Temporarily disable multiprocessing while scanning

Seems to have some sporadic consistency problems.

.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
+concurrency =
+	thread
+	multiprocessing
+source =
+	src/ocrmypdf
+omit =
+	tests/spoof/*
 
 [report]
 exclude_lines =

.docker/Dockerfile

@@ -1,50 +1,81 @@
 # OCRmyPDF
 #
-FROM      ubuntu:18.04
+FROM ubuntu:19.10 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 \
+  python3 \
+  python3-distutils \
+  ca-certificates \
+  curl \
+  git
 
-# 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]
+# Get the latest pip (Ubuntu version doesn't support manylinux2010)
+RUN \
+  curl https://bootstrap.pypa.io/get-pip.py | python3
 
-# 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
+# Compile and install jbig2
+# Needs libleptonica-dev, zlib1g-dev
+RUN \
+  mkdir jbig2 \
+  && curl -L https://github.com/agl/jbig2enc/archive/0.29.tar.gz | \
+  tar xz -C jbig2 --strip-components=1 \
+  && cd jbig2 \
+  && ./autogen.sh && ./configure && make && make install \
+  && cd .. \
+  && rm -rf jbig2
 
-# 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
+COPY . /app
 
-RUN useradd docker \
-  && mkdir /home/docker \
-  && chown docker:docker /home/docker
+WORKDIR /app
 
-USER docker
-WORKDIR /home/docker
+RUN pip3 install --no-cache-dir \
+  -r requirements/main.txt \
+  -r requirements/webservice.txt \
+  -r requirements/test.txt \
+  -r requirements/watcher.txt \
+  .
 
-# Must use array form of ENTRYPOINT
-# Non-array form does not append other arguments, because that is "intuitive"
-ENTRYPOINT ["/application/.docker/docker-wrapper.sh"]
+FROM base
+
+ENV LANG=C.UTF-8
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+  ghostscript \
+  img2pdf \
+  liblept5 \
+  libsm6 libxext6 libxrender-dev \
+  zlib1g \
+  pngquant \
+  python3 \
+  qpdf \
+  tesseract-ocr \
+  tesseract-ocr-chi-sim \
+  tesseract-ocr-deu \
+  tesseract-ocr-eng \
+  tesseract-ocr-fra \
+  tesseract-ocr-por \
+  tesseract-ocr-spa \
+  unpaper
+
+WORKDIR /app
+
+COPY --from=builder /usr/local/lib/ /usr/local/lib/
+COPY --from=builder /usr/local/bin/ /usr/local/bin/
+
+COPY --from=builder /app/misc/webservice.py /app/
+COPY --from=builder /app/misc/watcher.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/local/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/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/

.gitattributes

@@ -9,5 +9,6 @@
 *.png	binary
 *.jpg	binary
 *.bin   binary
+*.afdesign  binary
 
 .git_archival.txt  export-subst

.github/FUNDING.yml

@@ -0,0 +1,12 @@
+# These are supported funding model platforms
+
+github: # Replace with up to 4 GitHub Sponsors-enabled usernames e.g., [user1, user2]
+patreon: # Replace with a single Patreon username
+open_collective: james-barlow
+ko_fi: # Replace with a single Ko-fi username
+tidelift: # Replace with a single Tidelift platform-name/package-name e.g., npm/babel
+community_bridge: # Replace with a single Community Bridge project-name e.g., cloud-foundry
+liberapay: # Replace with a single Liberapay username
+issuehunt: # Replace with a single IssueHunt username
+otechie: # Replace with a single Otechie username
+custom: # Replace with up to 4 custom sponsorship URLs e.g., ['link1', 'link2']

.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

@@ -1,45 +1,47 @@
 # Development environment
+.bash_history
+.pylintrc
+.pytest_cache/
+.ruffus_history.sqlite
+.venv*/
 *.pyc
 *.sublime-*
-venv*/
-pyvenv.cfg
-tasks.py
-.bash_history
-.ruffus_history.sqlite
-.idea/
-.pytest_cache/
+*.DS_Store
+.mypy_cache/
 
 # 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
+*.traineddata
+/private

.pre-commit-config.yaml

@@ -0,0 +1,23 @@
+repos:
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v2.4.0
+    hooks:
+      - id: check-case-conflict
+      - id: check-merge-conflict
+      - id: check-toml
+      - id: check-yaml
+      - id: debug-statements
+  - repo: https://github.com/asottile/seed-isort-config
+    rev: v1.9.3
+    hooks:
+      - id: seed-isort-config
+  - repo: https://github.com/pre-commit/mirrors-isort
+    rev: v4.3.21 # pick the isort version you'd like to use from https://github.com/pre-commit/mirrors-isort/releases
+    hooks:
+      - id: isort
+  - repo: https://github.com/psf/black
+    rev: stable
+    hooks:
+      - id: black
+        language_version: python3.7
+        exclude: ^src/ocrmypdf/lib/_leptonica.py

.readthedocs.yml

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

.travis.yml

@@ -1,116 +0,0 @@
-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
-    - 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
-      env: EXTRAS=
-    - os: linux
-      sudo: required
-      language: python
-      python: 3.6
-      env: EXTRAS=
-    - os: linux
-      sudo: required
-      language: python
-      python: 3.6
-      env: EXTRAS=[fitz]
-    - os: osx
-      osx_image: xcode8
-      language: generic
-      env: EXTRAS=
-    - os: osx
-      osx_image: xcode8
-      language: generic
-      env: EXTRAS=[fitz]
-
-before_cache:
-- rm -f $HOME/.cache/pip/log/debug.log
-
-before_install: |
-  mkdir -p bin
-  if [[ "$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
-  elif [[ "$TRAVIS_OS_NAME" == "osx" ]]; then
-    brew update && brew bundle --file=.travis/Brewfile
-    pip3 install --upgrade pip
-    pip3 install wheel
-  fi
-
-install:
-- export PATH=$PWD/bin:$PATH
-- pip3 install ".$EXTRAS"
-- pip3 install -r test_requirements.txt
-
-script:
-- tesseract --version
-- qpdf --version
-- pytest -n auto
-
-deploy:
-  # release for main pypi
-  # 3.6 w/o fitz is considered the build leader and does the deploy, otherwise there is
-  # a race and all versions will try to deploy
-  # OTOH if we ever need separate binary wheels then each version needs its
-  # own deploy
-- provider: pypi
-  user: ocrmypdf-travis
-  password:
-    secure: "DTFOmmNL6olA0+yXvp4u9jXZlZeqrJsJ0526jzqf4a3gZ6jnGTq5UI6WzRsslSyoMMfXKtHQebqHM6ogSgCZinyZ3ufHJo8fn9brxbEc2gsiWkbj5o3bGwdWMT1vNNE7XW0VCpw87rZ1EEwjl4FJHFudMlPR1yfU5+uq0k0PACo="
-  distributions: "sdist bdist_wheel"
-  on:
-    branch: master
-    tags: true
-    condition: $TRAVIS_PYTHON_VERSION == "3.6" && $TRAVIS_OS_NAME == "linux" && $EXTRAS == ""
-  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

@@ -1,15 +0,0 @@
-brew 'python'
-
-brew 'ghostscript'
-brew 'jbig2dec'
-brew 'jbig2enc'
-brew 'leptonica'
-brew 'libffi'
-brew 'libtiff'
-brew 'libxml2'
-brew 'mupdf-tools'
-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

@@ -1,49 +0,0 @@
-# requirements
-include requirements.txt
-include test_requirements.txt
-include dev_requirements.txt
-
-# git
-include .git_archival.txt
-
-# docker
-include .dockerignore
-recursive-include .docker *
-
-# tests
-include .coveragerc
-recursive-include tests *.jpg
-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-include docs *.py
-recursive-include docs *.rst
-recursive-include docs *.svg
-recursive-exclude docs/_build *
-
-
-# support files
-recursive-include src/ocrmypdf/data *
-include *.py
-exclude tasks.py
-recursive-exclude .travis *
-exclude .travis*
-
-
-# code
-recursive-include src/ocrmypdf *.py
-exclude ocrmypdf/lib/_leptonica.py
-
-
-

README.md

@@ -0,0 +1,152 @@
+<img src="docs/images/logo.svg" width="240" alt="OCRmyPDF">
+
+[![Build Status][azure]](https://dev.azure.com/jim0585/ocrmypdf/_build/latest?definitionId=2&branchName=master) [![PyPI version][pypi]](https://pypi.org/project/ocrmypdf/) ![Homebrew version][homebrew] ![ReadTheDocs][docs] ![Python versions][pyversions]
+
+[azure]: https://dev.azure.com/jim0585/ocrmypdf/_apis/build/status/jbarlow83.OCRmyPDF?branchName=master
+
+[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"
+
+[pyversions]: https://img.shields.io/pypi/pyversions/ocrmypdf "Supported Python versions"
+
+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: 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, Windows, macOS and FreeBSD are supported. Docker images are also available.
+
+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 Homebrew users (macOS, Linux, Windows Subsystem for Linux) 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 pack
+
+# Arch Linux users
+pacman -S tesseract-data-eng tesseract-data-deu # Example: Install the English and German language packs
+```
+
+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).
+
+Please report issues on our [GitHub issues](https://github.com/jbarlow83/OCRmyPDF/issues) page, and follow the issue template for quick response.
+
+Requirements
+------------
+
+In addition to the required Python version (3.6+), OCRmyPDF requires external program installations of Ghostscript, Tesseract OCR, QPDF, and Leptonica. OCRmyPDF is pure Python, but uses CFFI to portably generate library bindings. OCRmyPDF works on pretty much everything: Linux, macOS, Windows and FreeBSD.
+
+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](https://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](https://heise.de/-2356670)
+- [heise Durchsuchbare PDF-Dokumente mit OCRmyPDF erstellen](https://www.heise.de/ratgeber/Durchsuchbare-PDF-Dokumente-mit-OCRmyPDF-erstellen-4607592.html)
+
+Business enquiries
+------------------
+
+OCRmyPDF would not be the software that it is today 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

@@ -1,154 +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 3.6, and requires external program installations of Ghostscript, Tesseract OCR, QPDF, and Leptonica. ocrmypdf is pure Python, but uses CFFI to portably generate library bindings.
-
-Python 3.5 is also supported.
-
-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.

azure-pipelines.yml

@@ -0,0 +1,251 @@
+trigger:
+  tags:
+    include:
+      - v*
+  branches:
+    include:
+      - "*"
+    exclude:
+      - "travis"
+
+stages:
+  - stage: "Test"
+    jobs:
+      - job: Windows
+        pool:
+          vmImage: "vs2017-win2016"
+        strategy:
+          matrix:
+            Python36:
+              python.version: "3.6"
+            Python37:
+              python.version: "3.7"
+            Python38:
+              python.version: "3.8"
+        steps:
+          - task: UsePythonVersion@0
+            inputs:
+              versionSpec: "$(python.version)"
+          - pwsh: |
+              choco install --yes --no-progress --pre tesseract
+              choco install --yes --no-progress python3
+              choco install --yes --no-progress ghostscript
+              choco install --yes --no-progress pngquant
+            displayName: "Install system packages"
+          - pwsh: |
+              refreshenv
+              python -m pip install --upgrade pip wheel
+              python -m pip install -r requirements/main.txt -r requirements/test.txt .
+            displayName: "Install Python packages"
+          - pwsh: |
+              refreshenv
+              $env:pathext += ';.py'
+              # -n auto helps Windows
+              python -m pytest -n auto --junitxml=test.xml --cov=ocrmypdf --cov-report=xml
+            displayName: "Test"
+          - task: PublishTestResults@2
+            inputs:
+              testResultsFiles: "test.xml"
+              testRunTitle: "$(Agent.OS) - $(Build.DefinitionName) - Python $(python.version)"
+            condition: succeededOrFailed()
+      - job: "Ubuntu_1804"
+        pool:
+          vmImage: "ubuntu-18.04"
+        strategy:
+          matrix:
+            Python36:
+              python.version: "3.6"
+            Python37:
+              python.version: "3.7"
+            Python38:
+              python.version: "3.8"
+        steps:
+          - task: UsePythonVersion@0
+            inputs:
+              versionSpec: "$(python.version)"
+          - bash: |
+              sudo apt-get update
+              sudo apt-get install -y --no-install-recommends \
+                python3-software-properties \
+                curl \
+                ghostscript \
+                img2pdf \
+                libexempi3 \
+                libffi-dev \
+                liblept5 \
+                libsm6 libxext6 libxrender-dev \
+                pngquant \
+                poppler-utils \
+                tesseract-ocr \
+                tesseract-ocr-deu \
+                tesseract-ocr-eng \
+                unpaper \
+                zlib1g
+            displayName: "Install system packages"
+          - bash: |
+              curl https://bootstrap.pypa.io/get-pip.py | python3
+              pip3 install -r requirements/main.txt -r requirements/test.txt .
+            displayName: "Install Python packages"
+          - bash: |
+              tesseract --version
+            displayName: "Record versions"
+          - bash: |
+              # -n auto is slower on Linux and breaks on Python 3.8
+              pytest -n0 --junitxml=test.xml --cov=ocrmypdf --cov-report=xml
+            displayName: "Test"
+          - task: PublishTestResults@2
+            inputs:
+              testResultsFiles: "test.xml"
+              testRunTitle: "$(Agent.OS) - $(Build.DefinitionName) - Python $(python.version)"
+            condition: succeededOrFailed()
+      - job: "Ubuntu_1604"
+        pool:
+          vmImage: "ubuntu-16.04"
+        strategy:
+          matrix:
+            Python36:
+              python.version: "3.6"
+        steps:
+          - task: UsePythonVersion@0
+            inputs:
+              versionSpec: "$(python.version)"
+          - bash: |
+              sudo apt-get update
+              sudo apt-get install -y --no-install-recommends \
+                software-properties-common
+              sudo add-apt-repository -y ppa:alex-p/tesseract-ocr
+              sudo apt-get update
+              sudo apt-get install -y --no-install-recommends \
+                ghostscript \
+                img2pdf \
+                libexempi3 \
+                libffi-dev \
+                liblept5 \
+                libsm6 libxext6 libxrender-dev \
+                pngquant \
+                poppler-utils \
+                tesseract-ocr \
+                tesseract-ocr-deu \
+                tesseract-ocr-eng \
+                unpaper \
+                zlib1g
+            displayName: "Install system packages"
+          - bash: |
+              curl https://bootstrap.pypa.io/get-pip.py | python3
+              pip3 install -r requirements/main.txt -r requirements/test.txt .
+            displayName: "Install Python packages"
+          - bash: |
+              tesseract --version
+            displayName: "Record versions"
+          - bash: |
+              # -n auto is slower on Linux and breaks on Python 3.8
+              pytest -n0 --junitxml=test.xml --cov=ocrmypdf --cov-report=xml
+            displayName: "Test"
+          - task: PublishTestResults@2
+            inputs:
+              testResultsFiles: "test.xml"
+              testRunTitle: "$(Agent.OS) - $(Build.DefinitionName) - Python $(python.version)"
+            condition: succeededOrFailed()
+      - job: "macOS_Mojave"
+        pool:
+          vmImage: "macos-10.14"
+        strategy:
+          matrix:
+            Python37:
+              python.version: "3.7"
+            Python38:
+              python.version: "3.8"
+        steps:
+          - task: UsePythonVersion@0
+            inputs:
+              versionSpec: "$(python.version)"
+          - bash: |
+              brew update
+              brew unlink python@2
+              brew install \
+                exempi \
+                ghostscript \
+                jbig2enc \
+                leptonica \
+                openjpeg \
+                pngquant \
+                tesseract \
+                unpaper
+            displayName: "Install system packages"
+          - bash: |
+              pip3 install --upgrade pip
+              pip3 install -r requirements/main.txt -r requirements/test.txt .
+            displayName: "Install Python packages"
+          - bash: |
+              tesseract --version
+            displayName: "Record versions"
+          - bash: pytest -nauto --junitxml=test.xml --cov=ocrmypdf --cov-report=xml
+            displayName: "Test"
+          - task: PublishTestResults@2
+            inputs:
+              testResultsFiles: "test.xml"
+              testRunTitle: "$(Agent.OS) - $(Build.DefinitionName) - Python $(python.version)"
+            condition: succeededOrFailed()
+          - task: PublishCodeCoverageResults@1
+            inputs:
+              codeCoverageTool: Cobertura
+              summaryFileLocation: "$(System.DefaultWorkingDirectory)/**/coverage.xml"
+
+  - stage: "Artifacts"
+    jobs:
+      - job: "sdist_wheel"
+        pool:
+          vmImage: "ubuntu-18.04"
+        steps:
+          - task: UsePythonVersion@0
+            inputs:
+              versionSpec: "3.7"
+          - bash: |
+              python -m pip install --upgrade pip wheel
+              python setup.py sdist bdist_wheel
+          - publish: dist
+            artifact: sdist_wheel
+
+  - stage: "Deploy"
+    jobs:
+      - deployment: "PyPI"
+        pool:
+          vmImage: "ubuntu-18.04"
+        environment: "deploy"
+        strategy:
+          runOnce:
+            deploy:
+              steps:
+                - download: current
+                  artifact: sdist_wheel
+                - script: |
+                    mkdir -p dist
+                    mv $(Pipeline.Workspace)/sdist_wheel/* dist
+                  displayName: "Move dist files"
+                - task: UsePythonVersion@0
+                  inputs:
+                    versionSpec: "3.8"
+                    architecture: x64
+                - script: |
+                    pip install --upgrade twine
+                  displayName: "Generate artifacts"
+                - script: |
+                    cat <<FILE >.pypirc
+                    [distutils]
+                    index-servers =
+                        pypi
+
+                    [pypi]
+                    username: __token__
+                    password: $(TOKEN_PYPI)
+
+                    FILE
+                  displayName: "Generate PyPI auth file"
+                - script: |
+                    python -m twine upload --config-file .pypirc dist/*
+                  displayName: "Upload to PyPI"
+                  condition: and(succeeded(), startsWith(variables['Build.SourceBranch'], 'refs/tags/'))
+                - script: |
+                    curl -X POST -d "token=$(TOKEN_RTD)" https://readthedocs.org/api/v2/webhook/pikepdf/39557/
+                  displayName: "Trigger ReadTheDocs"
+                  condition: and(succeeded(), or(startsWith(variables['Build.SourceBranch'], 'refs/tags/'), startsWith(variables['Build.SourceBranch'], 'refs/heads/master')))

dev_requirements.txt

@@ -1,4 +0,0 @@
-check-manifest >= 0.35
-twine >= 1.8.1
-coverage >= 4.4
-GitPython == 2.1.3

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.  
+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.
+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
@@ -151,39 +260,85 @@ user interface.
         - Name
         - Interpretation
     *	- 0
-        - ``ocrmypdf.exceptions.ExitCode.ok``
+        - ``ExitCode.ok``
         - Everything worked as expected.
     *	- 1
-        - ``ocrmypdf.exceptions.ExitCode.bad_args``
+        - ``ExitCode.bad_args``
         - Invalid arguments, exited with an error.
     *	- 2
-        - ``ocrmypdf.exceptions.ExitCode.input_file``
+        - ``ExitCode.input_file``
         - The input file does not seem to be a valid PDF.
     *	- 3
-        - ``ocrmypdf.exceptions.missing_dependency``
+        - ``ExitCode.missing_dependency``
         - An external program required by OCRmyPDF is missing.
     *	- 4
-        - ``ocrmypdf.exceptions.invalid_output_pdf``
-        - An output file was created, but it does not seem to be a valid PDF or
-          PDF/A. The file will be available.
+        - ``ExitCode.invalid_output_pdf``
+        - An output file was created, but it does not seem to be a valid PDF. The file will be available.
     *	- 5
-        - ``ocrmypdf.exceptions.file_access_error``
+        - ``ExitCode.file_access_error``
         - The user running OCRmyPDF does not have sufficient permissions to read the input file and write the output file.
     *	- 6
-        - ``ocrmypdf.exceptions.already_done_ocr``
+        - ``ExitCode.already_done_ocr``
         - The file already appears to contain text so it may not need OCR. See output message.
     *	- 7
-        - ``ocrmypdf.exceptions.child_process_error``
+        - ``ExitCode.child_process_error``
         - An error occurred in an external program (child process) and OCRmyPDF cannot continue.
     *	- 8
-        - ``ocrmypdf.exceptions.encrypted_pdf``
+        - ``ExitCode.encrypted_pdf``
         - The input PDF is encrypted. OCRmyPDF does not read encrypted PDFs. Use another program such as ``qpdf`` to remove encryption.
     *	- 9
-        - ``ocrmypdf.exceptions.invalid_config``
+        - ``ExitCode.invalid_config``
         - A custom configuration file was forwarded to Tesseract using ``--tesseract-config``, and Tesseract rejected this file.
+    *   - 10
+        - ``ExitCode.pdfa_conversion_failed``
+        - A valid PDF was created, PDF/A conversion failed. The file will be available.
     *	- 15
-        - ``ocrmypdf.exceptions.other_error``
+        - ``ExitCode.other_error``
         - Some other error occurred.
     *	- 130
-        - ``ocrmypdf.exceptions.ctrl_c``
+        - ``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:
+
+-  ``_rasterize.png`` - what the input page looks like
+-  ``_ocr.png`` - the file that is sent to Tesseract for OCR; depending
+   on arguments this may differ from the presentation image
+-  ``_pp_deskew.png`` - the image, after deskewing
+-  ``_pp_clean.png`` - the image, after cleaning with unpaper
+-  ``_ocr_tess.pdf`` - the OCR file; appears as a blank page with invisible
+   text embedded
+-  ``_ocr_tess.txt`` - the OCR text (not necessarily all text on the page,
+   if the page is mixed format)
+-  ``fix_docinfo.pdf`` - a temporary file created to fix the PDF DocumentInfo
+   data structure
+-  ``graft_layers.pdf`` - the rendered PDF with OCR layers grafted on
+-  ``pdfa.pdf`` - ``graft_layers.pdf`` after conversion to PDF/A
+-  ``pdfa.ps`` - a PostScript file used by Ghostscript for PDF/A conversion
+-  ``optimize.pdf`` - the PDF generated before optimization
+-  ``optimize.out.pdf`` - the PDF generated by optimization
+-  ``origin`` - the input file
+-  ``origin.pdf`` - the input file or the input image converted to PDF
+-  ``images/*`` - images extracted during the optimization process; here
+   the prefix indicates a PDF object ID not a page number

docs/api.rst

@@ -0,0 +1,112 @@
+======================
+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.ocr('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.ocr` 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
+- execute other subprocesses (forking and executing other programs)
+
+The Python process that calls ``ocrmypdf.ocr()`` 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.ocr()`` is suggested. That
+way your application will survive and remain interactive even if
+OCRmyPDF does not.
+
+.. warning::
+
+    On Windows, the script that calls ``ocrmypdf.ocr()`` must be protected
+    by an "ifmain" guard (``if __name__ == '__main__'``) or you must use
+    ``ocrmypdf.ocr(...use_threads=True)``. If you do not take at least one
+    of these steps, Windows fork semantics will prevent OCRmyPDF from working
+    correct.
+
+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. Use ``ocrmypdf.ocr(...progress_bar=False)`` to disable
+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.ocr
+
+.. autoclass:: ocrmypdf.Verbosity
+    :members:
+    :undoc-members:
+
+.. autoclass:: ocrmypdf.ExitCode
+    :members:
+    :undoc-members:
+
+.. autofunction:: ocrmypdf.configure_logging

docs/batch.rst

@@ -1,214 +1,224 @@
+================
 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 '<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 '{}' '{}'
+
+In a Windows batch file, use
+
+.. code-block:: bat
+
+   for /r %%f in (*.pdf) do ocrmypdf %%f %%f
 
 Sample script
-"""""""""""""
+-------------
 
-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
-
-	import logging
-	import os
-	import subprocess
-	import sys
-
-	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) > 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.)
+This user contributed script also provides an example of batch
+processing.
 
+.. literalinclude:: ../misc/batch.py
+    :caption: misc/batch.py
 
 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.
-
-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
-
-	# 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
-
-	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')
-
-
-
-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.
-
-
-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).
+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.
 
-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.
+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.
+
+.. literalinclude:: ../misc/synology.py
+    :caption: misc/synology.py - Sample script for Synology DiskStations
+
+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.
+
+Hot (watched) folders
+=====================
+
+Watched folders with watcher.py
+-------------------------------
+
+OCRmyPDF has a folder watcher called watcher.py, which is currently included in source
+distributions but not part of the main program. It may be used natively or may run
+in a Docker container. Native instances tend to give better performance. watcher.py
+works on all platforms.
+
+Users may need to customize the script to meet their requirements.
 
 .. code-block:: bash
 
-	pip install watchdog
+    pip3 install -r requirements/watcher.txt
 
-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.
+    env OCR_INPUT_DIRECTORY=/mnt/input-pdfs \
+        OCR_OUTPUT_DIRECTORY=/mnt/output-pdfs \
+        OCR_OUTPUT_DIRECTORY_YEAR_MONTH=1 \
+        python3 watcher.py
+
+.. csv-table:: watcher.py environment variables
+    :header: "Environment variable", "Description"
+    :widths: 50, 50
+
+    "OCR_INPUT_DIRECTORY", "Set input directory to monitor (recursive)"
+    "OCR_OUTPUT_DIRECTORY", "Set output directory (should not be under input)"
+    "OCR_ON_SUCCESS_DELETE", "This will delete the input file if the exit code is 0 (OK)"
+    "OCR_OUTPUT_DIRECTORY_YEAR_MONTH", "This will place files in the output in ``{output}/{year}/{month}/{filename}``"
+    "OCR_DESKEW", "Apply deskew to crooked input PDFs"
+    "OCR_JSON_SETTINGS", "A JSON string specifying any other arguments for ``ocrmypdf.ocr``"
+    "OCR_POLL_NEW_FILE_SECONDS", "Polling interval"
+    "OCR_LOGLEVEL", "Level of log messages to report"
+
+One could configure a networked scanner or scanning computer to drop files in the
+watched folder.
+
+Watched folders with Docker
+---------------------------
+
+The watcher service is included in the OCRmyPDF Docker image. To run it:
 
 .. 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
+    docker run \
+        -v <path to files to convert>:/input \
+        -v <path to store results>:/output \
+        -e OCR_OUTPUT_DIRECTORY_YEAR_MONTH=1 \
+        -e OCR_ON_SUCCESS_DELETE=1 \
+        -e OCR_DESKEW=1 \
+        -e PYTHONUNBUFFERED=1 \
+        -it --entrypoint python3 \
+        jbarlow83/ocrmypdf \
+        watcher.py
 
-For more complex behavior you can write a Python script around to use the watchdog API.
+This service will watch for a file that matches ``/input/\*.pdf`` and will
+convert it to a OCRed PDF in ``/output/``. The parameters to this image are:
 
-On file servers, you could configure watchmedo as a system service so it will run all the time.
+.. csv-table:: watcher.py parameters for Docker
+    :header: "Parameter", "Description"
+    :widths: 50, 50
+
+    "``-v <path to files to convert>:/input``", "Files placed in this location will be OCRed"
+    "``-v <path to store results>:/output``", "This is where OCRed files will be stored"
+    "``-e OCR_OUTPUT_DIRECTORY_YEAR_MONTH=1``", "Define environment variable OCR_OUTPUT_DIRECTORY_YEAR_MONTH=1"
+    "``-e OCR_ON_SUCCESS_DELETE=1``", "Define environment variable"
+    "``-e OCR_DESKEW=1``", "Define environment variable"
+    "``-e PYTHONBUFFERED=1``", "This will force STDOUT to be unbuffered and allow you to see messages in docker logs"
+
+This service relies on polling to check for changes to the filesystem. It
+may not be suitable for some environments, such as filesystems shared on a
+slow network.
+
+A configuration manager such as Docker Compose could be used to ensure that the
+service is always available.
+
+.. literalinclude:: ../misc/docker-compose.example.yml
+    :language: yaml
+    :caption: misc/docker-compose.example.yml
 
 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``.
+-  On Linux, `systemd user services <https://wiki.archlinux.org/index.php/Systemd/User>`__
+   can be configured to automatically perform OCR on a collection of files.
 
+-  `Watchman <https://facebook.github.io/watchman/>`__ is a more
+   powerful alternative to ``watchmedo``.
 
+AWS Lambda is not viable
+------------------------
+
+AWS Lambda and its equivalents have low limits on execution time and payload
+size, relative to OCRmyPDF's needs. As of this writing, the request/response
+payload for AWS Lambda was 6 MB, which means many PDFs will not fit.
+
+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:
+
+.. figure:: images/macos-workflow.png
+    :alt: Example macOS Automator workflow
+
+You may customize the command sent to ocrmypdf.

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/contributing.rst

@@ -0,0 +1,57 @@
+=======================
+Contributing guidelines
+=======================
+
+Contributions are welcome!
+
+Big changes
+===========
+
+Please open a new issue to discuss or propose a major change. Not only is it fun
+to discuss big ideas, but we might save each other's time too. Perhaps some of the
+work you're contemplating is already half-done in a development branch.
+
+Code style
+==========
+
+We use PEP8, ``black`` for code formatting and ``isort`` for import sorting. The
+settings for these programs are in ``pyproject.toml`` and ``setup.cfg``. Pull
+requests should follow the style guide. One difference we use from "black" style
+is that strings shown to the user are always in double quotes (``"``) and strings
+for internal uses are in single quotes (``'``).
+
+Tests
+=====
+
+New features should come with tests that confirm their correctness.
+
+New Python dependencies
+=======================
+
+If you are proposing a change that will require a new Python dependency, we
+prefer dependencies that are already packaged by Debian or Red Hat. This makes
+life much easier for our downstream package maintainers.
+
+Python dependencies must also be GPLv3 compatible.
+
+New non-Python dependencies
+===========================
+
+OCRmyPDF uses several external programs (Tesseract, Ghostscript and others) for
+its functionality. In general we prefer to avoid adding new external programs.
+
+Style guide: Is it OCRmyPDF or ocrmypdf?
+========================================
+
+The program/project is OCRmyPDF and the name of the executable or library is ocrmypdf.
+
+Known ports/packagers
+=====================
+
+OCRmyPDF has been ported to many platforms already. If you are interesting in
+porting to a new platform, check with
+`Repology <https://repology.org/projects/?search=ocrmypdf>`__ to see the status
+of that platform.
+
+Packager maintainers, please ensure that the command line completion scripts in
+``misc/`` are installed.

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.
 
@@ -45,45 +45,70 @@ 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
 
     ocrmypdf --sidecar output.txt input.pdf output.pdf
 
+.. note::
+
+    The sidecar file contains the **OCR text** found by OCRmyPDF. If the document
+    contains pages that already have text, that text will not appear in the
+    sidecar. If the option ``--pages`` is used, only those pages on which OCR
+    was performed will be included in the sidecar. If certain pages were skipped
+    because of options like ``--skip-big`` or ``--tesseract-timeout``, those pages
+    will not be in the sidecar.
+
+    To extract all text from a PDF, whether generated from OCR or otherwise,
+    use a program like Poppler's ``pdftotext`` or ``pdfgrep``.
+
 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 +119,88 @@ 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.org/skew-measurement.html>`__ algorithm as
+   implemented in `Leptonica <http://www.leptonica.org/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 +208,148 @@ 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
 
+Optimize images without performing 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 ``--force-ocr`` argument. Normally, OCRmyPDF does not modify files that already appear to contain OCR text.
+You can also optimize all images without performing any OCR:
 
 .. code-block:: bash
 
-    ocrmypdf --force-ocr input.pdf output.pdf
+    ocrmypdf --tesseract-timeout=0 --optimize 3 --skip-text input.pdf output.pdf
 
-Note that the method above will force rasterization of all pages, potentially reducing quality or losing vector content. 
+Perform OCR only certain pages
+------------------------------
 
-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.
+You can ask OCRmyPDF to only apply OCR to certain pages.
 
 .. 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
+    ocrmypdf --pages 2,3,13-17 input.pdf output.pdf
 
-``--output-type pdf`` is used here to avoid using Ghostscript which will also rasterize images.
+Hyphens denote a range of pages and commas separate page numbers. If you prefer
+to use spaces, quote all of the page numbers: ``--pages '2, 3, 5, 7'``.
 
+OCRmyPDF will warn if your list of page numbers contains duplicates or
+overlap pages. OCRmyPDF does not currently account for document page numbers,
+such as an introduction section of a book that uses Roman numerals. It simply
+counts the number of virtual pieces of paper since the start.
+
+Regardless of the argument to ``--pages``, OCRmyPDF will optimize all pages in
+the file and convert it to PDF/A, unless you disable those options. In this
+example, we want to OCR only the title and otherwise change the PDF as little
+as possible:
+
+.. code-block:: bash
+
+    ocrmypdf --pages 1 --output-type pdf --optimize 0 input.pdf output.pdf
+
+Redo existing 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.)
+
+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 --redo-ocr input.pdf output.pdf
+
+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.
+
+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 +1,172 @@
+=====================
+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.
+
+If you can run this command successfully, your system is ready to download and
+execute the image:
+
+.. code-block:: bash
+
+   docker run hello-world
+
+The recommended OCRmyPDF Docker image is currently named ``ocrmypdf``:
+
+.. code-block:: bash
+
+   docker pull jbarlow83/ocrmypdf
+
+
+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")
+
+See the Docker documentation for
+`adjusting memory and CPU on other platforms <https://docs.docker.com/config/containers/resource_constraints/>`__.
+
+Using the Docker image on the command line
+==========================================
+
+**Unlike typical Docker containers**, in this section the OCRmyPDF Docker
+container is emphemeral – it runs for one OCR job and terminates, just like a
+command line program. We are using Docker to deliver an application (as opposed
+to the more conventional case, where a Docker container runs as a server).
+
+To start a Docker container (instance of the image):
+
+.. code-block:: bash
+
+   docker tag jbarlow83/ocrmypdf 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 as stdin and read the output from
+stdout – **this avoids the messy permission issues with Docker entirely**.
+
+.. code-block:: bash
+
+   alias docker_ocrmypdf='docker run --rm -i ocrmypdf'
+   docker_ocrmypdf --version  # runs docker version
+   docker_ocrmypdf - - <input.pdf >output.pdf
+
+Or in the wonderful `fish shell <https://fishshell.com/>`__:
+
+.. code-block:: fish
+
+   alias docker_ocrmypdf 'docker run --rm ocrmypdf'
+   funcsave docker_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
+
+(However, when done this way, ``output.pdf`` may be owned by the root
+user.)
+
+.. _docker-lang-packs:
+
+Adding languages to the Docker image
+====================================
+
+By default the Docker image includes English, German, Simplified Chinese,
+French, Portuguese and Spanish, 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
+
+   # Add French
+   RUN apt install tesseract-ocr-fra
+
+You can also copy training data to ``/usr/share/tesseract-ocr/<tesseract version>/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 -m pytest
+
+Accessing the shell
+===================
+
+To use the bash shell in the Docker image:
+
+.. code-block:: bash
+
+   docker run -it --entrypoint bash  jbarlow83/ocrmypdf
+
+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 webservice.py
+
+This will configure the machine to listen on port 5000. On Linux machines
+this is port 5000 of localhost. On macOS or Windows machines running
+Docker, this is port 5000 of the virtual machine that runs your Docker
+images. You can find its IP address using the command ``docker-machine ip``.
+
+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>`.

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/images/logo.svg

@@ -0,0 +1,75 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg width="100%" height="100%" viewBox="0 0 503 227" version="1.1" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" xml:space="preserve" xmlns:serif="http://www.serif.com/" style="fill-rule:evenodd;clip-rule:evenodd;stroke-linecap:round;stroke-linejoin:round;stroke-miterlimit:1.5;">
+    <g id="svg" transform="matrix(0.965977,0,0,0.807602,0,0)">
+        <rect x="0" y="0" width="520" height="280" style="fill:white;"/>
+        <g transform="matrix(1.03522,0,0,1.23823,-69.7528,-83.422)">
+            <g transform="matrix(1,0,0,1,243.977,20.0703)">
+                <g id="Page">
+                    <g transform="matrix(0.961773,0,0,1.05962,6.19811,-3.01071)">
+                        <path d="M328.5,97.682C328.5,96.465 327.983,95.296 327.056,94.418C320.026,87.758 289.442,58.78 282.228,51.944C281.251,51.019 279.901,50.496 278.49,50.496C264.493,50.496 188.083,50.496 167.339,50.496C164.468,50.496 162.141,52.609 162.141,55.214C162.141,83.051 162.141,225.565 162.141,253.4C162.141,256.005 164.468,258.117 167.338,258.117C192.242,258.117 299.159,258.117 323.538,258.117C326.278,258.117 328.5,256.101 328.5,253.613C328.5,229.268 328.5,113.896 328.5,97.682Z" style="fill:rgb(253,253,253);stroke:rgb(51,51,51);stroke-width:3.95px;"/>
+                    </g>
+                    <g id="Dog-ear" serif:id="Dog ear" transform="matrix(1,0,0,1,-4,2)">
+                        <path d="M277.072,48.496L277.072,93.848C277.072,95.172 277.598,96.441 278.534,97.377C279.47,98.313 280.739,98.839 282.063,98.839C294.548,98.839 326.141,98.839 326.141,98.839" style="fill:rgb(245,245,245);stroke:rgb(51,51,51);stroke-width:4px;"/>
+                    </g>
+                </g>
+                <g transform="matrix(1,0,0,1,-29.6816,-0.395178)">
+                    <g transform="matrix(1.00243,0,0,1.11818,-144.72,-8.80181)">
+                        <path d="M465.73,119.654C465.73,117.605 463.874,115.941 461.588,115.941L310.259,115.941C307.973,115.941 306.117,117.605 306.117,119.654L306.117,183.108C306.117,185.157 307.973,186.821 310.259,186.821L461.588,186.821C463.874,186.821 465.73,185.157 465.73,183.108L465.73,119.654Z" style="fill:rgb(248,0,0);stroke:white;stroke-width:3.77px;"/>
+                    </g>
+                    <g transform="matrix(1.24571,0,0,1.35864,116.812,84.3924)">
+                        <g transform="matrix(64,0,0,64,42.1437,77.6203)">
+                            <path d="M0.084,0L0.084,-0.68L0.297,-0.68C0.371,-0.68 0.434,-0.663 0.487,-0.63C0.54,-0.596 0.566,-0.54 0.566,-0.462C0.566,-0.385 0.538,-0.328 0.481,-0.292C0.424,-0.255 0.36,-0.237 0.288,-0.237L0.213,-0.237L0.213,0L0.084,0ZM0.293,-0.572L0.213,-0.572L0.213,-0.344L0.295,-0.344C0.334,-0.344 0.365,-0.353 0.389,-0.371C0.413,-0.388 0.426,-0.416 0.429,-0.454C0.429,-0.498 0.417,-0.529 0.393,-0.546C0.369,-0.563 0.336,-0.572 0.293,-0.572Z" style="fill:white;fill-rule:nonzero;"/>
+                        </g>
+                        <g transform="matrix(64,0,0,64,79.7117,77.6203)">
+                            <path d="M0.332,0L0.084,0L0.084,-0.68L0.336,-0.68C0.441,-0.68 0.518,-0.648 0.569,-0.585C0.62,-0.522 0.645,-0.441 0.645,-0.344C0.645,-0.239 0.618,-0.155 0.563,-0.093C0.508,-0.031 0.431,0 0.332,0ZM0.337,-0.57L0.213,-0.57L0.213,-0.109L0.33,-0.109C0.385,-0.109 0.429,-0.127 0.462,-0.163C0.495,-0.199 0.511,-0.259 0.511,-0.344C0.511,-0.415 0.497,-0.47 0.469,-0.51C0.441,-0.55 0.397,-0.57 0.337,-0.57Z" style="fill:white;fill-rule:nonzero;"/>
+                        </g>
+                        <g transform="matrix(64,0,0,64,123.424,77.6203)">
+                            <path d="M0.405,-0.288L0.213,-0.288L0.213,0L0.084,0L0.084,-0.68L0.469,-0.68L0.489,-0.578L0.213,-0.578L0.213,-0.389L0.386,-0.389L0.405,-0.288Z" style="fill:white;fill-rule:nonzero;"/>
+                        </g>
+                    </g>
+                </g>
+            </g>
+            <g transform="matrix(1,0,0,1.52217,67.3796,10.7507)">
+                <rect x="23.501" y="81.3" width="162.305" height="61.77" style="fill:rgb(180,213,255);"/>
+            </g>
+            <g transform="matrix(0.967536,0,0,0.961535,5.90498,47.9703)">
+                <g transform="matrix(90.4804,0,0,90.4804,82.6698,167.705)">
+                    <path d="M0.057,-0.337C0.057,-0.442 0.084,-0.527 0.139,-0.594C0.194,-0.66 0.271,-0.694 0.37,-0.696C0.477,-0.696 0.556,-0.662 0.607,-0.593C0.658,-0.524 0.684,-0.441 0.684,-0.344C0.684,-0.239 0.657,-0.153 0.602,-0.086C0.547,-0.019 0.469,0.014 0.37,0.014C0.264,0.014 0.185,-0.02 0.134,-0.089C0.083,-0.157 0.057,-0.24 0.057,-0.337ZM0.192,-0.338C0.192,-0.267 0.206,-0.208 0.235,-0.163C0.264,-0.118 0.308,-0.095 0.369,-0.095C0.424,-0.095 0.467,-0.115 0.5,-0.156C0.533,-0.197 0.549,-0.259 0.549,-0.344C0.549,-0.415 0.535,-0.473 0.506,-0.518C0.477,-0.563 0.433,-0.586 0.372,-0.586C0.319,-0.586 0.275,-0.564 0.242,-0.519C0.209,-0.474 0.192,-0.414 0.192,-0.338Z" style="fill:rgb(51,51,51);fill-rule:nonzero;"/>
+                </g>
+                <g transform="matrix(90.4804,0,0,90.4804,147.906,167.705)">
+                    <path d="M0.505,-0.557C0.473,-0.567 0.448,-0.574 0.429,-0.579C0.41,-0.583 0.388,-0.585 0.361,-0.585C0.307,-0.585 0.265,-0.563 0.236,-0.519C0.207,-0.475 0.192,-0.415 0.192,-0.338C0.192,-0.272 0.204,-0.215 0.229,-0.167C0.254,-0.119 0.295,-0.095 0.353,-0.095C0.382,-0.095 0.409,-0.098 0.434,-0.104C0.459,-0.11 0.481,-0.117 0.502,-0.126L0.551,-0.03C0.525,-0.017 0.494,-0.006 0.457,0.002C0.42,0.01 0.388,0.014 0.36,0.014C0.254,0.014 0.177,-0.02 0.129,-0.088C0.081,-0.156 0.057,-0.239 0.057,-0.337C0.057,-0.442 0.084,-0.527 0.137,-0.594C0.19,-0.661 0.266,-0.694 0.365,-0.694C0.385,-0.694 0.413,-0.691 0.448,-0.684C0.483,-0.677 0.516,-0.666 0.545,-0.65L0.505,-0.557Z" style="fill:rgb(51,51,51);fill-rule:nonzero;"/>
+                </g>
+                <g transform="matrix(90.4804,0,0,90.4804,199.751,167.705)">
+                    <path d="M0.293,-0.572L0.213,-0.572L0.213,-0.364L0.295,-0.364C0.334,-0.364 0.366,-0.372 0.391,-0.388C0.416,-0.403 0.429,-0.429 0.429,-0.465C0.429,-0.503 0.417,-0.53 0.393,-0.547C0.369,-0.564 0.336,-0.572 0.293,-0.572ZM0.479,0L0.335,-0.26C0.328,-0.259 0.32,-0.259 0.312,-0.259C0.304,-0.258 0.296,-0.258 0.288,-0.258L0.213,-0.258L0.213,0L0.084,0L0.084,-0.68L0.297,-0.68C0.371,-0.68 0.434,-0.663 0.487,-0.629C0.54,-0.595 0.566,-0.542 0.566,-0.471C0.566,-0.429 0.555,-0.393 0.534,-0.363C0.512,-0.332 0.484,-0.309 0.45,-0.292L0.617,0L0.479,0Z" style="fill:rgb(51,51,51);fill-rule:nonzero;"/>
+                </g>
+            </g>
+            <g transform="matrix(0.916882,0,0,1,121.475,-32.6535)">
+                <g transform="matrix(86.953,0,0,86.953,152.996,241.878)">
+                    <path d="M0.479,-0.428C0.5,-0.451 0.527,-0.47 0.562,-0.484C0.596,-0.497 0.627,-0.504 0.654,-0.504C0.72,-0.504 0.767,-0.485 0.795,-0.446C0.822,-0.407 0.836,-0.36 0.836,-0.304L0.836,0L0.705,0L0.705,-0.298C0.705,-0.329 0.698,-0.352 0.683,-0.369C0.668,-0.385 0.647,-0.393 0.619,-0.393C0.6,-0.393 0.581,-0.388 0.56,-0.378C0.539,-0.368 0.521,-0.357 0.504,-0.344C0.505,-0.337 0.505,-0.331 0.506,-0.324C0.507,-0.317 0.507,-0.311 0.507,-0.304L0.507,0L0.376,0L0.376,-0.298C0.376,-0.329 0.369,-0.352 0.354,-0.369C0.339,-0.385 0.318,-0.393 0.291,-0.393C0.274,-0.393 0.258,-0.39 0.241,-0.383C0.224,-0.376 0.207,-0.367 0.192,-0.356L0.192,0L0.062,0L0.062,-0.485L0.13,-0.485L0.162,-0.441C0.184,-0.461 0.211,-0.476 0.242,-0.488C0.273,-0.499 0.3,-0.504 0.325,-0.504C0.363,-0.504 0.395,-0.497 0.42,-0.484C0.445,-0.47 0.465,-0.451 0.479,-0.428Z" style="fill:rgb(51,51,51);fill-rule:nonzero;"/>
+                </g>
+                <g transform="matrix(86.953,0,0,86.953,228.906,241.878)">
+                    <path d="M0.156,0.023L0.179,-0.034L0.006,-0.467L0.14,-0.485L0.252,-0.191L0.358,-0.485L0.495,-0.485L0.278,0.064C0.263,0.103 0.236,0.137 0.197,0.165C0.158,0.193 0.118,0.212 0.075,0.222L0.029,0.115C0.052,0.105 0.077,0.093 0.104,0.079C0.13,0.064 0.147,0.046 0.156,0.023Z" style="fill:rgb(51,51,51);fill-rule:nonzero;"/>
+                </g>
+            </g>
+            <g id="Selectors" transform="matrix(0.965977,0,0,0.807602,67.3796,67.3718)">
+                <g id="Right-selector" serif:id="Right selector">
+                    <g transform="matrix(1.03522,0,0,1.23823,2.07044,0)">
+                        <path d="M185.806,161.156L185.806,67.132" style="fill:none;stroke:rgb(76,159,255);stroke-width:4px;stroke-linecap:butt;"/>
+                    </g>
+                    <g transform="matrix(1.03522,0,0,1.23823,161.788,169.469)">
+                        <circle cx="31.523" cy="34.314" r="10.021" style="fill:rgb(76,159,255);stroke:rgb(76,159,255);stroke-width:4px;stroke-linecap:butt;"/>
+                    </g>
+                </g>
+                <g id="Left-selector" serif:id="Left selector">
+                    <g transform="matrix(1.03522,0,0,1.23823,-170.092,0)">
+                        <path d="M185.806,161.156L185.806,67.132" style="fill:none;stroke:rgb(76,159,255);stroke-width:4px;stroke-linecap:butt;"/>
+                    </g>
+                    <g transform="matrix(1.03522,0,0,1.23823,-10.3742,28.2274)">
+                        <circle cx="31.523" cy="34.314" r="10.021" style="fill:rgb(76,159,255);stroke:rgb(76,159,255);stroke-width:4px;stroke-linecap:butt;"/>
+                    </g>
+                </g>
+            </g>
+        </g>
+    </g>
+</svg>

docs/index.rst

@@ -1,15 +1,12 @@
-.. ocrmypdf documentation master file, created by
-   sphinx-quickstart on Sun Sep  4 14:29:43 2016.
-   You can adapt this file completely to your liking, but it should at least
-   contain the root `toctree` directive.
-
 OCRmyPDF documentation
 ======================
 
-OCRmyPDF adds an OCR text layer to scanned PDF files, allowing them to
-be searched.
+OCRmyPDF adds an optical charcter recognition (OCR) text layer to scanned PDF
+files, allowing them to be searched.
 
-PDF is the best format for storing and exchanging scanned documents.  Unfortunately, PDFs can be difficult to modify. OCRmyPDF makes it easy to apply image processing and OCR to existing PDFs.
+PDF is the best format for storing and exchanging scanned documents.
+Unfortunately, PDFs can be difficult to modify. OCRmyPDF makes it easy to apply
+image processing and OCR to existing PDFs.
 
 .. toctree::
    :maxdepth: 1
@@ -17,18 +14,27 @@ PDF is the best format for storing and exchanging scanned documents.  Unfortunat
    introduction
    release_notes
    installation
+   optimizer
    languages
+   jbig2
 
 .. toctree::
    :caption: Usage
    :maxdepth: 2
 
    cookbook
+   docker
    advanced
    batch
    security
    errors
 
+.. toctree::
+   :caption: Developers
+   :maxdepth: 2
+
+   api
+   contributing
 
 Indices and tables
 ==================
@@ -36,4 +42,3 @@ Indices and tables
 * :ref:`genindex`
 * :ref:`modindex`
 * :ref:`search`
-

docs/introduction.rst

@@ -1,115 +1,228 @@
+============
 Introduction
 ============
 
-OCRmyPDF is a Python 3 package that adds OCR layers to PDFs.
-
+OCRmyPDF is a Python 3 application and library 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 selected, 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 between 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`` 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 +1,73 @@
 .. _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 all languages.
+On most platforms, English is installed with Tesseract by default, but not always.
 
-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>`__.
+Languages are identified by standardized three-letter codes (called ISO 639-2 Alpha-3).
+Tesseract's documentation also lists the three-letter code for your language.
+Some are anglicized, e.g. Spanish is ``spa`` rather than ``esp``, while others
+are not, e.g. German is ``deu``.
 
-For Linux users, you can often find packages that provide language packs:
+After you have installed a language pack, you can use it ``ocrmypdf -l <language>``,
+for example ``ocrmypdf -l spa``. For multilingual documents, you can specify
+all languages to be expected, e.g. ``ocrmypdf -l eng+fra`` for English and French.
+English is assumed by default unless other language(s) are specified.
+
+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.
+Users of the OCRmyPDF Docker image should install language packs into a
+derived Docker image as
+:ref:`described in that section <docker-lang-packs>`.
 
-Adding individual language packs to a Docker image
-""""""""""""""""""""""""""""""""""""""""""""""""""
+Windows users
+=============
 
-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
+The Tesseract installer provided by Chocolatey already includes 100 languages.

docs/optimizer.rst

@@ -0,0 +1,75 @@
+================
+PDF optimization
+================
+
+OCRmyPDF includes an image-oriented PDF optimizer. By default, the optimizer
+runs with safe settings with the goal of improving compression at no loss of
+quality. At higher optimization levels, lossy optimizations may be applied and
+tuned. Optimization occurs after OCR, and only if OCR succeeded.  It does not
+perform other possible optimizations such as deduplicating resources,
+consolidating fonts, simplifying vector drawings, or anything of that nature.
+
+Optimization ranges from ``-O0`` through ``-O3``, where ``0`` disables
+optimization and ``3`` implements all options. ``1``, the default, performs only
+safe and lossless optimizations. (This is similar to GCC's optimization
+parameter.) The exact type of optimizations performed will vary over time.
+
+PDF optimization requires third-party, optional tools for certain optimizations.
+If these are not installed or cannot be found by OCRmyPDF, optimization will not
+be as good.
+
+Optimizations that always occurs
+================================
+
+OCRmyPDF will automatically replace obsolete or inferior compression schemes
+such as RLE or LZW with superior schemes such as Deflate and converting
+monochrome images to CCITT G4. Since this is harmless it always occurs and there
+is no way to disable it. Other non-image compressed objects are compressed as
+well.
+
+Fast web view
+=============
+
+OCRmyPDF automatically optimizes PDFs for "fast web view" in Adobe Acrobat's
+parlance, or equivalently, linearizes PDFs so that the resources they reference
+are presented in the order a viewer needs them for sequential display. This
+reduces the latency of viewing a PDF both online and from local storage. This
+actually slightly increases the file size.
+
+To disable this optimization and all others, use ``ocrmypdf --optimize 0 ...``
+or the shorthand ``-O0``.
+
+Lossless optimizations
+======================
+
+At optimization level ``-O1`` (the default), OCRmyPDF will also attempt lossless
+image optimization.
+
+If a JBIG2 encoder is available, then monochrome images will be converted to
+JBIG2, with the potential for huge savings on large black and white images,
+since JBIG2 is far more efficient than any other monochrome (bi-level)
+compression. (All known US patents related to JBIG2 have probably expired, but
+it remains the responsibility of the user to supply a JBIG2 encoder such as
+`jbig2enc <https://github.com/agl/jbig2enc>`__. OCRmyPDF does not implement
+JBIG2 encoding on its own.)
+
+OCRmyPDF currently does not attempt to recompress losslessly compressed objects
+more aggressively.
+
+Lossy optimizations
+===================
+
+At optimization level ``-O2`` and ``-O3``, OCRmyPDF will some attempt lossy
+image optimization.
+
+If ``pngquant`` is installed, OCRmyPDF will use it to perform quantize paletted
+images to reduce their size.
+
+The quality of JPEGs may be lowered, on the assumption that a lower quality
+image may be suitable for storage after OCR.
+
+It is not possible to optimize all image types. Uncommon image types may be
+skipped by the optimizer.
+
+OCRmyPDF provides :ref:`lossy mode JBIG2 <jbig2-lossy>` as an advanced feature
+that additional requires the argument ``--jbig2-lossy``.

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/batch.py

@@ -0,0 +1,50 @@
+#!/usr/bin/env python3
+# Original version by DeliciousPickle@github; modified
+
+# This script must be edited to meet your needs.
+
+import logging
+import os
+import sys
+
+import ocrmypdf
+
+# pylint: disable=logging-format-interpolation
+# pylint: disable=logging-not-lazy
+
+script_dir = os.path.dirname(os.path.realpath(__file__))
+print(script_dir + '/batch.py: Start')
+
+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'
+
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(asctime)s %(message)s',
+    filename=log_file,
+    filemode='w',
+)
+
+ocrmypdf.configure_logging(ocrmypdf.Verbosity.default)
+
+for dir_name, subdirs, file_list in os.walk(start_dir):
+    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)
+            result = ocrmypdf.ocr(filename, filename, deskew=True)
+            if result == ocrmypdf.ExitCode.already_done_ocr:
+                print("Skipped document because it already contained text")
+            elif result == ocrmypdf.ExitCode.ok:
+                print("OCR complete")
+            logging.info(result)

misc/completion/ocrmypdf.bash

@@ -0,0 +1,99 @@
+# ocrmypdf completion                                     -*- shell-script -*-
+
+set -o errexit
+
+_ocrmypdf()
+{
+    local cur prev cword words split
+
+    # Homebrew on Macs have version 1.3 of bash-completion which doesn't include - see #502
+    if declare -F _init_completions >/dev/null 2>&1; then
+        _init_completion -s || return
+    else
+        COMPREPLY=()
+        _get_comp_words_by_ref cur prev words cword
+    fi
+
+    if [[ $cur == -* ]]; then
+        COMPREPLY=( $( compgen -W '--language --image-dpi --output-type
+            --sidecar --version --jobs --quiet --verbose --title --author
+            --subject --keywords --rotate-pages --remove-background --deskew
+            --clean --clean-final --unpaper-args --oversample --remove-vectors
+            --threshold --force-ocr --skip-text --redo-ocr
+            --skip-big --jpeg-quality --png-quality --jbig2-lossy
+            --max-image-mpixels --tesseract-config --tesseract-pagesegmode
+            --help --tesseract-oem --pdf-renderer --tesseract-timeout
+            --rotate-pages-threshold --pdfa-image-compression --user-words
+            --user-patterns --keep-temporary-files --output-type
+            --no-progress-bar --pages --fast-web-view' \
+            --  "$cur" ) )
+        return
+    else
+        _filedir
+        return
+    fi
+
+    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=( $( compgen -W '{0..2}' -- "$cur" ) ) # max level ?
+            return
+            ;;
+        --tesseract-pagesegmode)
+            COMPREPLY=( $( compgen -W '{1..13}' -- "$cur" ) )
+            return
+            ;;
+        --sidecar|--title|--author|--subject|--keywords|--unpaper-args|--pages|--fast-web-view)
+            # argument required but no completions available
+            return
+            ;;
+    esac
+
+    $split && return
+} &&
+complete -F _ocrmypdf ocrmypdf
+
+set +o errexit
+
+# ex: filetype=sh

misc/completion/ocrmypdf.fish

@@ -0,0 +1,118 @@
+complete -c ocrmypdf -x -n '__fish_is_first_arg' -l version
+complete -c ocrmypdf -x -n '__fish_is_first_arg' -s h -s "?" -l help
+
+complete -c ocrmypdf -r -l sidecar -d "write OCR to text file"
+complete -c ocrmypdf -x -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 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)"
+
+function __fish_ocrmypdf_languages
+    set langs (tesseract --list-langs ^/dev/null)
+    set arr (string split '\n' $langs)
+    for lang in $arr[2..-1]
+        echo $lang
+    end
+end
+complete -c ocrmypdf -x -s l -l language -a '(__fish_ocrmypdf_languages)' -d "language"
+
+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-renderer -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"
+
+function __fish_ocrmypdf_verbose
+    echo -e "0\t"(_ "standard output messages")
+    echo -e "1\t"(_ "troubleshooting output messages")
+    echo -e "2\t"(_ "debugging output messages")
+end
+complete -c ocrmypdf -x -s v -l verbose -a '(__fish_ocrmypdf_verbose)' -d "set verbosity level"
+
+complete -c ocrmypdf -x -l no-progress-bar -d "disable the progress bar"
+
+function __fish_ocrmypdf_pdfa_compression
+    echo -e "auto\t"(_ "let Ghostscript decide how to compress images")
+    echo -e "jpeg\t"(_ "convert color and grayscale images to JPEG")
+    echo -e "lossless\t"(_ "convert color and grayscale images to lossless (PNG)")
+end
+complete -c ocrmypdf -x -l pdfa-image-compression -a '(__fish_ocrmypdf_pdfa_compression)' -d "set PDF/A image compression options"
+
+complete -c ocrmypdf -x -s j -l jobs -d "how many worker processes to use"
+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 pages -d "apply OCR to only the specified pages"
+complete -c ocrmypdf -x -l tesseract-config -d "set custom tesseract config file"
+
+function __fish_ocrmypdf_tesseract_pagesegmode
+  echo -e "0\t"(_ "orientation and script detection (OSD) only")
+  echo -e "1\t"(_ "automatic page segmentation with OSD")
+  echo -e "2\t"(_ "automatic page segmentation, but no OSD, or OCR")
+  echo -e "3\t"(_ "fully automatic page segmentation, but no OSD (default)")
+  echo -e "4\t"(_ "assume a single column of text of variable sizes")
+  echo -e "5\t"(_ "assume a single uniform block of vertically aligned text")
+  echo -e "6\t"(_ "assume a single uniform block of text")
+  echo -e "7\t"(_ "treat the image as a single text line")
+  echo -e "8\t"(_ "treat the image as a single word")
+  echo -e "9\t"(_ "treat the image as a single word in a circle")
+  echo -e "10\t"(_ "treat the image as a single character")
+  echo -e "11\t"(_ "sparse text - find as much text as possible in no particular order")
+  echo -e "12\t"(_ "sparse text with OSD")
+  echo -e "13\t"(_ "raw line - treat the image as a single text line")
+end
+complete -c ocrmypdf -x -l tesseract-pagesegmode -a '(__fish_ocrmypdf_tesseract_pagesegmode)' -d "set tesseract --psm"
+
+function __fish_ocrmypdf_tesseract_oem
+    echo -e "0\t"(_ "legacy engine only")
+    echo -e "1\t"(_ "neural nets LSTM engine only")
+    echo -e "2\t"(_ "legacy + LSTM engines")
+    echo -e "3\t"(_ "default, based on what is available")
+end
+complete -c ocrmypdf -x -l tesseract-oem -a '(__fish_ocrmypdf_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 -r -l user-words -d "specify location of user words file"
+complete -c ocrmypdf -r -l user-patterns -d "specify location of user patterns file"
+complete -c ocrmypdf -x -l fast-web-view -d "if file size if above this amount in MB, linearize PDF"
+
+complete -c ocrmypdf -x -a "(__fish_complete_suffix .pdf)"

misc/docker-compose.example.yml

@@ -0,0 +1,15 @@
+---
+version: "3.3"
+services:
+  ocrmypdf:
+    restart: always
+    container_name: ocrmypdf
+    image: jbarlow83/ocrmypdf
+    volumes:
+      - "/media/scan:/input"
+      - "/mnt/scan:/output"
+    environment:
+      - OCR_OUTPUT_DIRECTORY_YEAR_MONT=0
+    user: "<SET TO YOUR USER ID>:<SET TO YOUR GROUP ID>"
+    entrypoint: python3
+    command: watcher.py

misc/synology.py

@@ -0,0 +1,72 @@
+#!/bin/env python3
+# Contributed by github.com/Enantiomerie
+
+# This script must be edited to meet your needs.
+
+import logging
+import os
+import shutil
+import subprocess
+import sys
+import time
+
+# pylint: disable=logging-format-interpolation
+# pylint: disable=logging-not-lazy
+
+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(dir_name)
+    os.chdir(dir_name)
+    for filename in file_list:
+        file_stem, file_ext = os.path.splitext(filename)
+        if file_ext != '.pdf':
+            continue
+        full_path = os.path.join(dir_name, filename)
+        timestamp_ocr = time.strftime("%Y-%m-%d-%H%M_OCR_")
+        filename_ocr = timestamp_ocr + file_stem + '.pdf'
+        # create string for pdf processing
+        # the script is processed as root user via chron
+        cmd = [
+            'docker',
+            'run',
+            '--rm',
+            '-i',
+            'jbarlow83/ocrmypdf',
+            '--deskew',
+            '-',
+            '-',
+        ]
+        logging.info(cmd)
+        full_path_ocr = os.path.join(dir_name, filename_ocr)
+        with open(filename, 'rb') as input_file, open(
+            full_path_ocr, 'wb'
+        ) as output_file:
+            proc = subprocess.run(
+                cmd,
+                stdin=input_file,
+                stdout=output_file,
+                stderr=subprocess.PIPE,
+                check=False,
+            )
+        logging.info(proc.stderr.read())
+        os.chmod(full_path_ocr, 0o664)
+        os.chmod(full_path, 0o664)
+        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')

misc/watcher.py

@@ -0,0 +1,156 @@
+# Copyright (C) 2019 Ian Alexander: https://github.com/ianalexander
+# Copyright (C) 2020 James R Barlow: https://github.com/jbarlow83
+#
+# This program 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.
+#
+# 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.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+import json
+import logging
+import os
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+import pikepdf
+from watchdog.events import PatternMatchingEventHandler
+from watchdog.observers import Observer
+from watchdog.observers.polling import PollingObserver
+
+import ocrmypdf
+
+# pylint: disable=logging-format-interpolation
+
+INPUT_DIRECTORY = os.getenv('OCR_INPUT_DIRECTORY', '/input')
+OUTPUT_DIRECTORY = os.getenv('OCR_OUTPUT_DIRECTORY', '/output')
+OUTPUT_DIRECTORY_YEAR_MONTH = bool(os.getenv('OCR_OUTPUT_DIRECTORY_YEAR_MONTH', False))
+ON_SUCCESS_DELETE = bool(os.getenv('OCR_ON_SUCCESS_DELETE', False))
+DESKEW = bool(os.getenv('OCR_DESKEW', False))
+OCR_JSON_SETTINGS = json.loads(os.getenv('OCR_JSON_SETTINGS', '{}'))
+POLL_NEW_FILE_SECONDS = os.getenv('OCR_POLL_NEW_FILE_SECONDS', 1)
+USE_POLLING = bool(os.getenv('OCR_USE_POLLING', False))
+LOGLEVEL = os.getenv('OCR_LOGLEVEL', 'INFO').upper()
+PATTERNS = ['*.pdf']
+
+log = logging.getLogger('ocrmypdf-watcher')
+
+
+def get_output_dir(root, basename):
+    if OUTPUT_DIRECTORY_YEAR_MONTH:
+        today = datetime.today()
+        output_directory_year_month = (
+            Path(root) / str(today.year) / f'{today.month:02d}'
+        )
+        if not output_directory_year_month.exists():
+            output_directory_year_month.mkdir(parents=True, exist_ok=True)
+        output_path = Path(output_directory_year_month) / basename
+    else:
+        output_path = Path(OUTPUT_DIRECTORY) / basename
+    return output_path
+
+
+def wait_for_file_ready(file_path):
+    # This loop waits to make sure that the file is completely loaded on
+    # disk before attempting to read. Docker sometimes will publish the
+    # watchdog event before the file is actually fully on disk, causing
+    # pikepdf to fail.
+
+    retries = 5
+    while retries:
+        try:
+            pdf = pikepdf.open(file_path)
+        except (FileNotFoundError, pikepdf.PdfError) as e:
+            log.info(f"File {file_path} is not ready yet")
+            log.debug("Exception was", exc_info=e)
+            time.sleep(POLL_NEW_FILE_SECONDS)
+            retries -= 1
+        else:
+            pdf.close()
+            return True
+
+    return False
+
+
+def execute_ocrmypdf(file_path):
+    file_path = Path(file_path)
+    output_path = get_output_dir(OUTPUT_DIRECTORY, file_path.name)
+
+    log.info("-" * 20)
+    log.info(f'New file: {file_path}. Waiting until fully loaded...')
+    if not wait_for_file_ready(file_path):
+        log.info(f"Gave up waiting for {file_path} to become ready")
+        return
+    log.info(f'Attempting to OCRmyPDF to: {output_path}')
+    exit_code = ocrmypdf.ocr(
+        input_file=file_path,
+        output_file=output_path,
+        deskew=DESKEW,
+        **OCR_JSON_SETTINGS,
+    )
+    if exit_code == 0 and ON_SUCCESS_DELETE:
+        log.info(f'OCR is done. Deleting: {file_path}')
+        file_path.unlink()
+    else:
+        log.info('OCR is done')
+
+
+class HandleObserverEvent(PatternMatchingEventHandler):
+    def on_any_event(self, event):
+        if event.event_type in ['created']:
+            execute_ocrmypdf(event.src_path)
+
+
+def main():
+    ocrmypdf.configure_logging(
+        verbosity=ocrmypdf.Verbosity.default, manage_root_logger=True
+    )
+    log.setLevel(LOGLEVEL)
+    log.info(
+        f"Starting OCRmyPDF watcher with config:\n"
+        f"Input Directory: {INPUT_DIRECTORY}\n"
+        f"Output Directory: {OUTPUT_DIRECTORY}\n"
+        f"Output Directory Year & Month: {OUTPUT_DIRECTORY_YEAR_MONTH}"
+    )
+    log.debug(
+        f"INPUT_DIRECTORY: {INPUT_DIRECTORY}\n"
+        f"OUTPUT_DIRECTORY: {OUTPUT_DIRECTORY}\n"
+        f"OUTPUT_DIRECTORY_YEAR_MONTH: {OUTPUT_DIRECTORY_YEAR_MONTH}\n"
+        f"ON_SUCCESS_DELETE: {ON_SUCCESS_DELETE}\n"
+        f"DESKEW: {DESKEW}\n"
+        f"ARGS: {OCR_JSON_SETTINGS}\n"
+        f"POLL_NEW_FILE_SECONDS: {POLL_NEW_FILE_SECONDS}\n"
+        f"USE_POLLING: {USE_POLLING}\n"
+        f"LOGLEVEL: {LOGLEVEL}\n"
+    )
+
+    if 'input_file' in OCR_JSON_SETTINGS or 'output_file' in OCR_JSON_SETTINGS:
+        log.error('OCR_JSON_SETTINGS should not specify input file or output file')
+        sys.exit(1)
+
+    handler = HandleObserverEvent(patterns=PATTERNS)
+    if USE_POLLING:
+        observer = PollingObserver()
+    else:
+        observer = Observer()
+    observer.schedule(handler, INPUT_DIRECTORY, recursive=True)
+    observer.start()
+    try:
+        while True:
+            time.sleep(1)
+    except KeyboardInterrupt:
+        observer.stop()
+    observer.join()
+
+
+if __name__ == "__main__":
+    main()

misc/webservice.py

@@ -0,0 +1,126 @@
+# 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.
+"""
+
+import os
+import shlex
+from subprocess import PIPE, run
+from tempfile import TemporaryDirectory
+
+from flask import (
+    Flask,
+    Response,
+    abort,
+    flash,
+    redirect,
+    request,
+    send_from_directory,
+    url_for,
+)
+from werkzeug.utils import secure_filename
+
+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 +1,34 @@
+[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
+target-version = ["py36",
+"py37", "py38"]
+skip-string-normalization = true
+include = '\.pyi?$'
+exclude = '''
+/(
+    \.eggs
+  | \.git
+  | \.hg
+  | \.mypy_cache
+  | \.tox
+  | \.venv
+  | _build
+  | buck-out
+  | build
+  | dist
+  | docs
+  | misc
+  | \.egg-info
+  | src/ocrmypdf/lib/_leptonica.py
+)/
+'''

requirements/dev.txt

@@ -0,0 +1,2 @@
+twine >= 1.8.1
+coverage >= 4.5

requirements/main.txt

@@ -1,10 +1,10 @@
 # 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
-ruffus == 2.6.3
-Pillow == 5.1.0
-reportlab == 3.4.0
-PyPDF2 == 1.26.0
-img2pdf == 0.2.4
-cffi == 1.11.5
-PyMuPDF == 1.12.5
+cffi == 1.14.0
+img2pdf == 0.3.3
+pdfminer.six == 20200124
+pikepdf == 1.10.2
+Pillow == 7.0.0
+reportlab == 3.5.34
+tqdm == 4.42.1

requirements/test.txt

@@ -0,0 +1,7 @@
+pytest >= 5.0.0
+pytest-helpers-namespace >= 2019.1.8
+pytest-xdist >= 1.31.0
+pytest-cov >= 2.8.0
+python-xmp-toolkit == 2.0.1  # requires apt-get install libexempi3
+                     # or brew install exempi
+#PyMuPDF == 1.13.4    # optional

requirements/watcher.txt

@@ -0,0 +1 @@
+watchdog >= 0.8.2, < 1.0

requirements/webservice.txt

@@ -0,0 +1 @@
+Flask >= 1, < 2

setup.cfg

@@ -11,6 +11,19 @@ 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
+known_first_party = ocrmypdf
+known_third_party = PIL,_cffi_backend,cffi,flask,gs,img2pdf,pdfminer,pikepdf,pkg_resources,pytest,reportlab,setuptools,sphinx_rtd_theme,tqdm,watchdog,werkzeug
 
 [metadata]
-license_file = LICENSE
+license_file = LICENSE

setup.py

@@ -20,198 +20,46 @@
 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)
+
+from setuptools import find_packages, setup
+
+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
 
-
-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',
@@ -219,8 +67,9 @@ 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",
+        "Programming Language :: Python :: 3.8",
         "Development Status :: 5 - Production/Stable",
         "Environment :: Console",
         "Intended Audience :: End Users/Desktop",
@@ -228,44 +77,41 @@ setup(
         "Intended Audience :: System Administrators",
         "License :: OSI Approved :: GNU General Public License v3 (GPLv3)",
         "Operating System :: MacOS :: MacOS X",
+        "Operating System :: Microsoft :: Windows :: Windows 10",
         "Operating System :: POSIX",
         "Operating System :: POSIX :: BSD",
         "Operating System :: POSIX :: Linux",
         "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
-        'defusedxml >= 0.5.0',    # pure Python, so track HEAD closely
-        'img2pdf >= 0.2.4',       # pure Python, so track HEAD closely
-        'pikepdf',
-        '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
-        'PyPDF2 >= 1.26',         # pure Python, so track HEAD closely
-        'reportlab >= 3.3.0',     # oldest released version with sane image handling
-        'ruffus == 2.6.3',        # pinned - ocrmypdf implements a 2.6.3 workaround
+        '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, <= 20200124',
+        'pikepdf >= 1.8.1, < 2',
+        'Pillow >= 6.2.0',
+        'reportlab >= 3.3.0',  # oldest released version with sane image handling
+        'tqdm >= 4',
     ],
-    extras_require={
-        'fitz': ['PyMuPDF >= 1.12.5, != 1.13.3'],
-    },
     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

@@ -15,11 +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/>.
 
-import pkg_resources
-
-PROGRAM_NAME = 'ocrmypdf'
-
-# Official PEP 396
-__version__ = pkg_resources.get_distribution('ocrmypdf').version
-
-VERSION = __version__
+from . import helpers, hocrtransform, leptonica, pdfa, pdfinfo
+from ._version import PROGRAM_NAME, __version__
+from .api import Verbosity, configure_logging, ocr
+from .exceptions import (
+    BadArgsError,
+    DpiError,
+    EncryptedPdfError,
+    ExitCode,
+    ExitCodeException,
+    InputFileError,
+    MissingDependencyError,
+    OutputFileAccessError,
+    PdfMergeFailedError,
+    PriorOcrFoundError,
+    SubprocessOutputError,
+    TesseractConfigError,
+    UnsupportedImageFormatError,
+)

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.
 #
@@ -16,937 +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 multiprocessing
-import atexit
-import textwrap
 import logging
-import argparse
+import os
+import sys
 
-import PIL
+from . import __version__
+from ._jobcontext import make_logger
+from ._sync import run_pipeline
+from ._validation import check_closed_streams, check_options
+from .api import Verbosity, configure_logging
+from .cli import parser
+from .exceptions import BadArgsError, ExitCode, 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 .lib import fitz
-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"
-    )
-)
-
-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.")
-advanced.add_argument(
-    '--skip-repair', action='store_true',
-    help="Normally OCRmyPDF automatically repairs PDFs using qpdf before "
-         "processing.  If you have already run qpdf or a similar program "
-         "that repairs PDF errors, you can tell OCRmyPDF to skip repair with "
-         "this option.  This may be helpful in batch processing where all "
-         "files are repaired prior to OCR occurs, since repair is single "
-         "threaded and time consuming for large files.")
-
-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
-    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.TypeError':
-        # Even though repair_pdf will fail, ruffus will still try
-        # to call split_pages with no input files, likely due to a bug
-        if task_name == 'split_pages':
-            log.error("Input file '{0}' is not a valid PDF".format(
-                options.input_file))
-            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)]]
-
-    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())
-    if fitz:
-        _log.debug('PyMuPDF ' + fitz.version[0])
-        _log.debug('libmupdf ' + fitz.version[1])
-    else:
-        _log.debug('PyMuPDF not installed')
-
-
-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
-        exitcode = traverse_ruffus_exception(e.args, 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.invalid_output_pdf
-        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/_concurrent.py

@@ -0,0 +1,136 @@
+# © 2020 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 multiprocessing import Pool as ProcessPool
+from multiprocessing.dummy import Pool as ThreadPool
+from typing import Callable, Iterable, Optional
+
+from tqdm import tqdm
+
+
+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 process_init(queue, userfn, *userargs):
+    """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)
+
+    if userfn:
+        userfn(*userargs)
+
+
+def thread_init(_queue, userfn, *userargs):
+    if userfn:
+        userfn(*userargs)
+
+
+def exec_progress_pool(
+    *,
+    use_threads: bool,
+    max_workers: int,
+    tqdm_kwargs: dict,
+    task_initializer: Optional[Callable] = None,
+    task_initargs: Optional[tuple] = None,
+    task: Optional[Callable] = None,
+    task_arguments: Optional[Iterable] = None,
+    task_finished: Optional[Callable] = None,
+):
+    log_queue = multiprocessing.Queue(-1)
+    listener = threading.Thread(target=log_listener, args=(log_queue,))
+
+    if not task_initargs:
+        task_initargs = tuple()
+
+    if use_threads:
+        pool_class = ThreadPool
+        initializer = thread_init
+    else:
+        pool_class = ProcessPool
+        initializer = process_init
+    listener.start()
+
+    with tqdm(**tqdm_kwargs) as pbar:
+        pool = pool_class(
+            processes=max_workers,
+            initializer=initializer,
+            initargs=(log_queue, task_initializer, *task_initargs),
+        )
+        try:
+            results = pool.imap_unordered(task, task_arguments)
+            while True:
+                try:
+                    result = results.next()
+                    task_finished(result, pbar)
+                except StopIteration:
+                    break
+        except KeyboardInterrupt:
+            # Terminate pool so we exit instantly
+            pool.terminate()
+            # Don't try listener.join() here, will deadlock
+            raise
+        except Exception:
+            if not os.environ.get("PYTEST_CURRENT_TEST", ""):
+                # Unless inside pytest, exit immediately because no one wants
+                # to wait for child processes to finalize results that will be
+                # thrown away. Inside pytest, we want child processes to exit
+                # cleanly so that they output an error messages or coverage data
+                # we need from them.
+                pool.terminate()
+            raise
+        finally:
+            # Terminate log listener
+            log_queue.put_nowait(None)
+            pool.close()
+            pool.join()
+
+    listener.join()

src/ocrmypdf/_graft.py

@@ -0,0 +1,284 @@
+# © 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
+
+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)
+    corner = pikepdf.PdfMatrix().translated(mediabox[0], mediabox[1])
+    # -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, and
+    # finally move the lower left corner to match the mediabox
+    ctm = translate @ rotate @ scale @ untranslate @ corner
+
+    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(
+            f"Rotations for page {pageno}: [text, auto, misalign, content] = "
+            f"{text_rotation}, {autorotate_correction}, "
+            f"{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 os
 import shutil
-
-from .pdfinfo import PdfInfo
+import sys
 
 
-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['input_filename'], msg), kwargs
+        return '%s' % (msg,), kwargs
+
+
+class LogNamePageAdapter(logging.LoggerAdapter):
+    def process(self, msg, kwargs):
+        return (
+            #'[%s:%05u] %s' % (self.extra['input_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(input_filename=filename, page=page))
+    elif filename:
+        adapter = LogNameAdapter(log, dict(input_filename=filename))
+    else:
+        adapter = log
+    return adapter

src/ocrmypdf/_optimize.py

@@ -1,365 +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 subprocess import CalledProcessError
-import concurrent.futures
-from collections import defaultdict, namedtuple
-import struct
-import logging
-import sys
-
-from io import BytesIO
-from PIL import Image
-
-import pikepdf
-
-from ._jobcontext import JobContext
-from . import leptonica
-from .helpers import re_symlink
-from .exec import pngquant, jbig2enc
-
-PAGE_GROUP_SIZE = 10
-JPEG_QUALITY = 75
-PNG_QUALITY = (65, 75)
-
-
-def img_name(root, xref, ext):
-    return str(root / '{:08d}{}'.format(xref, ext))
-
-
-def png_name(root, xref):
-    return img_name(root, xref, '.png')
-
-
-def jpg_name(root, xref):
-    return img_name(root, xref, '.jpg')
-
-
-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
-    if image.Length < 100:
-        log.debug("Skipping small image, xref {}".format(xref))
-        return False
-
-    pim = pikepdf.PdfImage(image)
-
-    if len(pim.filter_decodeparms) > 1:
-        log.debug("Skipping multiply filtered, xref {}".format(xref))
-        return False
-    filtdp = pim.filter_decodeparms[0]
-
-    if pim.bits_per_component > 8:
-        return False  # Don't mess with wide gamut images
-
-    if filtdp[0] == '/JPXDecode':
-        return False  # Don't do JPEG2000
-
-    if pim.bits_per_component == 1 \
-            and filtdp != '/JBIG2Decode' \
-            and jbig2enc.available():
-        try:
-            imgname = Path(root / '{:08d}'.format(xref))
-            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:
-        # 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
-        # anything.
-        # 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
-
-        # We could get the ICC profile here, but there's no need to look at it
-        # for quality transcoding
-        # if icc:
-        #     stream = BytesIO(raw_jpeg.read_raw_bytes())
-        #     iccbytes = icc.read_bytes()
-        #     with Image.open(stream) as im:
-        #         im.save(jpg_name(root, xref), icc_profile=iccbytes)
-        try:
-            imgname = Path(root / '{:08d}'.format(xref))
-            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:
-        # 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)
-    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 True
-
-
-def extract_images(pike, root, log, options):
-    # Extract images we can improve
-    changed_xrefs = set()
-    jbig2_groups = defaultdict(lambda: [])
-    jpegs = []
-    pngs = []
-    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():
-            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
-
-    # 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 changed_xrefs, jbig2_groups, jpegs, pngs
-
-
-def convert_to_jbig2(pike, jbig2_groups, root, log, options):
-    """
-    Convert a group of JBIG2 images and insert into PDF.
-
-    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)
-            future = executor.submit(
-                jbig2enc.convert_group,
-                cwd=str(root),
-                infiles=(img_name(root, xref, ext) for xref, ext in xref_exts),
-                out_prefix=prefix
-            )
-            futures.append(future)
-        for future in concurrent.futures.as_completed(futures):
-            proc = future.result()
-            log.debug(proc.stderr.decode())
-
-    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)
-
-        for n, xref_ext in enumerate(xref_exts):
-            xref, ext = xref_ext
-            jbig2_im_file = root / (prefix + '.{:04d}'.format(n))
-            jbig2_im_data = jbig2_im_file.read_bytes()
-            im_obj = pike._get_object_id(xref, 0)
-            im_obj.write(
-                jbig2_im_data, pikepdf.Name('/JBIG2Decode'),
-                pikepdf.Dictionary({
-                    '/JBIG2Globals': jbig2_globals
-                })
-            )
-
-
-def transcode_jpegs(pike, jpegs, root, log, options):
-    for xref in jpegs:
-        in_jpg = Path(jpg_name(root, xref))
-        opt_jpg = in_jpg.with_suffix('.opt.jpg')
-
-        # This produces a debug warning from PIL
-        # DEBUG:PIL.Image:Error closing: 'NoneType' object has no attribute
-        # 'close'.  Seems to be mostly harmless
-        # https://github.com/python-pillow/Pillow/issues/1144
-        with Image.open(str(in_jpg)) as im:
-            im.save(str(opt_jpg),
-                    optimize=True,
-                    quality=JPEG_QUALITY)
-        if opt_jpg.stat().st_size > in_jpg.stat().st_size:
-            log.debug("xref {}, jpeg, made larger - skip".format(xref))
-            continue
-
-        compdata = leptonica.CompressedData.open(opt_jpg)
-        im_obj = pike._get_object_id(xref, 0)
-        im_obj.write(
-            compdata.read(), pikepdf.Name('/DCTDecode'),
-            pikepdf.Null()
-        )
-
-
-def transcode_pngs(pike, pngs, root, log, options):
-    if options.optimize >= 2:
-        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])
-
-    for xref in pngs:
-        im_obj = pike._get_object_id(xref, 0)
-
-        # Open, transcode (!), package for PDF
-        try:
-            pix = leptonica.Pix.open(png_name(root, xref))
-            compdata = pix.generate_pdf_ci_data(
-                leptonica.lept.L_FLATE_ENCODE, 0
-            )
-        except leptonica.LeptonicaError as e:
-            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 len(compdata) > int(im_obj.stream_dict.Length):
-            continue  # If we produced a larger image, don't use
-
-        predictor = pikepdf.Null()
-        if compdata.predictor > 0:
-            predictor = pikepdf.Dictionary({'/Predictor': compdata.predictor})
-
-        im_obj.BitsPerComponent = compdata.bps
-        im_obj.Width = compdata.w
-        im_obj.Height = compdata.h
-
-        if compdata.ncolors > 0:
-            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]
-            cs = palette
-        else:
-            if compdata.spp == 1:
-                cs = pikepdf.Name('/DeviceGray')
-            elif compdata.spp == 3:
-                cs = pikepdf.Name('/DeviceRGB')
-            elif compdata.spp == 4:
-                cs = pikepdf.Name('/DeviceCMYK')
-        im_obj.ColorSpace = cs
-        im_obj.write(compdata.read(), pikepdf.Name('/FlateDecode'), predictor)
-
-
-def optimize(
-    input_file,
-    output_file,
-    log,
-    context):
-
-    options = context.get_options()
-    if options.optimize == 0:
-        re_symlink(input_file, output_file, log)
-        return
-
-    global PNG_QUALITY
-    global JPEG_QUALITY
-
-    if options.optimize == 3:
-        PNG_QUALITY = (20, 40)
-        JPEG_QUALITY = 40
-
-    pike = pikepdf.Pdf.open(input_file)
-
-    root = Path(output_file).parent / 'images'
-    root.mkdir(exist_ok=True)
-    changed_xrefs, jbig2_groups, jpegs, pngs = extract_images(
-        pike, 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)
-
-    # Not object_stream_mode + preserve_pdfa generates noncompliant PDFs
-    target_file = output_file + '_opt.pdf'
-    pike.save(target_file, preserve_pdfa=True)
-
-    input_size = Path(input_file).stat().st_size
-    output_size = Path(target_file).stat().st_size
-    ratio = input_size / output_size
-    savings = 1 - output_size / input_size
-    log.info("Optimize ratio: {:.2f} savings: {:.1f}%".format(
-        ratio, 100 * savings))
-
-    if savings < 0:
-        log.info("Optimize did not improve the file - discarded")
-        re_symlink(input_file, output_file, log)
-    else:
-        re_symlink(target_file, output_file, log)
-
-
-def main(infile, outfile, level, jobs=1):
-    Options = namedtuple('Options', 'jobs optimize')
-
-    logging.basicConfig(level=logging.DEBUG)
-    log = logging.getLogger()
-
-    ctx = JobContext()
-    options = Options(jobs=jobs, optimize=int(level))
-    ctx.set_options(options)
-
-    optimize(infile, outfile, log, ctx)
-
-
-if __name__ == '__main__':
-    main(sys.argv[1], sys.argv[2], sys.argv[3])

src/ocrmypdf/_pipeline.py

@@ -0,0 +1,830 @@
+# © 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 pathlib import Path
+from shutil import copyfileobj
+
+import img2pdf
+import pikepdf
+from pikepdf.models.metadata import encode_pdf_date
+from PIL import Image
+
+from . import leptonica
+from ._version import PROGRAM_NAME
+from ._version import __version__ as VERSION
+from .exceptions import (
+    DpiError,
+    EncryptedPdfError,
+    InputFileError,
+    PriorOcrFoundError,
+    UnsupportedImageFormatError,
+)
+from .exec import ghostscript, tesseract
+from .helpers import safe_symlink
+from .hocrtransform import HocrTransform
+from .optimize import optimize
+from .pdfa import generate_pdfa_ps
+from .pdfinfo import Colorspace, Encoding, PdfInfo
+
+VECTOR_PAGE_DPI = 400
+
+
+def triage_image_file(input_file, output_file, options, log):
+    log.info("Input file is not a PDF, checking if it is an image...")
+    try:
+        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
+
+    with im:
+        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.error("Input CMYK image has no ICC profile, not usable")
+                raise UnsupportedImageFormatError()
+
+    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(original_filename, input_file, output_file, options, log):
+    try:
+        if _pdf_guess_version(input_file):
+            if options.image_dpi:
+                log.warning(
+                    "Argument --image-dpi is being ignored because the "
+                    "input file is a PDF, not an image."
+                )
+            # Origin file is a pdf create a symlink with pdf extension
+            safe_symlink(input_file, output_file)
+            return output_file
+    except EnvironmentError as e:
+        log.debug(f"Temporary file was at: {input_file}")
+        msg = str(e).replace(input_file, original_filename)
+        raise InputFileError(msg) from e
+
+    triage_image_file(input_file, output_file, options, log)
+    return output_file
+
+
+def get_pdfinfo(input_file, progbar=False):
+    try:
+        return PdfInfo(input_file, 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 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.warning(
+                "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 describe_rotation(page_context, orient_conf, correction):
+    """
+    Describe the page rotation we are going to perform.
+    """
+    direction = {0: '⇧', 90: '⇨', 180: '⇩', 270: '⇦'}
+    turns = {0: ' ', 90: '⬏', 180: '↻', 270: '⬑'}
+
+    existing_rotation = page_context.pageinfo.rotation
+    action = ''
+    if orient_conf.confidence >= page_context.options.rotate_pages_threshold:
+        if correction != 0:
+            action = 'will rotate ' + turns[correction]
+        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 = f"with existing rotation {direction.get(existing_rotation, '?')}, "
+    facing += f"page is facing {direction.get(orient_conf.angle, '?')}"
+
+    return f"{facing}, confidence {orient_conf.confidence:.2f} - {action}"
+
+
+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,
+    )
+
+    correction = orient_conf.angle % 360
+    page_context.log.info(describe_rotation(page_context, orient_conf, correction))
+    if (
+        orient_conf.confidence >= page_context.options.rotate_pages_threshold
+        and correction != 0
+    ):
+        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()
+
+        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 == Encoding.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
+    fix_docinfo_file = context.get_path('fix_docinfo.pdf')
+    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.
+    modified = False
+    with pikepdf.open(input_pdf) as pdf_file:
+        try:
+            len(pdf_file.docinfo)
+        except TypeError:
+            context.log.error(
+                "File contains a malformed DocumentInfo block - continuing anyway"
+            )
+        else:
+            if pdf_file.docinfo:
+                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(fix_docinfo_file)
+        else:
+            safe_symlink(input_pdf, fix_docinfo_file)
+
+    ghostscript.generate_pdfa(
+        pdf_version=input_pdfinfo.min_version,
+        pdf_pages=[fix_docinfo_file, 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 should_linearize(working_file, context):
+    filesize = os.stat(working_file).st_size
+    if filesize > (context.options.fast_web_view * 1_000_000):
+        return True
+    return False
+
+
+def metadata_fixup(working_file, context):
+    output_file = context.get_path('metafix.pdf')
+    options = context.options
+
+    def report_on_metadata(missing):
+        if not missing:
+            return
+        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", missing
+            )
+        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", missing
+            )
+
+    with pikepdf.open(context.origin) as original, pikepdf.open(working_file) as pdf:
+        docinfo = get_docinfo(original, options)
+        with pdf.open_metadata() as meta:
+            meta.load_from_docinfo(docinfo, delete_missing=False, raise_failure=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()
+            missing = set(meta_original.keys()) - set(meta.keys())
+            report_on_metadata(missing)
+
+        pdf.save(
+            output_file,
+            compress_streams=True,
+            preserve_pdfa=True,
+            object_stream_mode=pikepdf.ObjectStreamMode.generate,
+            linearize=(  # Don't linearize if optimize() will be linearizing too
+                should_linearize(working_file, context)
+                if options.optimize == 0
+                else False
+            ),
+        )
+
+    return output_file
+
+
+def optimize_pdf(input_file, context):
+    output_file = context.get_path('optimize.pdf')
+    save_settings = dict(
+        compress_streams=True,
+        preserve_pdfa=True,
+        object_stream_mode=pikepdf.ObjectStreamMode.generate,
+        linearize=should_linearize(input_file, context),
+    )
+    optimize(input_file, output_file, context, save_settings)
+    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/_sync.py

@@ -0,0 +1,366 @@
+# © 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
+from collections import namedtuple
+from pathlib import Path
+from tempfile import mkdtemp
+
+import PIL
+
+from ._concurrent import exec_progress_pool
+from ._graft import OcrGrafter
+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 .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(max_pixels):
+    # In Windows, child process will not inherit our change to this value in
+    # the parent process, so ensure workers get it set. Not needed when running
+    # threaded, but harmless to set again.
+    PIL.Image.MAX_IMAGE_PIXELS = max_pixels
+
+
+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 concurrently", max_workers)
+
+    # Tesseract 4.x can be multithreaded, and we also run multiple workers. We want
+    # to manage how many threads it uses to avoid creating total threads than cores.
+    # 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.
+    # As of Tesseract 4.1, 3 threads is the most effective on a 4 core/8 thread system.
+    tess_threads = min(3, 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))
+    try:
+        tess_threads = int(context.options.tesseract_env['OMP_THREAD_LIMIT'])
+    except ValueError:  # OMP_THREAD_LIMIT initialized to non-numeric
+        context.log.error("Environment variable OMP_THREAD_LIMIT is not numeric")
+    if tess_threads > 1:
+        context.log.info("Using Tesseract OpenMP thread limit %d", tess_threads)
+
+    sidecars = [None] * len(context.pdfinfo)
+    ocrgraft = OcrGrafter(context)
+
+    def update_page(result, pbar):
+        sidecars[result.pageno] = result.text
+        pbar.update()
+        ocrgraft.graft_page(result)
+        pbar.update()
+
+    exec_progress_pool(
+        use_threads=context.options.use_threads,
+        max_workers=max_workers,
+        tqdm_kwargs=dict(
+            total=(2 * len(context.pdfinfo)),
+            desc='OCR',
+            unit='page',
+            unit_scale=0.5,
+            disable=not context.options.progress_bar,
+        ),
+        task_initializer=worker_init,
+        task_initargs=(PIL.Image.MAX_IMAGE_PIXELS,),
+        task=exec_page_sync,
+        task_arguments=context.get_page_contexts(),
+        task_finished=update_page,
+    )
+
+    # 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  # pylint: disable=unnecessary-pass
+
+
+def samefile(f1, f2):
+    if os.name == 'nt':
+        return f1 == f2
+    else:
+        return os.path.samefile(f1, f2)
+
+
+def configure_debug_logging(log_filename, prefix=''):
+    log_file_handler = logging.FileHandler(log_filename, delay=True)
+    log_file_handler.setLevel(logging.DEBUG)
+    formatter = logging.Formatter(
+        '[%(asctime)s] - %(name)s - %(levelname)7s - %(message)s'
+    )
+    log_file_handler.setFormatter(formatter)
+    logging.getLogger(prefix).addHandler(log_file_handler)
+    return log_file_handler
+
+
+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.")
+    debug_log_handler = None
+    if (options.keep_temporary_files or options.verbose >= 1) and not os.environ.get(
+        'PYTEST_CURRENT_TEST', ''
+    ):
+        debug_log_handler = configure_debug_logging(Path(work_folder) / "debug.log")
+
+    try:
+        check_requested_output_file(options)
+        start_input_file, original_filename = create_input_file(options, work_folder)
+
+        # Triage image or pdf
+        origin_pdf = triage(
+            original_filename,
+            start_input_file,
+            os.path.join(work_folder, 'origin.pdf'),
+            options,
+            log,
+        )
+
+        # Gather pdfinfo and create context
+        pdfinfo = get_pdfinfo(origin_pdf, 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)
+
+        if options.output_file == '-':
+            log.info("Output sent to stdout")
+        elif 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']:
+                    log.info(
+                        "Output file is a %s (as expected)", pdfa_info['conformance']
+                    )
+                else:
+                    log.warning(
+                        "Output file is okay but is not PDF/A (seems to be %s)",
+                        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, start_input_file, options.output_file)
+
+    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:
+        if debug_log_handler:
+            try:
+                debug_log_handler.close()
+                log.removeHandler(debug_log_handler)
+            except EnvironmentError as e:
+                print(e, file=sys.stderr)
+        cleanup_working_files(work_folder, options)
+
+    return ExitCode.ok

src/ocrmypdf/_unicodefun.py

@@ -33,15 +33,21 @@
 # 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
+    if sys.version_info[0:3] >= (3, 7, 0):
+        return
+
     try:
         import locale
+
         fs_enc = codecs.lookup(locale.getpreferredencoding()).name
     except Exception:
         fs_enc = 'ascii'
@@ -51,8 +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
 
@@ -103,6 +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 +1,481 @@
+#!/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 locale
+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, is_iterable_notstr, monotonic, safe_symlink
+
+# -------------
+# External dependencies
+
+HOCR_OK_LANGS = frozenset(['eng', 'deu', 'spa', 'ita', 'por'])
+DEFAULT_LANGUAGE = 'eng'  # Enforce English hegemony
+
+log = logging.getLogger(__name__)
+
+
+# --------
+# Critical environment tests
+verify_python3_env()
+
+
+def check_platform():
+    if os.name == 'nt' and sys.maxsize <= 2 ** 32:  # pragma: no cover
+        # 32-bit interpreter on Windows
+        log.error(
+            "You are running OCRmyPDF in a 32-bit (x86) Python interpreter."
+            "Please use a 64-bit (x86-64) version of Python."
+        )
+
+
+def check_options_languages(options):
+    if not options.language:
+        options.language = [DEFAULT_LANGUAGE]
+        system_lang = locale.getlocale()[0]
+        if system_lang and not system_lang.startswith('en'):
+            log.debug("No language specified; assuming --language %s", DEFAULT_LANGUAGE)
+
+    # 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, languages
+    ):
+        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:
+            try:
+                pages.extend(range(int(start) - 1, int(end)))
+            except ValueError:
+                raise BadArgsError("invalid page range")
+
+    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 not tesseract.has_user_words(options.tesseract_env) and (
+        options.user_words or options.user_patterns
+    ):
+        log.warning(
+            "Tesseract 4.0 ignores --user-words and --user-patterns, so these "
+            "arguments have 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_platform()
+    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):  # pragma: no cover
+    """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, "<stdin>"
+    else:
+        try:
+            target = os.path.join(work_folder, 'origin')
+            safe_symlink(options.input_file, target)
+            return target, os.fspath(options.input_file)
+        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 options.optimize == 0:
+        reasons.append("Optimization was disabled.")
+    else:
+        image_optimizers = {
+            'jbig2': jbig2enc.available(),
+            'pngquant': pngquant.available(),
+        }
+        for name, available in image_optimizers.items():
+            if not available:
+                reasons.append(
+                    f"The optional dependency '{name}' was not found, so some image "
+                    f"optimizations could not be attempted."
+                )
+
+    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={'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
+    )
+    gs_version = ghostscript.version()
+    if gs_version in ('9.24', '9.51'):
+        raise MissingDependencyError(
+            f"Ghostscript {gs_version} contains serious regressions and is not "
+            "supported. Please upgrade to a newer version, or downgrade to the "
+            "previous version."
+        )
+    check_external_program(
+        program='qpdf',
+        package='qpdf',
+        version_checker=qpdf.version,
+        need_version='8.0.2',
+    )

src/ocrmypdf/_version.py

@@ -0,0 +1,24 @@
+# © 2017 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 pkg_resources
+
+PROGRAM_NAME = 'ocrmypdf'
+
+# Official PEP 396
+__version__ = pkg_resources.get_distribution('ocrmypdf').version

src/ocrmypdf/_weave.py

@@ -1,343 +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].decode())
-                for v in range(4)]
-    wt, ht = mediabox[2] - mediabox[0], mediabox[3] - mediabox[1]
-
-    mediabox = [float(base_page.MediaBox[v].decode())
-                for v in range(4)]
-    wp, hp = mediabox[2] - mediabox[0], mediabox[3] - mediabox[1]
-
-    translate = pikepdf.PdfMatrix((1, 0, 0, 1, -wt / 2, -ht / 2))
-    untranslate = pikepdf.PdfMatrix((1, 0, 0, 1, wp / 2, hp / 2))
-    # -rotation because the input is a clockwise angle and this formula
-    # uses CCW
-    rotation = -rotation % 360
-    if rotation == 0:
-        c, s = 1, 0
-    elif rotation == 90:
-        c, s = 0, 1
-    elif rotation == 180:
-        c, s = -1, 0
-    elif rotation == 270:
-        c, s = 0, -1
-    else:
-        raise NotImplementedError("rotation to arbitrary angle")
-
-    rotate = pikepdf.PdfMatrix((c, s, -s, c, 0, 0))
-
-    # 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((scale_x, 0, 0, scale_y, 0, 0))
-
-    # 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_id(*new_objgen)
-
-    queue.add(pdf_base.root.Outlines._objgen)
-    while queue:
-        objgen = queue.pop()
-        visited.add(objgen)
-        node = pdf_base._get_object_id(*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 +1,267 @@
+# © 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 contextlib import suppress
+from enum import IntEnum
+from pathlib import Path
+from typing import Dict, List
+
+from tqdm import tqdm
+
+from ._sync import run_pipeline
+from ._validation import check_options
+from .cli import parser
+
+
+class TqdmConsole:
+    """Wrapper to log messages in a way that is compatible with tqdm progress bar
+
+    This routes log messages through tqdm so that it can print them above the
+    progress bar, and then refresh the progress bar, rather than overwriting
+    it which looks messy.
+
+    For some reason Python 3.6 prints extra empty messages from time to time,
+    so we suppress those.
+    """
+
+    def __init__(self, file):
+        self.file = file
+        self.py36 = sys.version_info[0:2] == (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):
+        with suppress(AttributeError):
+            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: Verbosity,
+    progress_bar_friendly: bool = True,
+    manage_root_logger: bool = 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
+
+    Returns:
+        The toplevel logger for ocrmypdf (or the root logger, if we are managing it).
+    """
+
+    prefix = '' if manage_root_logger else 'ocrmypdf'
+    log = logging.getLogger(prefix)
+    log.setLevel(logging.DEBUG)
+
+    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 >= 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)
+
+    return log
+
+
+def create_options(*, input_file: os.PathLike, output_file: os.PathLike, **kwargs):
+    cmdline = []
+    deferred = []
+
+    for arg, val in kwargs.items():
+        if val is None:
+            continue
+
+        # These arguments with special handling for which we bypass
+        # argparse
+        if arg in {'tesseract_env', 'progress_bar'}:
+            deferred.append((arg, val))
+            continue
+
+        cmd_style_arg = arg.replace('_', '-')
+
+        # Booleans are special: add only if True, omit for False
+        if isinstance(val, bool):
+            if val:
+                cmdline.append(f"--{cmd_style_arg}")
+            continue
+
+        # We have a parameter
+        cmdline.append(f"--{cmd_style_arg}")
+        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'] = os.fspath(input_file)
+
+    return options
+
+
+def ocr(  # pylint: disable=unused-argument
+    input_file: os.PathLike,
+    output_file: os.PathLike,
+    *,
+    language: List[str] = None,
+    image_dpi: int = None,
+    output_type=None,
+    sidecar: os.PathLike = None,
+    jobs: int = None,
+    use_threads: bool = None,
+    title: str = None,
+    author: str = None,
+    subject: str = None,
+    keywords: str = None,
+    rotate_pages: bool = None,
+    remove_background: bool = None,
+    deskew: bool = None,
+    clean: bool = None,
+    clean_final: bool = None,
+    unpaper_args: str = None,
+    oversample: int = None,
+    remove_vectors: bool = None,
+    threshold: bool = None,
+    force_ocr: bool = None,
+    skip_text: bool = None,
+    redo_ocr: bool = None,
+    skip_big: float = None,
+    optimize: int = None,
+    jpg_quality: int = None,
+    png_quality: int = None,
+    jbig2_lossy: bool = None,
+    jbig2_page_group_size: int = None,
+    pages: str = None,
+    max_image_mpixels: float = None,
+    tesseract_config: List[str] = None,
+    tesseract_pagesegmode: int = None,
+    tesseract_oem: int = None,
+    pdf_renderer=None,
+    tesseract_timeout: float = None,
+    rotate_pages_threshold: float = None,
+    pdfa_image_compression=None,
+    user_words: os.PathLike = None,
+    user_patterns: os.PathLike = None,
+    fast_web_view: float = None,
+    keep_temporary_files: bool = None,
+    progress_bar: bool = None,
+    tesseract_env: Dict[str, str] = 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 +1,493 @@
+# © 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 ._version import PROGRAM_NAME as _PROGRAM_NAME
+from ._version import __version__ as _VERSION
+
+__all__ = ['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
+
+
+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=numeric(int, 0, 2),
+    default=0,
+    const=1,
+    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.",
+)
+advanced.add_argument(
+    '--fast-web-view',
+    type=numeric(float, 0),
+    default=1.0,
+    metavar="MEGABYTES",
+    help="If the size of file is more than this threshold (in MB), then "
+    "linearize the PDF for fast web viewing. This allows the PDF to be "
+    "displayed before it is fully downloaded in web browsers, but increases "
+    "the space required slightly. By default we skip this for small files "
+    "which do not benefit. If the threshold is 0 it will be apply to all files. "
+    "Set the threshold very high to disable.",
+)
+
+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
@@ -29,12 +31,20 @@ class ExitCode(IntEnum):
     child_process_error = 7
     encrypted_pdf = 8
     invalid_config = 9
+    pdfa_conversion_failed = 10
     other_error = 15
     ctrl_c = 130
 
 
 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):
@@ -43,6 +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):
@@ -75,7 +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

@@ -17,40 +17,262 @@
 
 """Wrappers to manage subprocess calls"""
 
+import logging
 import os
 import re
+import shutil
 import sys
-from subprocess import run, STDOUT, PIPE, CalledProcessError
-from ..exceptions import MissingDependencyError
+from collections.abc import Mapping
+from functools import lru_cache
+from subprocess import PIPE, STDOUT, CalledProcessError
+from subprocess import run as subprocess_run
+
+from ..exceptions import ExitCode, MissingDependencyError
+
+log = logging.getLogger(__name__)
 
 
+def _get_program(args, env=None):
+    program = args[0]
+    test_path = env.get('_OCRMYPDF_TEST_PATH', '')
+    if test_path:
+        program = shutil.which(program, path=test_path)
+    return program
 
-def get_version(program, *, 
-        version_arg='--version', regex=r'(\d+(\.\d+)*)'):
-    "Get the version of the specified program"
-    args_prog = [
-        program,
-        version_arg
-    ]
+
+def run(args, *, env=None, **kwargs):
+    """Wrapper around subprocess.run()
+
+    The main purpose of this wrapper is to allow us to substitute the main program
+    for a spoof in the test suite. The hidden variable _OCRMYPDF_TEST_PATH replaces
+    the main PATH as a location to check for programs to run.
+
+    Secondly we have to account for behavioral differences in Windows in particular.
+    Creating symbolic links in Windows requires administrator privileges and
+    may not work if for some reason we're using a FAT file system or the temporary
+    folder is on a different drive from the working folder. The test suite
+    works around this by creating shim Python scripts that perform the same function
+    as a symbolic link, but those shims require support on this side, to ensure
+    we call them with Python.
+
+    """
+    if not env:
+        env = os.environ
+
+    # Search in spoof path if necessary
+    program = _get_program(args, env)
+
+    # If we are running a .py on Windows, ensure we call it with this Python
+    # (to support test suite shims)
+    if os.name == 'nt' and program.lower().endswith('.py'):
+        args = [sys.executable, program] + args[1:]
+    else:
+        args = [program] + args[1:]
+
+    if os.name == 'nt':
+        paths = os.pathsep.join(os.get_exec_path(env))
+        if not shutil.which(args[0], path=paths):
+            shimmed_path = shim_paths_with_program_files(env)
+            new_args0 = shutil.which(args[0], path=shimmed_path)
+            if new_args0:
+                args[0] = new_args0
+
+    process_log = log.getChild(os.path.basename(program))
+    process_log.debug("Running: %s", args)
+    if sys.version_info < (3, 7) and os.name == 'nt':
+        # Can't use close_fds=True on Windows with Python 3.6 or older
+        # https://bugs.python.org/issue19575, etc.
+        kwargs['close_fds'] = False
+    proc = subprocess_run(args, env=env, **kwargs)
+    if process_log.isEnabledFor(logging.DEBUG):
+        try:
+            stderr = proc.stderr.decode('utf-8', 'replace')
+        except AttributeError:
+            stderr = proc.stderr
+        if stderr:
+            process_log.debug("stderr = %s", stderr)
+    return proc
+
+
+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]
     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 CalledProcessError as e:
-        if e.returncode < 0:
-            raise MissingDependencyError(
-                "Ran program '{}' but it exited with an error:\n{}".format(
-                    program, e.output)) from e
+    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:
+            raise MissingDependencyError(
+                f"Ran program '{program}' but it exited with an error:\n{e.output}"
+            ) from e
+        raise MissingDependencyError(
+            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
+
+
+def shim_paths_with_program_files(env=None):
+    if not env:
+        env = os.environ
+    program_files = env.get('PROGRAMFILES', '')
+    if not program_files:
+        return env.get('PATH', '')
+    paths = []
+    try:
+        for dirname in os.listdir(program_files):
+            if dirname.lower() == 'tesseract-ocr':
+                paths.append(os.path.join(program_files, dirname))
+            elif dirname.lower() == 'gs':
+                try:
+                    latest_gs = max(
+                        os.listdir(os.path.join(program_files, dirname)),
+                        key=lambda d: float(d[2:]),
+                    )
+                except (FileNotFoundError, NotADirectoryError):
+                    continue
+                paths.append(os.path.join(program_files, dirname, latest_gs, 'bin'))
+    except EnvironmentError:
+        pass
+    paths.extend(path for path in os.get_exec_path(env) if path not in set(paths))
+    return os.pathsep.join(paths)
+
+
+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}.
+'''
+
+windows_install_advice = '''
+If not already installed, install the Chocolatey package manager. Then use
+a command prompt to install the missing package:
+    choco install {package}
+'''
+
+
+def _get_platform():
+    if sys.platform.startswith('freebsd'):
+        return 'freebsd'
+    elif sys.platform.startswith('linux'):
+        return 'linux'
+    elif sys.platform.startswith('win'):
+        return 'windows'
+    return sys.platform
+
+
+def _error_trailer(program, package, **kwargs):
+    if isinstance(package, Mapping):
+        package = package.get(_get_platform(), program)
+
+    if _get_platform() == 'darwin':
+        log.info(osx_install_advice.format(**locals()))
+    elif _get_platform() == 'linux':
+        log.info(linux_install_advice.format(**locals()))
+    elif _get_platform() == 'windows':
+        log.info(windows_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

@@ -15,126 +15,195 @@
 # 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, CalledProcessError
-from shutil import copy
-from functools import lru_cache
+"""Interface to Ghostscript executable"""
+
+import logging
+import os
 import re
-import sys
+import warnings
+from contextlib import suppress
+from functools import lru_cache
+from io import BytesIO
+from os import fspath
+from pathlib import Path
+from shutil import which
+from subprocess import PIPE, CalledProcessError
+
 from PIL import Image
-from . import get_version
-from ..exceptions import SubprocessOutputError, MissingDependencyError
-from ..helpers import fspath
+
+from ..exceptions import MissingDependencyError, SubprocessOutputError
+from . import get_version, run
+
+gslog = logging.getLogger()
+
+GS = 'gs'
+if os.name == 'nt':
+    GS = which('gswin64c')
+    if not GS:
+        GS = which('gswin32c')
+        if not GS:
+            raise MissingDependencyError(
+                """
+                ---------------------------------------------------------------------
+                This error normally occurs when ocrmypdf can't Ghostscript.  Please
+                ensure Ghostscript is installed and its location is added to the
+                system PATH environment variable.
+
+                For details see:
+                    https://ocrmypdf.readthedocs.io/en/latest/installation.html
+                ---------------------------------------------------------------------
+                """
+            )
+    GS = Path(GS).stem
 
 
 @lru_cache(maxsize=1)
 def version():
-    return get_version('gs')
+    return get_version(GS)
 
 
 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):
     return re.search(r'error', stream, flags=re.IGNORECASE)
 
 
-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.
-    
-    The image is sized to match the integer pixels dimensions implied by 
+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
      be overridden with the values in page_dpi.
-    
+
     :param input_file: pathlike
     :param output_file: pathlike
     :param xres: resolution at which to rasterize page
-    :param yres: 
-    :param raster_device: 
-    :param log: 
+    :param yres:
+    :param raster_device:
+    :param log:
     :param pageno: page number to rasterize (beginning at page 1)
-    :param page_dpi: resolution tuple (x, y) overriding output image DPI 
-    :return: 
+    :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
+    if not log:
+        log = gslog
 
-    autorotate = '/PageByPage' if rotation is None else '/None'
-
-    with NamedTemporaryFile(delete=True) as tmp:
-        args_gs = [
-            'gs',
+    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)
+            f'-sDEVICE={raster_device}',
+            f'-dFirstPage={pageno}',
+            f'-dLastPage={pageno}',
+            f'-r{res[0]:f}x{res[1]:f}',
         ]
-        
-        log.debug(args_gs)
-        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)
+        + (['-dFILTERVECTOR'] if filter_vector else [])
+        + [
+            '-o',
+            '-',
+            '-sstdout=%stderr',
+            '-dAutoRotatePages=/None',  # Probably has no effect on raster
+            '-f',
+            fspath(input_file),
+        ]
+    )
 
-        if p.returncode != 0:
-            log.error('Ghostscript rasterizing failed')
-            raise SubprocessOutputError()
+    log.debug(args_gs)
+    try:
+        p = run(args_gs, stdout=PIPE, stderr=PIPE, check=True)
+    except CalledProcessError as e:
+        log.error(e.stderr.decode(errors='replace'))
+        raise SubprocessOutputError('Ghostscript rasterizing failed')
+    else:
+        stderr = p.stderr.decode(errors='replace')
+        if _gs_error_reported(stderr):
+            log.error(stderr)
+        elif stderr:
+            log.debug(stderr)
 
-        # 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
-
-        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 
-                # counterclockwise so this cancels out the rotation
-                if rotation == 90:
-                    im = im.transpose(Image.ROTATE_90)
-                elif rotation == 180:
-                    im = im.transpose(Image.ROTATE_180)
-                elif rotation == 270:
-                    im = im.transpose(Image.ROTATE_270)
-                if rotation % 180 == 90:
-                    page_dpi = page_dpi[1], page_dpi[0]
-            im.save(fspath(output_file), dpi=page_dpi)
+    with Image.open(BytesIO(p.stdout)) as im:
+        if rotation is not None:
+            log.debug("Rotating output by %i", rotation)
+            # rotation is a clockwise angle and Image.ROTATE_* is
+            # counterclockwise so this cancels out the rotation
+            if rotation == 90:
+                im = im.transpose(Image.ROTATE_90)
+            elif rotation == 180:
+                im = im.transpose(Image.ROTATE_180)
+            elif rotation == 270:
+                im = im.transpose(Image.ROTATE_270)
+            if rotation % 180 == 90:
+                page_dpi = page_dpi[1], page_dpi[0]
+        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=None,  # deprecated parameter
+    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
+    if threads is not None:
+        warnings.warn(
+            "use of deprecated parameter 'threads'", category=DeprecationWarning
+        )
+
     compression_args = []
     if compression == 'jpeg':
         compression_args = [
@@ -156,8 +225,8 @@ def generate_pdfa(pdf_pages, output_file, compression, log,
             "-dAutoFilterGrayImages=true",
         ]
 
-    # Older versions of Ghostscript expect a leading slash in 
-    # sColorConversionStrategy, newer ones should not have it. See Ghostscript 
+    # Older versions of Ghostscript expect a leading slash in
+    # sColorConversionStrategy, newer ones should not have it. See Ghostscript
     # git commit fe1c025d.
     strategy = 'RGB' if version() >= '9.19' else '/RGB'
 
@@ -167,46 +236,59 @@ def generate_pdfa(pdf_pages, output_file, compression, log,
         # https://bugs.ghostscript.com/show_bug.cgi?id=699216
         compression_args.append('-dPassThroughJPEGImages=false')
 
-    with NamedTemporaryFile(delete=True) as gs_pdf:
-        args_gs = [
-            "gs",
+    # 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",
+            "-dSAFER",
             "-dCompatibilityLevel=" + str(pdf_version),
-            "-dNumRenderingThreads=" + str(threads),
             "-sDEVICE=pdfwrite",
             "-dAutoRotatePages=/None",
             "-sColorConversionStrategy=" + strategy,
-            "-sProcessColorModel=DeviceRGB"
-        ] + compression_args + [
+        ]
+        + compression_args
+        + [
             "-dJPEGQ=95",
             "-dPDFA=" + pdfa_part,
             "-dPDFACompatibilityPolicy=1",
-            "-sOutputFile=" + gs_pdf.name,
+            "-o",
+            "-",
+            "-sstdout=%stderr",
         ]
-        args_gs.extend(pdf_pages)
-        log.debug(args_gs)
-        p = run(args_gs, stdout=PIPE, stderr=STDOUT,
-                universal_newlines=True)
-
-        if _gs_error_reported(p.stdout):
-            log.error(p.stdout)
-        elif 'overprint mode not set' in p.stdout:
+    )
+    args_gs.extend(fspath(s) for s in pdf_pages)  # Stringify Path objs
+    try:
+        with Path(output_file).open('wb') as output:
+            p = run(args_gs, stdout=output, stderr=PIPE, check=True)
+    except CalledProcessError as e:
+        # Ghostscript does not change return code when it fails to create
+        # PDF/A - check PDF/A status elsewhere
+        log.error(e.stderr.decode(errors='replace'))
+        raise SubprocessOutputError('Ghostscript PDF/A rendering failed')
+    else:
+        stderr = p.stderr.decode('utf-8', errors='replace')
+        if _gs_error_reported(stderr):
+            last_part = None
+            repcount = 0
+            for part in stderr.split('****'):
+                if part != last_part:
+                    if repcount > 1:
+                        log.error(f"(previous error message repeated {repcount} times)")
+                        repcount = 0
+                    log.error(part)
+                else:
+                    repcount += 1
+                last_part = part
+        elif 'overprint mode not set' in stderr:
             # Unless someone is going to print PDF/A documents on a
             # magical sRGB printer I can't see the removal of overprinting
             # being a problem....
             log.debug(
                 "Ghostscript had to remove PDF 'overprinting' from the "
                 "input file to complete PDF/A conversion. "
-                )
-        else:
-            log.debug(p.stdout)
-
-        if p.returncode == 0:
-            # Ghostscript does not change return code when it fails to create
-            # PDF/A - check PDF/A status elsewhere
-            copy(gs_pdf.name, output_file)
-        else:
-            log.error('Ghostscript PDF/A rendering failed')
-            raise SubprocessOutputError()
+            )

src/ocrmypdf/exec/jbig2enc.py

@@ -15,14 +15,13 @@
 # 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
+"""Interface to jbig2 executable"""
 
-from . import get_version
-from ..exceptions import ExitCode, MissingDependencyError
+from functools import lru_cache
+from subprocess import PIPE
+
+from ..exceptions import MissingDependencyError
+from . import get_version, run
 
 
 @lru_cache(maxsize=1)
@@ -30,7 +29,6 @@ def version():
     return get_version('jbig2', regex=r'jbig2enc (\d+(\.\d+)*).*')
 
 
-@lru_cache(maxsize=1)
 def available():
     try:
         version()
@@ -44,10 +42,20 @@ 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 +15,16 @@
 # 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
+"""Interface to pngquant executable"""
 
+from functools import lru_cache
+from subprocess import run
+from tempfile import NamedTemporaryFile
+
+from PIL import Image
+
+from ..exceptions import MissingDependencyError
 from . import get_version
-from ..exceptions import ExitCode
 
 
 @lru_cache(maxsize=1)
@@ -31,15 +32,40 @@ 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'):
+        with Image.open(input_file) as im, 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

@@ -15,211 +15,47 @@
 # 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
-import sys
-import os
-import re
-import resource
+"""Interface to qpdf executable"""
 
-from ..exceptions import InputFileError, SubprocessOutputError, \
-    MissingDependencyError, EncryptedPdfError
-from . import  get_version
-from ..helpers import re_symlink
+from io import StringIO
+
+import pikepdf
 
 
-@lru_cache(maxsize=1)
 def version():
-    return get_version('qpdf', regex=r'qpdf version (.+)')
+    return pikepdf.__libqpdf_version__
 
 
 def check(input_file, log=None):
-    args_qpdf = [
-        'qpdf',
-        '--check',
-        input_file
-    ]
-
-    if log is None:
-        import logging as log
-
+    pdf = None
     try:
-        run(args_qpdf, stderr=STDOUT, stdout=PIPE, universal_newlines=True,
-            check=True)
-    except CalledProcessError as e:
-        if e.returncode == 2:
-            log.error("{0}: not a valid PDF, and could not repair it.".format(
-                input_file))
-            log.error("Details:")
-            log.error(e.output)
-        elif e.returncode == 3:
-            log.info("qpdf --check returned warnings:")
-            log.info(e.output)
-        else:
-            log.warning(e.output)
+        pdf = pikepdf.open(input_file)
+    except pikepdf.PdfError as e:
+        if log:
+            log.error(e)
         return False
-    return True
+    else:
+        messages = pdf.check()
+        for msg in messages:
+            if 'error' in msg.lower():
+                log.error(msg)
+            else:
+                log.warning(msg)
 
-
-def _probably_encrypted(e):
-    """qpdf can report a false positive "file is encrypted" message for damaged
-    files - suppress this"""
-    return e.returncode == 2 and \
-        'invalid password' in e.output and \
-        'file is damaged' not in e.output
-
-
-def repair(input_file, output_file, log):
-    args_qpdf = [
-        'qpdf', input_file, output_file
-    ]
-    try:
-        run(args_qpdf, stderr=STDOUT, stdout=PIPE, universal_newlines=True,
-            check=True)
-    except CalledProcessError as e:
-        if e.returncode == 3 and e.output.find("operation succeeded"):
-            log.debug('qpdf found and fixed errors: ' + e.output)
-            return
-
-        if _probably_encrypted(e):
-            raise EncryptedPdfError() from e
-        elif e.returncode == 2:
-            log.error("{0}: not a valid PDF, and could not repair it.".format(
-                      input_file))
-            log.error("Details: " + e.output)
-            raise InputFileError() from e
+        sio = StringIO()
+        linearize = None
+        try:
+            pdf.check_linearization(sio)
+        except RuntimeError:
+            pass
         else:
-            log.error("{0}: unknown error".format(
-                      input_file))
-            log.error(e.output)
-            raise SubprocessOutputError() from e
+            linearize = sio.getvalue()
+            if linearize:
+                log.warning(linearize)
 
-
-def get_npages(input_file, log):
-    try:
-        pages = run(
-            ['qpdf', '--show-npages', input_file],
-            universal_newlines=True, check=True, stdout=PIPE, stderr=STDOUT)
-    except CalledProcessError as e:
-        if e.returncode == 2 and e.output.find('No such file'):
-            log.error(e.output)
-            raise InputFileError() from e
-    return int(pages)
-
-
-def split_pages(input_file, work_folder, npages):
-    """Split multipage PDF into individual pages.
-
-    Incredibly enough, this multiple process approach is about 70 times
-    faster than using Ghostscript.
-    """
-    for n in range(int(npages)):
-        args_qpdf = [
-            'qpdf', input_file,
-            '--pages', input_file, '{0}'.format(n + 1), '--',
-            os.path.join(work_folder, '{0:06d}.page.pdf'.format(n + 1))
-        ]
-        run(args_qpdf, check=True)
-
-
-def extract_page(input_file, output_file, pageno):
-    args_qpdf = [
-        'qpdf', input_file,
-        '--pages', input_file, '{0}'.format(pageno + 1), '--',
-        output_file
-    ]
-    run(args_qpdf, check=True)
-
-
-def _merge_inner(input_files, output_file, min_version=None, log=None):
-    """Merge the list of input files (all filenames) into the output file.
-
-    The input files may contain one or more pages.
-    """
-
-    # Single page 'merges' should still be attempted to that the same error
-    # checking is applied to single page case
-
-    version_arg = ['--min-version={}'.format(min_version)] \
-                  if min_version else []
-
-    if log is None:
-        import logging as log
-
-    args_qpdf = [
-        'qpdf'
-    ] + version_arg + [
-        input_files[0], '--pages'
-    ] + input_files + ['--', output_file]
-
-    try:
-        run(args_qpdf, check=True, stderr=PIPE, universal_newlines=True)
-    except CalledProcessError as e:
-        if e.returncode == 3 and \
-                e.stderr.find("unknown token while reading object") and \
-                e.stderr.find("operation succeeded"):
-            # Only whitelist the 'unknown token' problem (decimal/string issue)
-            # qpdf issue #165
-            log.warning('qpdf found and fixed errors: ' + e.stderr)
-            return
-        raise e from e
-
-
-def merge(input_files, output_file, min_version=None, log=None, max_files=None):
-    """Merge the list of input files (all filenames) into the output file.
-
-    The input files may contain one or more pages.
-
-    """
-    # qpdf requires that every file that contributes to the output has a file
-    # descriptor that remains open. That means, given our approach of one
-    # intermediate PDF per, we can practically hit the number of file
-    # descriptors.
-
-    if max_files is None or max_files < 2:
-        # Find out how many open file descriptors we can get away with
-        ulimits = resource.getrlimit(resource.RLIMIT_NOFILE)
-        max_open_files = ulimits[0]
-        max_files = max_open_files // 2  # Conservative guess
-
-    # We'll write things alongside the output file
-    output_dir = os.path.dirname(output_file)
-
-    import random
-    import string
-
-    def randstr():
-        return ''.join(random.sample(string.ascii_lowercase, 6))
-
-    # How many files to grab at once, merging all their contents
-    step_size = max_files
-
-    workqueue = input_files.copy()
-    counter = 1
-    next_workqueue = []
-    while len(workqueue) > 1 or len(next_workqueue) > 0:
-        # Take n files out of the queue
-        n = min(step_size, len(workqueue))
-        job = workqueue[0:n]
-        del workqueue[0:n]
-        log.debug('merging ' + repr(job))
-
-        # Merge them into 1 file, which will contain n^depth pages
-        merge_file = os.path.join(
-            output_dir, "merge-{:06d}-{}.pdf".format(counter, randstr()))
-        counter += 1
-        _merge_inner(job, merge_file, min_version=min_version, log=log)
-
-        # On the next
-        next_workqueue.append(merge_file)
-        log.debug('next_workqueue ' + repr(next_workqueue))
-
-        # If we're out of things to do in this queue, move on to the next
-        # queue. On the counter-th pass of the workqueue we can chew through
-        # (step_size)**N pages, so on most systems the second pass finishes
-        # the job.
-        if len(workqueue) == 0:
-            workqueue = next_workqueue
-            next_workqueue = []
-
-    re_symlink(workqueue.pop(), output_file, log)
+        if not messages and not linearize:
+            return True
+        return False
+    finally:
+        if pdf:
+            pdf.close()

src/ocrmypdf/exec/tesseract.py

@@ -15,24 +15,25 @@
 # 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
+"""Interface to Tesseract executable"""
+
+import logging
 import os
-import re
 import shutil
-from functools import lru_cache
 from collections import namedtuple
-from textwrap import dedent
-from subprocess import PIPE, CalledProcessError, \
-    TimeoutExpired, check_output, STDOUT
 from contextlib import suppress
+from os import fspath
+from subprocess import PIPE, STDOUT, CalledProcessError, TimeoutExpired
 
-from ..exceptions import MissingDependencyError, TesseractConfigError
-from ..helpers import page_number
-from . import get_version
+from ..exceptions import (
+    MissingDependencyError,
+    SubprocessOutputError,
+    TesseractConfigError,
+)
+from ..helpers import page_number, safe_symlink
+from . import get_version, run
 
-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"
@@ -41,8 +42,8 @@ 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'>
@@ -52,95 +53,125 @@ HOCR_TEMPLATE = """<?xml version="1.0" encoding="UTF-8"?>
 """
 
 
-@lru_cache(maxsize=1)
-def version():
-    return get_version('tesseract', regex=r'tesseract\s(.+)')
+class TesseractLoggerAdapter(logging.LoggerAdapter):
+    def process(self, msg, kwargs):
+        kwargs['extra'] = self.extra
+        return '[tesseract] %s' % (msg), kwargs
 
 
-def v4():
+def version(tesseract_env=None):
+    return get_version('tesseract', regex=r'tesseract\s(.+)', env=tesseract_env)
+
+
+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, langs=None):
     """Does Tesseract have textonly_pdf capability?
 
-    Available in 3.05.01, and v4.00.00alpha since January 2017. Best to
-    parse the parameter list
+    Available in v4.00.00alpha since January 2017. Best to
+    parse the parameter list.
     """
-    args_tess = [
-        'tesseract',
-        '--print-parameters'
-    ]
+    args_tess = tess_base_args(langs, engine_mode=None) + ['--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 has_user_words(tesseract_env=None):
+    """Does Tesseract have --user-words capability?
+
+    Not available in 4.0, but available in 4.1. Also available in 3.x, but
+    we no longer support 3.x.
+    """
+    return version(tesseract_env) >= '4.1'
 
 
-@lru_cache(maxsize=1)
-def languages():
-    args_tess = [
-        'tesseract',
-        '--list-langs'
-    ]
+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
+
+    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
+
+    for line in output.splitlines():
+        if line.startswith('Error'):
+            raise MissingDependencyError(lang_error(output))
+    header, *rest = output.splitlines()
+    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():
+    if engine_mode is not None:
         args.extend(['--oem', str(engine_mode)])
     return args
 
 
-def get_orientation(input_file, language: list, 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():
@@ -150,33 +181,21 @@ def get_orientation(input_file, language: list, engine_mode, timeout: float,
                 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))
+def tesseract_log_output(mainlog, stdout, input_file):
+    log = TesseractLoggerAdapter(
+        mainlog, extra=mainlog.extra if hasattr(mainlog, 'extra') else None
+    )
 
     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.")
-        text = stdout.decode('utf-8', 'backslashreplace')
+        text = stdout.decode('utf-8', 'ignore')
 
     lines = text.splitlines()
     for line in lines:
@@ -185,21 +204,31 @@ def tesseract_log_output(log, stdout, input_file):
         elif line.startswith("Warning in pixReadMem"):
             continue
         elif 'diacritics' in line:
-            log.warning(prefix + "lots of diacritics - possibly poor OCR")
+            log.warning("lots of diacritics - possibly poor OCR")
         elif line.startswith('OSD: Weak margin'):
-            log.warning(prefix + "unsure about page orientation")
+            log.warning("unsure about page orientation")
+        elif 'Error in pixScanForForeground' in line:
+            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(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())
+            log.error(line.strip())
         elif 'warning' in line.lower():
-            log.warning(prefix + line.strip())
+            log.warning(line.strip())
         elif 'read_params_file' in line.lower():
-            log.error(prefix + line.strip())
+            log.error(line.strip())
         else:
-            log.info(prefix + line.strip())
+            log.info(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")
 
 
@@ -208,8 +237,8 @@ def _generate_null_hocr(output_hocr, output_sidecar, image):
     the same size as the input image."""
     from PIL import Image
 
-    im = Image.open(image)
-    w, h = im.size
+    with Image.open(image) as im:
+        w, h = im.size
 
     with open(output_hocr, 'w', encoding="utf-8") as f:
         f.write(HOCR_TEMPLATE.format(w, h))
@@ -217,19 +246,28 @@ 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'))
+    output_hocr = next(o for o in output_files if fspath(o).endswith('.hocr'))
+    output_sidecar = next(o for o in output_files if fspath(o).endswith('.txt'))
     prefix = os.path.splitext(output_hocr)[0]
 
     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])
@@ -239,35 +277,30 @@ 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
@@ -284,7 +317,7 @@ def use_skip_page(text_only, skip_pdf, output_pdf, output_text):
         # Substitute a "skipped page"
         with suppress(FileNotFoundError):
             os.remove(output_pdf)  # In case it was partially created
-        os.symlink(skip_pdf, output_pdf)
+        safe_symlink(skip_pdf, output_pdf)
         return
 
     # Or normally, just write a 0 byte file to the output to indicate a skip
@@ -292,11 +325,24 @@ 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
@@ -308,14 +354,14 @@ 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, language):
         args_tesseract.extend(['-c', 'textonly_pdf=1'])
 
     if user_words:
@@ -329,31 +375,27 @@ 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 +18,19 @@
 # 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
+"""Interface to unpaper executable"""
+
 import os
+import shlex
 from functools import lru_cache
-from ..exceptions import MissingDependencyError
+from subprocess import PIPE, STDOUT, CalledProcessError
+from tempfile import TemporaryDirectory
+
+from PIL import Image
+
+from ..exceptions import MissingDependencyError, SubprocessOutputError
 from . import get_version
-
-
-try:
-    from PIL import Image
-except ImportError:
-    print("Could not find Python3 imaging library", file=sys.stderr)
-    raise
+from . import run as external_run
 
 
 @lru_cache(maxsize=1)
@@ -40,63 +39,89 @@ 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'}
 
-    im = Image.open(input_file)
-    if im.mode not in SUFFIXES.keys():
-        log.info("Converting image to other colorspace")
+    with TemporaryDirectory() as tmpdir, Image.open(input_file) as im:
+        if im.mode not in SUFFIXES.keys():
+            log.info("Converting image to other colorspace")
+            try:
+                if im.mode == 'P' and len(im.getcolors()) == 2:
+                    im = im.convert(mode='1')
+                else:
+                    im = im.convert(mode='RGB')
+            except IOError as e:
+                im.close()
+                raise MissingDependencyError(
+                    "Could not convert image with type " + im.mode
+                ) from e
+
         try:
-            if im.mode == 'P' and len(im.getcolors()) == 2:
-                im = im.convert(mode='1')
-            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
+            suffix = SUFFIXES[im.mode]
+        except KeyError:
+            raise MissingDependencyError(
+                "Failed to convert image to a supported format."
+            ) 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
-
-    with NamedTemporaryFile(suffix=suffix) as input_pnm, \
-            NamedTemporaryFile(suffix=suffix, mode="r+b") as output_pnm:
+        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 = external_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:
+                with Image.open(output_pnm) as imout:
+                    imout.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/helpers.py

@@ -15,64 +15,75 @@
 # 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
-from collections.abc import Iterable
-from contextlib import suppress, contextmanager
-from pathlib import Path
-import sys
-import os
+import logging
 import multiprocessing
+import os
+import shutil
 import warnings
+from collections.abc import Iterable
+from contextlib import suppress
+from functools import wraps
+from pathlib import Path
+
+log = logging.getLogger(__name__)
 
 
-def re_symlink(input_file, soft_link_name, log=None):
+def safe_symlink(input_file: os.PathLike, soft_link_name: os.PathLike, *args, **kwargs):
     """
     Helper function: relinks soft symbolic link if necessary
     """
-    input_file = str(input_file)  # For Py3.5
-    soft_link_name = str(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: safe_symlink(,log)")
+    if 'log' in kwargs:
+        log.warning('Deprecated: safe_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))
+    if os.name == 'nt':
+        # Don't actually use symlinks on Windows due to permission issues
+        shutil.copyfile(input_file, soft_link_name)
+        return
+
+    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 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])
+def monotonic(L: Iterable):
+    """Does list increase monotonically?"""
+    return all(b > a for a, b in zip(L, L[1:]))
+
+
+def page_number(input_file: os.PathLike):
+    """Get one-based page number implied by filename (000002.pdf -> 2)"""
+    return int(os.path.basename(os.fspath(input_file))[0:6])
 
 
 def available_cpu_count():
@@ -80,100 +91,63 @@ 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
 
 
-def is_file_writable(test_file):
+def is_file_writable(test_file: os.PathLike):
     """Intentionally racy test if target is writable.
 
     We intend to write to the output file if and only if we succeed and
     can replace it atomically. Before doing the OCR work, make sure
     the location is writable.
     """
-    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(str(p)))
+    try:
+        if not isinstance(test_file, Path):
+            p = Path(test_file)
         else:
+            p = test_file
+
+        if p.is_symlink():
             p = p.resolve(strict=False)
 
-    # p.is_file() throws an exception in some cases
-    if p.exists() and p.is_file():
-        return os.access(
-            str(p), os.W_OK,
-            effective_ids=(os.access in os.supports_effective_ids))
-    else:
-        try:
-            fp = p.open('wb')
-        except OSError:
-            return False
+        # p.is_file() throws an exception in some cases
+        if p.exists() and p.is_file():
+            return os.access(
+                os.fspath(p),
+                os.W_OK,
+                effective_ids=(os.access in os.supports_effective_ids),
+            )
         else:
-            fp.close()
-            with suppress(OSError):
-                p.unlink()
-        return True
-
-
-if sys.version_info[0:2] <= (3, 5):
-    def universal_open(p, *args, **kwargs):
-        "Work around Python 3.5's inability to open(pathlib.Path())"
-        try:
-            return p.open(*args, **kwargs)
-        except AttributeError:
-            return open(p, *args, **kwargs)
-
-
-    def fspath(path):
-        import pathlib
-        '''https://www.python.org/dev/peps/pep-0519/#os'''
-        if isinstance(path, (str, bytes)):
-            return path
-
-        # 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
+            try:
+                fp = p.open('wb')
+            except OSError:
+                return False
             else:
-                raise TypeError("expected __fspath__() to return str or bytes, "
-                                "not " + type(path).__name__)
-
-        raise TypeError(
-            "expected str, bytes, pathlib.Path or os.PathLike object, not "
-            + path_type.__name__)
-
-else:
-    universal_open = open
-    fspath = os.fspath
+                fp.close()
+                with suppress(OSError):
+                    p.unlink()
+            return True
+    except (EnvironmentError, RuntimeError) as e:
+        log.debug(e)
+        log.error(str(e))
+        return False
 
 
-def flatten_groups(groups):
-    for obj in groups:
-        if is_iterable_notstr(obj):
-            yield from obj
-        else:
-            yield obj
+def deprecated(func):
+    """Warn that function is deprecated"""
+
+    @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)
+
+    return new_func

src/ocrmypdf/hocrtransform.py

@@ -16,10 +16,10 @@
 # distribute, sublicense, and/or sell copies of the Software, and to
 # permit persons to whom the Software is furnished to do so, subject to
 # the following conditions:
-# 
+#
 # The above copyright notice and this permission notice shall be included
 # in all copies or substantial portions of the Software.
-# 
+#
 # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
 # OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 # MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
@@ -28,15 +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 PIL import Image
-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'])
 
@@ -45,7 +44,7 @@ class HocrTransformError(Exception):
     pass
 
 
-class HocrTransform():
+class HocrTransform:
 
     """
     A class for converting documents from the hOCR format.
@@ -54,17 +53,16 @@ class HocrTransform():
     """
 
     box_pattern = re.compile(r'bbox((\s+\d+){4})')
-    baseline_pattern = re.compile(r'''
-        baseline \s+ 
+    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
@@ -79,8 +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
@@ -91,7 +88,7 @@ class HocrTransform():
         if self.width is None or self.height is None:
             raise HocrTransformError("hocr file is missing page dimensions")
 
-    def __str__(self):
+    def __str__(self):  # pragma: no cover
         """
         Return the textual content of the HTML body
         """
@@ -139,14 +136,13 @@ class HocrTransform():
             matches = cls.baseline_pattern.search(element.attrib['title'])
             if matches:
                 return float(matches.group(1)), int(matches.group(2))
-        return (0, 0) 
+        return (0, 0)
 
     def pt_from_pixel(self, pxl):
         """
         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):
@@ -157,8 +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
@@ -169,17 +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:
@@ -189,40 +190,60 @@ class HocrTransform():
             pt = self.pt_from_pixel(pxl_coords)
 
             # draw the bbox border
-            if showBoundingboxes:
+            if showBoundingboxes:  # pragma: no cover
                 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):
+    def polyval(cls, poly, x):  # pragma: no cover
         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
@@ -248,30 +269,27 @@ class HocrTransform():
         # of the line box
         baseline_y2 = self.height - (line_box.y2 + intercept)
 
-        if showBoundingboxes:
+        if showBoundingboxes:  # pragma: no cover
             # draw the baseline in magenta, dashed
             pdf.setDash()
             pdf.setStrokeColorRGB(0.95, 0.65, 0.95)
             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)
@@ -282,28 +300,28 @@ class HocrTransform():
             box = self.pt_from_pixel(pxl_coords)
             if interwordSpaces:
                 # if  `--interword-spaces` is true, append a space
-                # to the end of each text element to allow simpler PDF viewers 
-                # such as PDF.js to better recognize words in search and copy 
+                # to the end of each text element to allow simpler PDF viewers
+                # such as PDF.js to better recognize words in search and copy
                 # and paste. Do not remove space from last word in line, even
                 # 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:
+            if showBoundingboxes:  # pragma: no cover
                 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:
@@ -312,14 +330,14 @@ class HocrTransform():
             # content stream while this issues a "offset" (Td) command.
             # .moveCursor() is relative to start of the text line, where the
             # "text line" means whatever reportlab defines it as. Do not use
-            # use .getCursor(), since moveCursor() rather unintuitively plans 
+            # use .getCursor(), since moveCursor() rather unintuitively plans
             # its moves relative to .getStartOfLine().
             # For skewed lines, in the text transform we set up a rotated
-            # coordinate system, so we don't have to account for the 
+            # coordinate system, so we don't have to account for the
             # incremental offset. Surprisingly most PDF viewers can handle this.
             cursor = text.getStartOfLine()
             dx = box.x1 - cursor[0]
-            dy = baseline_y2 - cursor[1]            
+            dy = baseline_y2 - cursor[1]
             text.moveCursor(dx, dy)
 
             # If reportlab tells us this word is 0 units wide, our best seems
@@ -332,19 +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/lib/__init__.py

@@ -16,13 +16,3 @@
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
 """Bindings to external libraries"""
-
-import os as _os
-
-try:
-    import fitz
-except ImportError:
-    fitz = None
-
-if _os.environ.get('_OCRMYPDF_NO_FITZ'):
-    fitz = None

src/ocrmypdf/lib/compile_leptonica.py

@@ -16,11 +16,13 @@
 # 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 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 +34,8 @@ 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 +65,28 @@ 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;
+
+/*! Array of compressed pix */
+struct PixaComp
+{
+    l_int32              n;         /*!< number of PixComp in ptr array    */
+    l_int32              nalloc;    /*!< number of PixComp ptrs allocated  */
+    l_int32              offset;    /*!< indexing offset into ptr array    */
+    struct PixComp     **pixc;      /*!< the array of ptrs to PixComp      */
+    struct Boxa         *boxa;      /*!< array of boxes                    */
+};
+typedef struct PixaComp PIXAC;
+
 struct Box
 {
     l_int32            x;
@@ -72,6 +98,25 @@ 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 +149,18 @@ 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  */
@@ -122,13 +179,59 @@ 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 * pixReadStream ( FILE *fp, l_int32 hint );
 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 );
+l_int32 getImpliedFileFormat ( const char *filename );
+l_ok pixWriteStream ( FILE *fp, PIX *pix, l_int32 format );
+l_ok pixWriteStreamJpeg ( FILE *fp, PIX *pixs, l_int32 quality, l_int32 progressive );
+l_ok pixWriteMem ( l_uint8 **pdata, size_t *psize, PIX *pix, l_int32 format );
+l_ok pixWriteMemJpeg ( l_uint8 **pdata, size_t *psize, PIX *pix, l_int32 quality, l_int32 progressive );
 l_int32
 pixWriteMemPng(l_uint8  **pdata,
                size_t    *psize,
@@ -137,6 +240,11 @@ pixWriteMemPng(l_uint8  **pdata,
 
 void pixDestroy ( PIX **ppix );
 
+l_ok
+pixEqual(PIX      *pix1,
+         PIX      *pix2,
+         l_int32 *psame);
+
 PIX *
 pixEndianByteSwapNew(PIX  *pixs);
 
@@ -157,6 +265,8 @@ 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 +292,18 @@ 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,
@@ -191,14 +313,12 @@ pixCleanBackgroundToWhite(PIX       *pixs,
                           l_int32    whiteval);
 
 BOX *
-pixFindPageForeground(PIX         *pixs,
-                      l_int32      threshold,
-                      l_int32      mindist,
-                      l_int32      erasedist,
-                      l_int32      pagenum,
-                      l_int32      showmorph,
-                      l_int32      display,
-                      const char  *pdfdir);
+pixFindPageForeground ( PIX *pixs,
+                        l_int32 threshold,
+                        l_int32 mindist,
+                        l_int32 erasedist,
+                        l_int32 showmorph,
+                        PIXAC *pixac );
 
 PIX *
 pixClipRectangle(PIX   *pixs,
@@ -265,13 +385,13 @@ 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 +409,116 @@ 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);
+
+PIX *
+pixGenHalftoneMask(PIX      *pixs,
+                        PIX     **ppixtext,
+                        l_int32  *phtfound,
+                        PIXA     *pixadb);
+
+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)
+    if Path('ocrmypdf/lib/_leptonica.py').exists() and Path('src/ocrmypdf').exists():
+        output = Path('ocrmypdf/lib/_leptonica.py')
+        output.rename('src/ocrmypdf/lib/_leptonica.py')
+        Path('ocrmypdf/lib').rmdir()
+        Path('ocrmypdf').rmdir()

src/ocrmypdf/optimize.py

@@ -0,0 +1,601 @@
+# © 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 concurrent.futures
+import sys
+import tempfile
+from collections import defaultdict
+from os import fspath
+from pathlib import Path
+
+import pikepdf
+from pikepdf import Dictionary, Name
+from PIL import Image
+from tqdm import tqdm
+
+from . import leptonica
+from ._jobcontext import PDFContext
+from .exceptions import OutputFileAccessError
+from .exec import jbig2enc, pngquant
+from .helpers import safe_symlink
+
+DEFAULT_JPEG_QUALITY = 75
+DEFAULT_PNG_QUALITY = 70
+
+
+def img_name(root, xref, ext):
+    return fspath(root / f'{xref:08d}{ext}')
+
+
+def png_name(root, xref):
+    return img_name(root, xref, '.png')
+
+
+def jpg_name(root, xref):
+    return img_name(root, xref, '.jpg')
+
+
+def tif_name(root, xref):
+    return img_name(root, xref, '.tif')
+
+
+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 %s", xref)
+        return None
+
+    pim = pikepdf.PdfImage(image)
+
+    if len(pim.filter_decodeparms) > 1:
+        log.debug("Skipping multiply filtered, xref %s", xref)
+        return None
+    filtdp = pim.filter_decodeparms[0]
+
+    if pim.bits_per_component > 8:
+        return None  # Don't mess with wide gamut images
+
+    if filtdp[0] == Name.JPXDecode:
+        return None  # Don't do JPEG2000
+
+    if Name.Decode in image:
+        return None  # Don't mess with custom Decode tables
+
+    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 / 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 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
+
+    # Don't try to PNG-optimize 1bpp images, since JBIG2 does it better.
+    if pim.bits_per_component == 1:
+        return None
+
+    try:
+        pim.indexed  # pikepdf 1.6.3 can't handle [/Indexed [/Array...]]
+    except NotImplementedError:
+        return None
+
+    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
+        # anything.
+        # 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 None
+
+        # We could get the ICC profile here, but there's no need to look at it
+        # for quality transcoding
+        # if icc:
+        #     stream = BytesIO(raw_jpeg.read_raw_bytes())
+        #     iccbytes = icc.read_bytes()
+        #     with Image.open(stream) as im:
+        #         im.save(jpg_name(root, xref), icc_profile=iccbytes)
+        try:
+            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 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))
+        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))
+        return xref, '.png'
+    elif (
+        not pim.indexed
+        and pim.colorspace == Name.ICCBased
+        and pim.bits_per_component == 1
+        and not options.jbig2_lossy
+    ):
+        # We can losslessly optimize 1-bit images to CCITT or JBIG2 without
+        # paying any attention to the ICC profile, provided we're not doing
+        # lossy JBIG2
+        pim.as_pil_image().save(png_name(root, xref))
+        return xref, '.png'
+
+    return None
+
+
+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):
+        try:
+            xobjs = page.Resources.XObject
+        except AttributeError:
+            continue
+        for _imname, image in dict(xobjs).items():
+            if image.objgen[1] != 0:
+                continue  # Ignore images in an incremental PDF
+            xref = image.objgen[0]
+            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: %s", (len(jbig2_groups),))
+    return jbig2_groups
+
+
+def _produce_jbig2_images(jbig2_groups, root, log, options):
+    """Produce JBIG2 images from their groups"""
+
+    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,
+            )
+            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()
+                if proc.stderr:
+                    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 = 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 + f'.{n:04d}')
+            jbig2_im_data = jbig2_im_file.read_bytes()
+            im_obj = pike.get_object(xref, 0)
+            im_obj.write(
+                jbig2_im_data, filter=Name.JBIG2Decode, decode_parms=jbig2_globals_dict
+            )
+
+
+def transcode_jpegs(pike, jpegs, root, log, options):
+    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')
+
+        # This produces a debug warning from PIL
+        # DEBUG:PIL.Image:Error closing: 'NoneType' object has no attribute
+        # '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)
+
+        if opt_jpg.stat().st_size > in_jpg.stat().st_size:
+            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(), filter=Name.DCTDecode)
+
+
+def transcode_pngs(pike, images, image_name_fn, root, log, options):
+    modified = set()
+    if options.optimize >= 2:
+        png_quality = (
+            max(10, options.png_quality - 10),
+            min(100, options.png_quality + 10),
+        )
+        with concurrent.futures.ThreadPoolExecutor(
+            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],
+                    )
+                )
+                modified.add(xref)
+            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 modified:
+        im_obj = pike.get_object(xref, 0)
+        try:
+            pix = leptonica.Pix.open(png_name(root, xref))
+            if pix.mode == '1':
+                compdata = pix.generate_pdf_ci_data(leptonica.lept.L_G4_ENCODE, 0)
+            else:
+                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
+
+        # 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):
+            log.debug(
+                f"pngquant: pngquant did not improve over original image "
+                f"{len(compdata)} > {int(im_obj.stream_dict.Length)}"
+            )
+            continue
+        if compdata.type == leptonica.lept.L_FLATE_ENCODE:
+            return rewrite_png(pike, im_obj, compdata, log)
+        elif compdata.type == leptonica.lept.L_G4_ENCODE:
+            return rewrite_png_as_g4(pike, im_obj, compdata, log)
+
+
+def rewrite_png_as_g4(pike, im_obj, compdata, log):
+    im_obj.BitsPerComponent = 1
+    im_obj.Width = compdata.w
+    im_obj.Height = compdata.h
+
+    im_obj.write(compdata.read())
+
+    log.debug(f"PNG to G4 {im_obj.objgen}")
+    if Name.Predictor in im_obj:
+        del im_obj.Predictor
+    if Name.DecodeParms in im_obj:
+        del im_obj.DecodeParms
+    im_obj.DecodeParms = Dictionary(
+        K=-1, BlackIs1=bool(compdata.minisblack), Columns=compdata.w
+    )
+
+    im_obj.Filter = Name.CCITTFaxDecode
+    return
+
+
+def rewrite_png(pike, im_obj, compdata, log):
+    # 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
+    # 10-14 means the actual predictor is specified in the data, so for any
+    # number >= 10 the PDF reader will use whatever the PNG data specifies.
+    # In practice Leptonica should use Paeth, 14, but 15 seems to be the
+    # designated value for "optimal". So we will use 15.
+    # See:
+    #   - PDF RM 7.4.4.4 Table 10
+    #   - https://github.com/DanBloomberg/leptonica/blob/master/src/pdfio2.c#L757
+    predictor = 15 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
+
+    log.debug(
+        f"PNG {im_obj.objgen}: palette={compdata.ncolors} spp={compdata.spp} bps={compdata.bps}"
+    )
+    if compdata.ncolors > 0:
+        # .ncolors is the number of colors in the palette, not the number of
+        # colors used in a true color image. The palette string is always
+        # given as RGB tuples even when the image is grayscale; see
+        # https://github.com/DanBloomberg/leptonica/blob/master/src/colormap.c#L2067
+        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 = [Name.Indexed, Name.DeviceRGB, compdata.ncolors - 1, palette_stream]
+        cs = palette
+    else:
+        # ncolors == 0 means we are using a colorspace without a palette
+        if compdata.spp == 1:
+            cs = Name.DeviceGray
+        elif compdata.spp == 3:
+            cs = Name.DeviceRGB
+        elif compdata.spp == 4:
+            cs = Name.DeviceCMYK
+    im_obj.ColorSpace = cs
+    im_obj.write(compdata.read(), filter=Name.FlateDecode, decode_parms=dparms)
+
+
+def optimize(input_file, output_file, context, save_settings):
+    log = context.log
+    options = context.options
+    if options.optimize == 0:
+        safe_symlink(input_file, output_file)
+        return
+
+    if options.jpeg_quality == 0:
+        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
+    if options.jbig2_page_group_size == 0:
+        options.jbig2_page_group_size = 10 if options.jbig2_lossy else 1
+
+    with pikepdf.Pdf.open(input_file) as pike:
+        root = Path(output_file).parent / 'images'
+        root.mkdir(exist_ok=True)
+
+        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)
+
+        jbig2_groups = extract_images_jbig2(pike, root, log, options)
+        convert_to_jbig2(pike, jbig2_groups, root, log, options)
+
+        target_file = Path(output_file).with_suffix('.opt.pdf')
+        pike.remove_unreferenced_resources()
+        pike.save(target_file, **save_settings)
+
+    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(f"Optimize ratio: {ratio:.2f} savings: {(100 * savings):.1f}%")
+
+    if savings < 0:
+        log.info("Image optimization did not improve the file - discarded")
+        # We still need to save the file
+        with pikepdf.open(input_file) as pike:
+            pike.remove_unreferenced_resources()
+            pike.save(output_file, **save_settings)
+    else:
+        safe_symlink(target_file, output_file)
+
+
+def main(infile, outfile, level, jobs=1):
+    from tempfile import TemporaryDirectory
+    from shutil import copy
+
+    class OptimizeOptions:
+        """Emulate ocrmypdf's options"""
+
+        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
+
+    options = OptimizeOptions(
+        input_file=infile,
+        jobs=jobs,
+        optimize_=int(level),
+        jpeg_quality=0,  # Use default
+        png_quality=0,
+        jb2lossy=False,
+    )
+
+    with TemporaryDirectory() as td:
+        context = PDFContext(options, td, infile, None)
+        tmpout = Path(td) / 'out.pdf'
+        optimize(
+            infile,
+            tmpout,
+            context,
+            dict(
+                compress_streams=True,
+                preserve_pdfa=True,
+                object_stream_mode=pikepdf.ObjectStreamMode.generate,
+            ),
+        )
+        copy(fspath(tmpout), fspath(outfile))
+
+
+if __name__ == '__main__':
+    main(sys.argv[1], sys.argv[2], sys.argv[3])

src/ocrmypdf/pdfa.py

@@ -15,55 +15,45 @@
 # 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
 
+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 base64
+from pathlib import Path
 from string import Template
-from binascii import hexlify
-from datetime import datetime
-from xml.parsers.expat import ExpatError
+
+import pikepdf
 import pkg_resources
-import PyPDF2 as pypdf
-import warnings
-from defusedxml.minidom import parseString as defused_parseString
-from unittest.mock import patch
 
 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)
+% Define an ICC profile :
+/ICCProfile $icc_profile
 def
 
-[$docinfo
-  /DOCINFO pdfmark
-
-% Define an ICC profile :
-
 [/_objdef {icc_PDFA} /type /stream /OBJ pdfmark
-[{icc_PDFA}
-<<
-  /N currentpagedevice /ProcessColorModel known {
-    currentpagedevice /ProcessColorModel get dup /DeviceGray eq
-    {pop 1} {
-      /DeviceRGB eq
-      {3}{4} ifelse
-    } ifelse
-  } {
-    (ERROR, unable to determine ProcessColorModel) == flush
-  } ifelse
->> /PUT pdfmark
-[{icc_PDFA} ICCProfile (r) file /PUT pdfmark
+[{icc_PDFA} << /N 3 >> /PUT pdfmark
+[{icc_PDFA} ICCProfile /PUT pdfmark
 
 % Define the output intent dictionary :
 
@@ -78,190 +68,60 @@ 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 ''
-
-    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.
+    :param target_filename: filename to save
+    :param icc: ICC identifier such as 'sRGB'
     """
-
-    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)
+    # Read the ICC profile, encode as ASCII85 and convert to a string which we
+    # will insert in the .ps file
+    bytes_icc_profile = Path(icc_profile).read_bytes()
+    icc_profile = base64.a85encode(bytes_icc_profile, adobe=True).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
-    with open(target_filename, 'w', encoding='ascii') as f:
-        f.write(ps)
+    Path(target_filename).write_text(ps, encoding='ascii')
+    return target_filename
 
 
 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.
     """
-    warnings.simplefilter('ignore', pypdf.utils.PdfReadWarning)
-    pdf = pypdf.PdfFileReader(filename)
-    try:
-        # Monkeypatch PyPDF2 to use defusedxml as its XML parser, for safety
-        with patch('xml.dom.minidom.parseString', new=defused_parseString):
-            xmp = pdf.getXmpMetadata()
-    except ExpatError:
-        return {'pass': False, 'output': 'pdf',
-                'conformance': 'Invalid XML metadata'}
-
-    try:
-        pdfa_nodes = xmp.getNodesInNamespace(
-            aboutUri='',
-            namespace='http://www.aiim.org/pdfa/ns/id/')
-    except AttributeError:
-        return {'pass': False, 'output': 'pdf',
-                'conformance': 'No XMP metadata'}
-
-    pdfa_dict = {attr.localName: attr.value for attr in pdfa_nodes}
-    if not pdfa_dict:
-        return {'pass': False, 'output': 'pdf',
-                'conformance': 'No XMP metadata'}
-
-    part_conformance = pdfa_dict['part'] + pdfa_dict['conformance']
-    valid_part_conforms = {'1A', '1B', '2A', '2B', '2U', '3A', '3B', '3U'}
-
-    conformance = 'PDF/A-{}'.format(
-        part_conformance)
-
-    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 Colorspace, Encoding, PdfInfo

src/ocrmypdf/pdfinfo/info.py

@@ -16,31 +16,31 @@
 # 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 Popen, PIPE
-from decimal import Decimal
-from math import hypot, isclose
+import logging
+import os
 import re
-import sys
-from collections import namedtuple
-from collections.abc import MutableMapping, Mapping
-import warnings
-from pathlib import Path
+from collections import defaultdict, namedtuple
+from decimal import Decimal
 from enum import Enum
-from contextlib import contextmanager
+from math import hypot, isclose
+from os import PathLike, fspath
+from pathlib import Path
+from warnings import warn
 
-from .lib import fitz
-from .helpers import universal_open, fspath
-
-from pikepdf import PdfMatrix
 import pikepdf
+from pikepdf import PdfMatrix
 
-Colorspace = Enum('Colorspace',
-                  'gray rgb cmyk lab icc index sep devn pattern jpeg2000')
+from ocrmypdf._concurrent import exec_progress_pool
+from ocrmypdf.exceptions import EncryptedPdfError
+from ocrmypdf.pdfinfo.layout import get_page_analysis, get_text_boxes
 
-Encoding = Enum('Encoding',
-                'ccitt jpeg jpeg2000 jbig2 asciihex ascii85 lzw flate ' + \
-                'runlength')
+logger = logging.getLogger()
 
+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,
@@ -71,7 +71,7 @@ FRIENDLY_ENCODING = {
     '/A85': Encoding.ascii85,
     '/LZW': Encoding.lzw,
     '/Fl': Encoding.flate,
-    '/RL': Encoding.runlength
+    '/RL': Encoding.runlength,
 }
 
 FRIENDLY_COMP = {
@@ -79,63 +79,48 @@ 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', 'name_index']
+)
+
+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
@@ -153,53 +138,69 @@ 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"""
+    name_index = defaultdict(lambda: [])
+    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':
+            name_index[image_name].append(settings)
+        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,
+        name_index=name_index,
+    )
 
 
 def _get_dpi(ctm_shorthand, image_size):
@@ -269,8 +270,7 @@ 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
@@ -295,7 +295,7 @@ class ImageInfo:
         self._bpc = int(pim.bits_per_component)
         try:
             self._enc = FRIENDLY_ENCODING.get(pim.filters[0], 'image')
-        except IndexError as e:
+        except IndexError:
             self._enc = '?'
 
         try:
@@ -305,14 +305,22 @@ class ImageInfo:
         if self._enc == Encoding.jpeg2000:
             self._color = Colorspace.jpeg2000
 
-        self._comp = FRIENDLY_COMP.get(self._color, '?')
+        if self._color == Colorspace.icc:
+            # Check the ICC profile to determine actual colorspace
+            pim_icc = pim.icc
+            if pim_icc.profile.xcolor_space == 'GRAY':
+                self._comp = 1
+            elif pim_icc.profile.xcolor_space == 'CMYK':
+                self._comp = 4
+            else:
+                self._comp = 3
+        else:
+            self._comp = FRIENDLY_COMP.get(self._color, '?')
 
-        # Bit of a hack... infer grayscale if component count is uncertain
-        # but encoding must be monochrome. This happens if a monochrome image
-        # has an ICC profile attached. Better solution would be to examine
-        # the ICC profile.
-        if self._comp == '?' and self._enc in (Encoding.ccitt, 'jbig2'):
-            self._comp = FRIENDLY_COMP[Colorspace.gray]
+            # Bit of a hack... infer grayscale if component count is uncertain
+            # but encoding only supports monochrome.
+            if self._comp == '?' and self._enc in (Encoding.ccitt, Encoding.jbig2):
+                self._comp = FRIENDLY_COMP[Colorspace.gray]
 
     @property
     def name(self):
@@ -355,19 +363,24 @@ 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):
@@ -391,6 +404,8 @@ 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)
@@ -403,17 +418,12 @@ 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):
-
-        # For each image that is drawn on this, check if we drawing the
-        # current image - yes this is O(n^2), but n == 1 almost always
-        for draw in contentsinfo.xobject_settings:
-            if draw.name != xobj:
-                continue
-
+        if xobj not in contentsinfo.name_index:
+            continue
+        for draw in contentsinfo.name_index[xobj]:
             if draw.stack_depth == 0 and _is_unit_square(draw.shorthand):
                 # At least one PDF in the wild (and test suite) draws an image
                 # when the graphics stack depth is 0, meaning that the image
@@ -422,8 +432,7 @@ 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):
@@ -453,11 +462,12 @@ 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.
@@ -466,7 +476,7 @@ 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
@@ -478,8 +488,7 @@ 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()
@@ -498,98 +507,76 @@ 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 _naive_find_text(*, pdf, page):
-    if not(page.get('/Type') == '/Page' and '/Contents' in page):
-        # Not a page, or has no /Contents => no text
-        return False
+def _page_has_text(text_blocks, page_width, page_height):
+    """Smarter text detection that ignores text in margins"""
 
-    # First we check the main content stream
-    contentsinfo = _interpret_contents(page, UNIT_SQUARE)
-    if contentsinfo.found_text:
-        return True
-
-    # Then see if there is a Form XObject with with a content stream
-    # that might have text.  For full completeness we should recursively
-    # search nested Form XObjects, as we do with images.  But that is
-    # rare.
-    if '/Resources' in page:
-        resources = page['/Resources']
-        if '/XObject' in resources:
-            xobjs = resources['/XObject'].as_dict()
-            for xobj in xobjs:
-                candidate = xobjs[xobj]
-                if candidate['/Subtype'] != '/Form':
-                    continue
-                form_xobject = candidate
-                # Content stream is attached to Form XObject dictionary
-                sub_contentsinfo = _interpret_contents(
-                    form_xobject, UNIT_SQUARE)
-                if sub_contentsinfo.found_text:
-                    return True
-    return False
-
-
-def _page_get_textblocks(infile, pageno):
-    "Smarter text detection"
-
-    doc = fitz.Document(infile)
-    if fitz.version[0] >= '1.13.0':
-        text = doc[pageno].getText('dict')
-    else:
-        import json
-        textjson = doc[pageno].getText('json')
-        text = json.loads(textjson)
-    if not text:
-        return
-
-    text['blocks'] = [blk for blk in text['blocks'] if blk['type'] == 0]
-    return text
-
-
-def _page_has_text(text):
-    "Smarter text detection that ignores text in margins"
-
-    pw, ph = text['width'], text['height']
+    pw, ph = float(page_width), float(page_height)
 
     margin_ratio = 0.125
-    interior_bbox = fitz.Rect(
-        margin_ratio * pw, margin_ratio * ph,
-        (1 - margin_ratio) * pw, (1 - margin_ratio) * ph
+    interior_bbox = (
+        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
+        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]
+
     has_text = False
-    for block in text['blocks']:
-        bbox = fitz.Rect(block['bbox'])
-        if bbox & interior_bbox:
+    for bbox in text_blocks:
+        if rects_intersect(bbox, interior_bbox):
             has_text = True
+            break
     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: PathLike):
     pageinfo = {}
     pageinfo['pageno'] = pageno
     pageinfo['images'] = []
 
     page = pdf.pages[pageno]
-
-    if fitz:
-        pageinfo['textinfo'] = _page_get_textblocks(str(infile), pageno)
-        pageinfo['has_text'] = _page_has_text(pageinfo['textinfo'])
-    else:
-        pageinfo['has_text'] = _naive_find_text(pdf=pdf, page=page)
-
-    mediabox = [Decimal(d.decode()) for d in page.MediaBox.as_list()]
+    mediabox = [Decimal(d) for d in page.MediaBox.as_list()]
     width_pt = mediabox[2] - mediabox[0]
     height_pt = mediabox[3] - mediabox[1]
 
+    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):
-        userunit = userunit.decode()
+        userunit = Decimal(userunit)
     pageinfo['userunit'] = userunit
     pageinfo['width_inches'] = width_pt * userunit / Decimal(72.0)
     pageinfo['height_inches'] = height_pt * userunit / Decimal(72.0)
@@ -600,24 +587,87 @@ 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))]
+worker_pdf = None
+
+
+def _pdf_pageinfo_sync(args):
+    pageno, infile = args
+    page = PageInfo(worker_pdf, pageno, infile)
+    return page
+
+
+def _pdf_pageinfo_concurrent(pdf, infile, progbar):
+    pages = [None] * len(pdf.pages)
+
+    def update_pageinfo(result, pbar):
+        page = result
+        pages[page.pageno] = page
+        pbar.update()
+
+    contexts = ((n, infile) for n in range(len(pdf.pages)))
+    global worker_pdf
+    worker_pdf = pdf
+
+    if os.name == 'nt' or True:
+        # We can't parallelize on Windows, because Windows cannot fork.
+        # We are trying to fork, then take advantage of the preloaded pikepdf.Pdf
+        # object in memory to save time reloading it, hence the silly global
+        # variable. Hey, it works. Threads are not helpful here because they
+        # will all just fight over the lock. So on Windows just run sequentially.
+        use_threads = True
+        max_workers = 1
+    else:
+        use_threads = False
+        max_workers = min(len(pages), 16)
+
+    exec_progress_pool(
+        use_threads=use_threads,
+        max_workers=max_workers,
+        tqdm_kwargs=dict(
+            total=len(pdf.pages), desc="Scan", unit='page', disable=not progbar
+        ),
+        task_initializer=None,
+        task_initargs=None,
+        task=_pdf_pageinfo_sync,
+        task_arguments=contexts,
+        task_finished=update_pageinfo,
+    )
+    return pages
+
+
+def _pdf_get_all_pageinfo(infile, log=None, progbar=False):
+    pdf = pikepdf.open(infile)  # Do not close in this function
+    try:
+        if pdf.is_encrypted:
+            raise EncryptedPdfError()  # Triggered by encryption with empty passwd
+        pages = _pdf_pageinfo_concurrent(pdf, infile, progbar)
+    except Exception:
+        pdf.close()
+        raise
+
+    return pages, pdf
 
 
 class PageInfo:
@@ -634,6 +684,14 @@ class PageInfo:
     def has_text(self):
         return self._pageinfo['has_text']
 
+    @property
+    def has_corrupt_text(self):
+        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']
@@ -665,10 +723,27 @@ class PageInfo:
     def images(self):
         return self._pageinfo['images']
 
-    def get_textareas(self):
-        if not fitz:
-            raise NotImplementedError("no impl without fitz")
-        yield from self._pageinfo['textinfo']['blocks']
+    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('Incomplete information on textboxes')
+            return self._pageinfo['bboxes']
+
+        return (
+            obj.bbox
+            for obj in self._pageinfo['textboxes']
+            if predicate(obj, visible, corrupt)
+        )
 
     @property
     def xres(self):
@@ -691,25 +766,32 @@ 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, log=logger, progbar=False):
         self._infile = infile
-        self._pages = _pdf_get_all_pageinfo(infile)
-        if fitz:
-            self._toc = fitz.Document(fspath(infile)).getToC()
-        else:
-            self._toc = []
+        self._pages, pdf = _pdf_get_all_pageinfo(infile, 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):
@@ -724,6 +806,10 @@ 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)):
@@ -731,8 +817,8 @@ class PdfInfo:
         return self._infile
 
     @property
-    def table_of_contents(self):
-        return self._toc
+    def needs_rendering(self):
+        return self._needs_rendering
 
     def __getitem__(self, item):
         return self._pages[item]
@@ -741,25 +827,23 @@ 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)
+    pagesinfo, pdfinfo = _pdf_get_all_pageinfo(args.infile)
     from pprint import pprint
-    pprint(info)
+
+    pprint(pdfinfo)
+    for page in pagesinfo:
+        pprint(page)
+        for im in page.images:
+            pprint(im)
 
 
 if __name__ == '__main__':

src/ocrmypdf/pdfinfo/layout.py

@@ -0,0 +1,306 @@
+# © 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
+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]+')
+
+#
+#  pdfminer 20181108 patches
+#
+
+if pdfminer.__version__ == '20181108':
+
+    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__
+
+#
+# end of pdfminer 20181108 patches
+#
+
+
+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
+        self.cur_item = None  # not defined in pdfminer code as it should be
+
+    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

src/ocrmypdf/quality.py

@@ -0,0 +1,60 @@
+# © 2020 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 typing import Iterable
+
+"""Utilities to measure OCR quality"""
+
+
+class OcrQualityDictionary:
+    """Manages a dictionary for simple OCR quality checks."""
+
+    def __init__(self, *, wordlist: Iterable[str] = []):
+        """Construct a dictionary from a list of words.
+
+        Words for which capitalization is important should be capitalized in the
+        dictionary. Words that contain spaces or other punctuation will never match.
+        """
+        self.dictionary = set()
+        self.dictionary.update(w for w in wordlist)
+
+    def measure_words_matched(self, ocr_text: str) -> float:
+        """Check how many unique words in the OCR text match a dictionary.
+
+        Words with mixed capitalized are only considered a match if the test word
+        matches that capitalization.
+
+        Returns:
+            number of words that match / number
+        """
+        text = re.sub(r"[0-9_]+", ' ', ocr_text)
+        text = re.sub(r'\W+', ' ', text)
+        text_words_list = re.split(r'\s+', text)
+        text_words = {w for w in text_words_list if len(w) >= 3}
+
+        matches = 0
+        for w in text_words:
+            if w in self.dictionary or (
+                w != w.lower() and w.lower() in self.dictionary
+            ):
+                matches += 1
+        if matches > 0:
+            hit_ratio = matches / len(text_words)
+        else:
+            hit_ratio = 0.0
+        return hit_ratio

test_requirements.txt

@@ -1,7 +0,0 @@
-pytest >= 3.2
-pytest-helpers-namespace
-pytest-xdist
-pytest-cov
-pytest-timeout
-python-xmp-toolkit   # requires apt-get install libexempi3
-                     # or brew install exempi

tests/cache/2400dpi/__-l__eng__-c__textonly_pdf=1__000001.ocr.png__000001.text__pdf__txt/stderr.bin

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

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