf 87'>Ultra—fast</span> <span class='ocrx_word' id='word_1_87' title='bbox 588 749 606 784; x_wconf 90'>3</span> <span class='ocrx_word' id='word_1_88' title='bbox 616 749 652 784; x_wconf 86'>1/2</span> <span class='ocrx_word' id='word_1_89' title='bbox 663 749 677 764; x_wconf 87'><em>”</em></span> <span class='ocrx_word' id='word_1_90' title='bbox 694 748 775 784; x_wconf 91'>disk</span> <span class='ocrx_word' id='word_1_91' title='bbox 790 748 887 784; x_wconf 89'>drive</span> <span class='ocrx_word' id='word_1_92' title='bbox 901 754 1012 785; x_wconf 85'>stores</span> <span class='ocrx_word' id='word_1_93' title='bbox 1027 748 1194 793; x_wconf 87'>complex</span> <span class='ocrx_word' id='word_1_94' title='bbox 1208 760 1316 793; x_wconf 87'>songs</span> <span class='ocrx_word' id='word_1_95' title='bbox 1330 748 1368 784; x_wconf 87'>in</span> <span class='ocrx_word' id='word_1_96' title='bbox 1382 748 1535 785; x_wconf 87'>seconds</span> <span class='ocrx_word' id='word_1_97' title='bbox 1550 748 1624 784; x_wconf 89'>and</span> <span class='ocrx_word' id='word_1_98' title='bbox 1638 748 1746 784; x_wconf 87'>holds</span> <span class='ocrx_word' id='word_1_99' title='bbox 1761 759 1844 784; x_wconf 85'>over</span> <span class='ocrx_word' id='word_1_100' title='bbox 1859 749 2000 791; x_wconf 88'>110,000</span> <span class='ocrx_word' id='word_1_101' title='bbox 2013 753 2117 784; x_wconf 89'>notes</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_6' title="bbox 383 797 554 842">
-    <p class='ocr_par' id='par_1_8' lang='eng' title="bbox 383 797 554 842">
-     <span class='ocr_line' id='line_1_11' title="bbox 383 797 554 842; baseline 0 -8; x_size 45; x_descenders 8; x_ascenders 12"><span class='ocrx_word' id='word_1_102' title='bbox 383 808 446 842; x_wconf 82'>per</span> <span class='ocrx_word' id='word_1_103' title='bbox 459 797 554 834; x_wconf 86'>disk!</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_7' title="bbox 349 855 2030 1016">
-    <p class='ocr_par' id='par_1_9' lang='eng' title="bbox 349 855 2030 1016">
-     <span class='ocr_line' id='line_1_12' title="bbox 350 855 1638 900; baseline 0.001 -9; x_size 45; x_descenders 9; x_ascenders 11"><span class='ocrx_word' id='word_1_104' title='bbox 350 863 367 881; x_wconf 71'><strong>0</strong></span> <span class='ocrx_word' id='word_1_105' title='bbox 383 856 464 891; x_wconf 90'>One</span> <span class='ocrx_word' id='word_1_106' title='bbox 478 866 520 891; x_wconf 87'>or</span> <span class='ocrx_word' id='word_1_107' title='bbox 534 855 580 891; x_wconf 88'>all</span> <span class='ocrx_word' id='word_1_108' title='bbox 594 856 712 892; x_wconf 89'>tracks</span> <span class='ocrx_word' id='word_1_109' title='bbox 726 867 811 900; x_wconf 90'>may</span> <span class='ocrx_word' id='word_1_110' title='bbox 823 856 869 892; x_wconf 88'>be</span> <span class='ocrx_word' id='word_1_111' title='bbox 882 856 1212 892; x_wconf 90'>TRANSPOSED</span> <span class='ocrx_word' id='word_1_112' title='bbox 1227 861 1264 892; x_wconf 90'>at</span> <span class='ocrx_word' id='word_1_113' title='bbox 1277 856 1338 892; x_wconf 85'>the</span> <span class='ocrx_word' id='word_1_114' title='bbox 1351 856 1463 892; x_wconf 90'>touch</span> <span class='ocrx_word' id='word_1_115' title='bbox 1477 856 1520 892; x_wconf 89'>of</span> <span class='ocrx_word' id='word_1_116' title='bbox 1531 867 1554 892; x_wconf 90'>a</span> <span class='ocrx_word' id='word_1_117' title='bbox 1568 856 1638 900; x_wconf 89'>key.</span> 
-     </span>
-     <span class='ocr_line' id='line_1_13' title="bbox 350 913 1535 958; baseline 0.001 -9; x_size 44; x_descenders 8; x_ascenders 11"><span class='ocrx_word' id='word_1_118' title='bbox 350 921 367 939; x_wconf 70'><strong>0</strong></span> <span class='ocrx_word' id='word_1_119' title='bbox 383 913 568 950; x_wconf 84'>Exclusive</span> <span class='ocrx_word' id='word_1_120' title='bbox 581 913 756 950; x_wconf 87'>real—time</span> <span class='ocrx_word' id='word_1_121' title='bbox 769 914 929 950; x_wconf 91'>ERASE</span> <span class='ocrx_word' id='word_1_122' title='bbox 944 914 1114 950; x_wconf 87'>function</span> <span class='ocrx_word' id='word_1_123' title='bbox 1128 914 1251 950; x_wconf 88'>makes</span> <span class='ocrx_word' id='word_1_124' title='bbox 1266 914 1400 958; x_wconf 87'>editing</span> <span class='ocrx_word' id='word_1_125' title='bbox 1414 915 1535 950; x_wconf 91'>FAST.</span> 
-     </span>
-     <span class='ocr_line' id='line_1_14' title="bbox 349 971 2030 1016; baseline 0.001 -10; x_size 44; x_descenders 8; x_ascenders 11"><span class='ocrx_word' id='word_1_126' title='bbox 349 979 367 997; x_wconf 67'><strong>0</strong></span> <span class='ocrx_word' id='word_1_127' title='bbox 382 971 568 1007; x_wconf 83'>Exclusive</span> <span class='ocrx_word' id='word_1_128' title='bbox 582 972 773 1007; x_wconf 87'>REPEAT</span> <span class='ocrx_word' id='word_1_129' title='bbox 787 972 958 1008; x_wconf 86'>function</span> <span class='ocrx_word' id='word_1_130' title='bbox 971 972 1245 1016; x_wconf 89'>automatically</span> <span class='ocrx_word' id='word_1_131' title='bbox 1258 977 1396 1016; x_wconf 90'>repeats</span> <span class='ocrx_word' id='word_1_132' title='bbox 1410 983 1481 1016; x_wconf 88'>any</span> <span class='ocrx_word' id='word_1_133' title='bbox 1493 972 1578 1008; x_wconf 91'>held</span> <span class='ocrx_word' id='word_1_134' title='bbox 1592 977 1694 1008; x_wconf 91'>notes</span> <span class='ocrx_word' id='word_1_135' title='bbox 1709 977 1746 1008; x_wconf 85'>at</span> <span class='ocrx_word' id='word_1_136' title='bbox 1760 982 1783 1008; x_wconf 89'>a</span> <span class='ocrx_word' id='word_1_137' title='bbox 1796 972 2030 1016; x_wconf 87'>pre-selected</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_8' title="bbox 382 1021 689 1065">
-    <p class='ocr_par' id='par_1_10' lang='eng' title="bbox 382 1021 689 1065">
-     <span class='ocr_line' id='line_1_15' title="bbox 382 1021 689 1065; baseline 0.003 -8; x_size 45; x_descenders 8; x_ascenders 12"><span class='ocrx_word' id='word_1_138' title='bbox 382 1021 564 1065; x_wconf 87'>rhythmic</span> <span class='ocrx_word' id='word_1_139' title='bbox 577 1021 689 1058; x_wconf 86'>value.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_9' title="bbox 349 1080 2174 1125">
-    <p class='ocr_par' id='par_1_11' lang='eng' title="bbox 349 1080 2174 1125">
-     <span class='ocr_line' id='line_1_16' title="bbox 349 1080 2174 1125; baseline 0.001 -11; x_size 45; x_descenders 9; x_ascenders 11"><span class='ocrx_word' id='word_1_140' title='bbox 349 1087 367 1105; x_wconf 73'><strong>0</strong></span> <span class='ocrx_word' id='word_1_141' title='bbox 382 1080 567 1115; x_wconf 87'>TIMING</span> <span class='ocrx_word' id='word_1_142' title='bbox 582 1080 908 1116; x_wconf 90'>CORRECTION</span> <span class='ocrx_word' id='word_1_143' title='bbox 921 1080 1041 1116; x_wconf 90'>works</span> <span class='ocrx_word' id='word_1_144' title='bbox 1056 1080 1186 1124; x_wconf 86'>during</span> <span class='ocrx_word' id='word_1_145' title='bbox 1199 1080 1378 1124; x_wconf 89'>playback</span> <span class='ocrx_word' id='word_1_146' title='bbox 1392 1080 1466 1116; x_wconf 90'>and</span> <span class='ocrx_word' id='word_1_147' title='bbox 1480 1085 1644 1124; x_wconf 87'>operates</span> <span class='ocrx_word' id='word_1_148' title='bbox 1658 1080 1814 1116; x_wconf 91'>without</span> <span class='ocrx_word' id='word_1_149' title='bbox 1831 1080 2044 1125; x_wconf 87'>‘chopping’</span> <span class='ocrx_word' id='word_1_150' title='bbox 2061 1085 2174 1116; x_wconf 88'>notes.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_10' title="bbox 349 1137 1287 1182">
-    <p class='ocr_par' id='par_1_12' lang='eng' title="bbox 349 1137 1287 1182">
-     <span class='ocr_line' id='line_1_17' title="bbox 349 1137 1287 1182; baseline 0.001 -9; x_size 45; x_descenders 9; x_ascenders 11"><span class='ocrx_word' id='word_1_151' title='bbox 349 1145 367 1163; x_wconf 73'><strong>0</strong></span> <span class='ocrx_word' id='word_1_152' title='bbox 382 1137 560 1182; x_wconf 85'>Optional</span> <span class='ocrx_word' id='word_1_153' title='bbox 575 1138 739 1174; x_wconf 89'>SMPTE</span> <span class='ocrx_word' id='word_1_154' title='bbox 752 1138 839 1174; x_wconf 89'>time</span> <span class='ocrx_word' id='word_1_155' title='bbox 853 1138 945 1174; x_wconf 89'>code</span> <span class='ocrx_word' id='word_1_156' title='bbox 959 1138 1287 1182; x_wconf 88'>synchronization.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_11' title="bbox 349 1195 874 1240">
-    <p class='ocr_par' id='par_1_13' lang='eng' title="bbox 349 1195 874 1240">
-     <span class='ocr_line' id='line_1_18' title="bbox 349 1195 874 1240; baseline 0 -8; x_size 45; x_descenders 8; x_ascenders 12"><span class='ocrx_word' id='word_1_157' title='bbox 349 1203 367 1222; x_wconf 70'><strong>0</strong></span> <span class='ocrx_word' id='word_1_158' title='bbox 382 1195 560 1240; x_wconf 87'>Optional</span> <span class='ocrx_word' id='word_1_159' title='bbox 573 1201 709 1233; x_wconf 86'>remote</span> <span class='ocrx_word' id='word_1_160' title='bbox 723 1196 874 1233; x_wconf 87'>control.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_12' title="bbox 346 1288 1239 1491">
-    <p class='ocr_par' id='par_1_14' lang='eng' title="bbox 346 1288 749 1329">
-     <span class='ocr_line' id='line_1_19' title="bbox 346 1288 749 1329; baseline 0.002 -9; x_size 42; x_descenders 9; x_ascenders 10"><span class='ocrx_word' id='word_1_161' title='bbox 346 1288 535 1329; x_wconf 82'>Recording</span> <span class='ocrx_word' id='word_1_162' title='bbox 547 1298 567 1321; x_wconf 73'>3</span> <span class='ocrx_word' id='word_1_163' title='bbox 579 1288 749 1328; x_wconf 86'>Sequence</span> 
-     </span>
-    </p>
-
-    <p class='ocr_par' id='par_1_15' lang='eng' title="bbox 346 1339 1239 1491">
-     <span class='ocr_line' id='line_1_20' title="bbox 383 1339 1239 1373; baseline 0 -7; x_size 33; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_164' title='bbox 383 1340 420 1366; x_wconf 87'>To</span> <span class='ocrx_word' id='word_1_165' title='bbox 430 1339 524 1366; x_wconf 85'>record</span> <span class='ocrx_word' id='word_1_166' title='bbox 535 1347 551 1366; x_wconf 87'>a</span> <span class='ocrx_word' id='word_1_167' title='bbox 562 1347 704 1373; x_wconf 80'>sequence,</span> <span class='ocrx_word' id='word_1_168' title='bbox 716 1340 815 1373; x_wconf 91'>simply</span> <span class='ocrx_word' id='word_1_169' title='bbox 825 1348 898 1373; x_wconf 88'>press</span> <span class='ocrx_word' id='word_1_170' title='bbox 910 1340 1065 1367; x_wconf 90'>RECORD</span> <span class='ocrx_word' id='word_1_171' title='bbox 1075 1340 1131 1367; x_wconf 87'>and</span> <span class='ocrx_word' id='word_1_172' title='bbox 1142 1341 1239 1372; x_wconf 87'>PLAY,</span> 
-     </span>
-     <span class='ocr_line' id='line_1_21' title="bbox 346 1378 1205 1412; baseline 0.001 -7; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_173' title='bbox 346 1379 411 1406; x_wconf 91'>then</span> <span class='ocrx_word' id='word_1_174' title='bbox 422 1378 483 1412; x_wconf 89'>play</span> <span class='ocrx_word' id='word_1_175' title='bbox 493 1387 562 1412; x_wconf 87'>your</span> <span class='ocrx_word' id='word_1_176' title='bbox 572 1379 659 1405; x_wconf 89'>MIDI</span> <span class='ocrx_word' id='word_1_177' title='bbox 671 1379 810 1412; x_wconf 86'>keyboard</span> <span class='ocrx_word' id='word_1_178' title='bbox 821 1379 848 1406; x_wconf 90'>in</span> <span class='ocrx_word' id='word_1_179' title='bbox 858 1379 923 1406; x_wconf 90'>time</span> <span class='ocrx_word' id='word_1_180' title='bbox 934 1384 963 1406; x_wconf 90'>to</span> <span class='ocrx_word' id='word_1_181' title='bbox 974 1379 1019 1406; x_wconf 92'>the</span> <span class='ocrx_word' id='word_1_182' title='bbox 1030 1379 1205 1412; x_wconf 88'>Sequencer’s</span> 
-     </span>
-     <span class='ocr_line' id='line_1_22' title="bbox 347 1417 1232 1452; baseline 0.002 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_183' title='bbox 347 1417 414 1445; x_wconf 86'>click</span> <span class='ocrx_word' id='word_1_184' title='bbox 425 1418 506 1445; x_wconf 84'>track.</span> <span class='ocrx_word' id='word_1_185' title='bbox 519 1418 609 1445; x_wconf 85'>When</span> <span class='ocrx_word' id='word_1_186' title='bbox 620 1419 665 1445; x_wconf 89'>the</span> <span class='ocrx_word' id='word_1_187' title='bbox 676 1426 810 1452; x_wconf 86'>sequence</span> <span class='ocrx_word' id='word_1_188' title='bbox 821 1419 901 1452; x_wconf 88'>loops</span> <span class='ocrx_word' id='word_1_189' title='bbox 912 1419 983 1446; x_wconf 86'>back</span> <span class='ocrx_word' id='word_1_190' title='bbox 995 1419 1101 1446; x_wconf 90'>around</span> <span class='ocrx_word' id='word_1_191' title='bbox 1112 1423 1141 1446; x_wconf 91'>to</span> <span class='ocrx_word' id='word_1_192' title='bbox 1152 1419 1201 1446; x_wconf 87'>bar</span> <span class='ocrx_word' id='word_1_193' title='bbox 1213 1419 1232 1450; x_wconf 88'>1,</span> 
-     </span>
-     <span class='ocr_line' id='line_1_23' title="bbox 346 1457 1223 1491; baseline 0.001 -7; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_194' title='bbox 346 1457 430 1490; x_wconf 84'>you’ll</span> <span class='ocrx_word' id='word_1_195' title='bbox 441 1457 506 1485; x_wconf 93'>hear</span> <span class='ocrx_word' id='word_1_196' title='bbox 517 1458 590 1485; x_wconf 79'>what</span> <span class='ocrx_word' id='word_1_197' title='bbox 600 1466 654 1491; x_wconf 88'>you</span> <span class='ocrx_word' id='word_1_198' title='bbox 666 1458 865 1491; x_wconf 77'>played—only</span> <span class='ocrx_word' id='word_1_199' title='bbox 876 1458 910 1485; x_wconf 90'>all</span> <span class='ocrx_word' id='word_1_200' title='bbox 921 1458 1018 1491; x_wconf 84'>timing</span> <span class='ocrx_word' id='word_1_201' title='bbox 1029 1466 1115 1485; x_wconf 84'>errors</span> <span class='ocrx_word' id='word_1_202' title='bbox 1126 1458 1179 1485; x_wconf 91'>will</span> <span class='ocrx_word' id='word_1_203' title='bbox 1189 1458 1223 1485; x_wconf 91'>be</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_13' title="bbox 346 1497 1245 1531">
-    <p class='ocr_par' id='par_1_16' lang='eng' title="bbox 346 1497 1245 1531">
-     <span class='ocr_line' id='line_1_24' title="bbox 346 1497 1245 1531; baseline 0.001 -7; x_size 35; x_descenders 7; x_ascenders 9"><span class='ocrx_word' id='word_1_204' title='bbox 346 1497 494 1524; x_wconf 86'>corrected!</span> <span class='ocrx_word' id='word_1_205' title='bbox 508 1497 628 1530; x_wconf 85'>(Timing</span> <span class='ocrx_word' id='word_1_206' title='bbox 638 1497 791 1525; x_wconf 82'>correction</span> <span class='ocrx_word' id='word_1_207' title='bbox 802 1506 864 1531; x_wconf 88'>may</span> <span class='ocrx_word' id='word_1_208' title='bbox 875 1498 909 1525; x_wconf 91'>be</span> <span class='ocrx_word' id='word_1_209' title='bbox 920 1497 1047 1531; x_wconf 81'>adjusted</span> <span class='ocrx_word' id='word_1_210' title='bbox 1058 1505 1089 1525; x_wconf 92'>or</span> <span class='ocrx_word' id='word_1_211' title='bbox 1099 1497 1245 1531; x_wconf 87'>defeated).</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_14' title="bbox 346 1536 1205 1965">
-    <p class='ocr_par' id='par_1_17' lang='eng' title="bbox 346 1536 1186 1610">
-     <span class='ocr_line' id='line_1_25' title="bbox 346 1536 1186 1570; baseline 0.001 -7; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_212' title='bbox 346 1537 409 1569; x_wconf 90'>Any</span> <span class='ocrx_word' id='word_1_213' title='bbox 419 1536 571 1563; x_wconf 86'>additional</span> <span class='ocrx_word' id='word_1_214' title='bbox 582 1541 659 1564; x_wconf 85'>notes</span> <span class='ocrx_word' id='word_1_215' title='bbox 670 1537 766 1570; x_wconf 87'>played</span> <span class='ocrx_word' id='word_1_216' title='bbox 777 1537 830 1563; x_wconf 88'>will</span> <span class='ocrx_word' id='word_1_217' title='bbox 840 1537 874 1564; x_wconf 92'>be</span> <span class='ocrx_word' id='word_1_218' title='bbox 885 1537 976 1564; x_wconf 88'>added</span> <span class='ocrx_word' id='word_1_219' title='bbox 986 1537 1045 1564; x_wconf 89'>into</span> <span class='ocrx_word' id='word_1_220' title='bbox 1055 1537 1101 1564; x_wconf 89'>the</span> <span class='ocrx_word' id='word_1_221' title='bbox 1111 1537 1186 1564; x_wconf 82'>track</span> 
-     </span>
-     <span class='ocr_line' id='line_1_26' title="bbox 347 1575 1052 1610; baseline 0.001 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_222' title='bbox 347 1575 495 1609; x_wconf 86'>—existing</span> <span class='ocrx_word' id='word_1_223' title='bbox 505 1580 582 1603; x_wconf 89'>notes</span> <span class='ocrx_word' id='word_1_224' title='bbox 593 1584 637 1603; x_wconf 90'>are</span> <span class='ocrx_word' id='word_1_225' title='bbox 648 1580 696 1603; x_wconf 88'>not</span> <span class='ocrx_word' id='word_1_226' title='bbox 707 1576 800 1603; x_wconf 90'>erased</span> <span class='ocrx_word' id='word_1_227' title='bbox 811 1576 890 1603; x_wconf 87'>while</span> <span class='ocrx_word' id='word_1_228' title='bbox 900 1576 1052 1610; x_wconf 87'>recording!</span> 
-     </span>
-    </p>
-
-    <p class='ocr_par' id='par_1_18' lang='eng' title="bbox 346 1616 1205 1965">
-     <span class='ocr_line' id='line_1_27' title="bbox 384 1616 1199 1648; baseline 0.001 -6; x_size 32; x_descenders 5; x_ascenders 8"><span class='ocrx_word' id='word_1_229' title='bbox 384 1616 471 1642; x_wconf 88'>FAST</span> <span class='ocrx_word' id='word_1_230' title='bbox 481 1616 671 1648; x_wconf 88'>FORWARD,</span> <span class='ocrx_word' id='word_1_231' title='bbox 684 1617 844 1648; x_wconf 87'>REWIND,</span> <span class='ocrx_word' id='word_1_232' title='bbox 857 1616 912 1643; x_wconf 89'>and</span> <span class='ocrx_word' id='word_1_233' title='bbox 924 1616 1068 1643; x_wconf 78'>LOCATE</span> <span class='ocrx_word' id='word_1_234' title='bbox 1079 1616 1199 1643; x_wconf 87'>controls</span> 
-     </span>
-     <span class='ocr_line' id='line_1_28' title="bbox 346 1655 1202 1689; baseline 0 -7; x_size 34; x_descenders 6; x_ascenders 9"><span class='ocrx_word' id='word_1_235' title='bbox 346 1663 409 1688; x_wconf 89'>may</span> <span class='ocrx_word' id='word_1_236' title='bbox 419 1655 453 1682; x_wconf 94'>be</span> <span class='ocrx_word' id='word_1_237' title='bbox 463 1655 530 1682; x_wconf 90'>used</span> <span class='ocrx_word' id='word_1_238' title='bbox 541 1659 569 1682; x_wconf 90'>at</span> <span class='ocrx_word' id='word_1_239' title='bbox 580 1663 632 1688; x_wconf 86'>any</span> <span class='ocrx_word' id='word_1_240' title='bbox 642 1655 707 1683; x_wconf 88'>time</span> <span class='ocrx_word' id='word_1_241' title='bbox 717 1659 746 1683; x_wconf 88'>to</span> <span class='ocrx_word' id='word_1_242' title='bbox 757 1655 867 1689; x_wconf 88'>quickly</span> <span class='ocrx_word' id='word_1_243' title='bbox 878 1664 968 1683; x_wconf 87'>access</span> <span class='ocrx_word' id='word_1_244' title='bbox 979 1664 1032 1689; x_wconf 87'>any</span> <span class='ocrx_word' id='word_1_245' title='bbox 1042 1655 1164 1683; x_wconf 90'>location</span> <span class='ocrx_word' id='word_1_246' title='bbox 1174 1655 1202 1682; x_wconf 94'>in</span> 
-     </span>
-     <span class='ocr_line' id='line_1_29' title="bbox 346 1694 1204 1728; baseline 0.001 -7; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_247' title='bbox 346 1702 414 1727; x_wconf 88'>your</span> <span class='ocrx_word' id='word_1_248' title='bbox 424 1702 558 1727; x_wconf 87'>sequence</span> <span class='ocrx_word' id='word_1_249' title='bbox 570 1694 612 1721; x_wconf 84'>for</span> <span class='ocrx_word' id='word_1_250' title='bbox 623 1695 847 1728; x_wconf 84'>spot—recording.</span> <span class='ocrx_word' id='word_1_251' title='bbox 860 1696 897 1722; x_wconf 89'>To</span> <span class='ocrx_word' id='word_1_252' title='bbox 908 1695 1028 1722; x_wconf 82'>overdub</span> <span class='ocrx_word' id='word_1_253' title='bbox 1039 1703 1056 1722; x_wconf 91'>a</span> <span class='ocrx_word' id='word_1_254' title='bbox 1066 1703 1125 1722; x_wconf 89'>new</span> <span class='ocrx_word' id='word_1_255' title='bbox 1135 1699 1204 1728; x_wconf 88'>part,</span> 
-     </span>
-     <span class='ocr_line' id='line_1_30' title="bbox 347 1733 1150 1768; baseline 0.001 -8; x_size 35; x_descenders 7; x_ascenders 9"><span class='ocrx_word' id='word_1_256' title='bbox 347 1733 426 1761; x_wconf 88'>select</span> <span class='ocrx_word' id='word_1_257' title='bbox 437 1741 453 1761; x_wconf 88'>a</span> <span class='ocrx_word' id='word_1_258' title='bbox 463 1733 589 1761; x_wconf 86'>different</span> <span class='ocrx_word' id='word_1_259' title='bbox 599 1734 674 1761; x_wconf 85'>track</span> <span class='ocrx_word' id='word_1_260' title='bbox 685 1734 741 1761; x_wconf 87'>and</span> <span class='ocrx_word' id='word_1_261' title='bbox 752 1738 818 1761; x_wconf 83'>start</span> <span class='ocrx_word' id='word_1_262' title='bbox 828 1734 1086 1768; x_wconf 63'>recording~while</span> <span class='ocrx_word' id='word_1_263' title='bbox 1096 1742 1150 1767; x_wconf 89'>you</span> 
-     </span>
-     <span class='ocr_line' id='line_1_31' title="bbox 346 1773 1203 1808; baseline 0.001 -8; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_264' title='bbox 346 1774 448 1806; x_wconf 86'>record,</span> <span class='ocrx_word' id='word_1_265' title='bbox 460 1774 506 1801; x_wconf 87'>the</span> <span class='ocrx_word' id='word_1_266' title='bbox 517 1773 662 1801; x_wconf 64'>first‘track</span> <span class='ocrx_word' id='word_1_267' title='bbox 673 1774 726 1801; x_wconf 91'>will</span> <span class='ocrx_word' id='word_1_268' title='bbox 736 1774 799 1807; x_wconf 89'>play</span> <span class='ocrx_word' id='word_1_269' title='bbox 809 1774 836 1801; x_wconf 91'>in</span> <span class='ocrx_word' id='word_1_270' title='bbox 847 1774 949 1808; x_wconf 84'>perfect</span> <span class='ocrx_word' id='word_1_271' title='bbox 961 1782 1026 1808; x_wconf 87'>sync</span> <span class='ocrx_word' id='word_1_272' title='bbox 1037 1774 1137 1808; x_wconf 89'>(unless</span> <span class='ocrx_word' id='word_1_273' title='bbox 1148 1782 1203 1807; x_wconf 93'>you</span> 
-     </span>
-     <span class='ocr_line' id='line_1_32' title="bbox 346 1813 1191 1847; baseline 0.002 -8; x_size 33; x_descenders 5; x_ascenders 9"><span class='ocrx_word' id='word_1_274' title='bbox 346 1813 454 1840; x_wconf 86'>MUTE</span> <span class='ocrx_word' id='word_1_275' title='bbox 464 1813 492 1845; x_wconf 88'>it,</span> <span class='ocrx_word' id='word_1_276' title='bbox 505 1821 537 1840; x_wconf 84'>or</span> <span class='ocrx_word' id='word_1_277' title='bbox 547 1813 642 1840; x_wconf 88'>SOLO</span> <span class='ocrx_word' id='word_1_278' title='bbox 653 1814 769 1841; x_wconf 87'>another</span> <span class='ocrx_word' id='word_1_279' title='bbox 779 1814 875 1847; x_wconf 86'>track).</span> <span class='ocrx_word' id='word_1_280' title='bbox 889 1814 920 1840; x_wconf 86'>In</span> <span class='ocrx_word' id='word_1_281' title='bbox 930 1813 984 1841; x_wconf 86'>this</span> <span class='ocrx_word' id='word_1_282' title='bbox 995 1822 1057 1847; x_wconf 87'>way,</span> <span class='ocrx_word' id='word_1_283' title='bbox 1069 1822 1106 1847; x_wconf 88'>up</span> <span class='ocrx_word' id='word_1_284' title='bbox 1117 1818 1146 1841; x_wconf 77'>to</span> <span class='ocrx_word' id='word_1_285' title='bbox 1158 1814 1191 1841; x_wconf 90'>32</span> 
-     </span>
-     <span class='ocr_line' id='line_1_33' title="bbox 346 1852 1205 1885; baseline 0.001 -6; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_286' title='bbox 346 1852 434 1879; x_wconf 84'>tracks</span> <span class='ocrx_word' id='word_1_287' title='bbox 445 1860 508 1885; x_wconf 90'>may</span> <span class='ocrx_word' id='word_1_288' title='bbox 518 1853 552 1879; x_wconf 93'>be</span> <span class='ocrx_word' id='word_1_289' title='bbox 562 1853 748 1880; x_wconf 90'>overdubbed!</span> <span class='ocrx_word' id='word_1_290' title='bbox 761 1853 808 1880; x_wconf 92'>All</span> <span class='ocrx_word' id='word_1_291' title='bbox 819 1854 905 1880; x_wconf 81'>MIDI</span> <span class='ocrx_word' id='word_1_292' title='bbox 917 1853 1011 1880; x_wconf 81'>effects</span> <span class='ocrx_word' id='word_1_293' title='bbox 1022 1861 1067 1880; x_wconf 86'>are</span> <span class='ocrx_word' id='word_1_294' title='bbox 1076 1853 1205 1880; x_wconf 87'>recorded</span> 
-     </span>
-     <span class='ocr_line' id='line_1_34' title="bbox 346 1891 1168 1926; baseline 0.002 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_295' title='bbox 346 1891 485 1925; x_wconf 84'>including</span> <span class='ocrx_word' id='word_1_296' title='bbox 495 1891 570 1925; x_wconf 87'>pitch</span> <span class='ocrx_word' id='word_1_297' title='bbox 580 1892 663 1924; x_wconf 84'>bend,</span> <span class='ocrx_word' id='word_1_298' title='bbox 675 1892 859 1924; x_wconf 86'>modulation,</span> <span class='ocrx_word' id='word_1_299' title='bbox 872 1892 991 1926; x_wconf 83'>velocity,</span> <span class='ocrx_word' id='word_1_300' title='bbox 1004 1892 1168 1924; x_wconf 85'>aftertouch,</span> 
-     </span>
-     <span class='ocr_line' id='line_1_35' title="bbox 346 1931 895 1965; baseline 0.002 -7; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_301' title='bbox 346 1931 448 1958; x_wconf 88'>sustain</span> <span class='ocrx_word' id='word_1_302' title='bbox 459 1931 549 1965; x_wconf 88'>pedal,</span> <span class='ocrx_word' id='word_1_303' title='bbox 561 1932 617 1959; x_wconf 90'>and</span> <span class='ocrx_word' id='word_1_304' title='bbox 628 1940 756 1965; x_wconf 88'>program</span> <span class='ocrx_word' id='word_1_305' title='bbox 766 1932 895 1965; x_wconf 87'>changes!</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_15' title="bbox 346 1998 1212 2163">
-    <p class='ocr_par' id='par_1_19' lang='eng' title="bbox 346 1998 479 2040">
-     <span class='ocr_line' id='line_1_36' title="bbox 346 1998 479 2040; baseline -0.008 -9; x_size 43; x_descenders 10; x_ascenders 11"><span class='ocrx_word' id='word_1_306' title='bbox 346 1998 479 2040; x_wconf 87'>Editing</span> 
-     </span>
-    </p>
-
-    <p class='ocr_par' id='par_1_20' lang='eng' title="bbox 346 2050 1212 2163">
-     <span class='ocr_line' id='line_1_37' title="bbox 383 2050 1186 2084; baseline 0.001 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_307' title='bbox 383 2050 419 2076; x_wconf 89'>To</span> <span class='ocrx_word' id='word_1_308' title='bbox 430 2057 503 2076; x_wconf 87'>erase</span> <span class='ocrx_word' id='word_1_309' title='bbox 514 2058 530 2077; x_wconf 91'>a</span> <span class='ocrx_word' id='word_1_310' title='bbox 540 2058 634 2083; x_wconf 88'>wrong</span> <span class='ocrx_word' id='word_1_311' title='bbox 644 2054 717 2082; x_wconf 84'>note,</span> <span class='ocrx_word' id='word_1_312' title='bbox 729 2050 829 2083; x_wconf 89'>simply</span> <span class='ocrx_word' id='word_1_313' title='bbox 839 2050 905 2077; x_wconf 88'>hold</span> <span class='ocrx_word' id='word_1_314' title='bbox 916 2051 1037 2077; x_wconf 85'>ERASE</span> <span class='ocrx_word' id='word_1_315' title='bbox 1048 2050 1103 2077; x_wconf 89'>and</span> <span class='ocrx_word' id='word_1_316' title='bbox 1113 2059 1186 2084; x_wconf 89'>press</span> 
-     </span>
-     <span class='ocr_line' id='line_1_38' title="bbox 346 2089 1212 2124; baseline 0.002 -8; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_317' title='bbox 346 2089 391 2116; x_wconf 86'>the</span> <span class='ocrx_word' id='word_1_318' title='bbox 402 2094 465 2117; x_wconf 86'>note</span> <span class='ocrx_word' id='word_1_319' title='bbox 475 2094 504 2117; x_wconf 86'>to</span> <span class='ocrx_word' id='word_1_320' title='bbox 515 2090 549 2117; x_wconf 93'>be</span> <span class='ocrx_word' id='word_1_321' title='bbox 559 2090 652 2117; x_wconf 89'>erased</span> <span class='ocrx_word' id='word_1_322' title='bbox 661 2090 718 2123; x_wconf 91'>just</span> <span class='ocrx_word' id='word_1_323' title='bbox 729 2090 822 2117; x_wconf 85'>before</span> <span class='ocrx_word' id='word_1_324' title='bbox 833 2090 852 2117; x_wconf 89'>it</span> <span class='ocrx_word' id='word_1_325' title='bbox 862 2090 937 2124; x_wconf 87'>plays</span> <span class='ocrx_word' id='word_1_326' title='bbox 947 2090 975 2117; x_wconf 92'>in</span> <span class='ocrx_word' id='word_1_327' title='bbox 986 2090 1032 2118; x_wconf 87'>the</span> <span class='ocrx_word' id='word_1_328' title='bbox 1043 2098 1212 2124; x_wconf 78'>sequence—</span> 
-     </span>
-     <span class='ocr_line' id='line_1_39' title="bbox 346 2129 1134 2163; baseline 0.003 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_329' title='bbox 346 2129 425 2156; x_wconf 92'>when</span> <span class='ocrx_word' id='word_1_330' title='bbox 435 2129 531 2162; x_wconf 91'>played</span> <span class='ocrx_word' id='word_1_331' title='bbox 542 2129 621 2161; x_wconf 82'>back,</span> <span class='ocrx_word' id='word_1_332' title='bbox 633 2129 652 2156; x_wconf 84'>it</span> <span class='ocrx_word' id='word_1_333' title='bbox 663 2129 716 2156; x_wconf 89'>will</span> <span class='ocrx_word' id='word_1_334' title='bbox 726 2130 760 2157; x_wconf 88'>be</span> <span class='ocrx_word' id='word_1_335' title='bbox 771 2138 849 2163; x_wconf 86'>gone.</span> <span class='ocrx_word' id='word_1_336' title='bbox 861 2131 947 2157; x_wconf 86'>Notes</span> <span class='ocrx_word' id='word_1_337' title='bbox 958 2138 1021 2163; x_wconf 88'>may</span> <span class='ocrx_word' id='word_1_338' title='bbox 1031 2130 1089 2157; x_wconf 87'>also</span> <span class='ocrx_word' id='word_1_339' title='bbox 1100 2130 1134 2157; x_wconf 91'>be</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_16' title="bbox 345 2168 1228 2242">
-    <p class='ocr_par' id='par_1_21' lang='eng' title="bbox 345 2168 1228 2242">
-     <span class='ocr_line' id='line_1_40' title="bbox 346 2168 1220 2203; baseline 0.001 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_340' title='bbox 346 2168 445 2200; x_wconf 86'>added,</span> <span class='ocrx_word' id='word_1_341' title='bbox 457 2168 560 2200; x_wconf 87'>erased,</span> <span class='ocrx_word' id='word_1_342' title='bbox 572 2177 604 2196; x_wconf 93'>or</span> <span class='ocrx_word' id='word_1_343' title='bbox 614 2169 739 2202; x_wconf 82'>changed</span> <span class='ocrx_word' id='word_1_344' title='bbox 749 2169 829 2203; x_wconf 86'>using</span> <span class='ocrx_word' id='word_1_345' title='bbox 839 2169 885 2196; x_wconf 85'>the</span> <span class='ocrx_word' id='word_1_346' title='bbox 896 2170 1031 2196; x_wconf 89'>SINGLE</span> <span class='ocrx_word' id='word_1_347' title='bbox 1042 2170 1131 2196; x_wconf 87'>STEP</span> <span class='ocrx_word' id='word_1_348' title='bbox 1143 2169 1220 2196; x_wconf 82'>func-</span> 
-     </span>
-     <span class='ocr_line' id='line_1_41' title="bbox 345 2207 1228 2242; baseline 0.002 -8; x_size 35; x_descenders 7; x_ascenders 9"><span class='ocrx_word' id='word_1_349' title='bbox 345 2207 412 2234; x_wconf 87'>tion.</span> <span class='ocrx_word' id='word_1_350' title='bbox 424 2208 461 2235; x_wconf 88'>To</span> <span class='ocrx_word' id='word_1_351' title='bbox 472 2208 592 2235; x_wconf 88'>overdub</span> <span class='ocrx_word' id='word_1_352' title='bbox 603 2212 680 2235; x_wconf 84'>notes</span> <span class='ocrx_word' id='word_1_353' title='bbox 691 2212 718 2235; x_wconf 89'>at</span> <span class='ocrx_word' id='word_1_354' title='bbox 729 2208 841 2242; x_wconf 83'>specific</span> <span class='ocrx_word' id='word_1_355' title='bbox 851 2209 943 2242; x_wconf 89'>points</span> <span class='ocrx_word' id='word_1_356' title='bbox 955 2208 1049 2236; x_wconf 87'>within</span> <span class='ocrx_word' id='word_1_357' title='bbox 1060 2217 1076 2236; x_wconf 87'>a</span> <span class='ocrx_word' id='word_1_358' title='bbox 1086 2216 1228 2242; x_wconf 86'>sequence,</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_17' title="bbox 345 2293 705 2326">
-    <p class='ocr_par' id='par_1_22' lang='eng' title="bbox 345 2293 705 2326">
-     <span class='ocr_line' id='line_1_42' title="bbox 345 2293 705 2326; baseline 0.003 -1; x_size 38.38298; x_descenders 5.3829789; x_ascenders 10"><span class='ocrx_word' id='word_1_359' title='bbox 345 2293 541 2326; x_wconf 86'>Additional</span> <span class='ocrx_word' id='word_1_360' title='bbox 553 2293 705 2326; x_wconf 83'>Features</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_18' title="bbox 1295 1289 2165 1593">
-    <p class='ocr_par' id='par_1_23' lang='eng' title="bbox 1296 1289 2165 1362">
-     <span class='ocr_line' id='line_1_43' title="bbox 1296 1289 2165 1322; baseline 0.001 -7; x_size 35.040108; x_descenders 6.9264708; x_ascenders 8.613636"><span class='ocrx_word' id='word_1_361' title='bbox 1296 1289 1394 1322; x_wconf 91'>simply</span> <span class='ocrx_word' id='word_1_362' title='bbox 1404 1297 1452 1316; x_wconf 87'>use</span> <span class='ocrx_word' id='word_1_363' title='bbox 1463 1290 1615 1321; x_wconf 86'>LOCATE,</span> <span class='ocrx_word' id='word_1_364' title='bbox 1628 1290 1716 1316; x_wconf 88'>FAST</span> <span class='ocrx_word' id='word_1_365' title='bbox 1726 1289 1917 1321; x_wconf 86'>FORWARD,</span> <span class='ocrx_word' id='word_1_366' title='bbox 1929 1297 1960 1317; x_wconf 92'>or</span> <span class='ocrx_word' id='word_1_367' title='bbox 1972 1290 2126 1316; x_wconf 91'>REWIND</span> <span class='ocrx_word' id='word_1_368' title='bbox 2136 1294 2165 1317; x_wconf 82'>to</span> 
-     </span>
-     <span class='ocr_line' id='line_1_44' title="bbox 1297 1328 2033 1362; baseline 0.001 -7; x_size 32; x_descenders 5; x_ascenders 8"><span class='ocrx_word' id='word_1_369' title='bbox 1297 1328 1356 1355; x_wconf 77'>find</span> <span class='ocrx_word' id='word_1_370' title='bbox 1366 1329 1412 1355; x_wconf 89'>the</span> <span class='ocrx_word' id='word_1_371' title='bbox 1423 1328 1527 1356; x_wconf 85'>desired</span> <span class='ocrx_word' id='word_1_372' title='bbox 1537 1329 1587 1356; x_wconf 88'>bar</span> <span class='ocrx_word' id='word_1_373' title='bbox 1598 1329 1720 1361; x_wconf 86'>number,</span> <span class='ocrx_word' id='word_1_374' title='bbox 1732 1329 1798 1356; x_wconf 78'>then</span> <span class='ocrx_word' id='word_1_375' title='bbox 1809 1333 1874 1356; x_wconf 87'>start</span> <span class='ocrx_word' id='word_1_376' title='bbox 1885 1329 2033 1362; x_wconf 85'>recording.</span> 
-     </span>
-    </p>
-
-    <p class='ocr_par' id='par_1_24' lang='eng' title="bbox 1295 1368 2160 1593">
-     <span class='ocr_line' id='line_1_45' title="bbox 1332 1368 2160 1402; baseline 0.002 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_377' title='bbox 1332 1368 1391 1395; x_wconf 90'>The</span> <span class='ocrx_word' id='word_1_378' title='bbox 1402 1368 1650 1395; x_wconf 86'>INSERT/COPY</span> <span class='ocrx_word' id='word_1_379' title='bbox 1662 1368 1789 1395; x_wconf 81'>function</span> <span class='ocrx_word' id='word_1_380' title='bbox 1800 1368 1893 1395; x_wconf 90'>allows</span> <span class='ocrx_word' id='word_1_381' title='bbox 1904 1376 1958 1402; x_wconf 91'>you</span> <span class='ocrx_word' id='word_1_382' title='bbox 1968 1373 1997 1395; x_wconf 91'>to</span> <span class='ocrx_word' id='word_1_383' title='bbox 2008 1377 2087 1396; x_wconf 91'>move</span> <span class='ocrx_word' id='word_1_384' title='bbox 2097 1369 2160 1396; x_wconf 89'>bars</span> 
-     </span>
-     <span class='ocr_line' id='line_1_46' title="bbox 1297 1407 2151 1441; baseline 0.002 -8; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_385' title='bbox 1297 1407 1369 1434; x_wconf 86'>from</span> <span class='ocrx_word' id='word_1_386' title='bbox 1380 1415 1433 1434; x_wconf 88'>one</span> <span class='ocrx_word' id='word_1_387' title='bbox 1443 1407 1565 1435; x_wconf 85'>location</span> <span class='ocrx_word' id='word_1_388' title='bbox 1576 1411 1605 1434; x_wconf 89'>to</span> <span class='ocrx_word' id='word_1_389' title='bbox 1616 1407 1796 1435; x_wconf 85'>another—in</span> <span class='ocrx_word' id='word_1_390' title='bbox 1806 1408 1852 1435; x_wconf 88'>the</span> <span class='ocrx_word' id='word_1_391' title='bbox 1863 1416 1937 1435; x_wconf 90'>same</span> <span class='ocrx_word' id='word_1_392' title='bbox 1948 1416 2083 1441; x_wconf 88'>sequence</span> <span class='ocrx_word' id='word_1_393' title='bbox 2093 1416 2125 1435; x_wconf 90'>or</span> <span class='ocrx_word' id='word_1_394' title='bbox 2135 1416 2151 1435; x_wconf 88'>a</span> 
-     </span>
-     <span class='ocr_line' id='line_1_47' title="bbox 1296 1447 2160 1481; baseline 0.001 -7; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_395' title='bbox 1296 1447 1421 1474; x_wconf 81'>different</span> <span class='ocrx_word' id='word_1_396' title='bbox 1431 1455 1491 1474; x_wconf 89'>one.</span> <span class='ocrx_word' id='word_1_397' title='bbox 1504 1448 1555 1474; x_wconf 87'>For</span> <span class='ocrx_word' id='word_1_398' title='bbox 1565 1448 1698 1481; x_wconf 84'>example,</span> <span class='ocrx_word' id='word_1_399' title='bbox 1710 1456 1764 1481; x_wconf 91'>you</span> <span class='ocrx_word' id='word_1_400' title='bbox 1775 1448 1863 1481; x_wconf 84'>might</span> <span class='ocrx_word' id='word_1_401' title='bbox 1873 1448 1955 1475; x_wconf 86'>insert</span> <span class='ocrx_word' id='word_1_402' title='bbox 1966 1456 1983 1475; x_wconf 92'>a</span> <span class='ocrx_word' id='word_1_403' title='bbox 1993 1456 2064 1481; x_wconf 88'>copy</span> <span class='ocrx_word' id='word_1_404' title='bbox 2074 1448 2106 1475; x_wconf 90'>of</span> <span class='ocrx_word' id='word_1_405' title='bbox 2115 1448 2160 1475; x_wconf 90'>the</span> 
-     </span>
-     <span class='ocr_line' id='line_1_48' title="bbox 1297 1486 2087 1521; baseline 0 -7; x_size 35; x_descenders 7; x_ascenders 9"><span class='ocrx_word' id='word_1_406' title='bbox 1297 1486 1355 1514; x_wconf 82'>first</span> <span class='ocrx_word' id='word_1_407' title='bbox 1366 1495 1439 1514; x_wconf 89'>verse</span> <span class='ocrx_word' id='word_1_408' title='bbox 1449 1487 1571 1514; x_wconf 89'>between</span> <span class='ocrx_word' id='word_1_409' title='bbox 1582 1487 1628 1514; x_wconf 88'>the</span> <span class='ocrx_word' id='word_1_410' title='bbox 1638 1487 1740 1514; x_wconf 87'>second</span> <span class='ocrx_word' id='word_1_411' title='bbox 1751 1487 1852 1514; x_wconf 83'>chorus</span> <span class='ocrx_word' id='word_1_412' title='bbox 1863 1487 1919 1514; x_wconf 90'>and</span> <span class='ocrx_word' id='word_1_413' title='bbox 1929 1487 1975 1514; x_wconf 88'>the</span> <span class='ocrx_word' id='word_1_414' title='bbox 1985 1487 2087 1521; x_wconf 85'>bridge.</span> 
-     </span>
-     <span class='ocr_line' id='line_1_49' title="bbox 1296 1527 2047 1560; baseline 0.003 -8; x_size 32; x_descenders 6; x_ascenders 7"><span class='ocrx_word' id='word_1_415' title='bbox 1296 1527 1441 1553; x_wconf 88'>DELETE</span> <span class='ocrx_word' id='word_1_416' title='bbox 1453 1527 1546 1553; x_wconf 90'>BARS</span> <span class='ocrx_word' id='word_1_417' title='bbox 1557 1531 1681 1559; x_wconf 84'>operates</span> <span class='ocrx_word' id='word_1_418' title='bbox 1691 1527 1737 1553; x_wconf 86'>the</span> <span class='ocrx_word' id='word_1_419' title='bbox 1748 1535 1823 1554; x_wconf 90'>same</span> <span class='ocrx_word' id='word_1_420' title='bbox 1833 1535 1891 1560; x_wconf 93'>way</span> <span class='ocrx_word' id='word_1_421' title='bbox 1901 1531 1930 1554; x_wconf 91'>to</span> <span class='ocrx_word' id='word_1_422' title='bbox 1940 1535 2047 1554; x_wconf 90'>remove</span> 
-     </span>
-     <span class='ocr_line' id='line_1_50' title="bbox 1295 1565 1577 1593; baseline 0.004 -1; x_size 34.748871; x_descenders 6.7488689; x_ascenders 9"><span class='ocrx_word' id='word_1_423' title='bbox 1295 1565 1441 1592; x_wconf 88'>unwanted</span> <span class='ocrx_word' id='word_1_424' title='bbox 1452 1565 1577 1593; x_wconf 84'>sections.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_19' title="bbox 1295 1638 2215 1960">
-    <p class='ocr_par' id='par_1_25' lang='eng' title="bbox 1296 1638 1589 1680">
-     <span class='ocr_line' id='line_1_51' title="bbox 1296 1638 1589 1680; baseline 0 -9; x_size 42; x_descenders 9; x_ascenders 10"><span class='ocrx_word' id='word_1_425' title='bbox 1296 1638 1452 1680; x_wconf 86'>Creating</span> <span class='ocrx_word' id='word_1_426' title='bbox 1465 1648 1485 1671; x_wconf 85'>a</span> <span class='ocrx_word' id='word_1_427' title='bbox 1497 1639 1589 1680; x_wconf 87'>Song</span> 
-     </span>
-    </p>
-
-    <p class='ocr_par' id='par_1_26' lang='eng' title="bbox 1295 1690 2215 1960">
-     <span class='ocr_line' id='line_1_52' title="bbox 1333 1690 2146 1723; baseline 0 -6; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_428' title='bbox 1333 1690 1394 1717; x_wconf 88'>One</span> <span class='ocrx_word' id='word_1_429' title='bbox 1404 1698 1462 1723; x_wconf 93'>way</span> <span class='ocrx_word' id='word_1_430' title='bbox 1472 1694 1500 1717; x_wconf 89'>to</span> <span class='ocrx_word' id='word_1_431' title='bbox 1511 1694 1598 1717; x_wconf 88'>create</span> <span class='ocrx_word' id='word_1_432' title='bbox 1608 1698 1625 1717; x_wconf 88'>a</span> <span class='ocrx_word' id='word_1_433' title='bbox 1635 1698 1704 1723; x_wconf 83'>song</span> <span class='ocrx_word' id='word_1_434' title='bbox 1715 1690 1736 1717; x_wconf 90'>is</span> <span class='ocrx_word' id='word_1_435' title='bbox 1747 1694 1776 1717; x_wconf 89'>to</span> <span class='ocrx_word' id='word_1_436' title='bbox 1787 1690 1880 1717; x_wconf 86'>record</span> <span class='ocrx_word' id='word_1_437' title='bbox 1891 1690 1958 1717; x_wconf 89'>each</span> <span class='ocrx_word' id='word_1_438' title='bbox 1969 1690 2044 1717; x_wconf 87'>track</span> <span class='ocrx_word' id='word_1_439' title='bbox 2056 1690 2090 1717; x_wconf 89'>all</span> <span class='ocrx_word' id='word_1_440' title='bbox 2101 1690 2146 1717; x_wconf 88'>the</span> 
-     </span>
-     <span class='ocr_line' id='line_1_53' title="bbox 1295 1729 2121 1762; baseline 0.001 -6; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_441' title='bbox 1295 1737 1353 1762; x_wconf 87'>way</span> <span class='ocrx_word' id='word_1_442' title='bbox 1362 1729 1481 1762; x_wconf 83'>through</span> <span class='ocrx_word' id='word_1_443' title='bbox 1493 1729 1541 1762; x_wconf 90'>(up</span> <span class='ocrx_word' id='word_1_444' title='bbox 1552 1733 1581 1756; x_wconf 89'>to</span> <span class='ocrx_word' id='word_1_445' title='bbox 1592 1730 1644 1756; x_wconf 89'>999</span> <span class='ocrx_word' id='word_1_446' title='bbox 1654 1729 1738 1762; x_wconf 90'>bars).</span> <span class='ocrx_word' id='word_1_447' title='bbox 1751 1729 1878 1757; x_wconf 89'>Another</span> <span class='ocrx_word' id='word_1_448' title='bbox 1888 1737 1945 1762; x_wconf 89'>way</span> <span class='ocrx_word' id='word_1_449' title='bbox 1956 1729 1977 1757; x_wconf 90'>is</span> <span class='ocrx_word' id='word_1_450' title='bbox 1987 1733 2016 1757; x_wconf 89'>to</span> <span class='ocrx_word' id='word_1_451' title='bbox 2027 1729 2121 1757; x_wconf 86'>record</span> 
-     </span>
-     <span class='ocr_line' id='line_1_54' title="bbox 1296 1768 2066 1802; baseline 0 -6; x_size 33; x_descenders 5; x_ascenders 9"><span class='ocrx_word' id='word_1_452' title='bbox 1296 1768 1362 1796; x_wconf 86'>each</span> <span class='ocrx_word' id='word_1_453' title='bbox 1373 1768 1448 1796; x_wconf 87'>basic</span> <span class='ocrx_word' id='word_1_454' title='bbox 1458 1768 1562 1796; x_wconf 83'>section</span> <span class='ocrx_word' id='word_1_455' title='bbox 1574 1769 1666 1802; x_wconf 89'>(verse,</span> <span class='ocrx_word' id='word_1_456' title='bbox 1679 1769 1788 1801; x_wconf 86'>chorus,</span> <span class='ocrx_word' id='word_1_457' title='bbox 1800 1769 1865 1802; x_wconf 83'>etc.)</span> <span class='ocrx_word' id='word_1_458' title='bbox 1876 1768 1904 1795; x_wconf 91'>in</span> <span class='ocrx_word' id='word_1_459' title='bbox 1914 1768 2066 1796; x_wconf 86'>individual</span> 
-     </span>
-     <span class='ocr_line' id='line_1_55' title="bbox 1296 1808 2215 1841; baseline 0 -6; x_size 32; x_descenders 5; x_ascenders 8"><span class='ocrx_word' id='word_1_460' title='bbox 1296 1816 1451 1841; x_wconf 85'>sequences,</span> <span class='ocrx_word' id='word_1_461' title='bbox 1463 1808 1528 1835; x_wconf 86'>then</span> <span class='ocrx_word' id='word_1_462' title='bbox 1538 1816 1587 1835; x_wconf 90'>use</span> <span class='ocrx_word' id='word_1_463' title='bbox 1597 1808 1643 1835; x_wconf 89'>the</span> <span class='ocrx_word' id='word_1_464' title='bbox 1653 1809 1799 1835; x_wconf 89'>CREATE</span> <span class='ocrx_word' id='word_1_465' title='bbox 1810 1808 1911 1835; x_wconf 89'>SONG</span> <span class='ocrx_word' id='word_1_466' title='bbox 1923 1808 2050 1836; x_wconf 85'>function</span> <span class='ocrx_word' id='word_1_467' title='bbox 2060 1812 2089 1835; x_wconf 89'>to</span> <span class='ocrx_word' id='word_1_468' title='bbox 2103 1808 2215 1836; x_wconf 87'>“chain”</span> 
-     </span>
-     <span class='ocr_line' id='line_1_56' title="bbox 1295 1847 2135 1881; baseline 0.001 -7; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_469' title='bbox 1295 1848 1370 1874; x_wconf 91'>them</span> <span class='ocrx_word' id='word_1_470' title='bbox 1381 1848 1508 1881; x_wconf 83'>together.</span> <span class='ocrx_word' id='word_1_471' title='bbox 1521 1848 1667 1875; x_wconf 87'>CREATE</span> <span class='ocrx_word' id='word_1_472' title='bbox 1678 1848 1779 1875; x_wconf 91'>SONG</span> <span class='ocrx_word' id='word_1_473' title='bbox 1789 1847 1842 1874; x_wconf 92'>will</span> <span class='ocrx_word' id='word_1_474' title='bbox 1853 1848 1918 1875; x_wconf 86'>then</span> <span class='ocrx_word' id='word_1_475' title='bbox 1929 1848 2135 1881; x_wconf 81'>automatically</span> 
-     </span>
-     <span class='ocr_line' id='line_1_57' title="bbox 1296 1887 2162 1920; baseline 0 -6; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_476' title='bbox 1296 1895 1366 1920; x_wconf 89'>copy</span> <span class='ocrx_word' id='word_1_477' title='bbox 1377 1887 1412 1914; x_wconf 88'>all</span> <span class='ocrx_word' id='word_1_478' title='bbox 1422 1887 1468 1914; x_wconf 85'>the</span> <span class='ocrx_word' id='word_1_479' title='bbox 1478 1891 1552 1920; x_wconf 86'>parts</span> <span class='ocrx_word' id='word_1_480' title='bbox 1563 1887 1621 1914; x_wconf 86'>into</span> <span class='ocrx_word' id='word_1_481' title='bbox 1632 1895 1649 1914; x_wconf 91'>a</span> <span class='ocrx_word' id='word_1_482' title='bbox 1659 1895 1718 1914; x_wconf 91'>new</span> <span class='ocrx_word' id='word_1_483' title='bbox 1729 1895 1870 1920; x_wconf 83'>sequence.</span> <span class='ocrx_word' id='word_1_484' title='bbox 1884 1887 1911 1914; x_wconf 84'>If</span> <span class='ocrx_word' id='word_1_485' title='bbox 1919 1887 2033 1919; x_wconf 89'>desired,</span> <span class='ocrx_word' id='word_1_486' title='bbox 2045 1895 2099 1920; x_wconf 92'>you</span> <span class='ocrx_word' id='word_1_487' title='bbox 2110 1895 2162 1914; x_wconf 85'>can</span> 
-     </span>
-     <span class='ocr_line' id='line_1_58' title="bbox 1296 1926 2169 1960; baseline 0.001 -7; x_size 34; x_descenders 7; x_ascenders 9"><span class='ocrx_word' id='word_1_488' title='bbox 1296 1934 1361 1953; x_wconf 89'>even</span> <span class='ocrx_word' id='word_1_489' title='bbox 1372 1930 1411 1953; x_wconf 90'>set</span> <span class='ocrx_word' id='word_1_490' title='bbox 1422 1926 1468 1953; x_wconf 88'>the</span> <span class='ocrx_word' id='word_1_491' title='bbox 1478 1926 1528 1953; x_wconf 89'>last</span> <span class='ocrx_word' id='word_1_492' title='bbox 1540 1926 1590 1953; x_wconf 88'>few</span> <span class='ocrx_word' id='word_1_493' title='bbox 1601 1927 1664 1953; x_wconf 89'>bars</span> <span class='ocrx_word' id='word_1_494' title='bbox 1675 1931 1704 1954; x_wconf 85'>to</span> <span class='ocrx_word' id='word_1_495' title='bbox 1715 1931 1806 1960; x_wconf 89'>repeat</span> <span class='ocrx_word' id='word_1_496' title='bbox 1816 1926 1955 1960; x_wconf 84'>infinitely,</span> <span class='ocrx_word' id='word_1_497' title='bbox 1968 1926 2011 1954; x_wconf 86'>for</span> <span class='ocrx_word' id='word_1_498' title='bbox 2022 1935 2038 1954; x_wconf 92'>a</span> <span class='ocrx_word' id='word_1_499' title='bbox 2049 1927 2169 1954; x_wconf 83'>fadeout.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_20' title="bbox 1293 2000 2179 2242">
-    <p class='ocr_par' id='par_1_27' lang='eng' title="bbox 1294 2000 1948 2040">
-     <span class='ocr_line' id='line_1_59' title="bbox 1294 2000 1948 2040; baseline 0.002 -8; x_size 40; x_descenders 8; x_ascenders 10"><span class='ocrx_word' id='word_1_500' title='bbox 1294 2000 1532 2040; x_wconf 84'>Composition</span> <span class='ocrx_word' id='word_1_501' title='bbox 1544 2000 1699 2033; x_wconf 85'>Without</span> <span class='ocrx_word' id='word_1_502' title='bbox 1711 2000 1948 2040; x_wconf 85'>Compromise</span> 
-     </span>
-    </p>
-
-    <p class='ocr_par' id='par_1_28' lang='eng' title="bbox 1293 2051 2179 2242">
-     <span class='ocr_line' id='line_1_60' title="bbox 1331 2051 2179 2085; baseline 0 -7; x_size 34; x_descenders 7; x_ascenders 9"><span class='ocrx_word' id='word_1_503' title='bbox 1331 2051 1389 2078; x_wconf 90'>The</span> <span class='ocrx_word' id='word_1_504' title='bbox 1399 2051 1563 2085; x_wconf 83'>technology</span> <span class='ocrx_word' id='word_1_505' title='bbox 1573 2059 1627 2084; x_wconf 92'>you</span> <span class='ocrx_word' id='word_1_506' title='bbox 1638 2060 1686 2079; x_wconf 86'>use</span> <span class='ocrx_word' id='word_1_507' title='bbox 1697 2051 1797 2078; x_wconf 89'>should</span> <span class='ocrx_word' id='word_1_508' title='bbox 1808 2059 1887 2078; x_wconf 87'>never</span> <span class='ocrx_word' id='word_1_509' title='bbox 1897 2052 1932 2079; x_wconf 91'>be</span> <span class='ocrx_word' id='word_1_510' title='bbox 1942 2059 1973 2079; x_wconf 89'>so</span> <span class='ocrx_word' id='word_1_511' title='bbox 1984 2051 2110 2085; x_wconf 85'>complex</span> <span class='ocrx_word' id='word_1_512' title='bbox 2120 2052 2179 2078; x_wconf 86'>that</span> 
-     </span>
-     <span class='ocr_line' id='line_1_61' title="bbox 1294 2090 2157 2124; baseline 0 -6; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_513' title='bbox 1294 2090 1313 2118; x_wconf 86'>it</span> <span class='ocrx_word' id='word_1_514' title='bbox 1323 2091 1459 2118; x_wconf 87'>interferes</span> <span class='ocrx_word' id='word_1_515' title='bbox 1470 2091 1535 2118; x_wconf 84'>with</span> <span class='ocrx_word' id='word_1_516' title='bbox 1545 2091 1591 2118; x_wconf 88'>the</span> <span class='ocrx_word' id='word_1_517' title='bbox 1602 2091 1715 2118; x_wconf 90'>creative</span> <span class='ocrx_word' id='word_1_518' title='bbox 1725 2099 1841 2124; x_wconf 83'>process.</span> <span class='ocrx_word' id='word_1_519' title='bbox 1854 2091 1947 2118; x_wconf 78'>That’s</span> <span class='ocrx_word' id='word_1_520' title='bbox 1957 2091 2086 2124; x_wconf 86'>precisely</span> <span class='ocrx_word' id='word_1_521' title='bbox 2096 2091 2157 2124; x_wconf 89'>why</span> 
-     </span>
-     <span class='ocr_line' id='line_1_62' title="bbox 1293 2130 2156 2164; baseline 0 -7; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_522' title='bbox 1293 2130 1339 2157; x_wconf 90'>the</span> <span class='ocrx_word' id='word_1_523' title='bbox 1350 2130 1576 2164; x_wconf 80'>LinnSequencer</span> <span class='ocrx_word' id='word_1_524' title='bbox 1586 2130 1607 2157; x_wconf 89'>is</span> <span class='ocrx_word' id='word_1_525' title='bbox 1619 2130 1747 2164; x_wconf 87'>designed</span> <span class='ocrx_word' id='word_1_526' title='bbox 1758 2134 1787 2157; x_wconf 87'>to</span> <span class='ocrx_word' id='word_1_527' title='bbox 1798 2130 1834 2157; x_wconf 89'>let</span> <span class='ocrx_word' id='word_1_528' title='bbox 1844 2138 1899 2163; x_wconf 93'>you</span> <span class='ocrx_word' id='word_1_529' title='bbox 1910 2138 2051 2164; x_wconf 88'>compose,</span> <span class='ocrx_word' id='word_1_530' title='bbox 2063 2130 2156 2157; x_wconf 87'>record</span> 
-     </span>
-     <span class='ocr_line' id='line_1_63' title="bbox 1294 2169 2145 2203; baseline 0 -6; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_531' title='bbox 1294 2170 1349 2197; x_wconf 90'>and</span> <span class='ocrx_word' id='word_1_532' title='bbox 1360 2169 1414 2197; x_wconf 83'>edit</span> <span class='ocrx_word' id='word_1_533' title='bbox 1425 2170 1503 2197; x_wconf 89'>while</span> <span class='ocrx_word' id='word_1_534' title='bbox 1514 2170 1643 2203; x_wconf 90'>devoting</span> <span class='ocrx_word' id='word_1_535' title='bbox 1653 2178 1721 2203; x_wconf 92'>your</span> <span class='ocrx_word' id='word_1_536' title='bbox 1731 2170 1882 2197; x_wconf 88'>undivided</span> <span class='ocrx_word' id='word_1_537' title='bbox 1893 2170 2027 2197; x_wconf 86'>attention</span> <span class='ocrx_word' id='word_1_538' title='bbox 2037 2174 2067 2197; x_wconf 91'>to</span> <span class='ocrx_word' id='word_1_539' title='bbox 2077 2178 2145 2203; x_wconf 90'>your</span> 
-     </span>
-     <span class='ocr_line' id='line_1_64' title="bbox 1294 2209 2126 2242; baseline 0 -6; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_540' title='bbox 1294 2209 1388 2236; x_wconf 88'>music.</span> <span class='ocrx_word' id='word_1_541' title='bbox 1402 2209 1451 2236; x_wconf 89'>See</span> <span class='ocrx_word' id='word_1_542' title='bbox 1460 2217 1529 2242; x_wconf 92'>your</span> <span class='ocrx_word' id='word_1_543' title='bbox 1540 2209 1611 2236; x_wconf 87'>Linn</span> <span class='ocrx_word' id='word_1_544' title='bbox 1621 2209 1712 2236; x_wconf 86'>dealer</span> <span class='ocrx_word' id='word_1_545' title='bbox 1722 2209 1806 2242; x_wconf 85'>today</span> <span class='ocrx_word' id='word_1_546' title='bbox 1817 2209 1859 2236; x_wconf 87'>for</span> <span class='ocrx_word' id='word_1_547' title='bbox 1870 2217 1887 2236; x_wconf 88'>a</span> <span class='ocrx_word' id='word_1_548' title='bbox 1897 2209 2126 2236; x_wconf 76'>demonstration!</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_21' title="bbox 347 2343 2171 2378">
-    <p class='ocr_par' id='par_1_29' lang='eng' title="bbox 347 2343 2171 2378">
-     <span class='ocr_line' id='line_1_65' title="bbox 347 2343 2171 2378; baseline 0.001 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_549' title='bbox 347 2350 361 2363; x_wconf 69'><strong>0</strong></span> <span class='ocrx_word' id='word_1_550' title='bbox 373 2343 483 2377; x_wconf 83'>Simple,</span> <span class='ocrx_word' id='word_1_551' title='bbox 495 2352 559 2377; x_wconf 89'>easy</span> <span class='ocrx_word' id='word_1_552' title='bbox 569 2348 598 2371; x_wconf 91'>to</span> <span class='ocrx_word' id='word_1_553' title='bbox 609 2344 683 2371; x_wconf 89'>learn</span> <span class='ocrx_word' id='word_1_554' title='bbox 694 2344 922 2378; x_wconf 80'>operation—the</span> <span class='ocrx_word' id='word_1_555' title='bbox 934 2345 967 2372; x_wconf 88'>32</span> <span class='ocrx_word' id='word_1_556' title='bbox 978 2345 1116 2372; x_wconf 86'>character</span> <span class='ocrx_word' id='word_1_557' title='bbox 1126 2345 1200 2372; x_wconf 92'>LCD</span> <span class='ocrx_word' id='word_1_558' title='bbox 1211 2345 1316 2378; x_wconf 90'>display</span> <span class='ocrx_word' id='word_1_559' title='bbox 1326 2345 1424 2378; x_wconf 88'>clearly</span> <span class='ocrx_word' id='word_1_560' title='bbox 1434 2345 1528 2378; x_wconf 86'>guides</span> <span class='ocrx_word' id='word_1_561' title='bbox 1539 2353 1594 2378; x_wconf 91'>you</span> <span class='ocrx_word' id='word_1_562' title='bbox 1604 2345 1724 2378; x_wconf 87'>through</span> <span class='ocrx_word' id='word_1_563' title='bbox 1735 2344 1770 2371; x_wconf 88'>all</span> <span class='ocrx_word' id='word_1_564' title='bbox 1781 2344 1947 2377; x_wconf 86'>operations.</span> <span class='ocrx_word' id='word_1_565' title='bbox 1961 2344 1989 2371; x_wconf 88'>If</span> <span class='ocrx_word' id='word_1_566' title='bbox 1997 2344 2112 2376; x_wconf 88'>needed,</span> <span class='ocrx_word' id='word_1_567' title='bbox 2125 2345 2171 2371; x_wconf 88'>the</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_22' title="bbox 373 2381 1083 2415">
-    <p class='ocr_par' id='par_1_30' lang='eng' title="bbox 373 2381 1083 2415">
-     <span class='ocr_line' id='line_1_66' title="bbox 373 2381 1083 2415; baseline 0.003 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_568' title='bbox 373 2381 472 2407; x_wconf 85'>HELP</span> <span class='ocrx_word' id='word_1_569' title='bbox 483 2381 583 2408; x_wconf 85'>button</span> <span class='ocrx_word' id='word_1_570' title='bbox 594 2381 711 2415; x_wconf 87'>displays</span> <span class='ocrx_word' id='word_1_571' title='bbox 722 2382 875 2409; x_wconf 89'>additional</span> <span class='ocrx_word' id='word_1_572' title='bbox 886 2382 1083 2415; x_wconf 87'>explanations.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_23' title="bbox 347 2427 2145 2507">
-    <p class='ocr_par' id='par_1_31' lang='eng' title="bbox 347 2427 2145 2507">
-     <span class='ocr_line' id='line_1_67' title="bbox 347 2427 1468 2461; baseline 0.002 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_573' title='bbox 347 2432 361 2446; x_wconf 69'><strong>0</strong></span> <span class='ocrx_word' id='word_1_574' title='bbox 373 2427 612 2454; x_wconf 88'>Non-destructive</span> <span class='ocrx_word' id='word_1_575' title='bbox 622 2427 914 2461; x_wconf 85'>recording—existing</span> <span class='ocrx_word' id='word_1_576' title='bbox 924 2432 1002 2455; x_wconf 90'>notes</span> <span class='ocrx_word' id='word_1_577' title='bbox 1013 2436 1057 2455; x_wconf 88'>are</span> <span class='ocrx_word' id='word_1_578' title='bbox 1068 2432 1116 2455; x_wconf 84'>not</span> <span class='ocrx_word' id='word_1_579' title='bbox 1127 2428 1220 2455; x_wconf 87'>erased</span> <span class='ocrx_word' id='word_1_580' title='bbox 1231 2428 1309 2455; x_wconf 91'>while</span> <span class='ocrx_word' id='word_1_581' title='bbox 1319 2428 1468 2461; x_wconf 87'>recording.</span> 
-     </span>
-     <span class='ocr_line' id='line_1_68' title="bbox 347 2473 2145 2507; baseline 0.001 -8; x_size 35; x_descenders 7; x_ascenders 9"><span class='ocrx_word' id='word_1_582' title='bbox 347 2478 361 2492; x_wconf 71'><strong>0</strong></span> <span class='ocrx_word' id='word_1_583' title='bbox 372 2473 433 2500; x_wconf 90'>Two</span> <span class='ocrx_word' id='word_1_584' title='bbox 444 2473 689 2500; x_wconf 89'>FOOTSWITCH</span> <span class='ocrx_word' id='word_1_585' title='bbox 701 2474 837 2501; x_wconf 79'>INPUTS</span> <span class='ocrx_word' id='word_1_586' title='bbox 848 2481 910 2507; x_wconf 89'>may</span> <span class='ocrx_word' id='word_1_587' title='bbox 921 2473 955 2500; x_wconf 91'>be</span> <span class='ocrx_word' id='word_1_588' title='bbox 966 2473 1091 2507; x_wconf 85'>assigned</span> <span class='ocrx_word' id='word_1_589' title='bbox 1101 2478 1130 2501; x_wconf 81'>to</span> <span class='ocrx_word' id='word_1_590' title='bbox 1141 2474 1271 2507; x_wconf 84'>remotely</span> <span class='ocrx_word' id='word_1_591' title='bbox 1281 2473 1387 2501; x_wconf 86'>control</span> <span class='ocrx_word' id='word_1_592' title='bbox 1398 2482 1481 2507; x_wconf 86'>many</span> <span class='ocrx_word' id='word_1_593' title='bbox 1491 2473 1523 2501; x_wconf 90'>of</span> <span class='ocrx_word' id='word_1_594' title='bbox 1531 2474 1578 2501; x_wconf 89'>the</span> <span class='ocrx_word' id='word_1_595' title='bbox 1588 2473 1750 2506; x_wconf 82'>commonly</span> <span class='ocrx_word' id='word_1_596' title='bbox 1760 2473 1827 2500; x_wconf 88'>used</span> <span class='ocrx_word' id='word_1_597' title='bbox 1840 2473 1991 2505; x_wconf 77'>functions,</span> <span class='ocrx_word' id='word_1_598' title='bbox 2003 2473 2145 2507; x_wconf 84'>including</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_24' title="bbox 372 2510 1090 2543">
-    <p class='ocr_par' id='par_1_32' lang='eng' title="bbox 372 2510 1090 2543">
-     <span class='ocr_line' id='line_1_69' title="bbox 372 2510 1090 2543; baseline 0.001 -6; x_size 35.625; x_descenders 8.90625; x_ascenders 8.90625"><span class='ocrx_word' id='word_1_599' title='bbox 372 2510 500 2542; x_wconf 88'>ERASE,</span> <span class='ocrx_word' id='word_1_600' title='bbox 513 2511 660 2542; x_wconf 86'>REPEAT,</span> <span class='ocrx_word' id='word_1_601' title='bbox 673 2511 883 2543; x_wconf 80'>PLAY/STOP,</span> <span class='ocrx_word' id='word_1_602' title='bbox 896 2519 927 2538; x_wconf 90'>or</span> <span class='ocrx_word' id='word_1_603' title='bbox 939 2511 1090 2538; x_wconf 83'>LOCATE.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_25' title="bbox 347 2556 1768 2590">
-    <p class='ocr_par' id='par_1_33' lang='eng' title="bbox 347 2556 1768 2590">
-     <span class='ocr_line' id='line_1_70' title="bbox 347 2556 1768 2590; baseline 0.001 -8; x_size 34; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_604' title='bbox 347 2561 361 2575; x_wconf 71'><strong>0</strong></span> <span class='ocrx_word' id='word_1_605' title='bbox 372 2556 433 2583; x_wconf 91'>Two</span> <span class='ocrx_word' id='word_1_606' title='bbox 443 2556 612 2583; x_wconf 83'>TRIGGER</span> <span class='ocrx_word' id='word_1_607' title='bbox 623 2556 797 2584; x_wconf 84'>OUTPUTS</span> <span class='ocrx_word' id='word_1_608' title='bbox 808 2565 871 2590; x_wconf 89'>may</span> <span class='ocrx_word' id='word_1_609' title='bbox 881 2557 915 2584; x_wconf 91'>be</span> <span class='ocrx_word' id='word_1_610' title='bbox 925 2557 1119 2590; x_wconf 84'>programmed</span> <span class='ocrx_word' id='word_1_611' title='bbox 1129 2561 1158 2584; x_wconf 93'>to</span> <span class='ocrx_word' id='word_1_612' title='bbox 1169 2561 1270 2590; x_wconf 87'>output</span> <span class='ocrx_word' id='word_1_613' title='bbox 1281 2557 1370 2590; x_wconf 88'>pulses</span> <span class='ocrx_word' id='word_1_614' title='bbox 1381 2561 1409 2584; x_wconf 92'>at</span> <span class='ocrx_word' id='word_1_615' title='bbox 1419 2565 1472 2590; x_wconf 90'>any</span> <span class='ocrx_word' id='word_1_616' title='bbox 1483 2557 1598 2584; x_wconf 86'>selected</span> <span class='ocrx_word' id='word_1_617' title='bbox 1608 2561 1673 2584; x_wconf 89'>note</span> <span class='ocrx_word' id='word_1_618' title='bbox 1683 2556 1768 2583; x_wconf 89'>value.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_26' title="bbox 347 2601 1226 2635">
-    <p class='ocr_par' id='par_1_34' lang='eng' title="bbox 347 2601 1226 2635">
-     <span class='ocr_line' id='line_1_71' title="bbox 347 2601 1226 2635; baseline 0.001 -7; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_619' title='bbox 347 2607 361 2620; x_wconf 65'><strong>0</strong></span> <span class='ocrx_word' id='word_1_620' title='bbox 372 2601 434 2628; x_wconf 88'>Will</span> <span class='ocrx_word' id='word_1_621' title='bbox 445 2609 510 2634; x_wconf 89'>sync</span> <span class='ocrx_word' id='word_1_622' title='bbox 521 2605 549 2628; x_wconf 84'>to</span> <span class='ocrx_word' id='word_1_623' title='bbox 561 2602 690 2629; x_wconf 88'>standard</span> <span class='ocrx_word' id='word_1_624' title='bbox 701 2602 864 2629; x_wconf 86'>LinnDrum</span> <span class='ocrx_word' id='word_1_625' title='bbox 875 2610 907 2629; x_wconf 91'>or</span> <span class='ocrx_word' id='word_1_626' title='bbox 918 2602 989 2629; x_wconf 89'>Linn</span> <span class='ocrx_word' id='word_1_627' title='bbox 1000 2603 1069 2629; x_wconf 90'>9000</span> <span class='ocrx_word' id='word_1_628' title='bbox 1080 2610 1145 2635; x_wconf 87'>sync</span> <span class='ocrx_word' id='word_1_629' title='bbox 1155 2607 1226 2630; x_wconf 89'>tone.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_27' title="bbox 347 2648 2100 2727">
-    <p class='ocr_par' id='par_1_35' lang='eng' title="bbox 347 2648 2100 2727">
-     <span class='ocr_line' id='line_1_72' title="bbox 347 2648 1664 2682; baseline 0.001 -7; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_630' title='bbox 347 2654 360 2667; x_wconf 75'>&#39;</span> <span class='ocrx_word' id='word_1_631' title='bbox 372 2648 483 2675; x_wconf 89'>Utilizes</span> <span class='ocrx_word' id='word_1_632' title='bbox 493 2648 564 2680; x_wconf 75'>ultra</span> <span class='ocrx_word' id='word_1_633' title='bbox 573 2648 744 2682; x_wconf 86'>high—speed,</span> <span class='ocrx_word' id='word_1_634' title='bbox 757 2649 772 2676; x_wconf 91'>8</span> <span class='ocrx_word' id='word_1_635' title='bbox 783 2649 862 2675; x_wconf 89'>MHz</span> <span class='ocrx_word' id='word_1_636' title='bbox 873 2649 954 2676; x_wconf 88'>80186</span> <span class='ocrx_word' id='word_1_637' title='bbox 965 2649 994 2676; x_wconf 91'>16</span> <span class='ocrx_word' id='word_1_638' title='bbox 1004 2648 1043 2676; x_wconf 90'>bit</span> <span class='ocrx_word' id='word_1_639' title='bbox 1054 2653 1197 2682; x_wconf 88'>computer</span> <span class='ocrx_word' id='word_1_640' title='bbox 1208 2649 1350 2682; x_wconf 90'>internally</span> <span class='ocrx_word' id='word_1_641' title='bbox 1361 2649 1403 2676; x_wconf 87'>for</span> <span class='ocrx_word' id='word_1_642' title='bbox 1414 2649 1502 2676; x_wconf 87'>FAST</span> <span class='ocrx_word' id='word_1_643' title='bbox 1512 2648 1664 2682; x_wconf 81'>operation.</span> 
-     </span>
-     <span class='ocr_line' id='line_1_73' title="bbox 347 2694 2100 2727; baseline 0.001 -7; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_644' title='bbox 347 2699 361 2713; x_wconf 70'><strong>0</strong></span> <span class='ocrx_word' id='word_1_645' title='bbox 372 2694 504 2721; x_wconf 88'>TEMPO</span> <span class='ocrx_word' id='word_1_646' title='bbox 515 2702 578 2727; x_wconf 92'>may</span> <span class='ocrx_word' id='word_1_647' title='bbox 589 2694 623 2721; x_wconf 90'>be</span> <span class='ocrx_word' id='word_1_648' title='bbox 633 2694 764 2727; x_wconf 87'>specified</span> <span class='ocrx_word' id='word_1_649' title='bbox 774 2694 802 2721; x_wconf 89'>in</span> <span class='ocrx_word' id='word_1_650' title='bbox 814 2695 1172 2722; x_wconf 84'>BEATS-PER—MINUTE</span> <span class='ocrx_word' id='word_1_651' title='bbox 1183 2703 1215 2722; x_wconf 93'>or</span> <span class='ocrx_word' id='word_1_652' title='bbox 1225 2695 1567 2722; x_wconf 84'>FRAMES-PER—BEAT</span> <span class='ocrx_word' id='word_1_653' title='bbox 1577 2698 1605 2721; x_wconf 89'>at</span> <span class='ocrx_word' id='word_1_654' title='bbox 1616 2695 1659 2726; x_wconf 91'>24,</span> <span class='ocrx_word' id='word_1_655' title='bbox 1672 2695 1716 2726; x_wconf 88'>25,</span> <span class='ocrx_word' id='word_1_656' title='bbox 1728 2702 1760 2721; x_wconf 90'>or</span> <span class='ocrx_word' id='word_1_657' title='bbox 1772 2695 1806 2721; x_wconf 88'>30</span> <span class='ocrx_word' id='word_1_658' title='bbox 1817 2694 1917 2721; x_wconf 86'>frames</span> <span class='ocrx_word' id='word_1_659' title='bbox 1928 2702 1976 2727; x_wconf 89'>per</span> <span class='ocrx_word' id='word_1_660' title='bbox 1987 2694 2100 2726; x_wconf 87'>second,</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_28' title="bbox 373 2731 652 2765">
-    <p class='ocr_par' id='par_1_36' lang='eng' title="bbox 373 2731 652 2765">
-     <span class='ocr_line' id='line_1_74' title="bbox 373 2731 652 2765; baseline 0 -7; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_661' title='bbox 373 2731 449 2764; x_wconf 87'>(even</span> <span class='ocrx_word' id='word_1_662' title='bbox 460 2731 529 2765; x_wconf 88'>drop</span> <span class='ocrx_word' id='word_1_663' title='bbox 541 2731 652 2764; x_wconf 81'>frame!)</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_29' title="bbox 347 2777 2174 2811">
-    <p class='ocr_par' id='par_1_37' lang='eng' title="bbox 347 2777 2174 2811">
-     <span class='ocr_line' id='line_1_75' title="bbox 347 2777 2174 2811; baseline 0.001 -8; x_size 33; x_descenders 5; x_ascenders 9"><span class='ocrx_word' id='word_1_664' title='bbox 347 2782 360 2796; x_wconf 72'><strong>0</strong></span> <span class='ocrx_word' id='word_1_665' title='bbox 372 2777 504 2804; x_wconf 82'>TEMPO</span> <span class='ocrx_word' id='word_1_666' title='bbox 515 2785 578 2810; x_wconf 91'>may</span> <span class='ocrx_word' id='word_1_667' title='bbox 588 2777 622 2804; x_wconf 92'>be</span> <span class='ocrx_word' id='word_1_668' title='bbox 633 2778 741 2804; x_wconf 84'>entered</span> <span class='ocrx_word' id='word_1_669' title='bbox 751 2777 934 2811; x_wconf 80'>numerically,</span> <span class='ocrx_word' id='word_1_670' title='bbox 946 2777 1101 2811; x_wconf 85'>adjustable</span> <span class='ocrx_word' id='word_1_671' title='bbox 1111 2777 1139 2804; x_wconf 92'>in</span> <span class='ocrx_word' id='word_1_672' title='bbox 1149 2778 1239 2805; x_wconf 88'>tenths</span> <span class='ocrx_word' id='word_1_673' title='bbox 1250 2778 1282 2805; x_wconf 85'>of</span> <span class='ocrx_word' id='word_1_674' title='bbox 1290 2786 1307 2805; x_wconf 94'>a</span> <span class='ocrx_word' id='word_1_675' title='bbox 1317 2777 1567 2805; x_wconf 85'>Beat-Per-Minute</span> <span class='ocrx_word' id='word_1_676' title='bbox 1577 2777 1748 2809; x_wconf 85'>increments,</span> <span class='ocrx_word' id='word_1_677' title='bbox 1760 2785 1792 2804; x_wconf 87'>or</span> <span class='ocrx_word' id='word_1_678' title='bbox 1803 2777 1839 2810; x_wconf 94'>by</span> <span class='ocrx_word' id='word_1_679' title='bbox 1849 2777 1964 2811; x_wconf 84'>tapping</span> <span class='ocrx_word' id='word_1_680' title='bbox 1975 2781 2086 2810; x_wconf 85'>quarter</span> <span class='ocrx_word' id='word_1_681' title='bbox 2097 2781 2174 2804; x_wconf 85'>notes</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_30' title="bbox 372 2815 808 2842">
-    <p class='ocr_par' id='par_1_38' lang='eng' title="bbox 372 2815 808 2842">
-     <span class='ocr_line' id='line_1_76' title="bbox 372 2815 808 2842; baseline 0.002 -1; x_size 32.351353; x_descenders 5.3513513; x_ascenders 9"><span class='ocrx_word' id='word_1_682' title='bbox 372 2822 410 2841; x_wconf 93'>on</span> <span class='ocrx_word' id='word_1_683' title='bbox 420 2815 466 2842; x_wconf 86'>the</span> <span class='ocrx_word' id='word_1_684' title='bbox 476 2815 545 2841; x_wconf 91'>TAP</span> <span class='ocrx_word' id='word_1_685' title='bbox 556 2815 689 2842; x_wconf 85'>TEMPO</span> <span class='ocrx_word' id='word_1_686' title='bbox 699 2815 808 2842; x_wconf 83'>button.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_31' title="bbox 347 2861 1792 2940">
-    <p class='ocr_par' id='par_1_39' lang='eng' title="bbox 347 2861 1792 2940">
-     <span class='ocr_line' id='line_1_77' title="bbox 347 2861 1792 2895; baseline 0.001 -8; x_size 33; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_687' title='bbox 347 2866 360 2880; x_wconf 68'><strong>0</strong></span> <span class='ocrx_word' id='word_1_688' title='bbox 372 2861 504 2887; x_wconf 87'>TEMPO</span> <span class='ocrx_word' id='word_1_689' title='bbox 515 2861 696 2888; x_wconf 86'>CHANGES</span> <span class='ocrx_word' id='word_1_690' title='bbox 707 2869 771 2894; x_wconf 92'>may</span> <span class='ocrx_word' id='word_1_691' title='bbox 781 2861 815 2888; x_wconf 92'>be</span> <span class='ocrx_word' id='word_1_692' title='bbox 825 2861 1019 2894; x_wconf 85'>programmed</span> <span class='ocrx_word' id='word_1_693' title='bbox 1030 2861 1087 2888; x_wconf 88'>into</span> <span class='ocrx_word' id='word_1_694' title='bbox 1099 2869 1115 2888; x_wconf 87'>a</span> <span class='ocrx_word' id='word_1_695' title='bbox 1126 2870 1268 2895; x_wconf 87'>sequence,</span> <span class='ocrx_word' id='word_1_696' title='bbox 1280 2861 1344 2888; x_wconf 89'>with</span> <span class='ocrx_word' id='word_1_697' title='bbox 1356 2862 1467 2888; x_wconf 88'>smooth</span> <span class='ocrx_word' id='word_1_698' title='bbox 1478 2861 1635 2888; x_wconf 85'>transitions</span> <span class='ocrx_word' id='word_1_699' title='bbox 1646 2861 1670 2887; x_wconf 90'>if</span> <span class='ocrx_word' id='word_1_700' title='bbox 1679 2861 1792 2888; x_wconf 89'>desired.</span> 
-     </span>
-     <span class='ocr_line' id='line_1_78' title="bbox 347 2906 1507 2940; baseline 0.002 -8; x_size 33; x_descenders 7; x_ascenders 8"><span class='ocrx_word' id='word_1_701' title='bbox 347 2911 360 2925; x_wconf 74'><strong><em>°</em></strong></span> <span class='ocrx_word' id='word_1_702' title='bbox 371 2906 434 2938; x_wconf 92'>Any</span> <span class='ocrx_word' id='word_1_703' title='bbox 444 2906 539 2932; x_wconf 85'>TIME</span> <span class='ocrx_word' id='word_1_704' title='bbox 550 2906 763 2933; x_wconf 88'>SIGNATURE</span> <span class='ocrx_word' id='word_1_705' title='bbox 773 2915 836 2939; x_wconf 88'>may</span> <span class='ocrx_word' id='word_1_706' title='bbox 846 2907 880 2933; x_wconf 91'>be</span> <span class='ocrx_word' id='word_1_707' title='bbox 891 2907 968 2938; x_wconf 89'>used,</span> <span class='ocrx_word' id='word_1_708' title='bbox 980 2907 1036 2934; x_wconf 89'>and</span> <span class='ocrx_word' id='word_1_709' title='bbox 1046 2915 1109 2940; x_wconf 91'>may</span> <span class='ocrx_word' id='word_1_710' title='bbox 1120 2907 1154 2934; x_wconf 90'>be</span> <span class='ocrx_word' id='word_1_711' title='bbox 1164 2907 1288 2940; x_wconf 84'>changed</span> <span class='ocrx_word' id='word_1_712' title='bbox 1299 2907 1393 2934; x_wconf 83'>within</span> <span class='ocrx_word' id='word_1_713' title='bbox 1404 2915 1420 2934; x_wconf 89'>a</span> <span class='ocrx_word' id='word_1_714' title='bbox 1431 2915 1507 2940; x_wconf 87'>song.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_32' title="bbox 1540 2970 1878 3140">
-    <p class='ocr_par' id='par_1_40' lang='eng' title="bbox 1540 2970 1878 3140">
-     <span class='ocr_line' id='line_1_79' title="bbox 1540 2970 1800 3089; baseline 0 -14; x_size 100.5; x_descenders 16.5; x_ascenders 28.499998"><span class='ocrx_word' id='word_1_715' title='bbox 1540 2970 1800 3089; x_wconf 74'>E1111</span> 
-     </span>
-     <span class='ocr_line' id='line_1_80' title="bbox 1554 3108 1878 3140; baseline 0 -5; x_size 32; x_descenders 5; x_ascenders 9"><span class='ocrx_word' id='word_1_716' title='bbox 1554 3108 1624 3135; x_wconf 91'>Linn</span> <span class='ocrx_word' id='word_1_717' title='bbox 1635 3108 1808 3140; x_wconf 87'>Electronics,</span> <span class='ocrx_word' id='word_1_718' title='bbox 1822 3109 1878 3136; x_wconf 84'>Inc.</span> 
-     </span>
-    </p>
-   </div>
-   <div class='ocr_carea' id='block_1_33' title="bbox 1554 3150 2188 3226">
-    <p class='ocr_par' id='par_1_41' lang='eng' title="bbox 1554 3150 2188 3226">
-     <span class='ocr_line' id='line_1_81' title="bbox 1557 3150 2163 3182; baseline 0 -5; x_size 32; x_descenders 5; x_ascenders 8"><span class='ocrx_word' id='word_1_719' title='bbox 1557 3150 1638 3177; x_wconf 88'>18720</span> <span class='ocrx_word' id='word_1_720' title='bbox 1648 3150 1761 3177; x_wconf 89'>Oxnard</span> <span class='ocrx_word' id='word_1_721' title='bbox 1772 3150 1866 3182; x_wconf 81'>Street,</span> <span class='ocrx_word' id='word_1_722' title='bbox 1878 3150 2006 3182; x_wconf 86'>Tarzana,</span> <span class='ocrx_word' id='word_1_723' title='bbox 2019 3150 2071 3177; x_wconf 93'>CA</span> <span class='ocrx_word' id='word_1_724' title='bbox 2082 3150 2163 3177; x_wconf 88'>91356</span> 
-     </span>
-     <span class='ocr_line' id='line_1_82' title="bbox 1554 3192 2188 3226; baseline 0 -7; x_size 33.5; x_descenders 5.5; x_ascenders 8.5"><span class='ocrx_word' id='word_1_725' title='bbox 1554 3193 1622 3226; x_wconf 89'>(818)</span> <span class='ocrx_word' id='word_1_726' title='bbox 1633 3193 1755 3219; x_wconf 60'>7088131</span> <span class='ocrx_word' id='word_1_727' title='bbox 1765 3193 1788 3219; x_wconf 91'>T</span> <span class='ocrx_word' id='word_1_728' title='bbox 1789 3193 1888 3219; x_wconf 87'>ELEX</span> <span class='ocrx_word' id='word_1_729' title='bbox 1899 3192 2022 3219; x_wconf 82'>#298949</span> <span class='ocrx_word' id='word_1_730' title='bbox 2033 3193 2125 3219; x_wconf 88'>LINN</span> <span class='ocrx_word' id='word_1_731' title='bbox 2135 3193 2188 3219; x_wconf 89'>UR</span> 
-     </span>
-    </p>
-   </div>
-  </div>
- </body>
-</html>