v8.2.3 notes

The LeptonicaErrorTrap was problematic for Celery and other
libraries that mess with stderr.

Closes #359

.docker/Dockerfile

@@ -3,38 +3,60 @@
 FROM      ubuntu:18.04
 
 RUN apt-get update && apt-get install -y --no-install-recommends \
+  build-essential autoconf automake libtool \
+  libleptonica-dev \
+  zlib1g-dev \
+  libexempi3 \
   ocrmypdf \
-  unpaper \
+  pngquant \
+  python3-pip \
+  python3-venv \
+  tesseract-ocr \
+  tesseract-ocr-chi-sim \
+  tesseract-ocr-deu \
   tesseract-ocr-eng \
   tesseract-ocr-fra \
-  tesseract-ocr-deu \
-  tesseract-ocr-spa \
   tesseract-ocr-por \
-  tesseract-ocr-chi-sim \
-  python3-venv \
-  python3-pip
+  tesseract-ocr-spa \
+  unpaper \
+  wget
+
 
 ENV LANG=C.UTF-8
 
+# Compile and install jbig2
+# Needs libleptonica-dev, zlib1g-dev
+RUN \
+  mkdir jbig2 \
+  && wget -q https://github.com/agl/jbig2enc/archive/0.29.tar.gz -O - | \
+      tar xz -C jbig2 --strip-components=1 \
+  && cd jbig2 \
+  && ./autogen.sh && ./configure && make && make install \
+  && cd .. \
+  && rm -rf jbig2
+
+RUN apt-get remove -y autoconf automake libtool
+
 RUN python3 -m venv --system-site-packages /appenv
 
 # 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 
+# build-essentials (gcc) to do a source installation
 # (i.e. "pip install ."). It's unclear to me why this is the case.
 RUN . /appenv/bin/activate; \
   pip install --upgrade pip \
-  && pip install ocrmypdf[fitz]
+  && pip install --upgrade ocrmypdf
 
 # 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
+  pip install -r /application/requirements/test.txt
 
 # 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 remove -y build-essential \
   && apt-get autoremove -y \
   && apt-get autoclean -y
 

.docker/alpine.dockerfile

@@ -0,0 +1,84 @@
+FROM alpine:3.9 as base
+
+FROM base as builder
+
+ENV LANG=C.UTF-8
+
+RUN \
+  echo '@testing http://nl.alpinelinux.org/alpine/edge/testing' >> /etc/apk/repositories \
+  # Add runtime dependencies
+  && apk add --update \
+    python3-dev \
+    py3-setuptools \
+    jbig2enc@testing \
+    ghostscript \
+    qpdf \
+    tesseract-ocr \
+    unpaper \
+    pngquant \
+    libxml2-dev \
+    libxslt-dev \
+    zlib-dev \
+    qpdf-dev \
+    libffi-dev \
+    leptonica-dev \
+    binutils \
+  # Install pybind11 for pikepdf
+  && pip3 install pybind11 \
+  # Install flask for the webservice
+  && pip3 install flask \
+  # Add build dependencies
+  && apk add --virtual build-dependencies \
+    build-base \
+    git
+
+COPY . /app
+
+WORKDIR /app
+
+RUN pip3 install .
+
+FROM base
+
+ENV LANG=C.UTF-8
+
+RUN \
+  echo '@testing http://nl.alpinelinux.org/alpine/edge/testing' >> /etc/apk/repositories \
+  # Add runtime dependencies
+  && apk add --update \
+    python3 \
+    jbig2enc@testing \
+    ghostscript \
+    qpdf \
+    tesseract-ocr \
+    tesseract-ocr-data-deu \
+    tesseract-ocr-data-chi_sim \
+    unpaper \
+    pngquant \
+    libxml2 \
+    libxslt \
+    zlib \
+    qpdf \
+    libffi \
+    leptonica-dev \
+    binutils \
+  && mkdir /app
+
+WORKDIR /app
+
+# Copy build artifacts (python site-packages9
+COPY --from=builder /usr/lib/python3.6/site-packages /usr/lib/python3.6/site-packages
+COPY --from=builder /usr/bin/ocrmypdf /usr/bin/dumppdf.py /usr/bin/latin2ascii.py /usr/bin/pdf2txt.py /usr/bin/img2pdf /usr/bin/chardetect /usr/bin/
+
+# Copy
+COPY --from=builder /app/.docker/webservice.py /app/
+
+# Copy minimal project files to get the test suite.
+COPY --from=builder /app/setup.cfg /app/setup.py /app/README.md /app/
+COPY --from=builder /app/requirements /app/requirements
+COPY --from=builder /app/tests /app/tests
+COPY --from=builder /app/src /app/src
+# Copy PKG-INFO from build artifact in app dir to make setuptools-scm happy
+RUN cp /usr/lib/python3.6/site-packages/ocrmypdf-*.egg-info/PKG-INFO /app
+
+ENTRYPOINT ["/usr/bin/ocrmypdf"]

.docker/webservice.dockerfile

@@ -0,0 +1,24 @@
+# OCRmyPDF webservice
+#
+FROM      jbarlow83/ocrmypdf-polyglot:latest
+
+USER root
+
+# Update system and install our dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+  python3-flask
+
+RUN apt-get autoremove -y && apt-get clean -y
+
+EXPOSE 5000
+
+COPY .docker/webservice.py /application
+
+USER docker
+
+VOLUME ["/config"]
+
+# This config file is optional
+ENV OCRMYPDF_WEBSERVICE_SETTINGS "/config/config.py"
+
+ENTRYPOINT ["python3", "/application/webservice.py"]

.docker/webservice.py

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

.dockerignore

@@ -5,6 +5,7 @@
 *.sublime*
 **/*.pyc
 .*/
+!.git/
 !.docker/
 .ruffus_history.sqlite
 bin/

.github/ISSUE_TEMPLATE

@@ -1,5 +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 for the OCRmyPDF maintainer only, see the [Wiki](https://github.com/jbarlow83/OCRmyPDF/wiki).

.github/issue_template.md

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

.gitignore

@@ -1,47 +1,43 @@
 # Development environment
+.bash_history
+.pylintrc
+.pytest_cache/
+.ruffus_history.sqlite
+.venv/
 *.pyc
 *.sublime-*
-venv*/
-.venv/
-pyvenv.cfg
-tasks.py
-.bash_history
-.ruffus_history.sqlite
-.idea/
-.pytest_cache/
-.pylintrc
 
 # Package building
-*.egg-info/
-.cache/
 .eggs/
+*.egg-info/
 build/
 dist/
 wheelhouse/
 
 # 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

.pre-commit-config.yaml

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

.readthedocs.yml

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

.travis.yml

@@ -1,64 +1,127 @@
-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
-    - poppler-utils
-    - qpdf
-    - tesseract-ocr
-    - tesseract-ocr-deu
-    - tesseract-ocr-eng
-    - tesseract-ocr-fra
-
 matrix:
   include:
     - os: linux
-      sudo: required
-      language: python
-      python: "3.5"
-    - os: linux
+      dist: trusty
       sudo: required
       language: python
       python: "3.6"
+      env:
+        - DIST=trusty
+      addons: &trusty_apt
+        apt:
+          update: true
+          sources:
+          - sourceline: 'ppa:alex-p/tesseract-ocr'
+          - sourceline: 'ppa:heyarje/libav-11'
+          - sourceline: 'ppa:vshn/ghostscript'
+          packages:
+          - ghostscript
+          - libavcodec56
+          - libavformat56
+          - libavutil54
+          - libexempi3
+          - libffi-dev
+          - pngquant
+          - poppler-utils
+          - qpdf
+          - tesseract-ocr
+          - tesseract-ocr-deu
+          - tesseract-ocr-eng
+          - tesseract-ocr-fra
     - os: linux
+      dist: xenial
       sudo: required
       language: python
-      python: "3.7-dev"
+      python: "3.7"
+      env:
+        - DIST=xenial
+      addons:
+        apt:
+          update: true
+          sources:
+          - sourceline: 'ppa:alex-p/tesseract-ocr'
+          packages:
+          - ghostscript
+          - libexempi3
+          - libffi-dev
+          - pngquant
+          - poppler-utils
+          - qpdf
+          - tesseract-ocr
+          - tesseract-ocr-deu
+          - tesseract-ocr-eng
+          - tesseract-ocr-fra
+          - unpaper
+    - os: osx
+      osx_image: xcode9.2
+      language: generic
+      addons:
+        homebrew:
+          update: true
+          packages:
+            - exempi
+            - ghostscript
+            - jbig2enc
+            - leptonica
+            - openjpeg
+            - pngquant
+            - python
+            - qpdf
+            - tesseract
+            - unpaper
+    - os: osx
+      osx_image: xcode9.2
+      language: generic
+      env:
+        - ADD_PDFMINER=1
+      addons:
+        homebrew:
+          update: true
+          packages:
+            - exempi
+            - ghostscript
+            - jbig2enc
+            - leptonica
+            - openjpeg
+            - pngquant
+            - python
+            - qpdf
+            - tesseract
+            - unpaper
 
 before_cache:
 - rm -f $HOME/.cache/pip/log/debug.log
 
 before_install: |
+  mkdir -p bin
   if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then
-    pip install --upgrade pip
-    mkdir -p packages
-    wget -q 'https://www.dropbox.com/s/vaq0kbwi6e6au80/unpaper_6.1-1.deb?raw=1' -O packages/unpaper_6.1-1.deb
-    sudo dpkg -i packages/unpaper_6.1-1.deb
+    pip3 install --upgrade pip
+    pip3 install --upgrade wheel
+    if [[ "$DIST" == "trusty" ]]; then
+      mkdir -p packages
+      wget -q 'https://www.dropbox.com/s/vaq0kbwi6e6au80/unpaper_6.1-1.deb?raw=1' -O packages/unpaper_6.1-1.deb
+      sudo dpkg -i packages/unpaper_6.1-1.deb
+    fi
   elif [[ "$TRAVIS_OS_NAME" == "osx" ]]; then
-    brew update && brew bundle --file=.travis/Brewfile
     pip3 install --upgrade pip
     pip3 install wheel
   fi
 
 install:
-- pip3 install pycparser # py3.7 workaround for https://github.com/eliben/pycparser/issues/251
-- pip3 install .
-- pip3 install -r test_requirements.txt
+- export PATH=$PWD/bin:$PATH
+- pip3 install pycparser  # py3.7 workaround for https://github.com/eliben/pycparser/issues/251
+- pip3 install -r requirements/main.txt
+- pip3 install --no-deps .
+- |
+  if [[ "$ADD_PDFMINER" == "1" ]]; then
+    pip3 install --no-deps .[pdfminer]
+  fi
+- pip3 install -r requirements/test.txt
 
 script:
 - tesseract --version
@@ -67,7 +130,7 @@ script:
 
 deploy:
   # release for main pypi
-  # 3.6 w/o fitz is considered the build leader and does the deploy, otherwise there is
+  # 3.6 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

.travis/Brewfile

@@ -1,13 +0,0 @@
-brew 'python'
-
-brew 'ghostscript'
-brew 'jbig2dec'
-brew 'leptonica'
-brew 'libffi'
-brew 'libtiff'
-brew 'libxml2'
-brew 'mupdf-tools'
-brew 'openjpeg'
-brew 'qpdf'
-brew 'tesseract'
-brew 'unpaper'

MANIFEST.in

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

README.md

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

README.rst

@@ -1,153 +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
--  Provides debug mode to enable easy verification of the OCR results
--  Processes pages in parallel when more than one CPU core is
-   available
--  Uses `Tesseract OCR <https://github.com/tesseract-ocr/tesseract>`_ engine
--  Supports more than `100 languages <https://github.com/tesseract-ocr/tessdata>`_ recognized by Tesseract
--  Battle-tested on thousands of PDFs, a test suite and continuous integration
-
-For details: please consult the `documentation <https://ocrmypdf.readthedocs.io/en/latest/>`_.
-
-Motivation
-----------
-
-I searched the web for a free command line tool to OCR PDF files on
-Linux/UNIX: I found many, but none of them were really satisfying.
-
--  Either they produced PDF files with misplaced text under the image (making copy/paste impossible)
--  Or they did not handle accents and multilingual characters
--  Or they changed the resolution of the embedded images
--  Or they generated ridiculously large PDF files
--  Or they crashed when trying to OCR some of my PDF files
--  Or they did not produce valid PDF files (even though they were readable with my current PDF reader)
--  On top of that none of them produced PDF/A files (format dedicated for long time storage)
-
-...so I decided to develop my own tool (using various existing scripts
-as an inspiration).
-
-Installation
-------------
-
-Linux, UNIX, and macOS are supported. Windows is not directly supported but there is a Docker image available that runs on Windows.
-
-Users of Debian 9 or later or Ubuntu 16.10 or later may simply
-
-.. code-block:: bash
-
-   apt-get install ocrmypdf
-
-and macOS users with Homebrew may simply
-
-.. code-block:: bash
-
-   brew install ocrmypdf
-
-For everyone else, `see our documentation <https://ocrmypdf.readthedocs.io/en/latest/installation.html>`_ for installation steps.
-
-Languages
----------
-
-OCRmyPDF uses Tesseract for OCR, and relies on its language packs. For Linux users,
-you can often find packages that provide language packs:
-
-.. code-block:: bash
-
-   # Display a list of all Tesseract language packs
-   apt-cache search tesseract-ocr
-
-   # Debian/Ubuntu users
-   apt-get install tesseract-ocr-chi-sim  # Example: Install Chinese Simplified language back
-
-You can then pass the ``-l LANG`` argument to OCRmyPDF to give a hint as to what languages it should search for. Multiple
-languages can be requested.
-
-Documentation and support
--------------------------
-
-Once ocrmypdf is installed, the built-in help which explains the command syntax and options can be accessed via:
-
-.. code-block:: bash
-
-   ocrmypdf --help
-
-Our `documentation is served on Read the Docs <https://ocrmypdf.readthedocs.io/en/latest/index.html>`_.
-
-If you detect an issue, please:
-
--  Check whether your issue is already known
--  If no problem report exists on github, please create one here:
-   https://github.com/jbarlow83/OCRmyPDF/issues
--  Describe your problem thoroughly
--  Append the console output of the script when running the debug mode
-   (``-v 1`` option)
--  If possible provide your input PDF file as well as the content of the
-   temporary folder (using a file sharing service like Dropbox)
-
-Requirements
-------------
-
-Runs on CPython 3.5, 3.6 and 3.7. Requires external program installations of Ghostscript, Tesseract OCR, QPDF, and Leptonica. ocrmypdf is pure Python, but uses CFFI to portably generate library bindings.
-
-Press & Media
--------------
-
--  `c't 1-2014, page 59 <http://heise.de/-2279695>`_:
-   Detailed presentation of OCRmyPDF v1.0 in the leading German IT
-   magazine c't
--  `heise Open Source, 09/2014: Texterkennung mit
-   OCRmyPDF <http://heise.de/-2356670>`_
-
-License
--------
-
-The OCRmyPDF software is licensed under the GNU GPLv3. Certain files are covered by other licenses, as noted in their source files.
-
-The license for each test file varies, and is noted in tests/resources/README.rst. The documentation is licensed under Creative Commons Attribution-ShareAlike 4.0 (CC-BY-SA 4.0).
-
-OCRmyPDF versions prior to 6.0 were licensed under the MIT License.
-
-Disclaimer
-----------
-
-The software is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
-CONDITIONS OF ANY KIND, either express or implied.

docs/advanced.rst

@@ -1,6 +1,34 @@
 Advanced features
 =================
 
+Control of unpaper
+------------------
+
+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
 ----------------------
 
@@ -9,17 +37,18 @@ OCRmyPDF provides many features to control the behavior of the OCR engine, Tesse
 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 ``--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.
 
@@ -27,35 +56,34 @@ If you want to adjust the amount of time spent on OCR, change ``--tesseract-time
 
 .. code-block:: bash
 
-	# Allow 300 seconds for OCR; skip any page larger than 50 megapixels
-	ocrmypdf --tesseract-timeout 300 --skip-big 50 bigfile.pdf output.pdf
+    # Allow 300 seconds for OCR; skip any page larger than 50 megapixels
+    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:
 
 .. 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:$PATH \
-		TESSDATA_PREFIX=/home/user/src/tesseract4 \
-		ocrmypdf --tesseract-oem 2 input.pdf output.pdf
-
-In this example ``TESSDATA_PREFIX`` directs Tesseract 4.0 to use LSTM training data. ``--tesseract-oem 1`` requests tesseract 4.0's new LSTM engine. (Tesseract 4.0 only.)
+    env \
+        PATH=/home/user/src/tesseract/api:$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
 """""""""""""""""""""""""""""""""
@@ -80,19 +108,19 @@ Create a file named "no-dict.cfg" with these contents:
 
 ::
 
-	load_system_dawg 0
-	language_model_penalty_non_dict_word 0
-	language_model_penalty_non_freq_dict_word 0
+    load_system_dawg 0
+    language_model_penalty_non_dict_word 0
+    language_model_penalty_non_freq_dict_word 0
 
 then run ocrmypdf as follows (along with any other desired arguments):
 
 .. code-block:: bash
 
-	ocrmypdf --tesseract-config no-dict.cfg input.pdf output.pdf
+    ocrmypdf --tesseract-config no-dict.cfg input.pdf output.pdf
 
 .. 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
@@ -105,7 +133,7 @@ rendering
   Creating a new PDF from other data (such as an existing PDF).
 
 
-OCRmyPDF has three PDF renderers: ``sandwich``, ``hocr``, ``tesseract``. 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
 """""""""""""""""""""""""
@@ -116,8 +144,6 @@ Currently this is the best renderer for most uses, however it is implemented in
 
 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.
 
-This renderer requires Tesseract 3.05.01 or newer.
-
 The ``hocr`` renderer
 """""""""""""""""""""
 
@@ -132,8 +158,82 @@ This works in all versions of Tesseract.
 The ``tesseract`` renderer
 """"""""""""""""""""""""""
 
-The ``tesseract`` renderer creates a PDF with the image and text layers precomposed, meaning that it always transcodes, loses image quality and rasterizes any vector objects. It does a better job on non-Latin text and document structure than ``hocr``.
+The ``tesseract`` renderer was removed. OCRmyPDF's new approach to text layer grafting makes it functionally equivalent to ``sandwich``.
 
-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.
+Return code policy
+------------------
 
-*This renderer is deprecated and will be removed whenever support for older versions of Tesseract is dropped.*
+OCRmyPDF writes all messages to ``stderr``.  ``stdout`` is reserved for piping
+output files.  ``stdin`` is reserved for piping input files.
+
+The return codes generated by the OCRmyPDF are considered part of the stable
+user interface.  They may be imported from ``ocrmypdf.exceptions``.
+
+.. list-table:: Return codes
+    :widths: 5 35 60
+    :header-rows: 1
+
+    *	- Code
+        - Name
+        - Interpretation
+    *	- 0
+        - ``ExitCode.ok``
+        - Everything worked as expected.
+    *	- 1
+        - ``ExitCode.bad_args``
+        - Invalid arguments, exited with an error.
+    *	- 2
+        - ``ExitCode.input_file``
+        - The input file does not seem to be a valid PDF.
+    *	- 3
+        - ``ExitCode.missing_dependency``
+        - An external program required by OCRmyPDF is missing.
+    *	- 4
+        - ``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
+        - ``ExitCode.file_access_error``
+        - The user running OCRmyPDF does not have sufficient permissions to read the input file and write the output file.
+    *	- 6
+        - ``ExitCode.already_done_ocr``
+        - The file already appears to contain text so it may not need OCR. See output message.
+    *	- 7
+        - ``ExitCode.child_process_error``
+        - An error occurred in an external program (child process) and OCRmyPDF cannot continue.
+    *	- 8
+        - ``ExitCode.encrypted_pdf``
+        - The input PDF is encrypted. OCRmyPDF does not read encrypted PDFs. Use another program such as ``qpdf`` to remove encryption.
+    *	- 9
+        - ``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
+        - ``ExitCode.other_error``
+        - Some other error occurred.
+    *	- 130
+        - ``ExitCode.ctrl_c``
+        - The program was interrupted by pressing Ctrl+C.
+
+
+Debugging the intermediate files
+--------------------------------
+
+OCRmyPDF normally saves its intermediate results to a temporary folder and deletes this folder when it exits, whether it succeeded or failed.
+
+If the ``-k`` argument is issued on the command line, OCRmyPDF will keep the temporary folder and print the location, whether it succeeded or failed (provided the Python interpreter did not crash). An example message is:
+
+.. code-block:: none
+
+    Temporary working files saved at:
+    /tmp/com.github.ocrmypdf.u20wpz07
+
+The organization of this folder is an implementation detail and subject to change between releases. However the general organization is that working files on a per page basis have the page number as a prefix (starting with page 1), an infix indicates the processing stage, and a suffix indicates the file type. Some important files include:
+
+* ``.page.png`` - what the input page looks like
+* ``.image`` - the image we will show the user if we are in a mode that changes the final appearance; may be in one of several image formats
+* ``.text.pdf`` - the OCR file; this will load as a blank page but should have visible text if checked with a tool like pdftotext or pdfminder.six
+* ``.ocr.png`` - the file that is sent to Tesseract for OCR; depending on arguments this may differ from the presentation image
+* ``layers.rendered.pdf`` - the composite PDF, before metadata repair and optimization
+* ``images/*`` - images extracted during the optimization process; here the prefix indicates a PDF object ID not a page number

docs/batch.rst

@@ -18,16 +18,22 @@ The ``--tag`` argument tells parallel to print the filename as a prefix whenever
 
 	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 '{}' '{}' \;
+	
+Alternatively, with a docker container (mounts a volume to the container where the PDFs are stored):
+
+.. code-block:: bash
+
+	find . -printf '%p' -name '*.pdf' -exec docker run --rm -v <host dir>:<container dir> jbarlow83/ocrmypdf-alpine '<container dir>/{}' '<container dir>/{}' \;
 
 This only runs one ``ocrmypdf`` process at a time. This variation uses ``find`` to create a directory list and ``parallel`` to parallelize runs of ``ocrmypdf``, again updating files in place.
 
@@ -66,7 +72,7 @@ This user contributed script also provides an example of batch processing.
 	    log_file = script_dir + '/ocr-tree.log'
 
 	logging.basicConfig(
-			level=logging.INFO, format='%(asctime)s %(message)s', 
+			level=logging.INFO, format='%(asctime)s %(message)s',
 			filename=log_file, filemode='w')
 
 	for dir_name, subdirs, file_list in os.walk(start_dir):
@@ -80,9 +86,9 @@ This user contributed script also provides an example of batch processing.
 	            print(full_path)
 	            cmd = ["ocrmypdf",  "--deskew", filename, filename]
 	            logging.info(cmd)
-	            proc = subprocess.Popen(
+	            proc = subprocess.run(
 	            	cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
-	            result = proc.stdout.read()
+	            result = proc.stdout
 	            if proc.returncode == 6:
 	                print("Skipped document because it already contained text")
 	            elif proc.returncode == 0:
@@ -144,14 +150,14 @@ This is only possible for x86-based Synology products. Some Synology products us
 	            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 
+	# 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 
+	# 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)
+	            proc = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
 	            result = proc.stdout.read()
 	            logging.info(result)
 	            full_path_OCR = dir_name + '/' + filename_OCR
@@ -163,14 +169,11 @@ This is only possible for x86-based Synology products. Some Synology products us
 	            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
 ---------------------
 
@@ -211,4 +214,12 @@ Alternatives
 
 * `Watchman <https://facebook.github.io/watchman/>`_ is a more powerful alternative to ``watchmedo``.
 
+macOS Automator
+---------------
 
+You can use the Automator app with macOS, to create a Workflow or Quick Action. Use a *Run Shell Script* action in your workflow. In the context of Automator, the ``PATH`` may be set differently your Terminal's ``PATH``; you may need to explicitly set the PATH to include ``ocrmypdf``. The following example may serve as a starting point:
+
+.. image:: images/macos-workflow.png
+	:alt: Example macOS Automator script
+
+You may customize the command sent to ocrmypdf.

docs/conf.py

@@ -52,7 +52,9 @@ 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 +62,35 @@ 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 +159,7 @@ 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 +272,25 @@ 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 +330,7 @@ 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 +343,15 @@ man_pages = [
 # (source start file, target name, title, author,
 #  dir menu entry, description, category)
 texinfo_documents = [
-    (master_doc, 'ocrmypdf', 'ocrmypdf Documentation',
-     author, 'ocrmypdf', 'One line description of project.',
-     'Miscellaneous'),
+    (
+        master_doc,
+        'ocrmypdf',
+        'ocrmypdf Documentation',
+        author,
+        'ocrmypdf',
+        'One line description of project.',
+        'Miscellaneous',
+    )
 ]
 
 # Documents to append as an appendix to all manuals.

docs/cookbook.rst

@@ -5,7 +5,7 @@ Basic examples
 --------------
 
 Help!
-"""""
+^^^^^
 
 ocrmypdf has built-in help.
 
@@ -15,28 +15,28 @@ ocrmypdf has built-in 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.
 
@@ -45,7 +45,7 @@ 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.
 
@@ -57,11 +57,10 @@ You can increase (decrease) the parameter ``--rotate-pages-threshold`` to make p
 
 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
 
@@ -70,9 +69,10 @@ By default OCRmyPDF assumes the document is English.
 
 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".
 
@@ -81,9 +81,12 @@ This produces a file named "output.pdf" and a companion text file named "output.
     ocrmypdf --sidecar output.txt input.pdf output.pdf
 
 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
 
@@ -94,9 +97,12 @@ 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
 
@@ -104,6 +110,9 @@ Use a program like `img2pdf <https://gitlab.mister-muffin.de/josch/img2pdf>`_ to
 
 ``img2pdf`` is recommended because it does an excellent job at generating PDFs without transcoding images.
 
+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
@@ -112,10 +121,10 @@ For convenience, OCRmyPDF can also convert single images to PDFs on its own. If
 
 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
 ----------------
@@ -127,11 +136,13 @@ OCRmyPDF perform some image processing on each page of a PDF, if desired.  The s
 * ``--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.
 
+* ``--mask-barcodes`` will suppress any barcodes detected in a page image. Barcodes are known to confuse Tesseract OCR and interfere with the recognition of text on the same baseline as a barcode. The output file will contain the unaltered image of the barcode.
+
 .. note::
 
     In many cases image processing will rasterize PDF pages as images, potentially losing quality.
@@ -140,9 +151,8 @@ OCRmyPDF perform some image processing on each page of a PDF, if desired.  The s
 
     ``--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:
 
@@ -158,7 +168,7 @@ Image processing commands can be combined. The order in which options are given
 
 
 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.
 
@@ -167,28 +177,22 @@ If you set ``--tesseract-timeout 0`` OCRmyPDF will apply its image processing wi
     ocrmypdf --tesseract-timeout=0 --remove-background input.pdf output.pdf
 
 
-Redo OCR
-""""""""
+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 ``--force-ocr`` argument. Normally, OCRmyPDF does not modify files that already appear to contain OCR text.
+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 --force-ocr input.pdf output.pdf
+    ocrmypdf --redo-ocr input.pdf output.pdf
 
-Note that the method above will force rasterization of all pages, potentially reducing quality or losing vector content. 
+This method will replace OCR without rasterizing, reducing quality or removing vector content. If a file contains a mix of pure digital text and OCR, digital text will be ignored and OCR will be replaced. As such this mode is incompatible with image processing options, since they alter the appearance of the file.
 
-To ensure quality is preserved, one could extract all of the images and rebuild the PDF for a lossless transformation. This recipe does not work when PDFs contain multiple images per page, as many do in practice. It will also lose any page rotation information.
-
-.. code-block:: bash
-
-    pdfimages -all old-ocr.pdf prefix  # extract all images
-    img2pdf -o temp.pdf prefix*        # construct new PDF from the images
-    # review the new PDF to ensure it visually matches the old one
-    ocrmypdf --output-type pdf temp.pdf new-ocr.pdf
-
-``--output-type pdf`` is used here to avoid using Ghostscript which will also rasterize images.
+In some cases, existing OCR cannot be detected or replaced. Files produced by OCRmyPDF v2.2 or earlier, for example, are internally represented as having visible text with an opaque image drawn on top. This situation cannot be detected.
 
+If ``--redo-ocr`` does not work, you can use ``--force-ocr``, which will force rasterization of all pages, potentially reducing quality or losing vector content.
 
 Improving OCR quality
 ---------------------
@@ -199,3 +203,36 @@ Rotating pages and deskewing helps to ensure that the page orientation is correc
 
 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 +1,155 @@
+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, and temporary storage space.
+
+.. _docker-install:
+
+Installing the Docker image
+---------------------------
+
+If you have `Docker <https://docs.docker.com/>`_ installed on your system, you can install a Docker image of the latest release.
+
+The recommended OCRmyPDF Docker image is currently named ``ocrmypdf-alpine``:
+
+.. code-block:: bash
+
+    docker pull jbarlow83/ocrmypdf-alpine
+
+Follow the Docker installation instructions for your platform.  If you can run this command successfully, your system is ready to download and execute the image:
+
+.. code-block:: bash
+
+    docker run hello-world
+
+OCRmyPDF will use all available CPU cores.  By default, the VirtualBox machine instance on Windows and macOS has only a single CPU core enabled. Use the VirtualBox Manager to determine the name of your Docker engine host, and then follow these optional steps to enable multiple CPUs:
+
+.. code-block:: bash
+
+    # Optional step for Mac OS X users
+    docker-machine stop "yourVM"
+    VBoxManage modifyvm "yourVM" --cpus 2  # or whatever number of core is desired
+    docker-machine start "yourVM"
+    eval $(docker-machine env "yourVM")
+
+Using the Docker image on the command line
+------------------------------------------
+
+**Unlike typical Docker containers**, in this mode we are using the OCRmyPDF Docker container is intended to be emphemeral – it runs for one OCR job and then terminates, just like a command line program. We are using Docker as a way of delivering an application, not a server.
+
+To start a Docker container (instance of the image):
+
+.. code-block:: bash
+
+    docker tag jbarlow83/ocrmypdf-alpine ocrmypdf
+    docker run --rm ocrmypdf (... all other arguments here...)
+
+For convenience, create a shell alias to hide the Docker command:
+
+.. code-block:: bash
+
+    alias ocrmypdf='docker run --rm -v "$(pwd):/home/docker" ocrmypdf'
+    ocrmypdf --version  # runs docker version
+
+Or in the wonderful `fish shell <https://fishshell.com/>`_:
+
+.. code-block:: fish
+
+    alias ocrmypdf 'docker run --rm ocrmypdf'
+    funcsave ocrmypdf
+
+.. _docker-lang-packs:
+
+Adding languages to the Docker image
+------------------------------------
+
+By default the Docker image includes English, German and Simplified Chinese, the most popular languages for OCRmyPDF users based on feedback. You may add other languages by creating a new Dockerfile based on the public one:
+
+.. code-block:: dockerfile
+
+    FROM jbarlow83/ocrmypdf-alpine
+
+    # Add French
+    RUN apk add tesseract-ocr-data-fra
+
+Executing the test suite
+------------------------
+
+The OCRmyPDF test suite is installed with image.  To run it:
+
+.. code-block:: bash
+
+    docker run --entrypoint python3 jbarlow83/ocrmypdf-alpine setup.py test
+
+Using the OCRmyPDF web service wrapper
+--------------------------------------
+
+The OCRmyPDF Docker image includes an example, barebones HTTP web service. The webservice may be launched as follows:
+
+.. code-block:: bash
+
+    docker run --entrypoint python3 -p 5000:5000 jbarlow83/ocrmypdf-alpine webservice.py
+
+Unlike command line usage this program will open a socket and wait for connections.
+
+.. warning::
+
+    The OCRmyPDF web service wrapper is intended for demonstration or development. It provides no security, no authentication, no protection against denial of service attacks, and no load balancing. The default Flask WSGI server is used, which is intended for development only. The server is single-threaded and so can respond to only one client at a time. It cannot respond to clients while busy with OCR.
+
+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>`.
+
+Legacy Ubuntu Docker images
+---------------------------
+
+Previously OCRmyPDF was delivered in several Docker images for different purposes, based on Ubuntu.
+
+The Ubuntu-based images will be maintained for some time but should not be used for new deployments. They are as follows:
+
+.. list-table::
+    :widths: auto
+    :header-rows: 1
+
+    *   - Image name
+        - Download command
+        - Notes
+    *   - ocrmypdf
+        - ``docker pull jbarlow83/ocrmypdf``
+        - Latest ocrmypdf with Tesseract 4.0.0-beta1 on Ubuntu 18.04. Includes English, French, German, Spanish, Portugeuse and Simplified Chinese.
+    *   - ocrmypdf-polyglot
+        - ``docker pull jbarlow83/ocrmypdf-polyglot``
+        - As above, with all available language packs.
+    *   - ocrmypdf-webservice
+        - ``docker pull jbarlow83/ocrmypdf-webservice``
+        - All language packs, and a simple HTTP wrapper allowing OCRmyPDF to be used as a web service. Note that this component is licensed under AGPLv3.
+
+To execute the Ubuntu-based OCRmyPDF on a local file, you must `provide a writable volume to the Docker image <https://docs.docker.com/userguide/dockervolumes/>`_, and both the input and output file must be inside the writable volume. This limitation applies only to the legacy images.
+
+This example command uses the current working directory as the writable volume:
+
+.. code-block:: bash
+
+    docker run --rm -v "$(pwd):/home/docker" <other docker arguments>   ocrmypdf <your arguments to ocrmypdf>
+
+In this worked example, the current working directory contains an input file called ``test.pdf`` and the output will go to ``output.pdf``:
+
+.. code-block:: bash
+
+    docker run --rm -v "$(pwd):/home/docker"   ocrmypdf --skip-text test.pdf output.pdf
+
+.. note:: The working directory should be a writable local volume or Docker may not have permission to access it.
+
+Note that ``ocrmypdf`` has its own separate ``-v VERBOSITYLEVEL`` argument to control debug verbosity. All Docker arguments should before the ``ocrmypdf`` image name and all arguments to ``ocrmypdf`` should be listed after.
+
+In some environments the permissions associated with Docker can be complex to configure. The process that executes Docker may end up not having the permissions to write the specified file system. In that case one can stream the file into and out of the Docker process and avoid all permission hassles, using ``-`` as the input and output filename:
+
+.. code-block:: bash
+
+    docker run --rm -i   ocrmypdf <other arguments to ocrmypdf> - - <input.pdf >output.pdf

docs/index.rst

@@ -18,12 +18,14 @@ PDF is the best format for storing and exchanging scanned documents.  Unfortunat
    release_notes
    installation
    languages
+   jbig2
 
 .. toctree::
    :caption: Usage
    :maxdepth: 2
 
    cookbook
+   docker
    advanced
    batch
    security
@@ -36,4 +38,3 @@ Indices and tables
 * :ref:`genindex`
 * :ref:`modindex`
 * :ref:`search`
-

docs/installation.rst

@@ -1,16 +1,56 @@
-Installation
-============
+Installing OCRmyPDF
+===================
 
-The easiest way to install OCRmyPDF to follow the steps for your operating system/platform.
+.. |latest| image:: https://img.shields.io/pypi/v/ocrmypdf.svg
+    :alt: OCRmyPDF latest released version on PyPI
 
-If you want to use the latest version of OCRmyPDF, your best bet is to install the most recent version your platform provides, and then upgrade that version by installing the Python binary wheels.
+|latest|
+
+The easiest way to install OCRmyPDF is to follow the steps for your operating
+system/platform, although sometimes this version may be out of date.
+
+If you want to use the latest version of OCRmyPDF, your best bet is to install
+the most recent version your platform provides, and then upgrade that version by
+installing the Python binary wheels.
 
 .. contents:: Platform-specific steps
-    :depth: 1
+    :depth: 2
     :local:
 
-Installing on Debian and Ubuntu 16.10 or newer
-----------------------------------------------
+Installing on Linux
+-------------------
+
+Debian and Ubuntu 16.10 or newer
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. |deb-stable| image:: https://repology.org/badge/version-for-repo/debian_stable/ocrmypdf.svg
+    :alt: Debian 9 stable ("stretch")
+
+.. |deb-testing| image:: https://repology.org/badge/version-for-repo/debian_testing/ocrmypdf.svg
+    :alt: Debian 10 testing ("buster")
+
+.. |deb-unstable| image:: https://repology.org/badge/version-for-repo/debian_unstable/ocrmypdf.svg
+    :alt: Debian unstable
+
+.. |ubu-1710| image:: https://repology.org/badge/version-for-repo/ubuntu_17_10/ocrmypdf.svg
+    :alt: Ubuntu 17.10
+
+.. |ubu-1804| image:: https://repology.org/badge/version-for-repo/ubuntu_18_04/ocrmypdf.svg
+    :alt: Ubuntu 18.04 LTS
+
+.. |ubu-1810| image:: https://repology.org/badge/version-for-repo/ubuntu_18_10/ocrmypdf.svg
+    :alt: Ubuntu 18.10
+
+
++-------------------------------------------+
+| **OCRmyPDF versions in Debian & Ubuntu**  |
++-------------------------------------------+
+| |latest|                                  |
++-------------------------------------------+
+| |deb-stable| |deb-testing| |deb-unstable| |
++-------------------------------------------+
+| |ubu-1710| |ubu-1804| |ubu-1810|          |
++-------------------------------------------+
 
 Users of Debian 9 ("stretch") or later or Ubuntu 16.10 or later may simply
 
@@ -18,10 +58,235 @@ Users of Debian 9 ("stretch") or later or Ubuntu 16.10 or later may simply
 
     apt-get install ocrmypdf
 
-To see what versions are available, check the `Debian Package Tracker <https://tracker.debian.org/pkg/ocrmypdf>`_ or `Ubuntu launchpad.net <https://launchpad.net/ocrmypdf>`_.
+As indicated in the table above, Debian and Ubuntu releases may lag behind the latest version. If the version available for your platform is out of date, you could opt to install the latest version from source. See `Installing HEAD revision from sources`_.
 
-Installing on macOS with Homebrew
----------------------------------
+For full details on version availability for your platform, check the `Debian Package Tracker <https://tracker.debian.org/pkg/ocrmypdf>`_ or `Ubuntu launchpad.net <https://launchpad.net/ocrmypdf>`_.
+
+.. note::
+
+    OCRmyPDF for Debian and Ubuntu currently omit the JBIG2 encoder. OCRmyPDF works fine without it but will produce larger output files. If you build jbig2enc from source, ocrmypdf 7.0.0 and later will automatically detect it (specifically the ``jbig2`` binary) on the ``PATH``. To add JBIG2 encoding, see :ref:`jbig2`.
+
+Fedora 29 or newer
+^^^^^^^^^^^^^^^^^^
+
+.. |fedora-29| image:: https://repology.org/badge/version-for-repo/fedora29/ocrmypdf.svg
+    :alt: Fedora 29
+
+.. |fedora-rawhide| image:: https://repology.org/badge/version-for-repo/fedora_rawhide/ocrmypdf.svg
+    :alt: Fedore Rawhide
+
+
++------------------------------+
+| **OCRmyPDF version**         |
++------------------------------+
+| |latest|                     |
++------------------------------+
+| |fedora-29| |fedora-rawhide| |
++------------------------------+
+
+Users of Fedora 29 later may simply
+
+.. code-block:: bash
+
+    dnf install ocrmypdf
+
+For full details on version availability, check the `Fedora Package Tracker
+<https://apps.fedoraproject.org/packages/ocrmypdf>`_.
+
+If the version available for your platform is out of date, you could opt to
+install the latest version from source. See `Installing HEAD revision from
+sources`_.
+
+.. note::
+
+    OCRmyPDF for Fedora currently omits the JBIG2 encoder due to patent issues.
+    OCRmyPDF works fine without it but will produce larger output files. If you
+    build jbig2enc from source, ocrmypdf 7.0.0 and later will automatically
+    detect it on the ``PATH``. To add JBIG2 encoding, see `Installing the JBIG2
+    encoder <jbig2>`_.
+
+Installing the latest version on Ubuntu 18.04 LTS
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Ubuntu 18.04 includes ocrmypdf 6.1.2. To install a more recent version, first
+install the system version to get most of the dependencies:
+
+.. code-block:: bash
+
+    sudo apt-get update
+    sudo apt-get install \
+        ocrmypdf \
+        python3-pip
+
+There are a few dependency changes between ocrmypdf 6.1.2 and 7.x. Let's get
+these, too.
+
+.. code-block:: bash
+
+    sudo apt-get install \
+        libexempi3 \
+        pngquant
+
+Then install the most recent ocrmypdf for the local user and set the user's ``PATH`` to check for the user's Python packages.
+
+.. code-block:: bash
+
+    export PATH=$HOME/.local/bin:$PATH
+    pip3 install --user ocrmypdf
+
+To add JBIG2 encoding, see :ref:`jbig2`.
+
+Ubuntu 16.04 LTS
+^^^^^^^^^^^^^^^^
+
+No package is available for Ubuntu 16.04. OCRmyPDF 8.0 and newer require Python
+3.6. Ubuntu 16.04 ships Python 3.5, but you can install Python 3.6 on it. Or,
+you can skip Python 3.6 and install OCRmyPDF 7.x or older - for that procedure,
+please see the installation documentation for the version of OCRmyPDF you plan
+to use.
+
+**Install system packages for OCRmyPDF**
+
+.. code-block:: bash
+
+    sudo apt-get update
+    sudo apt-get install -y software-properties-common python-software-properties
+    sudo add-apt-repository -y \
+        ppa:jonathonf/python-3.6 \
+        ppa:alex-p/tesseract-ocr
+    sudo apt-get update
+    sudo apt-get install -y \
+        ghostscript \
+        libexempi3 \
+        libffi6 \
+        pngquant \
+        python3.6 \
+        qpdf \
+        tesseract-ocr \
+        unpaper
+
+This will install a Python 3.6 binary at ``/usr/bin/python3.6`` alongside the
+system's Python 3.5. Do not remove the system Python. This will also install
+Tesseract 4.0 from a PPA, since the version available in Ubuntu 16.04 is too old
+for OCRmyPDF.
+
+Now install pip for Python 3.6. This will install the Python 3.6 version of
+``pip`` at ``/usr/local/bin/pip``.
+
+.. code-block:: bash
+
+    curl https://bootstrap.pypa.io/get-pip.py | sudo python3.6
+
+**Install OCRmyPDF**
+
+OCRmyPDF requires the locale to be set for UTF-8. **On some minimal Ubuntu
+installations systems**, it may be necessary to set the locale.
+
+.. code-block:: bash
+
+    # Optional: Only need to set these if they are not already set
+    export LC_ALL=C.UTF-8
+    export LANG=C.UTF-8
+
+Now install OCRmyPDF for the current user, and ensure that the ``PATH``
+environment variable contains ``$HOME/.local/bin``.
+
+.. code-block:: bash
+
+    export PATH=$HOME/.local/bin:$PATH
+    pip3 install --user ocrmypdf
+
+To add JBIG2 encoding, see :ref:`jbig2`.
+
+Ubuntu 14.04 LTS
+^^^^^^^^^^^^^^^^
+
+Installing on Ubuntu 14.04 LTS (trusty) is more difficult than some other
+options, because of its age. Several backports are required. For explanations of
+some steps of this procedure, see the similar steps for Ubuntu 16.04.
+
+Install system dependencies:
+
+.. code-block:: bash
+
+    sudo apt-get update
+    sudo apt-get install \
+        software-properties-common python-software-properties \
+        zlib1g-dev \
+        libexempi3 \
+        libjpeg-dev \
+        libffi-dev \
+        pngquant \
+        qpdf
+
+We will need backports of Ghostscript 9.16, libav-11 (for unpaper 6.1),
+Tesseract 4.00 (alpha), and Python 3.6. This will replace Ghostscript and
+Tesseract 3.x on your system. Python 3.6 will be installed alongside the system
+Python 3.4.
+
+If you prefer to not modify your system in this matter, consider using a Docker
+container.
+
+.. code-block:: bash
+
+    sudo add-apt-repository ppa:vshn/ghostscript -y
+    sudo add-apt-repository ppa:heyarje/libav-11 -y
+    sudo add-apt-repository ppa:alex-p/tesseract-ocr -y
+    sudo add-apt-repository ppa:jonathonf/python-3.6 -y
+
+    sudo apt-get update
+
+    sudo apt-get install \
+        python3.6-dev \
+        ghostscript \
+        tesseract-ocr \
+        tesseract-ocr-eng \
+        libavformat56 libavcodec56 libavutil54 \
+        wget
+
+Now we need to install ``pip`` and let it install ocrmypdf:
+
+.. code-block:: bash
+
+    curl https://bootstrap.pypa.io/ez_setup.py -o - | python3.6 && python3.6 -m easy_install pip
+    pip3.6 install ocrmypdf
+
+These installation instructions omit the optional dependency ``unpaper``, which is only available at version 0.4.2 in Ubuntu 14.04. The author could not find a backport of ``unpaper``, and created a .deb package to do the job of installing unpaper 6.1 (for x86 64-bit only):
+
+.. code-block:: bash
+
+    wget -q 'https://www.dropbox.com/s/vaq0kbwi6e6au80/unpaper_6.1-1.deb?raw=1' -O unpaper_6.1-1.deb
+    sudo dpkg -i unpaper_6.1-1.deb
+
+To add JBIG2 encoding, see :ref:`jbig2`.
+
+ArchLinux (AUR)
+^^^^^^^^^^^^^^^
+
+.. image:: https://repology.org/badge/version-for-repo/aur/ocrmypdf.svg
+    :alt: ArchLinux
+    :target: https://repology.org/metapackage/ocrmypdf
+
+There is an `ArchLinux User Repository package for ocrmypdf <https://aur.archlinux.org/packages/ocrmypdf/>`_. You can use the following command.
+
+.. code-block:: bash
+
+    yaourt -S ocrmypdf
+
+If you have any difficulties with installation, check the repository package page.
+
+Other Linux packages
+^^^^^^^^^^^^^^^^^^^^
+
+See the `Repology <https://repology.org/metapackage/ocrmypdf/versions>`_ page.
+
+In general, first install the OCRmyPDF package for your system, then optionally use the procedure `Installing with Python pip`_ to install a more recent version.
+
+Installing on macOS
+-------------------
+
+Homebrew
+^^^^^^^^
 
 .. image:: https://img.shields.io/homebrew/v/ocrmypdf.svg
     :alt: homebrew
@@ -33,6 +298,12 @@ OCRmyPDF is now a standard `Homebrew <https://brew.sh>`_ formula. To install on
 
     brew install ocrmypdf
 
+This will include only the English language pack. If you need other languages you can optionally install them all:
+
+.. code-block:: bash
+
+    brew install tesseract-lang  # Optional: Install all language packs
+
 .. note::
 
     Users who previously installed OCRmyPDF on macOS using ``pip install ocrmypdf`` should remove the pip version (``pip3 uninstall ocrmypdf``) before switching to the Homebrew version.
@@ -41,111 +312,8 @@ OCRmyPDF is now a standard `Homebrew <https://brew.sh>`_ formula. To install on
 
     Users who previously installed OCRmyPDF from the private tap should switch to the mainline version (``brew untap jbarlow83/ocrmypdf``) and install from there.
 
-.. _Docker-install:
-
-Installing the Docker image
----------------------------
-
-For many users, installing the Docker image will be easier than installing all of OCRmyPDF's dependencies. For Windows, it is the only option.
-
-If you have `Docker <https://docs.docker.com/>`_ installed on your system, you can install a Docker image of the latest release.
-
-Follow the Docker installation instructions for your platform.  If you can run this command successfully, your system is ready to download and execute the image:
-
-.. code-block:: bash
-
-    docker run hello-world
-
-OCRmyPDF will use all available CPU cores.  By default, the VirtualBox machine instance on Windows and macOS has only a single CPU core enabled. Use the VirtualBox Manager to determine the name of your Docker engine host, and then follow these optional steps to enable multiple CPUs:
-
-.. code-block:: bash
-
-    # Optional step for Mac OS X users
-    docker-machine stop "yourVM"
-    VBoxManage modifyvm "yourVM" --cpus 2  # or whatever number of core is desired
-    docker-machine start "yourVM"
-    eval $(docker-machine env "yourVM")
-
-Assuming you have a Docker engine running, you can download one of the three available images:
-
-.. list-table::
-    :widths: auto
-    :header-rows: 1
-
-    *   - Image name
-        - Download command
-        - Notes
-    *   - ocrmypdf
-        - ``docker pull jbarlow83/ocrmypdf``
-        - Latest ocrmypdf with Tesseract 4.0.0-beta1 on Ubuntu 18.04. Includes English, French, German, Spanish, Portugeuse and Simplified Chinese.
-    *   - ocrmypdf-polyglot
-        - ``docker pull jbarlow83/ocrmypdf-polyglot``
-        - As above, with all available language packs.
-
-For example:
-
-.. code-block:: bash
-
-    docker pull jbarlow83/ocrmypdf
-
-Then tag it to give a more convenient name, just ocrmypdf:
-
-.. code-block:: bash
-
-    docker tag jbarlow83/ocrmypdf ocrmypdf
-
-.. _docker-polyglot:
-
-The alternative "polyglot" image provides `all available language packs <https://github.com/tesseract-ocr/tesseract/blob/master/doc/tesseract.1.asc#languages>`_.
-
-You can then run ocrmypdf using the command:
-
-.. code-block:: bash
-
-    docker run --rm ocrmypdf --help
-
-To execute the OCRmyPDF on a local file, you must `provide a writable volume to the Docker image <https://docs.docker.com/userguide/dockervolumes/>`_, and both the input and output file must be inside the writable volume.  This example command uses the current working directory as the writable volume:
-
-.. code-block:: bash
-
-    docker run --rm -v "$(pwd):/home/docker" <other docker arguments>   ocrmypdf <your arguments to ocrmypdf>
-
-In this worked example, the current working directory contains an input file called ``test.pdf`` and the output will go to ``output.pdf``:
-
-.. code-block:: bash
-
-    docker run --rm -v "$(pwd):/home/docker"   ocrmypdf --skip-text test.pdf output.pdf
-
-.. note:: The working directory should be a writable local volume or Docker may not have permission to access it.
-
-Note that ``ocrmypdf`` has its own separate ``-v VERBOSITYLEVEL`` argument to control debug verbosity. All Docker arguments should before the ``ocrmypdf`` image name and all arguments to ``ocrmypdf`` should be listed after.
-
-In some environments the permissions associated with Docker can be complex to configure. The process that executes Docker may end up not having the permissions to write the specified file system. In that case one can stream the file into and out of the Docker process and avoid all permission hassles, using ``-`` as the input and output filename:
-
-.. code-block:: bash
-
-    docker run --rm -i   ocrmypdf <other arguments to ocrmypdf> - - <input.pdf >output.pdf
-
-For convenience, a shell alias can hide the docker command:
-
-.. code-block:: bash
-
-    alias ocrmypdf='docker run --rm -v "$(pwd):/home/docker" ocrmypdf'
-    ocrmypdf --version  # runs docker version
-
-Or in the wonderful `fish shell <https://fishshell.com/>`_:
-
-.. code-block:: fish
-
-    alias ocrmypdf 'docker run --rm -v (pwd):/home/docker ocrmypdf'
-    funcsave ocrmypdf
-
-.. note::
-
-    The ocrmypdf Docker containers are designed to be used for a single OCR job. The ``docker run --rm`` argument tells Docker to delete temporary storage associated with container when it is done executing.
-
 Manual installation on macOS
-----------------------------
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
 These instructions probably work on all macOS supported by Homebrew.
 
@@ -157,24 +325,14 @@ Update Homebrew:
 
     brew update
 
-Install or upgrade the required Homebrew packages, if any are missing:
+Install or upgrade the required Homebrew packages, if any are missing. To do this, download the ``Brewfile`` that lists all of the dependencies to the current directory, and run ``brew bundle`` to process them (installing or upgrading as needed). ``Brewfile`` is a plain text file.
 
 .. code-block:: bash
 
-    brew install libpng openjpeg jbig2dec libtiff     # image libraries
-    brew install qpdf
-    brew install ghostscript
-    brew install python3
-    brew install libxml2 libffi leptonica
-    brew install unpaper   # optional
+    wget https://github.com/jbarlow83/OCRmyPDF/raw/master/.travis/Brewfile
+    brew bundle
 
-Python 3.5, 3.6 and 3.7 are supported.
-
-Install the required Tesseract OCR engine with the language packs you plan to use:
-
-.. code-block:: bash
-
-    brew install tesseract                       # Option 1: for English, French, German, Spanish
+This will include the English, French, German and Spanish language packs. If you need other languages you can optionally install them all:
 
 .. _macos-all-languages:
 
@@ -206,160 +364,28 @@ The command line program should now be available:
 
     ocrmypdf --help
 
-Installing the latest version on Ubuntu 18.04 LTS
--------------------------------------------------
+Installing the Docker image
+---------------------------
 
-Ubuntu 18.04 includes ocrmypdf 6.1.2. To install a more recent version, first
-install the system version to get all the dependencies:
-
-.. code-block:: bash
-
-    sudo apt-get update
-    sudo apt-get install \
-        ocrmypdf \
-        python3-pip
-
-Then install ocrmypdf 6.1.5 for the local user and set the user's ``PATH`` to check for the user's Python packages.
-
-.. code-block:: bash
-
-    export PATH=$HOME/.local/bin:$PATH
-    pip3 install --user ocrmypdf
-
-
-Installing on Ubuntu 16.04 LTS
-------------------------------
-
-No package is currently available for Ubuntu 16.04, but you can install the dependencies manually:
-
-.. code-block:: bash
-
-    sudo apt-get update
-    sudo apt-get install \
-        unpaper \
-        ghostscript \
-        tesseract-ocr \
-        qpdf \
-        python3-pip \
-        python3-cffi
-
-If you wish install OCRmyPDF for the current user:
-
-.. code-block:: bash
-
-    pip3 install --user ocrmypdf
-
-Alternately, system-wide. Note that this may modify the system Python environment:
-
-.. code-block:: bash
-
-    sudo pip3 install ocrmypdf
-
-If you wish to install OCRmyPDF to a virtual environment to isolate the system Python, you can follow these steps.
-
-.. code-block:: bash
-
-    python3 -m venv venv-ocrmypdf
-    source venv-ocrmypdf/bin/activate
-    pip3 install ocrmypdf
-
-
-Installing on Ubuntu 14.04 LTS
-------------------------------
-
-Installing on Ubuntu 14.04 LTS (trusty) is more difficult than some other options, because it is older and does not provide ``pip``.
-
-Update apt-get:
-
-.. code-block:: bash
-
-    sudo apt-get update
-
-Install system dependencies:
-
-.. code-block:: bash
-
-    sudo apt-get install \
-        software-properties-common python-software-properties \
-        zlib1g-dev \
-        libjpeg-dev \
-        libffi-dev \
-        qpdf
-
-We will need backports of Ghostscript 9.16, libav-11 (for unpaper 6.1), Tesseract 4.00 (alpha), and Python 3.6. This will replace Ghostscript and Tesseract 3.x on your system. Python 3.6 will be installed alongside the system Python 3.
-
-If you prefer to not modify your system in this matter, consider using a Docker container.
-
-.. code-block:: bash
-
-    sudo add-apt-repository ppa:vshn/ghostscript -y
-    sudo add-apt-repository ppa:heyarje/libav-11 -y
-    sudo add-apt-repository ppa:alex-p/tesseract-ocr -y
-    sudo add-apt-repository ppa:jonathonf/python-3.6 -y
-
-    sudo apt-get update
-
-    sudo apt-get install \
-        python3.6-dev \
-        ghostscript \
-        tesseract-ocr \
-        tesseract-ocr-eng \
-        libavformat56 libavcodec56 libavutil54 \
-        wget
-
-Now we need to install ``pip`` and let it install ocrmypdf:
-
-.. code-block:: bash
-
-    curl https://bootstrap.pypa.io/ez_setup.py -o - | python3.6 && python3.6 -m easy_install pip
-    pip3.6 install ocrmypdf
-
-The ``wget`` command will download a program and run it.
-
-These installation instructions omit the optional dependency ``unpaper``, which is only available at version 0.4.2 in Ubuntu 14.04. The author could not find a backport of ``unpaper``, and created a .deb package to do the job of installing unpaper 6.1 (for x86 64-bit only):
-
-.. code-block:: bash
-
-    wget -q 'https://www.dropbox.com/s/vaq0kbwi6e6au80/unpaper_6.1-1.deb?raw=1' -O unpaper_6.1-1.deb
-    sudo dpkg -i unpaper_6.1-1.deb
-
-
-Installing on ArchLinux
------------------------
-
-The author is aware of an `ArchLinux package for ocrmypdf <https://aur.archlinux.org/packages/ocrmypdf/>`_. It seems like the following command might work.
-
-.. code-block:: bash
-
-    pacman -S ocrmypdf
+For some users, installing the Docker image will be easier than installing all of OCRmyPDF's dependencies. For Windows, it is the only option.
 
+See `OCRmyPDF Docker Image <docker>`_ for more information.
 
 Installing on Windows
 ---------------------
 
-Direct installation on Windows is not possible.  Install the _`Docker` container as described above.  Ensure that your command prompt can run the docker "hello world" container.
-
-It would probably not be too difficult to run on Windows.  The main reason this has been avoided is the difficulty of packaging and installing the various non-Python dependencies: Tesseract, QPDF, Ghostscript, Leptonica.  Pull requests to add or improve Windows support would be quite welcome.
-
-
-Running on Windows
-~~~~~~~~~~~~~~~~~~
-
-The command line syntax to run ocrmypdf from a command prompt will resemble:
-
-.. code-block:: bat
-
-    docker run -v /c/Users/sampleuser:/home/docker ocrmypdf --skip-text test.pdf output.pdf
-
-where /c/Users/sampleuser is a Unix representation of the Windows path C:\\Users\\sampleuser, assuming a user named "sampleuser" is running ocrmypdf on a file in their home directory, and the files "test.pdf" and "output.pdf" are in the sampleuser folder. The Windows user must have read and write permissions.
-
-`Bash on Ubuntu on Windows <https://github.com/Microsoft/BashOnWindows>`_ should also be a viable route for running the OCRmyPDF Docker container.
+Direct installation on Windows is not possible.  `Install the Docker <docker-install>`_ container as described above.  Ensure that your command prompt can run the docker "hello world" container.
 
+It would probably not be too difficult to port on Windows.  The main reason this has been avoided is the difficulty of packaging and installing the various non-Python dependencies: Tesseract, QPDF, Ghostscript, Leptonica.  Pull requests to add or improve Windows support would be quite welcome.
 
 Installing with Python pip
 --------------------------
 
-First, install `your platform's version <https://repology.org/metapackage/ocrmypdf/versions>`_ of ``ocrmypdf``, if available, as a way of ensuring that external dependencies are (mostly) satisified, even though the platform version may be out of date. Use ``ocrmypdf --version`` to confirm what version was installed.
+OCRmyPDF is delivered by PyPI because it is a convenient way to install the latest version. However, PyPI and ``pip`` cannot address the fact that ``ocrmypdf`` depends on certain non-Python system libraries and programs being instsalled.
+
+For best results, first install `your platform's version <https://repology.org/metapackage/ocrmypdf/versions>`_ of ``ocrmypdf``, using the instructions elsewhere in this document. Then you can use ``pip`` to get the latest version if your platform version is out of date. Chances are that this will satisfy most dependencies.
+
+Use ``ocrmypdf --version`` to confirm what version was installed.
 
 Then you can install the latest OCRmyPDF from the Python wheels. First try:
 
@@ -376,32 +402,42 @@ Since ``pip3 install --user`` does not work correctly on some platforms, notably
     pip3 install ocrmypdf
 
 Requirements for pip and HEAD install
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
-OCRmyPDF currently requires these external programs to be installed:
+OCRmyPDF currently requires these external programs and libraries to be installed, and must be satisfied using the operating system package manager. ``pip`` cannot provide them.
 
-- Python 3.5 or newer
-- Tesseract 3.04 or newer
+- Python 3.6 or newer
 - Ghostscript 9.15 or newer
-- qpdf 7.0.0 or newer
+- qpdf 8.1.0 or newer
+- Tesseract 4.0.0-alpha or newer
 
-The following dependencies are recommended:
+As of ocrmypdf 7.2.1, the following versions are recommended:
 
-- Python 3.6
-- Tesseract 4.00 or newer
-- Ghostscript 9.22 or newer
-- qpdf 8.0.2 or newer
+- Python 3.7
+- Ghostscript 9.23 or newer
+- qpdf 8.2.1
+- Tesseract 4.0.0 or newer
+- jbig2enc 0.29 or newer
+- pngquant 2.5 or newer
 - unpaper 6.1
 
-These are in addition to the Python packaging dependencies, meaning that unfortunately, the ``pip install`` command cannot satisfy all of them.
+jbig2enc, pngquant, and unpaper are optional. If missing certain features are disabled. OCRmyPDF will discover them as soon as they are available.
 
-Python 3.6 and Tesseract 4.0.0-beta.1 are recommended for best OCR results and best performance.
+**jbig2enc**, if present, will be used to optimize the encoding of monochrome images.  This can significantly reduce the file size of the output file.  It is not required.  `jbig2enc <https://github.com/agl/jbig2enc>`_ is not generally available for Ubuntu or Debian due to lingering concerns about patent issues, but can easily be built from source. To add JBIG2 encoding, see :ref:`jbig2`.
+
+**pngquant**, if present, is optionally used to optimize the encoding of PNG-style images in PDFs (actually, any that are that losslessly encoded) by lossily quantizing to a smaller color palette. It is only activated then the ``--optimize`` argument is ``2`` or ``3``.
+
+**unpaper**, if present, enables the ``--clean`` and ``--clean-final`` command line options.
+
+These are in addition to the Python packaging dependencies, meaning that unfortunately, the ``pip install`` command cannot satisfy all of them.
 
 
 Installing HEAD revision from sources
 -------------------------------------
 
-If you have ``git`` and Python 3.5 or newer installed, you can install from source. When the ``pip`` installer runs, it will alert you if dependencies are missing.
+If you have ``git`` and Python 3.6 or newer installed, you can install from source. When the ``pip`` installer runs, it will alert you if dependencies are missing.
+
+If you prefer to build every from source, you will need to `build pikepdf from source <https://pikepdf.readthedocs.io/en/latest/installation.html#building-from-source>`_. First ensure you can build and install pikepdf.
 
 To install the HEAD revision from sources in the current Python 3 environment:
 
@@ -425,8 +461,8 @@ You may find it easiest to install in a virtual environment, rather than system-
     cd OCRmyPDF
     pip3 install .
 
-However, ``ocrmypdf`` will only be accessible on the system PATH after
-you activate the virtual environment.
+However, ``ocrmypdf`` will only be accessible on the system PATH
+when you activate the virtual environment.
 
 To run the program:
 
@@ -439,8 +475,18 @@ need to be installed. The script requires specific versions of the
 dependencies. Older version than the ones mentioned in the release notes
 are likely not to be compatible to OCRmyPDF.
 
+For development
+^^^^^^^^^^^^^^^
 
-Other Linux packages
---------------------
+To install all of the development and test requirements:
 
-See the `Repology <https://repology.org/metapackage/ocrmypdf/versions>`_ page.
+.. code-block:: bash
+
+    git clone -b master https://github.com/jbarlow83/OCRmyPDF.git
+    python3 -m venv
+    source venv/bin/activate
+    cd OCRmyPDF
+    pip install -e .
+    pip install -r requirements/dev.txt -r requirements/test.txt
+
+To add JBIG2 encoding, see :ref:`jbig2`.

docs/introduction.rst

@@ -1,9 +1,7 @@
 Introduction
 ============
-
 OCRmyPDF is a Python 3 package that adds OCR layers to PDFs.
 
-
 About OCR
 ---------
 
@@ -18,7 +16,7 @@ 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.
 
-.. image:: bitmap_vs_svg.svg
+.. image:: images/bitmap_vs_svg.svg
 
 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.
 
@@ -66,6 +64,8 @@ In the case of a PDF that is nothing other than a container of images (no rotati
 
 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
 -----------
@@ -76,24 +76,26 @@ OCRmyPDF is limited by the Tesseract OCR engine.  As such it experiences these l
 * 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.
+* 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.
-* 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.
 
 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.
+* Because some popular open source PDF viewers have a particularly hard time with spaces betweem words, OCRmyPDF appends a space to each text element as a workaround (when using ``--pdf-renderer hocr``). While this mixes document structure with graphical information that ideally should be left to the PDF viewer to interpret, it improves compatibility with some viewers and does not cause problems for better ones.
 
 Ghostscript also imposes some limitations:
 
 * PDFs containing JBIG2-encoded content will be converted to CCITT Group4 encoding, which has lower compression ratios, if Ghostscript PDF/A is enabled.
 * PDFs containing JPEG 2000-encoded content will be converted to JPEG encoding, which may introduce compression artifacts, if Ghostscript PDF/A is enabled.
-* Ghostscript may transcode grayscale and color images, either lossy to lossless or lossless to lossy, based on an internal algorithm. This behavior can be suppressed by setting ``--pdfa-image-compression`` to ``jpeg`` or ``lossless`` to set all images to one type or the other. Ghostscript has no option to maintain the input image's format.
-  
-OCRmyPDF is currently not designed to be used as a Python API; it is designed to be run as a command line tool. ``import ocrmypf`` currently attempts to process the command line on ``sys.argv`` at import time so it has side effects that will interfere with its use as a package. The API it presents should not be considered stable.
+* 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
+* The Python API exported by ``import ocrmypdf`` is design to help scripts that use OCRmyPDF but is not currently capable of running OCRmyPDF jobs due to limitations in an underlying library.
 
 Similar programs
 ----------------
@@ -108,8 +110,10 @@ To the author's knowledge, OCRmyPDF is the most feature-rich and thoroughly test
 Web front-ends
 --------------
 
+The Docker image ``ocrmypdf-alpine`` provides a web service front-end that allows files to submitted over HTTP and the results "downloaded". This is an HTTP server intended to simplify web services deployments; it is not intended to be deployed on the public internet and no real security measures to speak of.
+
+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-web <https://github.com/sseemayer/OCRmyPDF-web>`_, a micro web-frontend for OCRmyPDF (third-party, not actively maintained)
-
-Bear in mind that OCRmyPDF is not designed to be secure against malware-bearing PDFs (see `Using OCRmyPDF online`_).
 
+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.

docs/jbig2.rst

@@ -0,0 +1,35 @@
+.. _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

@@ -3,7 +3,7 @@
 Installing additional language packs
 ====================================
 
-OCRmyPDF uses Tesseract for OCR, and relies on its language packs for languages other than English. 
+OCRmyPDF uses Tesseract for OCR, and relies on its language packs for languages other than English.
 
 Tesseract supports `most languages <https://github.com/tesseract-ocr/tesseract/blob/master/doc/tesseract.1.asc#languages>`_.
 
@@ -17,12 +17,27 @@ Debian and Ubuntu users
    # 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
-   
+   # 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
 -----------
 
@@ -31,46 +46,4 @@ You can install additional language packs by :ref:`installing Tesseract using Ho
 Docker users
 ------------
 
-Users of the Docker image may use the alternative :ref:`"polyglot" container <docker-polyglot>` which includes all languages.
-
-Adding individual language packs to a Docker image
-""""""""""""""""""""""""""""""""""""""""""""""""""
-
-If you wish to add a single language pack, you could do the following:
-
-* Download the desired ``.trainedata`` file from the `tessdata <https://github.com/tesseract-ocr/tessdata>`_ repository. Let's use Hebrew in this example (``heb.traineddata``)
-
-* Copy the file to ``/home/user/downloads/heb.traineddata``.
-
-* Create a new container based on the ocrmypdf-tess4 image and jump into it with a terminal:
-
-.. code-block:: bash
-
-	host$ docker run  -v /home/user/downloads:/home/docker -it --entrypoint /bin/bash ocrmypdf-tess4
-
-* Put the file where Tesseract expects it:
-
-.. code-block:: bash
-
-	docker$ cp /home/docker/heb.traineddata /usr/share/tesseract-ocr/tessdata
-
-* Note the container id, and save it as a new image (in this example, ``ocrmypdf-tess4-heb``)
-
-.. code-block:: bash
-
-    host$ docker commit <container_id> ocrmypdf-tess4-heb
-
-
-Known limitations
------------------
-
-As of v4.2, users of ocrmypdf working languages outside the Latin alphabet should use the following syntax:
-
-.. code-block:: bash
-
-	ocrmypdf -l eng+gre --output-type pdf --pdf-renderer tesseract
-
-The reasons for this are:
-
-* The latest version of Ghostscript (9.19 as of this writing) has unfixed bugs in Unicode handling that generate invalid character maps, so Ghostscript cannot be used for PDF/A conversion
-* The default "hocr" PDF renderer does not handle Asian fonts properly
+Users of the OCRmyPDF Docker image should install language packs into a derived Docker image as :ref:`described in that section <docker-lang-packs>`.

docs/pipeline.svg

@@ -4,550 +4,389 @@
 <!-- Generated by graphviz version 2.40.1 (20161225.0304)
  -->
 <!-- Title: Pipeline: Pages: 1 -->
-<svg width="1952pt" height="1277pt"
- viewBox="0.00 0.00 1952.00 1277.18" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
-<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 1273.1825)">
+<svg width="1484pt" height="1277pt"
+ viewBox="0.00 0.00 1484.00 1277.20" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 1273.1977)">
 <title>Pipeline:</title>
-<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-1273.1825 1948,-1273.1825 1948,4 -4,4"/>
+<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-1273.1977 1480,-1273.1977 1480,4 -4,4"/>
 <g id="clust1" class="cluster">
 <title>clustertasks</title>
-<polygon fill="none" stroke="#000000" points="8,-8 8,-1261.1825 1936,-1261.1825 1936,-8 8,-8"/>
-<text text-anchor="middle" x="972" y="-1233.1825" font-family="Times,serif" font-size="30.00" fill="#ff3232">Pipeline:</text>
+<polygon fill="none" stroke="#000000" points="8,-8 8,-1261.1977 1468,-1261.1977 1468,-8 8,-8"/>
+<text text-anchor="middle" x="738" y="-1233.1977" font-family="Times,serif" font-size="30.00" fill="#ff3232">Pipeline:</text>
 </g>
 <!-- t0 -->
 <g id="node1" class="node">
 <title>t0</title>
-<polygon fill="#efa03b" stroke="#000000" points="876.7053,-1215.1825 667.2947,-1215.1825 663.2947,-1211.1825 663.2947,-1179.1825 872.7053,-1179.1825 876.7053,-1183.1825 876.7053,-1215.1825"/>
-<polyline fill="none" stroke="#000000" points="872.7053,-1211.1825 663.2947,-1211.1825 "/>
-<polyline fill="none" stroke="#000000" points="872.7053,-1211.1825 872.7053,-1179.1825 "/>
-<polyline fill="none" stroke="#000000" points="872.7053,-1211.1825 876.7053,-1215.1825 "/>
-<text text-anchor="middle" x="770" y="-1191.1825" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.triage</text>
+<polygon fill="#efa03b" stroke="#000000" points="1215.7053,-1215.1977 1006.2947,-1215.1977 1002.2947,-1211.1977 1002.2947,-1179.1977 1211.7053,-1179.1977 1215.7053,-1183.1977 1215.7053,-1215.1977"/>
+<polyline fill="none" stroke="#000000" points="1211.7053,-1211.1977 1002.2947,-1211.1977 "/>
+<polyline fill="none" stroke="#000000" points="1211.7053,-1211.1977 1211.7053,-1179.1977 "/>
+<polyline fill="none" stroke="#000000" points="1211.7053,-1211.1977 1215.7053,-1215.1977 "/>
+<text text-anchor="middle" x="1109" y="-1191.1977" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.triage</text>
 </g>
 <!-- t1 -->
 <g id="node2" class="node">
 <title>t1</title>
-<polygon fill="#efa03b" stroke="#000000" points="941.0433,-1157.1825 602.9567,-1157.1825 598.9567,-1153.1825 598.9567,-1121.1825 937.0433,-1121.1825 941.0433,-1125.1825 941.0433,-1157.1825"/>
-<polyline fill="none" stroke="#000000" points="937.0433,-1153.1825 598.9567,-1153.1825 "/>
-<polyline fill="none" stroke="#000000" points="937.0433,-1153.1825 937.0433,-1121.1825 "/>
-<polyline fill="none" stroke="#000000" points="937.0433,-1153.1825 941.0433,-1157.1825 "/>
-<text text-anchor="middle" x="770" y="-1133.1825" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.repair_and_parse_pdf</text>
+<polygon fill="#efa03b" stroke="#000000" points="1280.0433,-1157.1977 941.9567,-1157.1977 937.9567,-1153.1977 937.9567,-1121.1977 1276.0433,-1121.1977 1280.0433,-1125.1977 1280.0433,-1157.1977"/>
+<polyline fill="none" stroke="#000000" points="1276.0433,-1153.1977 937.9567,-1153.1977 "/>
+<polyline fill="none" stroke="#000000" points="1276.0433,-1153.1977 1276.0433,-1121.1977 "/>
+<polyline fill="none" stroke="#000000" points="1276.0433,-1153.1977 1280.0433,-1157.1977 "/>
+<text text-anchor="middle" x="1109" y="-1133.1977" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.repair_and_parse_pdf</text>
 </g>
 <!-- t0&#45;&gt;t1 -->
 <g id="edge1" class="edge">
 <title>t0&#45;&gt;t1</title>
-<path fill="none" stroke="#0044a0" d="M770,-1179.1488C770,-1175.4743 770,-1171.5242 770,-1167.5888"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="773.5001,-1167.4047 770,-1157.4047 766.5001,-1167.4047 773.5001,-1167.4047"/>
+<path fill="none" stroke="#0044a0" d="M1109,-1179.1641C1109,-1175.4895 1109,-1171.5395 1109,-1167.604"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1112.5001,-1167.4199 1109,-1157.4199 1105.5001,-1167.42 1112.5001,-1167.4199"/>
 </g>
 <!-- t2 -->
 <g id="node3" class="node">
 <title>t2</title>
-<polygon fill="#efa03b" stroke="#000000" points="1043.9447,-1078.2585 787,-1095.158 530.0553,-1078.2585 530.2951,-1050.9147 1043.7049,-1050.9147 1043.9447,-1078.2585"/>
-<polygon fill="none" stroke="#000000" points="1047.9814,-1081.9991 787,-1099.164 526.0186,-1081.9991 526.3263,-1046.9173 1047.6737,-1046.9173 1047.9814,-1081.9991"/>
-<text text-anchor="middle" x="787" y="-1064.7009" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.pre_split_pages</text>
+<polygon fill="#efa03b" stroke="#000000" points="1030.0926,-1078.2732 785,-1095.1731 539.9074,-1078.2732 540.1362,-1050.9285 1029.8638,-1050.9285 1030.0926,-1078.2732"/>
+<polygon fill="none" stroke="#000000" points="1034.1249,-1082.0021 785,-1099.1801 535.8751,-1082.0021 536.1685,-1046.931 1033.8315,-1046.931 1034.1249,-1082.0021"/>
+<text text-anchor="middle" x="785" y="-1064.7153" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.marker_pages</text>
 </g>
 <!-- t1&#45;&gt;t2 -->
 <g id="edge2" class="edge">
 <title>t1&#45;&gt;t2</title>
-<path fill="none" stroke="#0044a0" d="M774.5557,-1120.8306C775.4962,-1117.0421 776.5194,-1112.92 777.5618,-1108.7212"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="780.986,-1109.4542 779.9984,-1098.9055 774.1922,-1107.7677 780.986,-1109.4542"/>
+<path fill="none" stroke="#0044a0" d="M1023.449,-1121.1152C984.5695,-1112.8974 938.0675,-1103.0685 896.4957,-1094.2816"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="897.1066,-1090.8335 886.5989,-1092.1898 895.6589,-1097.6822 897.1066,-1090.8335"/>
 </g>
-<!-- t20 -->
-<g id="node22" class="node">
-<title>t20</title>
-<polygon fill="#efa03b" stroke="#000000" points="649.5156,-1020.2193 286.4844,-1020.2193 282.4844,-1016.2193 282.4844,-984.2193 645.5156,-984.2193 649.5156,-988.2193 649.5156,-1020.2193"/>
-<polyline fill="none" stroke="#000000" points="645.5156,-1016.2193 282.4844,-1016.2193 "/>
-<polyline fill="none" stroke="#000000" points="645.5156,-1016.2193 645.5156,-984.2193 "/>
-<polyline fill="none" stroke="#000000" points="645.5156,-1016.2193 649.5156,-1020.2193 "/>
-<text text-anchor="middle" x="466" y="-996.2193" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.generate_postscript_stub</text>
+<!-- t16 -->
+<g id="node17" class="node">
+<title>t16</title>
+<polygon fill="#00cc66" stroke="#000000" points="1215.5059,-328.1861 942.4941,-328.1861 938.4941,-324.1861 938.4941,-292.1861 1211.5059,-292.1861 1215.5059,-296.1861 1215.5059,-328.1861"/>
+<polyline fill="none" stroke="#000000" points="1211.5059,-324.1861 938.4941,-324.1861 "/>
+<polyline fill="none" stroke="#000000" points="1211.5059,-324.1861 1211.5059,-292.1861 "/>
+<polyline fill="none" stroke="#000000" points="1211.5059,-324.1861 1215.5059,-328.1861 "/>
+<text text-anchor="middle" x="1077" y="-304.1861" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.weave_layers</text>
 </g>
-<!-- t1&#45;&gt;t20 -->
-<g id="edge31" class="edge">
-<title>t1&#45;&gt;t20</title>
-<path fill="none" stroke="#0044a0" d="M598.785,-1123.5822C563.105,-1117.7214 531.8209,-1109.8012 517,-1099.1825 493.9382,-1082.6593 480.4369,-1052.4741 473.1842,-1030.1561"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="476.5199,-1029.095 470.2797,-1020.5329 469.8185,-1031.1177 476.5199,-1029.095"/>
+<!-- t1&#45;&gt;t16 -->
+<g id="edge24" class="edge">
+<title>t1&#45;&gt;t16</title>
+<path fill="none" stroke="#0044a0" d="M1097.2373,-1120.9576C1077.987,-1088.7184 1043.7317,-1019.7372 1063,-963.2656 1106.87,-834.6914 1482.8098,-573.3183 1409,-459.2656 1362.7595,-387.8137 1273.2613,-350.2799 1198.8305,-330.7528"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1199.4555,-327.3005 1188.9015,-328.2301 1197.7316,-334.0849 1199.4555,-327.3005"/>
 </g>
-<!-- t24 -->
-<g id="node24" class="node">
-<title>t24</title>
-<polygon fill="#efa03b" stroke="#000000" points="20.4213,-156.4247 242,-129.112 463.5787,-156.4247 463.3719,-200.6177 20.6281,-200.6177 20.4213,-156.4247"/>
-<polygon fill="none" stroke="#000000" points="16.4083,-152.8864 242,-125.079 467.5917,-152.8864 467.3496,-204.6205 16.6504,-204.6205 16.4083,-152.8864"/>
-<text text-anchor="middle" x="242" y="-162.6394" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.merge_pages_mupdf</text>
+<!-- t17 -->
+<g id="node18" class="node">
+<title>t17</title>
+<polygon fill="#efa03b" stroke="#000000" points="1439.5156,-1009.7492 1076.4844,-1009.7492 1072.4844,-1005.7492 1072.4844,-973.7492 1435.5156,-973.7492 1439.5156,-977.7492 1439.5156,-1009.7492"/>
+<polyline fill="none" stroke="#000000" points="1435.5156,-1005.7492 1072.4844,-1005.7492 "/>
+<polyline fill="none" stroke="#000000" points="1435.5156,-1005.7492 1435.5156,-973.7492 "/>
+<polyline fill="none" stroke="#000000" points="1435.5156,-1005.7492 1439.5156,-1009.7492 "/>
+<text text-anchor="middle" x="1256" y="-985.7492" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.generate_postscript_stub</text>
 </g>
-<!-- t1&#45;&gt;t24 -->
-<g id="edge41" class="edge">
-<title>t1&#45;&gt;t24</title>
-<path fill="none" stroke="#0044a0" d="M598.6617,-1124.8567C376.4578,-1103.9215 17,-1060.9746 17,-1002.2193 17,-1002.2193 17,-1002.2193 17,-335.7123 17,-276.266 68.2524,-235.5065 121.8346,-209.1377"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="123.6227,-212.1629 131.1479,-204.7049 120.6143,-205.8423 123.6227,-212.1629"/>
+<!-- t1&#45;&gt;t17 -->
+<g id="edge25" class="edge">
+<title>t1&#45;&gt;t17</title>
+<path fill="none" stroke="#0044a0" d="M1127.0251,-1121.1176C1153.0342,-1095.0292 1201.2007,-1046.7157 1230.7429,-1017.0834"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1233.2458,-1019.5302 1237.8275,-1009.9772 1228.2885,-1014.5879 1233.2458,-1019.5302"/>
 </g>
-<!-- t23 -->
-<g id="node25" class="node">
-<title>t23</title>
-<polygon fill="#efa03b" stroke="#000000" points="1006.4902,-156.4239 1218,-129.1092 1429.5098,-156.4239 1429.3124,-200.6199 1006.6876,-200.6199 1006.4902,-156.4239"/>
-<polygon fill="none" stroke="#000000" points="1002.4742,-152.9063 1218,-125.073 1433.5258,-152.9063 1433.2948,-204.623 1002.7052,-204.623 1002.4742,-152.9063"/>
-<text text-anchor="middle" x="1218" y="-162.6394" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.merge_pages_qpdf</text>
+<!-- t18 -->
+<g id="node19" class="node">
+<title>t18</title>
+<polygon fill="#efa03b" stroke="#000000" points="1070.189,-214.4288 1263,-187.1099 1455.811,-214.4288 1455.631,-258.6317 1070.369,-258.6317 1070.189,-214.4288"/>
+<polygon fill="none" stroke="#000000" points="1066.1715,-210.9543 1263,-183.0662 1459.8285,-210.9543 1459.6181,-262.6353 1066.3819,-262.6353 1066.1715,-210.9543"/>
+<text text-anchor="middle" x="1263" y="-220.6461" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.metadata_fixup</text>
 </g>
-<!-- t1&#45;&gt;t23 -->
-<g id="edge46" class="edge">
-<title>t1&#45;&gt;t23</title>
-<path fill="none" stroke="#0044a0" d="M941.3419,-1136.2202C1263.4143,-1128.3561 1927,-1100.2573 1927,-1002.2193 1927,-1002.2193 1927,-1002.2193 1927,-335.7123 1927,-234.7734 1651.3069,-194.7548 1443.7522,-178.9295"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1443.8443,-175.4267 1433.6111,-178.1712 1443.3223,-182.4072 1443.8443,-175.4267"/>
+<!-- t1&#45;&gt;t18 -->
+<g id="edge28" class="edge">
+<title>t1&#45;&gt;t18</title>
+<path fill="none" stroke="#0044a0" d="M1241.1115,-1121.1192C1311.0076,-1105.2558 1393.5851,-1075.5895 1449,-1020.2329 1458.4922,-1010.7507 1459,-1005.1661 1459,-991.7492 1459,-991.7492 1459,-991.7492 1459,-393.7258 1459,-341.9727 1441.8377,-325.2216 1402,-292.1861 1390.9869,-283.0534 1378.4885,-274.871 1365.6104,-267.6415"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1367.1079,-264.4723 1356.6476,-262.7888 1363.775,-270.6279 1367.1079,-264.4723"/>
 </g>
 <!-- t3 -->
 <g id="node4" class="node">
 <title>t3</title>
-<polygon fill="#efa03b" stroke="#000000" points="1055.3851,-1020.2193 808.6149,-1020.2193 804.6149,-1016.2193 804.6149,-984.2193 1051.3851,-984.2193 1055.3851,-988.2193 1055.3851,-1020.2193"/>
-<polyline fill="none" stroke="#000000" points="1051.3851,-1016.2193 804.6149,-1016.2193 "/>
-<polyline fill="none" stroke="#000000" points="1051.3851,-1016.2193 1051.3851,-984.2193 "/>
-<polyline fill="none" stroke="#000000" points="1051.3851,-1016.2193 1055.3851,-1020.2193 "/>
-<text text-anchor="middle" x="930" y="-996.2193" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.split_page</text>
+<polygon fill="#efa03b" stroke="#000000" points="1015.5194,-999.3074 785,-1016.2081 554.4806,-999.3074 554.6958,-971.9616 1015.3042,-971.9616 1015.5194,-999.3074"/>
+<polygon fill="none" stroke="#000000" points="1019.5491,-1003.0204 785,-1020.2165 550.4509,-1003.0204 550.7267,-967.9639 1019.2733,-967.9639 1019.5491,-1003.0204"/>
+<text text-anchor="middle" x="785" y="-985.7492" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.ocr_or_skip</text>
 </g>
 <!-- t2&#45;&gt;t3 -->
 <g id="edge3" class="edge">
 <title>t2&#45;&gt;t3</title>
-<path fill="none" stroke="#0044a0" d="M837.0145,-1046.7494C851.946,-1039.5987 868.2558,-1031.7881 882.9879,-1024.733"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="884.9035,-1027.6964 892.4109,-1020.2204 881.88,-1021.383 884.9035,-1027.6964"/>
+<path fill="none" stroke="#0044a0" d="M785,-1046.5938C785,-1041.5635 785,-1036.139 785,-1030.7273"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="788.5001,-1030.4427 785,-1020.4427 781.5001,-1030.4428 788.5001,-1030.4427"/>
 </g>
 <!-- t4 -->
 <g id="node5" class="node">
 <title>t4</title>
-<polygon fill="#efa03b" stroke="#000000" points="1194.5194,-941.2939 964,-958.1945 733.4806,-941.2939 733.6958,-913.948 1194.3042,-913.948 1194.5194,-941.2939"/>
-<polygon fill="none" stroke="#000000" points="1198.5491,-945.0069 964,-962.203 729.4509,-945.0069 729.7267,-909.9503 1198.2733,-909.9503 1198.5491,-945.0069"/>
-<text text-anchor="middle" x="964" y="-927.7357" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.ocr_or_skip</text>
+<polygon fill="#efa03b" stroke="#000000" points="889.3166,-941.2656 582.6834,-941.2656 578.6834,-937.2656 578.6834,-905.2656 885.3166,-905.2656 889.3166,-909.2656 889.3166,-941.2656"/>
+<polyline fill="none" stroke="#000000" points="885.3166,-937.2656 578.6834,-937.2656 "/>
+<polyline fill="none" stroke="#000000" points="885.3166,-937.2656 885.3166,-905.2656 "/>
+<polyline fill="none" stroke="#000000" points="885.3166,-937.2656 889.3166,-941.2656 "/>
+<text text-anchor="middle" x="734" y="-917.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.rasterize_preview</text>
 </g>
 <!-- t3&#45;&gt;t4 -->
 <g id="edge4" class="edge">
 <title>t3&#45;&gt;t4</title>
-<path fill="none" stroke="#0044a0" d="M939.1114,-983.867C941.197,-979.6659 943.4863,-975.0549 945.8053,-970.3839"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="948.9439,-971.9327 950.2559,-961.4194 942.6741,-968.8199 948.9439,-971.9327"/>
+<path fill="none" stroke="#0044a0" d="M767.1627,-967.797C762.8396,-961.9919 758.1927,-955.7519 753.7858,-949.8343"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="756.3465,-947.4129 747.5666,-941.483 750.7323,-951.5939 756.3465,-947.4129"/>
 </g>
 <!-- t5 -->
 <g id="node6" class="node">
 <title>t5</title>
-<polygon fill="#efa03b" stroke="#000000" points="1119.3166,-883.252 812.6834,-883.252 808.6834,-879.252 808.6834,-847.252 1115.3166,-847.252 1119.3166,-851.252 1119.3166,-883.252"/>
-<polyline fill="none" stroke="#000000" points="1115.3166,-879.252 808.6834,-879.252 "/>
-<polyline fill="none" stroke="#000000" points="1115.3166,-879.252 1115.3166,-847.252 "/>
-<polyline fill="none" stroke="#000000" points="1115.3166,-879.252 1119.3166,-883.252 "/>
-<text text-anchor="middle" x="964" y="-859.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.rasterize_preview</text>
+<polygon fill="#efa03b" stroke="#000000" points="1028.0824,-883.2656 769.9176,-883.2656 765.9176,-879.2656 765.9176,-847.2656 1024.0824,-847.2656 1028.0824,-851.2656 1028.0824,-883.2656"/>
+<polyline fill="none" stroke="#000000" points="1024.0824,-879.2656 765.9176,-879.2656 "/>
+<polyline fill="none" stroke="#000000" points="1024.0824,-879.2656 1024.0824,-847.2656 "/>
+<polyline fill="none" stroke="#000000" points="1024.0824,-879.2656 1028.0824,-883.2656 "/>
+<text text-anchor="middle" x="897" y="-859.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.orient_page</text>
+</g>
+<!-- t3&#45;&gt;t5 -->
+<g id="edge6" class="edge">
+<title>t3&#45;&gt;t5</title>
+<path fill="none" stroke="#0044a0" d="M869.1177,-967.8063C880.553,-961.0264 890.81,-952.3527 898,-941.2656 907.042,-927.3226 906.9457,-908.7053 904.5139,-893.387"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="907.914,-892.5383 902.5408,-883.4071 901.0469,-893.8961 907.914,-892.5383"/>
 </g>
 <!-- t4&#45;&gt;t5 -->
 <g id="edge5" class="edge">
 <title>t4&#45;&gt;t5</title>
-<path fill="none" stroke="#0044a0" d="M964,-909.7834C964,-904.5184 964,-898.8955 964,-893.4799"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="967.5001,-893.4695 964,-883.4695 960.5001,-893.4695 967.5001,-893.4695"/>
+<path fill="none" stroke="#0044a0" d="M784.6807,-905.2319C801.0865,-899.3943 819.4462,-892.8614 836.4913,-886.7963"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="838.1026,-889.938 846.3506,-883.288 835.7559,-883.343 838.1026,-889.938"/>
 </g>
 <!-- t6 -->
 <g id="node7" class="node">
 <title>t6</title>
-<polygon fill="#efa03b" stroke="#000000" points="1269.0824,-825.252 1010.9176,-825.252 1006.9176,-821.252 1006.9176,-789.252 1265.0824,-789.252 1269.0824,-793.252 1269.0824,-825.252"/>
-<polyline fill="none" stroke="#000000" points="1265.0824,-821.252 1006.9176,-821.252 "/>
-<polyline fill="none" stroke="#000000" points="1265.0824,-821.252 1265.0824,-789.252 "/>
-<polyline fill="none" stroke="#000000" points="1265.0824,-821.252 1269.0824,-825.252 "/>
-<text text-anchor="middle" x="1138" y="-801.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.orient_page</text>
-</g>
-<!-- t4&#45;&gt;t6 -->
-<g id="edge7" class="edge">
-<title>t4&#45;&gt;t6</title>
-<path fill="none" stroke="#0044a0" d="M1072.6871,-909.8532C1097.768,-902.2558 1119.5887,-893.2079 1128,-883.252 1138.9559,-870.2843 1141.4763,-851.3696 1141.2475,-835.6545"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1144.7291,-835.2052 1140.7199,-825.3983 1137.7384,-835.5649 1144.7291,-835.2052"/>
+<polygon fill="#efa03b" stroke="#000000" points="1064.7325,-825.2656 689.2675,-825.2656 685.2675,-821.2656 685.2675,-789.2656 1060.7325,-789.2656 1064.7325,-793.2656 1064.7325,-825.2656"/>
+<polyline fill="none" stroke="#000000" points="1060.7325,-821.2656 685.2675,-821.2656 "/>
+<polyline fill="none" stroke="#000000" points="1060.7325,-821.2656 1060.7325,-789.2656 "/>
+<polyline fill="none" stroke="#000000" points="1060.7325,-821.2656 1064.7325,-825.2656 "/>
+<text text-anchor="middle" x="875" y="-801.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.rasterize_with_ghostscript</text>
 </g>
 <!-- t5&#45;&gt;t6 -->
-<g id="edge6" class="edge">
+<g id="edge7" class="edge">
 <title>t5&#45;&gt;t6</title>
-<path fill="none" stroke="#0044a0" d="M1018.1009,-847.2184C1035.9411,-841.2716 1055.9459,-834.6034 1074.4267,-828.4431"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1075.5526,-831.7572 1083.9326,-825.2745 1073.3389,-825.1165 1075.5526,-831.7572"/>
-</g>
-<!-- t7 -->
-<g id="node8" class="node">
-<title>t7</title>
-<polygon fill="#efa03b" stroke="#000000" points="917.7325,-767.252 542.2675,-767.252 538.2675,-763.252 538.2675,-731.252 913.7325,-731.252 917.7325,-735.252 917.7325,-767.252"/>
-<polyline fill="none" stroke="#000000" points="913.7325,-763.252 538.2675,-763.252 "/>
-<polyline fill="none" stroke="#000000" points="913.7325,-763.252 913.7325,-731.252 "/>
-<polyline fill="none" stroke="#000000" points="913.7325,-763.252 917.7325,-767.252 "/>
-<text text-anchor="middle" x="728" y="-743.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.rasterize_with_ghostscript</text>
-</g>
-<!-- t6&#45;&gt;t7 -->
-<g id="edge8" class="edge">
-<title>t6&#45;&gt;t7</title>
-<path fill="none" stroke="#0044a0" d="M1010.5209,-789.2184C964.4341,-782.6988 912.2171,-775.312 865.3222,-768.6781"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="865.792,-765.2098 855.4003,-767.2745 864.8114,-772.1408 865.792,-765.2098"/>
-</g>
-<!-- t17 -->
-<g id="node15" class="node">
-<title>t17</title>
-<polygon fill="#ff69b4" stroke="#000000" points="1224.2696,-447.252 843.7304,-447.252 839.7304,-443.252 839.7304,-411.252 1220.2696,-411.252 1224.2696,-415.252 1224.2696,-447.252"/>
-<polyline fill="none" stroke="#000000" points="1220.2696,-443.252 839.7304,-443.252 "/>
-<polyline fill="none" stroke="#000000" points="1220.2696,-443.252 1220.2696,-411.252 "/>
-<polyline fill="none" stroke="#000000" points="1220.2696,-443.252 1224.2696,-447.252 "/>
-<text text-anchor="middle" x="1032" y="-423.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.ocr_tesseract_textonly_pdf</text>
-</g>
-<!-- t6&#45;&gt;t17 -->
-<g id="edge16" class="edge">
-<title>t6&#45;&gt;t17</title>
-<path fill="none" stroke="#0044a0" d="M1158.4397,-788.9107C1179.7261,-767.6923 1210,-730.5987 1210,-691.252 1210,-691.252 1210,-691.252 1210,-575.252 1210,-532.3545 1227.3937,-512.2638 1200,-479.252 1190.5284,-467.8379 1163.9108,-457.8666 1134.5783,-449.8944"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1135.1079,-446.4151 1124.5475,-447.2681 1133.3349,-453.1869 1135.1079,-446.4151"/>
-</g>
-<!-- t14 -->
-<g id="node17" class="node">
-<title>t14</title>
-<polygon fill="#00cc66" stroke="#000000" points="1565,-457.252 1242.4141,-429.252 1565,-401.252 1887.5859,-429.252 1565,-457.252"/>
-<text text-anchor="middle" x="1565" y="-423.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.select_image_layer</text>
-</g>
-<!-- t6&#45;&gt;t14 -->
-<g id="edge22" class="edge">
-<title>t6&#45;&gt;t14</title>
-<path fill="none" stroke="#0044a0" d="M1211.6459,-789.1984C1287.5889,-768.3738 1395,-731.7602 1395,-691.252 1395,-691.252 1395,-691.252 1395,-575.252 1395,-519.4485 1448.3697,-480.3436 1495.3026,-456.7437"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1497.1229,-459.7505 1504.5857,-452.2301 1494.0619,-453.4552 1497.1229,-459.7505"/>
-</g>
-<!-- t21 -->
-<g id="node20" class="node">
-<title>t21</title>
-<polygon fill="#efa03b" stroke="#000000" points="1414.7716,-270.1725 1169.2284,-270.1725 1165.2284,-266.1725 1165.2284,-234.1725 1410.7716,-234.1725 1414.7716,-238.1725 1414.7716,-270.1725"/>
-<polyline fill="none" stroke="#000000" points="1410.7716,-266.1725 1165.2284,-266.1725 "/>
-<polyline fill="none" stroke="#000000" points="1410.7716,-266.1725 1410.7716,-234.1725 "/>
-<polyline fill="none" stroke="#000000" points="1410.7716,-266.1725 1414.7716,-270.1725 "/>
-<text text-anchor="middle" x="1290" y="-246.1725" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.skip_page</text>
-</g>
-<!-- t6&#45;&gt;t21 -->
-<g id="edge28" class="edge">
-<title>t6&#45;&gt;t21</title>
-<path fill="none" stroke="#0044a0" d="M1269.322,-800.6326C1488.9576,-787.7851 1907,-754.9633 1907,-691.252 1907,-691.252 1907,-691.252 1907,-429.252 1907,-329.7211 1606.6136,-282.8737 1425.0278,-263.647"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1425.1684,-260.1428 1414.8592,-262.586 1424.4419,-267.105 1425.1684,-260.1428"/>
-</g>
-<!-- t19 -->
-<g id="node21" class="node">
-<title>t19</title>
-<polygon fill="#66ccff" stroke="#000000" points="465.6739,-447.252 60.3261,-447.252 56.3261,-443.252 56.3261,-411.252 461.6739,-411.252 465.6739,-415.252 465.6739,-447.252"/>
-<polyline fill="none" stroke="#000000" points="461.6739,-443.252 56.3261,-443.252 "/>
-<polyline fill="none" stroke="#000000" points="461.6739,-443.252 461.6739,-411.252 "/>
-<polyline fill="none" stroke="#000000" points="461.6739,-443.252 465.6739,-447.252 "/>
-<text text-anchor="middle" x="261" y="-423.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.ocr_tesseract_and_render_pdf</text>
-</g>
-<!-- t6&#45;&gt;t19 -->
-<g id="edge30" class="edge">
-<title>t6&#45;&gt;t19</title>
-<path fill="none" stroke="#0044a0" d="M1006.6775,-802.4744C883.1032,-797.0687 693.0454,-786.4114 529,-767.252 327.1018,-743.6717 78,-894.5226 78,-691.252 78,-691.252 78,-691.252 78,-575.252 78,-532.3545 60.4757,-512.1551 88,-479.252 97.684,-467.6756 124.6714,-457.7191 154.5327,-449.8074"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="155.4872,-453.176 164.3106,-447.311 153.7556,-446.3936 155.4872,-453.176"/>
-</g>
-<!-- t8 -->
-<g id="node9" class="node">
-<title>t8</title>
-<polygon fill="#efa03b" stroke="#000000" points="942.0801,-709.252 517.9199,-709.252 513.9199,-705.252 513.9199,-673.252 938.0801,-673.252 942.0801,-677.252 942.0801,-709.252"/>
-<polyline fill="none" stroke="#000000" points="938.0801,-705.252 513.9199,-705.252 "/>
-<polyline fill="none" stroke="#000000" points="938.0801,-705.252 938.0801,-673.252 "/>
-<polyline fill="none" stroke="#000000" points="938.0801,-705.252 942.0801,-709.252 "/>
-<text text-anchor="middle" x="728" y="-685.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.preprocess_remove_background</text>
-</g>
-<!-- t7&#45;&gt;t8 -->
-<g id="edge9" class="edge">
-<title>t7&#45;&gt;t8</title>
-<path fill="none" stroke="#0044a0" d="M728,-731.2184C728,-727.5438 728,-723.5938 728,-719.6583"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="731.5001,-719.4742 728,-709.4743 724.5001,-719.4743 731.5001,-719.4742"/>
+<path fill="none" stroke="#0044a0" d="M890.1597,-847.2319C888.6839,-843.3412 887.0909,-839.1417 885.5115,-834.9776"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="888.731,-833.5965 881.9119,-825.4878 882.186,-836.0791 888.731,-833.5965"/>
 </g>
 <!-- t13 -->
 <g id="node16" class="node">
 <title>t13</title>
-<polygon fill="#efa03b" stroke="#000000" points="481,-535.252 97.1874,-507.252 481,-479.252 864.8126,-507.252 481,-535.252"/>
-<text text-anchor="middle" x="481" y="-501.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.select_visible_page_image</text>
+<polygon fill="#00cc66" stroke="#000000" points="1077,-515.2656 754.4141,-487.2656 1077,-459.2656 1399.5859,-487.2656 1077,-515.2656"/>
+<text text-anchor="middle" x="1077" y="-481.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.select_image_layer</text>
 </g>
-<!-- t7&#45;&gt;t13 -->
+<!-- t5&#45;&gt;t13 -->
 <g id="edge20" class="edge">
-<title>t7&#45;&gt;t13</title>
-<path fill="none" stroke="#0044a0" d="M592.7552,-731.2177C555.8805,-724.9263 521.5464,-717.3855 505,-709.252 470.5499,-692.3178 456.557,-686.3456 441,-651.252 425.2764,-615.7827 441.5685,-572.6191 457.8398,-542.8513"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="460.9365,-544.4864 462.8533,-534.0664 454.8568,-541.0167 460.9365,-544.4864"/>
+<title>t5&#45;&gt;t13</title>
+<path fill="none" stroke="#0044a0" d="M1000.2704,-847.241C1031.2359,-840.6873 1060.5415,-832.9862 1074,-825.2656 1114.0954,-802.2644 1145,-795.4899 1145,-749.2656 1145,-749.2656 1145,-749.2656 1145,-633.2656 1145,-590.368 1154.8687,-575.2844 1135,-537.2656 1131.571,-530.7041 1126.8893,-524.6314 1121.6622,-519.1451"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1123.8989,-516.4405 1114.2626,-512.0367 1119.0494,-521.4885 1123.8989,-516.4405"/>
+</g>
+<!-- t7 -->
+<g id="node8" class="node">
+<title>t7</title>
+<polygon fill="#efa03b" stroke="#000000" points="856.0801,-767.2656 431.9199,-767.2656 427.9199,-763.2656 427.9199,-731.2656 852.0801,-731.2656 856.0801,-735.2656 856.0801,-767.2656"/>
+<polyline fill="none" stroke="#000000" points="852.0801,-763.2656 427.9199,-763.2656 "/>
+<polyline fill="none" stroke="#000000" points="852.0801,-763.2656 852.0801,-731.2656 "/>
+<polyline fill="none" stroke="#000000" points="852.0801,-763.2656 856.0801,-767.2656 "/>
+<text text-anchor="middle" x="642" y="-743.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.preprocess_remove_background</text>
+</g>
+<!-- t6&#45;&gt;t7 -->
+<g id="edge8" class="edge">
+<title>t6&#45;&gt;t7</title>
+<path fill="none" stroke="#0044a0" d="M802.5546,-789.2319C777.7884,-783.067 749.9071,-776.1266 724.4138,-769.7806"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="724.95,-766.3073 714.4006,-767.288 723.259,-773.1 724.95,-766.3073"/>
+</g>
+<!-- t12 -->
+<g id="node15" class="node">
+<title>t12</title>
+<polygon fill="#efa03b" stroke="#000000" points="742,-593.2656 358.1874,-565.2656 742,-537.2656 1125.8126,-565.2656 742,-593.2656"/>
+<text text-anchor="middle" x="742" y="-559.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.select_visible_page_image</text>
+</g>
+<!-- t6&#45;&gt;t12 -->
+<g id="edge18" class="edge">
+<title>t6&#45;&gt;t12</title>
+<path fill="none" stroke="#0044a0" d="M875.1443,-789.1129C874.6867,-773.2956 872.5831,-750.0189 865,-731.2656 844.2991,-680.0716 804.0126,-630.4165 775.1838,-598.9641"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="777.4937,-596.3085 768.1202,-591.3701 772.3682,-601.076 777.4937,-596.3085"/>
+</g>
+<!-- t8 -->
+<g id="node9" class="node">
+<title>t8</title>
+<polygon fill="#efa03b" stroke="#000000" points="732.6759,-709.2656 413.3241,-709.2656 409.3241,-705.2656 409.3241,-673.2656 728.6759,-673.2656 732.6759,-677.2656 732.6759,-709.2656"/>
+<polyline fill="none" stroke="#000000" points="728.6759,-705.2656 409.3241,-705.2656 "/>
+<polyline fill="none" stroke="#000000" points="728.6759,-705.2656 728.6759,-673.2656 "/>
+<polyline fill="none" stroke="#000000" points="728.6759,-705.2656 732.6759,-709.2656 "/>
+<text text-anchor="middle" x="571" y="-685.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.preprocess_deskew</text>
+</g>
+<!-- t7&#45;&gt;t8 -->
+<g id="edge9" class="edge">
+<title>t7&#45;&gt;t8</title>
+<path fill="none" stroke="#0044a0" d="M619.9243,-731.2319C613.9709,-726.3685 607.4267,-721.0226 601.115,-715.8666"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="603.2653,-713.1037 593.3065,-709.4878 598.8367,-718.5248 603.2653,-713.1037"/>
+</g>
+<!-- t7&#45;&gt;t12 -->
+<g id="edge17" class="edge">
+<title>t7&#45;&gt;t12</title>
+<path fill="none" stroke="#0044a0" d="M717.0346,-731.1646C726.8757,-725.6884 735.7103,-718.5557 742,-709.2656 763.0291,-678.205 759.4083,-634.0888 752.8082,-602.8898"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="756.1481,-601.7976 750.487,-592.842 749.3278,-603.3733 756.1481,-601.7976"/>
 </g>
 <!-- t9 -->
 <g id="node10" class="node">
 <title>t9</title>
-<polygon fill="#efa03b" stroke="#000000" points="773.6759,-651.252 454.3241,-651.252 450.3241,-647.252 450.3241,-615.252 769.6759,-615.252 773.6759,-619.252 773.6759,-651.252"/>
-<polyline fill="none" stroke="#000000" points="769.6759,-647.252 450.3241,-647.252 "/>
-<polyline fill="none" stroke="#000000" points="769.6759,-647.252 769.6759,-615.252 "/>
-<polyline fill="none" stroke="#000000" points="769.6759,-647.252 773.6759,-651.252 "/>
-<text text-anchor="middle" x="612" y="-627.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.preprocess_deskew</text>
+<polygon fill="#efa03b" stroke="#000000" points="712.8828,-651.2656 411.1172,-651.2656 407.1172,-647.2656 407.1172,-615.2656 708.8828,-615.2656 712.8828,-619.2656 712.8828,-651.2656"/>
+<polyline fill="none" stroke="#000000" points="708.8828,-647.2656 407.1172,-647.2656 "/>
+<polyline fill="none" stroke="#000000" points="708.8828,-647.2656 708.8828,-615.2656 "/>
+<polyline fill="none" stroke="#000000" points="708.8828,-647.2656 712.8828,-651.2656 "/>
+<text text-anchor="middle" x="560" y="-627.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.preprocess_clean</text>
 </g>
 <!-- t8&#45;&gt;t9 -->
 <g id="edge10" class="edge">
 <title>t8&#45;&gt;t9</title>
-<path fill="none" stroke="#0044a0" d="M691.9327,-673.2184C680.9122,-667.7081 668.6526,-661.5783 657.1104,-655.8072"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="658.5545,-652.6162 648.045,-651.2745 655.424,-658.8772 658.5545,-652.6162"/>
+<path fill="none" stroke="#0044a0" d="M567.5798,-673.2319C566.8624,-669.4493 566.0897,-665.3747 565.3216,-661.3247"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="568.7581,-660.6605 563.4559,-651.4878 561.8807,-661.9649 568.7581,-660.6605"/>
 </g>
-<!-- t8&#45;&gt;t13 -->
-<g id="edge19" class="edge">
-<title>t8&#45;&gt;t13</title>
-<path fill="none" stroke="#0044a0" d="M763.3121,-673.1037C771.2083,-667.2594 778.5531,-659.9847 783,-651.252 790.2604,-636.9942 792.214,-628.3326 783,-615.252 759.3602,-581.6921 662.7714,-550.9641 585.1057,-530.9385"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="585.806,-527.5052 575.2515,-528.4283 584.078,-534.2886 585.806,-527.5052"/>
+<!-- t8&#45;&gt;t12 -->
+<g id="edge16" class="edge">
+<title>t8&#45;&gt;t12</title>
+<path fill="none" stroke="#0044a0" d="M670.0045,-673.2618C693.2528,-667.2338 713.6919,-659.8307 722,-651.2656 734.1971,-638.6912 739.4901,-620.2051 741.6123,-603.5914"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="745.1041,-603.8427 742.5574,-593.5585 738.1349,-603.1862 745.1041,-603.8427"/>
 </g>
 <!-- t10 -->
 <g id="node11" class="node">
 <title>t10</title>
-<polygon fill="#efa03b" stroke="#000000" points="1117.8828,-593.252 816.1172,-593.252 812.1172,-589.252 812.1172,-557.252 1113.8828,-557.252 1117.8828,-561.252 1117.8828,-593.252"/>
-<polyline fill="none" stroke="#000000" points="1113.8828,-589.252 812.1172,-589.252 "/>
-<polyline fill="none" stroke="#000000" points="1113.8828,-589.252 1113.8828,-557.252 "/>
-<polyline fill="none" stroke="#000000" points="1113.8828,-589.252 1117.8828,-593.252 "/>
-<text text-anchor="middle" x="965" y="-569.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.preprocess_clean</text>
+<polygon fill="#efa03b" stroke="#000000" points="340.1094,-583.2656 35.8906,-583.2656 31.8906,-579.2656 31.8906,-547.2656 336.1094,-547.2656 340.1094,-551.2656 340.1094,-583.2656"/>
+<polyline fill="none" stroke="#000000" points="336.1094,-579.2656 31.8906,-579.2656 "/>
+<polyline fill="none" stroke="#000000" points="336.1094,-579.2656 336.1094,-547.2656 "/>
+<polyline fill="none" stroke="#000000" points="336.1094,-579.2656 340.1094,-583.2656 "/>
+<text text-anchor="middle" x="186" y="-559.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.select_ocr_image</text>
 </g>
 <!-- t9&#45;&gt;t10 -->
 <g id="edge11" class="edge">
 <title>t9&#45;&gt;t10</title>
-<path fill="none" stroke="#0044a0" d="M721.7564,-615.2184C760.9379,-608.7806 805.2658,-601.4973 845.2464,-594.9283"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="846.0113,-598.3496 855.3115,-593.2745 844.8763,-591.4422 846.0113,-598.3496"/>
+<path fill="none" stroke="#0044a0" d="M460.7568,-615.2213C409.7279,-605.9434 347.4237,-594.6153 295.0537,-585.0935"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="295.5223,-581.6214 285.0575,-583.276 294.2701,-588.5085 295.5223,-581.6214"/>
 </g>
-<!-- t9&#45;&gt;t13 -->
-<g id="edge18" class="edge">
-<title>t9&#45;&gt;t13</title>
-<path fill="none" stroke="#0044a0" d="M593.2309,-615.1993C573.0988,-595.8356 540.6771,-564.6514 515.8453,-540.7673"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="518.0381,-538.0203 508.4046,-533.6106 513.1856,-543.0654 518.0381,-538.0203"/>
+<!-- t9&#45;&gt;t12 -->
+<g id="edge15" class="edge">
+<title>t9&#45;&gt;t12</title>
+<path fill="none" stroke="#0044a0" d="M608.2948,-615.2213C627.0226,-608.2242 648.8712,-600.0609 669.4134,-592.3859"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="670.9148,-595.5613 679.0573,-588.7826 668.4648,-589.004 670.9148,-595.5613"/>
 </g>
 <!-- t11 -->
 <g id="node12" class="node">
 <title>t11</title>
-<polygon fill="#efa03b" stroke="#000000" points="1191.1094,-525.252 886.8906,-525.252 882.8906,-521.252 882.8906,-489.252 1187.1094,-489.252 1191.1094,-493.252 1191.1094,-525.252"/>
-<polyline fill="none" stroke="#000000" points="1187.1094,-521.252 882.8906,-521.252 "/>
-<polyline fill="none" stroke="#000000" points="1187.1094,-521.252 1187.1094,-489.252 "/>
-<polyline fill="none" stroke="#000000" points="1187.1094,-521.252 1191.1094,-525.252 "/>
-<text text-anchor="middle" x="1037" y="-501.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.select_ocr_image</text>
+<polygon fill="#00cc66" stroke="#000000" points="334.0996,-505.2656 19.9004,-505.2656 15.9004,-501.2656 15.9004,-469.2656 330.0996,-469.2656 334.0996,-473.2656 334.0996,-505.2656"/>
+<polyline fill="none" stroke="#000000" points="330.0996,-501.2656 15.9004,-501.2656 "/>
+<polyline fill="none" stroke="#000000" points="330.0996,-501.2656 330.0996,-469.2656 "/>
+<polyline fill="none" stroke="#000000" points="330.0996,-501.2656 334.0996,-505.2656 "/>
+<text text-anchor="middle" x="175" y="-481.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.ocr_tesseract_hocr</text>
 </g>
 <!-- t10&#45;&gt;t11 -->
 <g id="edge12" class="edge">
 <title>t10&#45;&gt;t11</title>
-<path fill="none" stroke="#0044a0" d="M984.2947,-557.0293C992.2846,-549.4832 1001.6851,-540.6049 1010.318,-532.4517"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1012.7851,-534.9359 1017.6521,-525.5251 1007.9787,-529.8468 1012.7851,-534.9359"/>
-</g>
-<!-- t10&#45;&gt;t13 -->
-<g id="edge17" class="edge">
-<title>t10&#45;&gt;t13</title>
-<path fill="none" stroke="#0044a0" d="M836.5676,-557.2078C770.772,-547.9638 690.4915,-536.6847 622.8731,-527.1846"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="622.9531,-523.6616 612.5634,-525.7361 621.9792,-530.5935 622.9531,-523.6616"/>
-</g>
-<!-- t12 -->
-<g id="node13" class="node">
-<title>t12</title>
-<polygon fill="#00cc66" stroke="#000000" points="822.0996,-447.252 507.9004,-447.252 503.9004,-443.252 503.9004,-411.252 818.0996,-411.252 822.0996,-415.252 822.0996,-447.252"/>
-<polyline fill="none" stroke="#000000" points="818.0996,-443.252 503.9004,-443.252 "/>
-<polyline fill="none" stroke="#000000" points="818.0996,-443.252 818.0996,-411.252 "/>
-<polyline fill="none" stroke="#000000" points="818.0996,-443.252 822.0996,-447.252 "/>
-<text text-anchor="middle" x="663" y="-423.252" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.ocr_tesseract_hocr</text>
-</g>
-<!-- t11&#45;&gt;t12 -->
-<g id="edge13" class="edge">
-<title>t11&#45;&gt;t12</title>
-<path fill="none" stroke="#0044a0" d="M950.2568,-489.1612C893.2425,-477.2705 818.5749,-461.6981 759.6569,-449.4104"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="760.1,-445.9276 749.5961,-447.3122 758.6708,-452.7802 760.1,-445.9276"/>
-</g>
-<!-- t11&#45;&gt;t17 -->
-<g id="edge15" class="edge">
-<title>t11&#45;&gt;t17</title>
-<path fill="none" stroke="#0044a0" d="M1035.8403,-489.1612C1035.2388,-479.7779 1034.4904,-468.1019 1033.8144,-457.5566"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1037.3033,-457.2705 1033.1707,-447.5149 1030.3177,-457.7184 1037.3033,-457.2705"/>
+<path fill="none" stroke="#0044a0" d="M183.4487,-547.1748C182.1254,-537.7914 180.4788,-526.1154 178.9917,-515.5701"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="182.4378,-514.9417 177.5755,-505.5285 175.5064,-515.9193 182.4378,-514.9417"/>
 </g>
 <!-- t15 -->
 <g id="node14" class="node">
 <title>t15</title>
-<polygon fill="#00cc66" stroke="#000000" points="1229.0433,-353.7123 920.9567,-353.7123 916.9567,-349.7123 916.9567,-317.7123 1225.0433,-317.7123 1229.0433,-321.7123 1229.0433,-353.7123"/>
-<polyline fill="none" stroke="#000000" points="1225.0433,-349.7123 916.9567,-349.7123 "/>
-<polyline fill="none" stroke="#000000" points="1225.0433,-349.7123 1225.0433,-317.7123 "/>
-<polyline fill="none" stroke="#000000" points="1225.0433,-349.7123 1229.0433,-353.7123 "/>
-<text text-anchor="middle" x="1073" y="-329.7123" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.render_hocr_page</text>
+<polygon fill="#ff69b4" stroke="#000000" points="736.2696,-505.2656 355.7304,-505.2656 351.7304,-501.2656 351.7304,-469.2656 732.2696,-469.2656 736.2696,-473.2656 736.2696,-505.2656"/>
+<polyline fill="none" stroke="#000000" points="732.2696,-501.2656 351.7304,-501.2656 "/>
+<polyline fill="none" stroke="#000000" points="732.2696,-501.2656 732.2696,-469.2656 "/>
+<polyline fill="none" stroke="#000000" points="732.2696,-501.2656 736.2696,-505.2656 "/>
+<text text-anchor="middle" x="544" y="-481.2656" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.ocr_tesseract_textonly_pdf</text>
 </g>
-<!-- t12&#45;&gt;t15 -->
+<!-- t10&#45;&gt;t15 -->
 <g id="edge14" class="edge">
-<title>t12&#45;&gt;t15</title>
-<path fill="none" stroke="#0044a0" d="M754.3855,-411.2269C799.2649,-402.117 854.0781,-390.6067 903,-379.252 933.2642,-372.2277 966.3927,-363.8908 995.3012,-356.4015"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="996.5416,-359.6955 1005.3389,-353.7914 994.7799,-352.9208 996.5416,-359.6955"/>
+<title>t10&#45;&gt;t15</title>
+<path fill="none" stroke="#0044a0" d="M269.0322,-547.1748C323.4947,-535.3086 394.7854,-519.776 451.1284,-507.5002"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="452.0829,-510.8744 461.1086,-505.3257 450.5927,-504.0348 452.0829,-510.8744"/>
 </g>
-<!-- t16 -->
-<g id="node19" class="node">
-<title>t16</title>
-<polygon fill="#00cc66" stroke="#000000" points="651.9202,-270.1725 284.0798,-270.1725 280.0798,-266.1725 280.0798,-234.1725 647.9202,-234.1725 651.9202,-238.1725 651.9202,-270.1725"/>
-<polyline fill="none" stroke="#000000" points="647.9202,-266.1725 280.0798,-266.1725 "/>
-<polyline fill="none" stroke="#000000" points="647.9202,-266.1725 647.9202,-234.1725 "/>
-<polyline fill="none" stroke="#000000" points="647.9202,-266.1725 651.9202,-270.1725 "/>
-<text text-anchor="middle" x="466" y="-246.1725" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.render_hocr_debug_page</text>
+<!-- t14 -->
+<g id="node13" class="node">
+<title>t14</title>
+<polygon fill="#00cc66" stroke="#000000" points="330.0433,-411.7258 21.9567,-411.7258 17.9567,-407.7258 17.9567,-375.7258 326.0433,-375.7258 330.0433,-379.7258 330.0433,-411.7258"/>
+<polyline fill="none" stroke="#000000" points="326.0433,-407.7258 17.9567,-407.7258 "/>
+<polyline fill="none" stroke="#000000" points="326.0433,-407.7258 326.0433,-375.7258 "/>
+<polyline fill="none" stroke="#000000" points="326.0433,-407.7258 330.0433,-411.7258 "/>
+<text text-anchor="middle" x="174" y="-387.7258" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.render_hocr_page</text>
 </g>
-<!-- t12&#45;&gt;t16 -->
-<g id="edge27" class="edge">
-<title>t12&#45;&gt;t16</title>
-<path fill="none" stroke="#0044a0" d="M772.353,-411.2135C825.7002,-401.3017 880.0px85,-389.0648 888,-379.252 912.2169,-349.0628 914.0433,-320.8011 888,-292.1725 871.9898,-274.5731 764.4698,-264.5473 661.9779,-258.9344"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="661.9319,-255.4271 651.7593,-258.3883 661.5583,-262.4171 661.9319,-255.4271"/>
+<!-- t11&#45;&gt;t14 -->
+<g id="edge13" class="edge">
+<title>t11&#45;&gt;t14</title>
+<path fill="none" stroke="#0044a0" d="M174.8072,-469.2302C174.6642,-455.8584 174.4666,-437.3702 174.3027,-422.0357"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="177.8024,-421.987 174.1956,-412.025 170.8028,-422.0619 177.8024,-421.987"/>
 </g>
-<!-- t25 -->
-<g id="node27" class="node">
-<title>t25</title>
-<polygon fill="#efa03b" stroke="#000000" points="488.7205,-323.495 682,-296.1762 875.2795,-323.495 875.0991,-367.6976 488.9009,-367.6976 488.7205,-323.495"/>
-<polygon fill="none" stroke="#000000" points="484.7033,-320.0193 682,-292.1327 879.2967,-320.0193 879.0858,-371.7013 484.9142,-371.7013 484.7033,-320.0193"/>
-<text text-anchor="middle" x="682" y="-329.7123" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.merge_sidecars</text>
+<!-- t19 -->
+<g id="node22" class="node">
+<title>t19</title>
+<polygon fill="#efa03b" stroke="#000000" points="351.7205,-381.5085 545,-354.1897 738.2795,-381.5085 738.0991,-425.7112 351.9009,-425.7112 351.7205,-381.5085"/>
+<polygon fill="none" stroke="#000000" points="347.7033,-378.0329 545,-350.1463 742.2967,-378.0329 742.0858,-429.7148 347.9142,-429.7148 347.7033,-378.0329"/>
+<text text-anchor="middle" x="545" y="-387.7258" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.merge_sidecars</text>
 </g>
-<!-- t12&#45;&gt;t25 -->
-<g id="edge52" class="edge">
-<title>t12&#45;&gt;t25</title>
-<path fill="none" stroke="#0044a0" d="M666.6634,-411.2166C668.3795,-402.768 670.5105,-392.2769 672.6326,-381.8293"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="676.09,-382.3905 674.6507,-371.8939 669.2301,-380.9971 676.09,-382.3905"/>
+<!-- t11&#45;&gt;t19 -->
+<g id="edge32" class="edge">
+<title>t11&#45;&gt;t19</title>
+<path fill="none" stroke="#0044a0" d="M246.3397,-469.2302C287.966,-458.7066 342.1269,-445.0142 392.8772,-432.184"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="393.8058,-435.5594 402.6429,-429.7151 392.0901,-428.773 393.8058,-435.5594"/>
 </g>
-<!-- t18 -->
-<g id="node18" class="node">
-<title>t18</title>
-<polygon fill="#00cc66" stroke="#000000" points="1147.2989,-270.1725 856.7011,-270.1725 852.7011,-266.1725 852.7011,-234.1725 1143.2989,-234.1725 1147.2989,-238.1725 1147.2989,-270.1725"/>
-<polyline fill="none" stroke="#000000" points="1143.2989,-266.1725 852.7011,-266.1725 "/>
-<polyline fill="none" stroke="#000000" points="1143.2989,-266.1725 1143.2989,-234.1725 "/>
-<polyline fill="none" stroke="#000000" points="1143.2989,-266.1725 1147.2989,-270.1725 "/>
-<text text-anchor="middle" x="1000" y="-246.1725" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.combine_layers</text>
+<!-- t14&#45;&gt;t16 -->
+<g id="edge22" class="edge">
+<title>t14&#45;&gt;t16</title>
+<path fill="none" stroke="#0044a0" d="M230.6615,-375.7128C262.2207,-366.4581 302.4172,-355.9425 339,-350.1861 448.3957,-332.9725 745.7869,-320.7866 927.9215,-314.6743"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="928.248,-318.1655 938.1258,-314.3342 928.0148,-311.1693 928.248,-318.1655"/>
 </g>
-<!-- t15&#45;&gt;t18 -->
-<g id="edge24" class="edge">
-<title>t15&#45;&gt;t18</title>
-<path fill="none" stroke="#0044a0" d="M1057.1601,-317.5854C1047.0201,-305.9814 1033.7081,-290.7474 1022.4822,-277.9007"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1025.0247,-275.4911 1015.809,-270.264 1019.7536,-280.0972 1025.0247,-275.4911"/>
-</g>
-<!-- t17&#45;&gt;t18 -->
-<g id="edge23" class="edge">
-<title>t17&#45;&gt;t18</title>
-<path fill="none" stroke="#0044a0" d="M1160.5767,-411.2248C1195.7121,-403.5317 1227.1378,-393.066 1238,-379.252 1261.9222,-348.8288 1262.9959,-321.72 1238,-292.1725 1231.6408,-284.6553 1197.8632,-277.2656 1157.6598,-270.9231"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1158.0117,-267.4362 1147.596,-269.3761 1156.9481,-274.3549 1158.0117,-267.4362"/>
-</g>
-<!-- t17&#45;&gt;t25 -->
-<g id="edge51" class="edge">
-<title>t17&#45;&gt;t25</title>
-<path fill="none" stroke="#0044a0" d="M964.5165,-411.2166C925.304,-400.7368 874.3329,-387.1145 826.4989,-374.3305"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="827.2267,-370.9023 816.6621,-371.7016 825.4193,-377.6649 827.2267,-370.9023"/>
-</g>
-<!-- t13&#45;&gt;t14 -->
+<!-- t15&#45;&gt;t16 -->
 <g id="edge21" class="edge">
-<title>t13&#45;&gt;t14</title>
-<path fill="none" stroke="#0044a0" d="M672.8683,-493.1993C736.8314,-488.6166 808.4777,-483.5937 874,-479.252 1033.5051,-468.6828 1073.5671,-468.8587 1233,-457.252 1285.8429,-453.4051 1343.3743,-448.7264 1395.4741,-444.3037"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1395.81,-447.7879 1405.477,-443.4521 1395.2161,-440.8131 1395.81,-447.7879"/>
+<title>t15&#45;&gt;t16</title>
+<path fill="none" stroke="#0044a0" d="M630.0636,-469.2083C667.5329,-460.6467 711.76,-449.5912 751,-437.2656 850.4985,-406.0121 963.1653,-359.5222 1026.586,-332.2894"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1028.1566,-335.4238 1035.9553,-328.2523 1025.3866,-328.9951 1028.1566,-335.4238"/>
+</g>
+<!-- t15&#45;&gt;t19 -->
+<g id="edge31" class="edge">
+<title>t15&#45;&gt;t19</title>
+<path fill="none" stroke="#0044a0" d="M544.1928,-469.2302C544.2822,-460.8687 544.393,-450.5066 544.5035,-440.166"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="548.006,-439.9443 544.6132,-429.9075 541.0064,-439.8694 548.006,-439.9443"/>
+</g>
+<!-- t12&#45;&gt;t13 -->
+<g id="edge19" class="edge">
+<title>t12&#45;&gt;t13</title>
+<path fill="none" stroke="#0044a0" d="M833.9833,-543.8486C879.0975,-533.3444 933.4556,-520.6879 979.3116,-510.0109"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="980.3156,-513.3709 989.2614,-507.6943 978.7282,-506.5532 980.3156,-513.3709"/>
 </g>
 <!-- t13&#45;&gt;t16 -->
-<g id="edge26" class="edge">
+<g id="edge23" class="edge">
 <title>t13&#45;&gt;t16</title>
-<path fill="none" stroke="#0044a0" d="M479.4495,-479.3529C478.2545,-457.9761 476.5497,-427.7492 475,-401.252 472.546,-359.2945 469.5914,-310.7019 467.754,-280.6981"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="471.2362,-280.2988 467.1306,-270.5318 464.2493,-280.7273 471.2362,-280.2988"/>
+<path fill="none" stroke="#0044a0" d="M1077,-459.2068C1077,-426.2624 1077,-371.9187 1077,-338.7552"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1080.5001,-338.3818 1077,-328.3819 1073.5001,-338.3819 1080.5001,-338.3818"/>
 </g>
-<!-- t13&#45;&gt;t19 -->
+<!-- t16&#45;&gt;t18 -->
+<g id="edge27" class="edge">
+<title>t16&#45;&gt;t18</title>
+<path fill="none" stroke="#0044a0" d="M1117.3592,-292.0592C1133.8695,-284.6437 1153.6803,-275.7459 1173.2953,-266.9361"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1174.8979,-270.0531 1182.586,-262.7632 1172.0299,-263.6676 1174.8979,-270.0531"/>
+</g>
+<!-- t17&#45;&gt;t18 -->
+<g id="edge26" class="edge">
+<title>t17&#45;&gt;t18</title>
+<path fill="none" stroke="#0044a0" d="M1326.2292,-973.7283C1377.8362,-955.768 1439,-922.4247 1439,-865.2656 1439,-865.2656 1439,-865.2656 1439,-393.7258 1439,-338.7092 1393.2056,-296.3464 1348.3356,-268.0433"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1350.0403,-264.9831 1339.6842,-262.7465 1346.3852,-270.9531 1350.0403,-264.9831"/>
+</g>
+<!-- t20 -->
+<g id="node20" class="node">
+<title>t20</title>
+<polygon fill="#efa03b" stroke="#000000" points="1400.3184,-161.1062 1129.6816,-161.1062 1125.6816,-157.1062 1125.6816,-125.1062 1396.3184,-125.1062 1400.3184,-129.1062 1400.3184,-161.1062"/>
+<polyline fill="none" stroke="#000000" points="1396.3184,-157.1062 1125.6816,-157.1062 "/>
+<polyline fill="none" stroke="#000000" points="1396.3184,-157.1062 1396.3184,-125.1062 "/>
+<polyline fill="none" stroke="#000000" points="1396.3184,-157.1062 1400.3184,-161.1062 "/>
+<text text-anchor="middle" x="1263" y="-137.1062" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.optimize_pdf</text>
+</g>
+<!-- t18&#45;&gt;t20 -->
 <g id="edge29" class="edge">
-<title>t13&#45;&gt;t19</title>
-<path fill="none" stroke="#0044a0" d="M415.2907,-483.9551C385.6007,-473.4286 350.7427,-461.0699 321.6782,-450.7652"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="322.6542,-447.3978 312.0595,-447.3549 320.315,-453.9954 322.6542,-447.3978"/>
+<title>t18&#45;&gt;t20</title>
+<path fill="none" stroke="#0044a0" d="M1263,-182.9188C1263,-179.129 1263,-175.3607 1263,-171.739"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1266.5001,-171.4394 1263,-161.4394 1259.5001,-171.4394 1266.5001,-171.4394"/>
 </g>
-<!-- t14&#45;&gt;t18 -->
-<g id="edge25" class="edge">
-<title>t14&#45;&gt;t18</title>
-<path fill="none" stroke="#0044a0" d="M1519.9776,-405.138C1460.3688,-374.2505 1351.1978,-321.2463 1252,-292.1725 1221.8888,-283.3473 1189.1477,-276.3434 1157.5373,-270.8287"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1157.7641,-267.3171 1147.3187,-269.0901 1156.59,-274.2179 1157.7641,-267.3171"/>
+<!-- t21 -->
+<g id="node21" class="node">
+<title>t21</title>
+<polygon fill="#efa03b" stroke="#000000" points="1096.0652,-47.332 1263,-20.0049 1429.9348,-47.332 1429.779,-91.5483 1096.221,-91.5483 1096.0652,-47.332"/>
+<polygon fill="none" stroke="#000000" points="1092.0538,-43.9306 1263,-15.9468 1433.9462,-43.9306 1433.7643,-95.5532 1092.2357,-95.5532 1092.0538,-43.9306"/>
+<text text-anchor="middle" x="1263" y="-53.5531" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.copy_final</text>
 </g>
-<!-- t22 -->
-<g id="node23" class="node">
-<title>t22</title>
-<polygon fill="#efa03b" stroke="#000000" points="489.918,-156.4263 735,-129.1169 980.082,-156.4263 979.8532,-200.6137 490.1468,-200.6137 489.918,-156.4263"/>
-<polygon fill="none" stroke="#000000" points="485.9027,-152.8467 735,-125.09 984.0973,-152.8467 983.8293,-204.6159 486.1707,-204.6159 485.9027,-152.8467"/>
-<text text-anchor="middle" x="735" y="-162.6394" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.merge_pages_ghostscript</text>
-</g>
-<!-- t18&#45;&gt;t22 -->
-<g id="edge35" class="edge">
-<title>t18&#45;&gt;t22</title>
-<path fill="none" stroke="#0044a0" d="M942.8252,-234.1499C918.2191,-226.3936 888.4232,-217.0013 859.1708,-207.7804"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="859.8692,-204.3309 849.2796,-204.6625 857.7647,-211.007 859.8692,-204.3309"/>
-</g>
-<!-- t18&#45;&gt;t24 -->
-<g id="edge39" class="edge">
-<title>t18&#45;&gt;t24</title>
-<path fill="none" stroke="#0044a0" d="M852.4716,-245.1051C747.4645,-239.0474 603.1919,-228.6387 477,-212.1725 463.5912,-210.4229 449.788,-208.4132 435.908,-206.238"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="436.2183,-202.7434 425.7921,-204.6263 435.1168,-209.6562 436.2183,-202.7434"/>
-</g>
-<!-- t18&#45;&gt;t23 -->
-<g id="edge44" class="edge">
-<title>t18&#45;&gt;t23</title>
-<path fill="none" stroke="#0044a0" d="M1047.0344,-234.1499C1066.9243,-226.5285 1090.9364,-217.3275 1114.596,-208.2617"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1115.9033,-211.509 1123.9889,-204.6625 1113.3986,-204.9724 1115.9033,-211.509"/>
-</g>
-<!-- t16&#45;&gt;t22 -->
-<g id="edge36" class="edge">
-<title>t16&#45;&gt;t22</title>
-<path fill="none" stroke="#0044a0" d="M524.0378,-234.1499C549.1239,-226.3599 579.5241,-216.9196 609.3422,-207.6602"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="610.4833,-210.9707 618.9955,-204.6625 608.4073,-204.2857 610.4833,-210.9707"/>
-</g>
-<!-- t16&#45;&gt;t24 -->
-<g id="edge40" class="edge">
-<title>t16&#45;&gt;t24</title>
-<path fill="none" stroke="#0044a0" d="M417.6711,-234.1499C397.2337,-226.5285 372.5607,-217.3275 348.2499,-208.2617"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="349.1912,-204.8773 338.5986,-204.6625 346.7453,-211.4361 349.1912,-204.8773"/>
-</g>
-<!-- t16&#45;&gt;t23 -->
-<g id="edge45" class="edge">
-<title>t16&#45;&gt;t23</title>
-<path fill="none" stroke="#0044a0" d="M651.7596,-235.3105C656.5547,-234.9215 661.308,-234.5415 666,-234.1725 811.2137,-222.7545 848.5512,-230.9322 993,-212.1725 1006.176,-210.4614 1019.7432,-208.4558 1033.3753,-206.2619"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1034.0068,-209.7051 1043.3083,-204.6325 1032.8736,-202.7974 1034.0068,-209.7051"/>
-</g>
-<!-- t21&#45;&gt;t22 -->
-<g id="edge32" class="edge">
-<title>t21&#45;&gt;t22</title>
-<path fill="none" stroke="#0044a0" d="M1165.0941,-235.3973C1085.9035,-224.7541 997.501,-212.8552 993,-212.1725 980.474,-210.2728 967.615,-208.2703 954.6442,-206.2099"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="955.1339,-202.7439 944.7072,-204.6238 954.0305,-209.6564 955.1339,-202.7439"/>
-</g>
-<!-- t21&#45;&gt;t24 -->
-<g id="edge37" class="edge">
-<title>t21&#45;&gt;t24</title>
-<path fill="none" stroke="#0044a0" d="M1165.2215,-234.9906C1162.1187,-234.699 1159.041,-234.4254 1156,-234.1725 855.1023,-209.1529 777.1841,-244.6523 477,-212.1725 462.026,-210.5524 446.5737,-208.5164 431.0798,-206.2151"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="431.3094,-202.7099 420.8968,-204.6663 430.2567,-209.6303 431.3094,-202.7099"/>
-</g>
-<!-- t21&#45;&gt;t23 -->
-<g id="edge42" class="edge">
-<title>t21&#45;&gt;t23</title>
-<path fill="none" stroke="#0044a0" d="M1274.3771,-234.0471C1268.8557,-227.6412 1262.3807,-220.0px91 1255.8248,-212.5231"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1258.4543,-210.2128 1249.2743,-204.9233 1253.152,-214.783 1258.4543,-210.2128"/>
-</g>
-<!-- t19&#45;&gt;t22 -->
-<g id="edge34" class="edge">
-<title>t19&#45;&gt;t22</title>
-<path fill="none" stroke="#0044a0" d="M260.6159,-410.9505C259.8316,-364.871 259.0249,-246.8273 271,-234.1725 302.6435,-200.7331 431.3919,-218.4498 477,-212.1725 490.4888,-210.316 504.3558,-208.2974 518.3278,-206.1801"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="519.1602,-209.5936 528.5161,-204.6219 518.1019,-202.6741 519.1602,-209.5936"/>
-</g>
-<!-- t19&#45;&gt;t24 -->
-<g id="edge38" class="edge">
-<title>t19&#45;&gt;t24</title>
-<path fill="none" stroke="#0044a0" d="M259.0498,-411.2136C258.0595,-401.7112 256.8826,-389.8585 256,-379.252 251.2976,-322.7418 247.1648,-257.5963 244.6312,-214.9019"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="248.1192,-214.5935 244.0377,-204.8165 241.1313,-215.0048 248.1192,-214.5935"/>
-</g>
-<!-- t19&#45;&gt;t23 -->
-<g id="edge43" class="edge">
-<title>t19&#45;&gt;t23</title>
-<path fill="none" stroke="#0044a0" d="M282.1528,-411.0561C318.8331,-380.6455 397.8323,-320.0855 476,-292.1725 553.979,-264.3269 579.2078,-283.062 661,-270.1725 742.8817,-257.269 762.3086,-248.2304 844,-234.1725 909.9705,-222.82 927.0119,-223.4221 993,-212.1725 1003.5851,-210.368 1014.439,-208.4601 1025.3874,-206.4907"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1026.2474,-209.892 1035.4637,-204.666 1025,-203.004 1026.2474,-209.892"/>
-</g>
-<!-- t19&#45;&gt;t25 -->
-<g id="edge50" class="edge">
-<title>t19&#45;&gt;t25</title>
-<path fill="none" stroke="#0044a0" d="M342.173,-411.2166C389.8324,-400.6274 451.9326,-386.8297 509.989,-373.9305"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="511.0179,-377.2873 520.0207,-371.7016 509.4996,-370.4539 511.0179,-377.2873"/>
-</g>
-<!-- t20&#45;&gt;t22 -->
-<g id="edge33" class="edge">
-<title>t20&#45;&gt;t22</title>
-<path fill="none" stroke="#0044a0" d="M323.4472,-984.1943C198.6099,-964.2959 37,-926.6746 37,-865.252 37,-865.252 37,-865.252 37,-429.252 37,-211.1917 260.8906,-241.2767 477,-212.1725 490.808,-210.313 505.0108,-208.2741 519.3157,-206.1255"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="519.8697,-209.5816 529.2317,-204.6214 518.8199,-202.6608 519.8697,-209.5816"/>
-</g>
-<!-- t26 -->
-<g id="node26" class="node">
-<title>t26</title>
-<polygon fill="#efa03b" stroke="#000000" points="568.0652,-47.332 735,-20.0049 901.9348,-47.332 901.779,-91.5483 568.221,-91.5483 568.0652,-47.332"/>
-<polygon fill="none" stroke="#000000" points="564.0538,-43.9306 735,-15.9468 905.9462,-43.9306 905.7643,-95.5532 564.2357,-95.5532 564.0538,-43.9306"/>
-<text text-anchor="middle" x="735" y="-53.5531" font-family="Times,serif" font-size="20.00" fill="#000000">ocrmypdf.pipeline.copy_final</text>
-</g>
-<!-- t22&#45;&gt;t26 -->
-<g id="edge49" class="edge">
-<title>t22&#45;&gt;t26</title>
-<path fill="none" stroke="#0044a0" d="M735,-124.8984C735,-118.6708 735,-112.2475 735,-105.9711"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="738.5001,-105.6074 735,-95.6074 731.5001,-105.6075 738.5001,-105.6074"/>
-</g>
-<!-- t24&#45;&gt;t26 -->
-<g id="edge47" class="edge">
-<title>t24&#45;&gt;t26</title>
-<path fill="none" stroke="#0044a0" d="M368.6395,-140.6178C427.4047,-127.6148 498.1821,-111.9539 561.7955,-97.8781"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="562.9329,-101.2112 571.9406,-95.6333 561.4205,-94.3765 562.9329,-101.2112"/>
-</g>
-<!-- t23&#45;&gt;t26 -->
-<g id="edge48" class="edge">
-<title>t23&#45;&gt;t26</title>
-<path fill="none" stroke="#0044a0" d="M1095.1818,-140.9007C1037.3475,-127.8387 967.4279,-112.0473 904.6388,-97.8663"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="905.1511,-94.3939 894.6256,-95.6048 903.6089,-101.2219 905.1511,-94.3939"/>
+<!-- t20&#45;&gt;t21 -->
+<g id="edge30" class="edge">
+<title>t20&#45;&gt;t21</title>
+<path fill="none" stroke="#0044a0" d="M1263,-124.9764C1263,-119.37 1263,-112.9163 1263,-106.2938"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1266.5001,-105.8456 1263,-95.8457 1259.5001,-105.8457 1266.5001,-105.8456"/>
 </g>
 </g>
 </svg>

docs/release_notes.rst

@@ -1,14 +1,300 @@
 Release notes
 =============
 
-OCRmyPDF uses `semantic versioning <http://semver.org/>`_ for its command line interface.
+OCRmyPDF uses `semantic versioning <http://semver.org/>`_ for its command line interface and its public API.
 
-The OCRmyPDF package itself does not contain a public API, although it is fairly stable and breaking changes are usually timed with a major release. A future release will clearly define the stable public API.
+The ``ocrmypdf`` package may now be imported. The public API may be useful in scripts that launch OCRmyPDF processes or that wish to use some of its features for working with PDFs.
+
+Unfortunately, the public API does **not** expose the ability to actually OCR a PDF. This is due to a limitation in an underlying library (ruffus) that makes OCRmyPDF non-reentrant.
+
+Note that it is licensed under GPLv3, so scripts that ``import ocrmypdf`` and are released publicly should probably also be licensed under GPLv3.
 
 .. Issue regex
    find:    [^`]\#([0-9]{1,3})[^0-9]
    replace: `#$1 <https://github.com/jbarlow83/OCRmyPDF/issues/$1>`_
 
+v8.2.3
+------
+
+-   Fixed that ``--mask-barcodes`` would occasionally leave a unwanted temporary file named ``junkpixt`` in the current working folder.
+
+-   Fixed (hopefully) handling of Leptonica errors in an environment where a non-standard ``sys.stderr`` is present.
+
+-   Improved help text for ``--verbose``.
+
+v8.2.2
+------
+
+-   Fixed a regression from v8.2.0, an exception that occurred while attempting to report that ``unpaper`` or another optional dependency was unavailable.
+
+-   In some cases, ``ocrmypdf [-c|--clean]`` failed to exit with an error when ``unpaper`` is not installed.
+
+v8.2.1
+------
+
+-   This release was canceled.
+
+v8.2.0
+------
+
+-   A major improvement to our Docker image is now available thanks to hard work contributed by @mawi12345. The new Docker image, ocrmypdf-alpine, is based on Alpine Linux, and includes most of the functionality of three existed images in a smaller package. This image will replace the main Docker image eventually but for now all are being built. `See documentation for details <https://ocrmypdf.readthedocs.io/en/latest/docker.html>`_.
+
+-   Documentation reorganized especially around the use of Docker images.
+
+-   Fixed a problem with PDF image optimization, where the optimizer would unnecessarily decompress and recompress PNG images, in some cases losing the benefits of the quantization it just had just performed. The optimizer is now capable of embedding PNG images into PDFs without transcoding them.
+
+-   Fixed a minor regression with lossy JBIG2 image optimization. All JBIG2 candidates images were incorrectly placed into a single optimization group for the whole file, instead of grouping pages together. This usually makes a larger JBIG2Globals dictionary and results in inferior compression, so it worked less well than designed. However, quality would not be impacted. Lossless JBIG2 was entirely unaffected.
+
+-   Updated dependencies, including pikepdf to 1.1.0. This fixes `#358 <https://github.com/jbarlow83/OCRmyPDF/issues/358>`_.
+
+-   The install-time version checks for certain external programs have been removed from setup.py. These tests are now performed at run-time.
+
+-   The non-standard option to override install-time checks (``setup.py install --force``) is now deprecated and prints a warning. It will be removed in a future release.
+
+v8.1.0
+------
+
+-   Added a feature, ``--unpaper-args``, which allows passing arbitrary arguments to ``unpaper`` when using ``--clean`` or ``--clean-final``. The default, very conservative unpaper settings are suppressed.
+
+-   The argument ``--clean-final`` now implies ``--clean``. It was possible to issue ``--clean-final`` on its before this, but it would have no useful effect.
+
+-   Fixed an exception on traversing corrupt table of contents entries (specifically, those with invalid destination objects)
+
+-   Fixed an issue when using ``--tesseract-timeout`` and image processing features on a file with more than 100 pages. `#347 <https://github.com/jbarlow83/OCRmyPDF/issues/347>`_
+
+-   OCRmyPDF now always calls ``os.nice(5)`` to signal to operating systems that it is a background process.
+
+v8.0.1
+------
+
+-   Fixed an exception when parsing PDFs that are missing a required field. `#325 <https://github.com/jbarlow83/OCRmyPDF/issues/325>`_
+
+-   pikepdf 1.0.5 is now required, to address some other PDF parsing issues.
+
+v8.0.0
+------
+
+No major features. The intent of this release is to sever support for older versions of certain dependencies.
+
+**Breaking changes**
+
+-   Dropped support for Tesseract 3.x. Tesseract 4.0 or newer is now required.
+
+-   Dropped support for Python 3.5.
+
+-   Some ``ocrmypdf.pdfa`` APIs that were deprecated in v7.x were removed. This functionality has been moved to pikepdf.
+
+**Other changes**
+
+-   Fixed an unhandled exception when attempting to mask barcodes. `#322 <https://github.com/jbarlow83/OCRmyPDF/issues/322>`_
+
+-   It is now possible to use ocrmypdf without pdfminer.six, to support distributions that do not have it or cannot currently use it (e.g. Homebrew). Downstream maintainers should include pdfminer.six if possible.
+
+-   A warning is now issue when PDF/A conversion removes some XMP metadata from the input PDF. (Only a "whitelist" of certain XMP metadata types are allowed in PDF/A.)
+
+-   Fixed several issues that caused PDF/As to be produced with nonconforming XMP metadata (would fail validation with veraPDF).
+
+-   Fixed some instances where invalid DocumentInfo from a PDF cause XMP metadata creation to fail.
+
+-   Fixed a few documentation problems.
+
+-   pikepdf 1.0.2 is now required.
+
+v7.4.0
+------
+
+-   ``--force-ocr`` may now be used with the new ``--threshold`` and ``--mask-barcodes`` features
+
+-   pikepdf >= 0.9.1 is now required.
+
+-   Changed metadata handling to pikepdf 0.9.1. As a result, metadata handling of non-ASCII characters in Ghostscript 9.25 or later is fixed.
+
+-   chardet >= 3.0.4 is temporarily listed as required. pdfminer.six depends on it, but the most recent release does not specify this requirement. (`#326 <https://github.com/jbarlow83/OCRmyPDF/issues/326>`_)
+
+-   python-xmp-toolkit and libexempi are no longer required.
+
+-   A new Docker image is now being provided for users who wish to access OCRmyPDF over a simple HTTP interface, instead of the command line.
+
+-   Increase tolerance of PDFs that overflow or underflow the PDF graphics stack. (`#325 <https://github.com/jbarlow83/OCRmyPDF/issues/325>`_)
+
+v7.3.1
+------
+
+-   Fixed performance regression from v7.3.0; fast page analysis was not selected when it should be.
+
+-   Fixed a few exceptions related to the new ``--mask-barcodes`` feature and improved argument checking
+
+-   Added missing detection of TrueType fonts that lack a Unicode mapping
+
+
+v7.3.0
+------
+
+-   Added a new feature ``--redo-ocr`` to detect existing OCR in a file, remove it, and redo the OCR. This may be particularly helpful for anyone who wants to take advantage of OCR quality improvements in Tesseract 4.0. Note that OCR added by OCRmyPDF before version 3.0 cannot be detected since it was not properly marked as invisible text in the earliest versions. OCR that constructs a font from visible text, such as Adobe Acrobat's ClearScan.
+
+-   OCRmyPDF's content detection is generally more sophisticated. It learns more about the contents of each PDF and makes better recommendations:
+
+    -   OCRmyPDF can now detect when a PDF contains text that cannot be mapped to Unicode (meaning it is readable to human eyes but copy-pastes as gibberish). In these cases it recommends ``--force-ocr`` to make the text searchable.
+
+    -   PDFs containing vector objects are now rendered at more appropriate resolution for OCR.
+
+    -   We now exit with an error for PDFs that contain Adobe LiveCycle Designer's dynamic XFA forms. Currently the open source community does not have tools to work with these files.
+
+    -   OCRmyPDF now warns when a PDF that contains Adobe AcroForms, since such files probably do not need OCR. It can work with these files.
+
+-   Added three new **experimental** features to improve OCR quality in certain conditions. The name, syntax and behavior of these arguments is subject to change. They may also be incompatible with some other features.
+
+    -   ``--remove-vectors`` which strips out vector graphics. This can improve OCR quality since OCR will not search artwork for readable text; however, it currently removes "text as curves" as well.
+
+    -   ``--mask-barcodes`` to detect and suppress barcodes in files. We have observed that barcodes can interfere with OCR because they are "text-like" but not actually textual.
+
+    -   ``--threshold`` which uses a more sophisticated thresholding algorithm than is currently in use in Tesseract OCR. This works around a `known issue in Tesseract 4.0 <https://github.com/tesseract-ocr/tesseract/issues/1990>`_ with dark text on bright backgrounds.
+
+-   Fixed an issue where an error message was not reported when the installed Ghostscript was very old.
+
+-   The PDF optimizer now saves files with object streams enabled when the optimization level is ``--optimize 1`` or higher (the default). This makes files a little bit smaller, but requires PDF 1.5. PDF 1.5 was first released in 2003 and is broadly supported by PDF viewers, but some rudimentary PDF parsers such as PyPDF2 do not understand object streams. You can use the command line tool ``qpdf --object-streams=disable`` or `pikepdf <https://github.com/pikepdf/pikepdf>`_ library to remove them.
+
+-   New dependency: pdfminer.six 20181108. Note this is a fork of the Python 2-only pdfminer.
+
+-   Deprecation notice: At the end of 2018, we will be ending support for Python 3.5 and Tesseract 3.x. OCRmyPDF v7 will continue to work with older versions.
+
+v7.2.1
+------
+
+-   Fix compatibility with an API change in pikepdf 0.3.5.
+
+-   A kludge to support Leptonica versions older than 1.72 in the test suite was dropped. Older versions of Leptonica are likely still compatible. The only impact is that a portion of the test suite will be skipped.
+
+
+v7.2.0
+------
+
+**Lossy JBIG2 behavior change**
+
+A user reported that ocrmypdf was in fact using JBIG2 in **lossy** compression mode. This was not the intended behavior. 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 is a concern for their use case.
+
+JBIG2 lossy mode does achieve higher compression ratios than any other monochrome 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.
+
+Only users who have reviewed the concerns with JBIG2 in lossy mode should opt-in. As such, lossy mode JBIG2 is only turned on when the new argument ``--jbig2-lossy`` is issued. This is independent of the setting for ``--optimize``.
+
+Users who did not install an optional JBIG2 encoder are unaffected.
+
+(Thanks to user 'bsdice' for reporting this issue.)
+
+**Other issues**
+
+-   When the image optimizer quantizes an image to 1 bit per pixel, it will now attempt to further optimize that image as CCITT or JBIG2, instead of keeping it in the "flate" encoding which is not efficient for 1 bpp images. (`#297 <https://github.com/jbarlow83/OCRmyPDF/issues/297>`_)
+
+-   Images in PDFs that are used as soft masks (i.e. transparency masks or alpha channels) are now excluded from optimization.
+
+-   Fixed handling of Tesseract 4.0-rc1 which now accepts invalid Tesseract configuration files, which broke the test suite.
+
+v7.1.0
+------
+
+-   Improve the performance of initial text extraction, which is done to determine if a file contains existing text of some kind or not. On large files, this initial processing is now about 20x times faster. (`#299 <https://github.com/jbarlow83/OCRmyPDF/issues/299>`_)
+
+-   pikepdf 0.3.3 is now required.
+
+-   Fixed issue `#231 <https://github.com/jbarlow83/OCRmyPDF/issues/231>`_, a problem with JPEG2000 images where image metadata was only available inside the JPEG2000 file.
+
+-   Fixed some additional Ghostscript 9.25 compatibility issues.
+
+-   Improved handling of KeyboardInterrupt error messages. (`#301 <https://github.com/jbarlow83/OCRmyPDF/issues/301>`_)
+
+-   README.md is now served in GitHub markdown instead of reStructuredText.
+
+v7.0.6
+------
+
+-   Blacklist Ghostscript 9.24, now that 9.25 is available and fixes many regressions in 9.24.
+
+
+v7.0.5
+------
+
+-   Improve capability with Ghostscript 9.24, and enable the JPEG passthrough feature when this version in installed.
+
+-   Ghostscript 9.24 lost the ability to set PDF title, author, subject and keyword metadata to Unicode strings. OCRmyPDF will set ASCII strings and warn when Unicode is suppressed. Other software may be used to update metadata. This is a short term work around.
+
+-   PDFs generated by Kodak Capture Desktop, or generally PDFs that contain indirect references to null objects in their table of contents, would have an invalid table of contents after processing by OCRmyPDF that might interfere with other viewers. This has been fixed.
+
+-   Detect PDFs generated by Adobe LiveCycle, which can only be displayed in Adobe Acrobat and Reader currently. When these are encountered, exit with an error instead of performing OCR on the "Please wait" error message page.
+
+v7.0.4
+------
+
+-   Fix exception thrown when trying to optimize a certain type of PNG embedded in a PDF with the ``-O2``
+
+-   Update to pikepdf 0.3.2, to gain support for optimizing some additional image types that were previously excluded from optimization (CMYK and grayscale). Fixes `#285 <https://github.com/jbarlow83/OCRmyPDF/issues/285>`_.
+
+v7.0.3
+------
+
+-   Fix issue `#284 <https://github.com/jbarlow83/OCRmyPDF/issues/284>`_, an error when parsing inline images that have are also image masks, by upgrading pikepdf to 0.3.1
+
+v7.0.2
+------
+
+-   Fix a regression with ``--rotate-pages`` on pages that already had rotations applied. (`#279 <https://github.com/jbarlow83/OCRmyPDF/issues/279>`_)
+
+-   Improve quality of page rotation in some cases by rasterizing a higher quality preview image. (`#281 <https://github.com/jbarlow83/OCRmyPDF/issues/281>`_)
+
+v7.0.1
+------
+
+-   Fix compatibility with img2pdf >= 0.3.0 by rejecting input images that have an alpha channel
+
+-   Add forward compatibility for pikepdf 0.3.0 (unrelated to img2pdf)
+
+-   Various documentation updates for v7.0.0 changes
+
+v7.0.0
+------
+
+-   The core algorithm for combining OCR layers with existing PDF pages has been rewritten and improved considerably.  PDFs are no longer split into single page PDFs for processing; instead, images are rendered and the OCR results are grafted onto the input PDF.  The new algorithm uses less temporary disk space and is much more performant especially for large files.
+
+-   New dependency: `pikepdf <https://github.com/pikepdf/pikepdf>`_. pikepdf is a powerful new Python PDF library driving the latest OCRmyPDF features, built on the QPDF C++ library (libqpdf).
+
+-   New feature: PDF optimization with ``-O`` or ``--optimize``.  After OCR, OCRmyPDF will perform image optimizations relevant to OCR PDFs.
+
+    +   If a JBIG2 encoder is available, then monochrome images will be converted, 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.)
+
+    +   If ``pngquant`` is installed, OCRmyPDF will optionally use it to perform lossy quantization and compression of PNG images.
+
+    +   The quality of JPEGs can also be lowered, on the assumption that a lower quality image may be suitable for storage after OCR.
+
+    +   This image optimization component will eventually be offered as an independent command line utility.
+
+    +   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.
+
+-   Small amounts of text in the margins of a page, such as watermarks, page numbers, or digital stamps, will no longer prevent the rest of a page from being OCRed when ``--skip-text`` is issued. This behavior is based on a heuristic.
+
+-   Removed features
+
+    +   The deprecated ``--pdf-renderer tesseract`` PDF renderer was removed.
+
+    +   ``-g``, the option to generate debug text pages, was removed because it was a maintenance burden and only worked in isolated cases. HOCR pages can still be previewed by running the hocrtransform.py with appropriate settings.
+
+-   Removed dependencies
+
+    +   ``PyPDF2``
+
+    +   ``defusedxml``
+
+    +   ``PyMuPDF``
+
+-   The ``sandwich`` PDF renderer can be used with all supported versions of Tesseract, including that those prior to v3.05 which don't support ``-c textonly``. (Tesseract v4.0.0 is recommended and more efficient.)
+
+-   ``--pdf-renderer auto`` option and the diagnostics used to select a PDF renderer now work better with old versions, but may make different decisions than past versions.
+
+-   If everything succeeds but PDF/A conversion fails, a distinct return code is now returned (``ExitCode.pdfa_conversion_failed (10)``) where this situation previously returned ``ExitCode.invalid_output_pdf (4)``. The latter is now returned only if there is some indication that the output file is invalid.
+
+-   Notes for downstream packagers
+
+    +   There is also a new dependency on ``python-xmp-toolkit`` which in turn depends on ``libexempi3``.
+
+    +   It may be necessary to separately ``pip install pycparser`` to avoid `another Python 3.7 issue <https://github.com/eliben/pycparser/pull/135>`_.
 
 v6.2.5
 ------
@@ -19,7 +305,6 @@ v6.2.5
 
 -   This release branch no longer supports macOS. macOS users should upgrade to v7.x.
 
-
 v6.2.4
 ------
 
@@ -27,12 +312,11 @@ v6.2.4
 -   Backport blacklisting Ghostscript 9.24
 -   Older versions of Ghostscript are still supported
 
-
 v6.2.3
 ------
 
 -   Fix compatibility with img2pdf >= 0.3.0 by rejecting input images that have an alpha channel
-
+-   This version will be included in Ubuntu 18.10
 
 v6.2.2
 ------
@@ -46,7 +330,6 @@ v6.2.1
 
 -   Fix recent versions of Tesseract (after 4.0.0-beta1) not being detected as supporting the ``sandwich`` renderer (`#271 <https://github.com/ppjbarlow83/OCRmyPDF/issues/271>`_).
 
-
 v6.2.0
 ------
 
@@ -54,7 +337,7 @@ v6.2.0
 
 -   Creation of PDF/A-3 is now supported. However, there is no ability to attach files to PDF/A-3.
 
--   List more reasons why the file size might grow.
+-   Lists more reasons why the file size might grow.
 
 -   Fix issue `#262 <https://github.com/ppjbarlow83/OCRmyPDF/issues/262>`_, ``--remove-background`` error on PDFs contained colormapped (paletted) images.
 
@@ -115,7 +398,7 @@ v6.1.1
 v6.1.0
 ------
 
--   PyMuPDF is now an optional but recommended dependency, to alleviate installation difficulties on platforms that have less access to PyMuPDF than the author anticipated.  Install OCRmyPDF with ``pip install ocrmypdf[fitz]`` to use it to its full potential.
+-   PyMuPDF is now an optional but recommended dependency, to alleviate installation difficulties on platforms that have less access to PyMuPDF than the author anticipated.  (For version 6.x only) install OCRmyPDF with ``pip install ocrmypdf[fitz]`` to use it to its full potential.
 
 -   Fix ``FileExistsError`` that could occur if OCR timed out while it was generating the output file. (`#218 <https://github.com/jbarlow83/OCRmyPDF/issues/218>`_)
 
@@ -157,7 +440,7 @@ v6.0.0
 v5.7.0
 ------
 
--   Fixed an issue that caused poor CPU utilization on machines more than 4 cores when running Tesseract 4. (Related to issue `#217 <https://github.com/jbarlow83/OCRmyPDF/issues/217>`_.)
+-   Fixed an issue that caused poor CPU utilization on machines with more than 4 cores when running Tesseract 4. (Related to issue `#217 <https://github.com/jbarlow83/OCRmyPDF/issues/217>`_.)
 
 -   The 'hocr' renderer has been improved. The 'sandwich' and 'tesseract' renderers are still better for most use cases, but 'hocr' may be useful for people who work with the PDF.js renderer in English/ASCII languages. (`#225 <https://github.com/jbarlow83/OCRmyPDF/issues/225>`_)
 
@@ -296,7 +579,6 @@ v5.0.1
 
 -   Fixed issue `#169 <https://github.com/jbarlow83/OCRmyPDF/issues/169>`_, exception due to failure to create sidecar text files on some versions of Tesseract 3.04, including the jbarlow83/ocrmypdf Docker image
 
-
 v5.0
 ----
 
@@ -312,7 +594,6 @@ v5.0
 -   Fixed issue `#164 <https://github.com/jbarlow83/OCRmyPDF/issues/164>`_, a typo
 -   Removed the command line parameters ``-n`` and ``--just-print`` since they have not worked for some time (reported as Ubuntu bug `#1687308 <https://bugs.launchpad.net/ubuntu/+source/ocrmypdf/+bug/1687308>`_)
 
-
 v4.5.6
 ------
 
@@ -320,21 +601,18 @@ v4.5.6
 -   Fixed issue `#158 <https://github.com/jbarlow83/OCRmyPDF/issues/158>`_, ocrmypdf now stops and terminates if Ghostscript fails on an intermediate step, as it is not possible to proceed.
 -   Fixed issue `#160 <https://github.com/jbarlow83/OCRmyPDF/issues/160>`_, exception thrown on certain invalid arguments instead of error message
 
-
 v4.5.5
 ------
 
 -   Automated update of macOS homebrew tap
 -   Fixed issue `#154 <https://github.com/jbarlow83/OCRmyPDF/issues/154>`_, KeyError '/Contents' when searching for text on blank pages that have no /Contents record.  Note: incomplete fix for this issue.
 
-
 v4.5.4
 ------
 
 -   Fix ``--skip-big`` raising an exception if a page contains no images (`#152 <https://github.com/jbarlow83/OCRmyPDF/issues/152>`_) (thanks to @TomRaz)
 -   Fix an issue where pages with no images might trigger "cannot write mode P as JPEG" (`#151 <https://github.com/jbarlow83/OCRmyPDF/issues/151>`_)
 
-
 v4.5.3
 ------
 
@@ -343,7 +621,6 @@ v4.5.3
 -   Fixed an issue where the ``tess4`` renderer would duplicate content onto output pages if tesseract failed or timed out
 -   Fixed ``tess4`` renderer not recognized when lossless reconstruction is possible
 
-
 v4.5.2
 ------
 
@@ -351,21 +628,18 @@ v4.5.2
 -   Make "using Tesseract 4.0" warning less ominous
 -   Set up machinery for homebrew OCRmyPDF tap
 
-
 v4.5.1
 ------
 
 -   Fix issue `#137 <https://github.com/jbarlow83/OCRmyPDF/issues/137>`_, proportions of images with a non-square pixel aspect ratio would be distorted in output for ``--force-ocr`` and some other combinations of flags
 
-
 v4.5
 ----
 
--   Exotic PDFs containing "Form XObjects" are now supported (issue `#134 <https://github.com/jbarlow83/OCRmyPDF/issues/134>`_; PDF reference manual 8.10), and images they contain are taken into account when determining the resolution for rasterizing
+-   PDFs containing "Form XObjects" are now supported (issue `#134 <https://github.com/jbarlow83/OCRmyPDF/issues/134>`_; PDF reference manual 8.10), and images they contain are taken into account when determining the resolution for rasterizing
 -   The Tesseract 4 Docker image no longer includes all languages, because it took so long to build something would tend to fail
 -   OCRmyPDF now warns about using ``--pdf-renderer tesseract`` with Tesseract 3.04 or lower due to issues with Ghostscript corrupting the OCR text in these cases
 
-
 v4.4.2
 ------
 
@@ -376,7 +650,6 @@ v4.4.2
 
 -   OCRmyPDF now prevents running the Tesseract 4 renderer with Tesseract 3.04, which was permitted in v4.4 and v4.4.1 but will not work
 
-
 v4.4.1
 ------
 
@@ -403,32 +676,27 @@ v4.4
 
     + However, OCRmyPDF's dependency "ruffus" is not re-entrant, so no Python API is available. Scripts should continue to use the command line interface.
 
-
 v4.3.5
 ------
 
 -   Update documentation to confirm Python 3.6.0 compatibility. No code changes were needed, so many earlier versions are likely supported.
 
-
 v4.3.4
 ------
 
 -   Fixed "decimal.InvalidOperation: quantize result has too many digits" for high DPI images
 
-
 v4.3.3
 ------
 
 -   Fixed PDF/A creation with Ghostscript 9.20 properly
 -   Fixed an exception on inline stencil masks with a missing optional parameter
 
-
 v4.3.2
 ------
 
 -   Fixed a PDF/A creation issue with Ghostscript 9.20 (note: this fix did not actually work)
 
-
 v4.3.1
 ------
 
@@ -439,7 +707,6 @@ v4.3.1
 -   Cleaned up debug output
 -   Tesseract caching in test cases is now more cautious about false cache hits and reproducing exact output, not that any problems were observed
 
-
 v4.3
 ----
 
@@ -457,7 +724,6 @@ v4.2.5
 -   Fixed an issue (`#100 <https://github.com/jbarlow83/OCRmyPDF/issues/100>`_) with PDFs that omit the optional /BitsPerComponent parameter on images
 -   Removed non-free file milk.pdf
 
-
 v4.2.4
 ------
 
@@ -477,20 +743,17 @@ v4.2.3
 -   Added a test case for indirect objects and linearized PDFs
 -   Deprecated the OCRmyPDF.sh shell script
 
-
 v4.2.2
 ------
 
 -   Improvements to documentation
 
-
 v4.2.1
 ------
 
 -   Fixed an issue where PDF pages that contained stencil masks would report an incorrect DPI and cause Ghostscript to abort
 -   Implemented stdin streaming
 
-
 v4.2
 ----
 
@@ -529,7 +792,6 @@ v4.1.4
 
 -   Bug fix: monochrome images with an ICC profile attached were incorrectly converted to full color images if lossless reconstruction was not possible due to other settings; consequence was increased file size for these images
 
-
 v4.1.3
 ------
 
@@ -538,14 +800,12 @@ v4.1.3
 -   Fix order of operations for matrix multiplication (no effect on most users)
 -   Add a few leptonica wrapper functions (no effect on most users)
 
-
 v4.1.2
 ------
 
 -   Replace IEC sRGB ICC profile with Debian's sRGB (from icc-profiles-free) which is more compatible with the MIT license
 -   More helpful error message for an error related to certain types of malformed PDFs
 
-
 v4.1
 ----
 
@@ -553,20 +813,17 @@ v4.1
 -   Fixed problems in error checking if ``unpaper`` is uninstalled or missing at run-time
 -   Fixed problems with "RethrownJobError" errors during error handling that suppressed the useful error messages
 
-
 v4.0.7
 ------
 
 -   Minor correction to Ghostscript output settings
 
-
 v4.0.6
 ------
 
 -   Update install instructions
 -   Provide a sRGB profile instead of using Ghostscript's
 
-
 v4.0.5
 ------
 
@@ -580,18 +837,14 @@ v4.0.4
 
 Released with verbose debug message turned on. Do not use. Skip to v4.0.5.
 
-
 v4.0.3
 ------
 
 New features
-^^^^^^^^^^^^
 
 -   Page orientations detected are now reported in a summary comment
 
-
 Fixes
-^^^^^
 
 -   Show stack trace if unexpected errors occur
 -   Treat "too few characters" error message from Tesseract as a reason to skip that page rather than
@@ -603,7 +856,7 @@ v4.0.2
 ------
 
 Fixes
-^^^^^
+
 
 -   Fixed compatibility with Tesseract 3.04.01 release, particularly its different way of outputting
     orientation information
@@ -615,7 +868,7 @@ v4.0.1
 ------
 
 Fixes
-^^^^^
+
 
 -   Fixed a KeyError if tesseract fails to find page orientation information
 
@@ -624,7 +877,6 @@ v4.0
 ----
 
 New features
-^^^^^^^^^^^^
 
 -   Automatic page rotation (``-r``) is now available. It uses ignores any prior rotation information
     on PDFs and sets rotation based on the dominant orientation of detectable text. This feature is
@@ -634,7 +886,6 @@ New features
 
 
 Fixes
-^^^^^
 
 -   Fixed an issue where lossless reconstruction could cause some pages to be appear incorrectly
     if the page was rotated by the user in Acrobat after being scanned (specifically if it a /Rotate tag)
@@ -643,7 +894,6 @@ Fixes
 
 
 Changes
-^^^^^^^
 
 -   Logging output is now much easier to read
 -   ``--deskew`` is now performed by Leptonica instead of unpaper (`#25 <https://github.com/jbarlow83/OCRmyPDF/issues/25>`_)
@@ -659,7 +909,6 @@ v3.2.1
 ------
 
 Changes
-^^^^^^^
 
 -   Fixed issue `#47 <https://github.com/jbarlow83/OCRmyPDF/issues/47>`_ "convert() got and unexpected keyword argument 'dpi'" by upgrading to img2pdf 0.2
 -   Tweaked the Dockerfiles
@@ -669,7 +918,6 @@ v3.2
 ----
 
 New features
-^^^^^^^^^^^^
 
 -   Lossless reconstruction: when possible, OCRmyPDF will inject text layers without
     otherwise manipulating the content and layout of a PDF page. For example, a PDF containing a mix
@@ -681,7 +929,6 @@ New features
     for the polyglots among us. It is much larger.
 
 Changes
-^^^^^^^
 
 -   JPEG transcoding quality is now 95 instead of the default 75. Bigger file sizes for less degradation.
 
@@ -691,7 +938,6 @@ v3.1.1
 ------
 
 Changes
-^^^^^^^
 
 -   Fixed bug that caused incorrect page size and DPI calculations on documents with mixed page sizes
 
@@ -699,7 +945,6 @@ v3.1
 ----
 
 Changes
-^^^^^^^
 
 -   Default output format is now PDF/A-2b instead of PDF/A-1b
 -   Python 3.5 and macOS El Capitan are now supported platforms - no changes were
@@ -717,7 +962,6 @@ v3.0
 ----
 
 New features
-^^^^^^^^^^^^
 
 -   Easier installation with a Docker container or Python's ``pip`` package manager
 -   Eliminated many external dependencies, so it's easier to setup
@@ -740,7 +984,6 @@ New features
 -   Multiple images on the same PDF page are now supported
 
 Changes
-^^^^^^^
 
 -   New, robust rewrite in Python 3.4+ with ruffus_ pipelines
 -   Now uses Ghostscript 9.14's improved color conversion model to preserve PDF colors
@@ -777,8 +1020,7 @@ Changes
 .. _Unpaper: https://github.com/Flameeyes/unpaper
 .. _JHOVE: http://jhove.sourceforge.net/
 
-Release candidates
-^^^^^^^^^^^^^^^^^^
+Release candidates^
 
 -   rc9:
 
@@ -846,12 +1088,11 @@ where ``settings.txt`` contains *one argument per line*, for example:
 
 
 Fixes
-^^^^^
+
 
 -   Handling of filenames containing spaces: fixed
 
 Notes and known issues
-^^^^^^^^^^^^^^^^^^^^^^
 
 -   Some dependencies may work with lower versions than tested, so try
     overriding dependencies if they are "in the way" to see if they work.

docs/security.rst

@@ -1,7 +1,7 @@
 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.
 
@@ -19,20 +19,20 @@ This `article <https://theinvisiblethings.blogspot.ca/2013/02/converting-untrust
 How OCRmyPDF processes PDFs
 ---------------------------
 
-OCRmyPDF must open and interpret your PDF in order to insert an OCR layer. First, it runs all PDFs through `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.
-
-After qpdf, OCRmyPDF examines each page using `PyPDF2 <https://github.com/mstamy2/PyPDF2>`_. This library also has no warranties or guarantees. OCRmyPDF works with qpdf 5.0 and up, but version 7.0 is recommended because of known security vulnerabilities in early versions.
+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.
 
 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:
 
@@ -60,13 +60,13 @@ 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. 
+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.
 
@@ -76,4 +76,4 @@ After OCR is applied, password protection is not permitted on PDF/A documents bu
 
 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/completions.fish

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

pyproject.toml

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

requirements.txt

@@ -1,11 +0,0 @@
-# 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 == 2.7.0
-Pillow == 5.2.0
-reportlab == 3.4.0
-PyPDF2 == 1.26.0
-img2pdf == 0.2.4
-cffi == 1.11.5
-PyMuPDF == 1.12.5
-defusedxml == 0.5.0

requirements/dev.txt

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

requirements/main.txt

@@ -0,0 +1,13 @@
+# requirements.txt can be used to replicate the developer's build environment
+# setup.py lists a separate set of requirements that are looser to simplify
+# installation
+chardet == 3.0.4
+cffi == 1.12.2
+img2pdf == 0.3.3
+pdfminer.six == 20181108
+pikepdf == 1.1.0
+Pillow >= 5.0.0, != 5.1.0 ; sys_platform == "darwin"
+pycparser == 2.19
+python-xmp-toolkit == 2.0.1
+reportlab == 3.5.13
+ruffus == 2.8.1

requirements/test.txt

@@ -0,0 +1,8 @@
+pytest == 4.3.0
+pytest-helpers-namespace >= 2019.1.8
+pytest-xdist
+pytest-cov >= 2.6.1
+python-xmp-toolkit   # requires apt-get install libexempi3
+                     # or brew install exempi
+PyPDF2 >= 1.26.0
+#PyMuPDF == 1.13.4    # optional

setup.cfg

@@ -11,6 +11,15 @@ ignore =
 [tool:pytest]
 norecursedirs = lib .pc .git output cache resources
 testpaths = tests
+filterwarnings =
+	ignore:.*XMLParser.*:DeprecationWarning
+
+[isort]
+multi_line_output=3
+include_trailing_comma=True
+force_grid_wrap=0
+use_parentheses=True
+line_length=88
 
 [metadata]
-license_file = LICENSE
+license_file = LICENSE

setup.py

@@ -20,192 +20,48 @@
 from __future__ import print_function, unicode_literals
 
 import sys
-if sys.version_info < (3, 5):
-    print("Python 3.5 or newer is required", file=sys.stderr)
+
+if sys.version_info < (3, 6):
+    print("Python 3.6 or newer is required", file=sys.stderr)
     sys.exit(1)
 
-from setuptools import setup, find_packages  # nopep8
-from subprocess import STDOUT, check_output, CalledProcessError  # nopep8
-from collections.abc import Mapping  # nopep8
-import re  # nopep8
+from setuptools import setup, find_packages
+from subprocess import STDOUT, check_output, CalledProcessError
+from collections.abc import Mapping
+import re
 
-
-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='7.0.0', # test suite known to fail on 5.1.1
-        package='qpdf',
-        version_check_args=['--version']
-    )
-
 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',
@@ -213,7 +69,6 @@ setup(
     package_dir={'': 'src'},
     keywords=['PDF', 'OCR', 'optical character recognition', 'PDF/A', 'scanning'],
     classifiers=[
-        "Programming Language :: Python :: 3.5",
         "Programming Language :: Python :: 3.6",
         "Programming Language :: Python :: 3.7",
         "Development Status :: 5 - Production/Stable",
@@ -229,37 +84,37 @@ setup(
         "Topic :: Scientific/Engineering :: Image Recognition",
         "Topic :: Text Processing :: Indexing",
         "Topic :: Text Processing :: Linguistic",
-        ],
-    python_requires=' >= 3.5',
-    setup_requires=[
-        'cffi >= 1.9.1',        # to build the leptonica module
-        'pytest-runner',        # to enable python setup.py test
-        'setuptools_scm',       # so that version will work
+    ],
+    python_requires=' >= 3.6',
+    setup_requires=[  # can be removed whenever we can drop pip 9 support
+        'cffi >= 1.9.1',  # to build the leptonica module
+        'pytest-runner',  # to enable python setup.py test
+        'setuptools_scm',  # so that version will work
+        'setuptools_scm_git_archive',  # enable version from github tarballs
     ],
     use_scm_version={'version_scheme': 'post-release'},
-    cffi_modules=[
-        'src/ocrmypdf/lib/compile_leptonica.py:ffibuilder'
-    ],
+    cffi_modules=['src/ocrmypdf/lib/compile_leptonica.py:ffibuilder'],
     install_requires=[
-        'cffi >= 1.9.1',          # must be a setup and install requirement
-        'defusedxml >= 0.5.0',    # pure Python, so track HEAD closely
-        'img2pdf >= 0.2.4',       # pure Python, so track HEAD closely
+        'chardet >= 3.0.4, < 4',  # unlisted requirement of pdfminer.six 20181108
+        'cffi >= 1.9.1',  # must be a setup and install requirement
+        'img2pdf >= 0.3.0, < 0.4',  # pure Python, so track HEAD closely
+        'pdfminer.six == 20181108 ; sys_platform != "darwin"',
+        'pikepdf >= 1.1.0, < 2',
         'Pillow >= 4.0.0, != 5.1.0 ; sys_platform == "darwin"',
-                                  # Pillow < 4 has BytesIO/TIFF bug w/img2pdf 0.2.3
-                                  # block 5.1.0, broken wheels
-        'PyPDF2 >= 1.26',         # pure Python, so track HEAD closely
-        'reportlab >= 3.3.0',     # oldest released version with sane image handling
+        # Pillow < 4 has BytesIO/TIFF bug w/img2pdf 0.2.3
+        # block 5.1.0, broken wheels
+        'reportlab >= 3.3.0',  # oldest released version with sane image handling
         'ruffus >= 2.7.0',
     ],
-    extras_require={
-        'fitz': ['PyMuPDF >= 1.12.5']     # for table of contents bug
-    },
+    extras_require={'pdfminer': ['pdfminer.six == 20181108']},
     tests_require=tests_require,
-    entry_points={
-        'console_scripts': [
-            'ocrmypdf = ocrmypdf.__main__:run_pipeline'
-        ],
-    },
+    entry_points={'console_scripts': ['ocrmypdf = ocrmypdf.__main__:run_pipeline']},
     package_data={'ocrmypdf': ['data/sRGB.icc']},
     include_package_data=True,
-    zip_safe=False)
+    zip_safe=False,
+    project_urls={
+        'Documentation': 'https://ocrmypdf.readthedocs.io/',
+        'Source': 'https://github.com/jbarlow83/ocrmypdf',
+        'Tracker': 'https://github.com/jbarlow83/ocrmypdf/issues',
+    },
+)

src/ocrmypdf/__init__.py

@@ -23,3 +23,24 @@ PROGRAM_NAME = 'ocrmypdf'
 __version__ = pkg_resources.get_distribution('ocrmypdf').version
 
 VERSION = __version__
+
+from .exceptions import (
+    ExitCode,
+    BadArgsError,
+    PdfMergeFailedError,
+    MissingDependencyError,
+    UnsupportedImageFormatError,
+    DpiError,
+    OutputFileAccessError,
+    PriorOcrFoundError,
+    InputFileError,
+    SubprocessOutputError,
+    EncryptedPdfError,
+    TesseractConfigError,
+)
+
+from . import helpers
+from . import hocrtransform
+from . import leptonica
+from . import pdfa
+from . import pdfinfo

src/ocrmypdf/_jobcontext.py

@@ -0,0 +1,83 @@
+# © 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 shutil
+import sys
+from contextlib import suppress
+from multiprocessing.managers import SyncManager
+
+from .pdfinfo import PdfInfo
+
+
+class JobContext:
+    """Holds our context for a particular run of the pipeline
+
+    A multiprocessing manager effectively creates a separate process
+    that keeps the master job context object.  Other threads access
+    job context via multiprocessing proxy objects.
+
+    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 __init__(self):
+        self.pdfinfo = None
+        self.options = None
+        self.work_folder = None
+        self.rotations = {}
+
+    def generate_pdfinfo(self, infile):
+        self.pdfinfo = PdfInfo(infile)
+
+    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 cleanup_working_files(work_folder, options):
+    if options.keep_temporary_files:
+        print(f"Temporary working files saved at:\n{work_folder}", file=sys.stderr)
+    else:
+        with suppress(FileNotFoundError):
+            shutil.rmtree(work_folder)

src/ocrmypdf/_unicodefun.py

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

src/ocrmypdf/_weave.py

@@ -0,0 +1,419 @@
+# © 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 itertools import groupby
+from pathlib import Path
+import os
+
+import pikepdf
+
+from .exec import tesseract
+from .helpers import flatten_groups, page_number
+
+
+MAX_OPEN_PAGE_PDFS = int(os.environ.get('_OCRMYPDF_MAX_OPEN_PAGE_PDFS', 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, log):
+    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 _weave_layers_graft(
+    *, 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()
+
+    if not tesseract.has_textonly_pdf():
+        # If we don't have textonly_pdf, edit the stream to delete the
+        # instruction to draw the image Tesseract generated, which we do not
+        # use.
+        stream = bytearray(pdf_text_contents)
+        pattern = b'/Im1 Do'
+        idx = stream.find(pattern)
+        stream[idx : (idx + len(pattern))] = b' ' * len(pattern)
+        pdf_text_contents = bytes(stream)
+
+    base_page = pdf_base.pages.p(page_num)
+
+    # The text page always will be oriented up by this stage but the original
+    # content may have a rotation applied. Wrap the text stream with a rotation
+    # so it will be oriented the same way as the rest of the page content.
+    # (Previous versions OCRmyPDF rotated the content layer to match the text.)
+    mediabox = [float(pdf_text.pages[0].MediaBox[v]) for v in range(4)]
+    wt, ht = mediabox[2] - mediabox[0], mediabox[3] - mediabox[1]
+
+    mediabox = [float(base_page.MediaBox[v]) for v in range(4)]
+    wp, hp = mediabox[2] - mediabox[0], mediabox[3] - mediabox[1]
+
+    translate = pikepdf.PdfMatrix().translated(-wt / 2, -ht / 2)
+    untranslate = pikepdf.PdfMatrix().translated(wp / 2, hp / 2)
+    # -rotation because the input is a clockwise angle and this formula
+    # uses CCW
+    rotation = -rotation % 360
+    rotate = pikepdf.PdfMatrix().rotated(rotation)
+
+    # Because of rounding of DPI, we might get a text layer that is not
+    # identically sized to the target page. Scale to adjust. Normally this
+    # is within 0.998.
+    if rotation in (90, 270):
+        wt, ht = ht, wt
+    scale_x = wp / wt
+    scale_y = hp / ht
+
+    log.debug('%r', (scale_x, scale_y))
+    scale = pikepdf.PdfMatrix().scaled(scale_x, scale_y)
+
+    # Translate the text so it is centered at (0, 0), rotate it there, adjust
+    # for a size different between initial and text PDF, then untranslate
+    ctm = translate @ rotate @ scale @ untranslate
+
+    pdf_text_contents = b'q %s cm\n' % ctm.encode() + pdf_text_contents + b'\nQ\n'
+
+    new_text_layer = pikepdf.Stream(pdf_base, pdf_text_contents)
+
+    if strip_old_text:
+        strip_invisible_text(pdf_base, base_page, log)
+
+    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 _traverse_toc(pdf_base, visitor_fn, log):
+    """
+    Walk the table of contents, calling visitor_fn() at each node
+
+    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.
+
+    """
+
+    visited = set()
+    queue = set()
+    link_keys = ('/Parent', '/First', '/Last', '/Prev', '/Next')
+
+    if not '/Outlines' in pdf_base.root:
+        return
+
+    queue.add(pdf_base.root.Outlines.objgen)
+    while queue:
+        objgen = queue.pop()
+        visited.add(objgen)
+        node = pdf_base.get_object(objgen)
+        log.debug('fix toc: exploring outline entries at %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:
+                # Direct references are not allowed here, but it's not clear
+                # what we should do if we find any. Removing them is an option:
+                # node[key] = pdf_base.make_indirect(None)
+                continue
+            objgen = item.objgen
+            if objgen not in visited:
+                queue.add(objgen)
+
+        if visitor_fn:
+            visitor_fn(pdf_base, node, log)
+
+
+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.)
+
+    It may ultimately be better to find a way to rebuild a page in place.
+
+    """
+
+    if not pageref_remap:
+        return
+
+    def remap_dest(dest_node):
+        """
+        Inner helper function: change the objgen for any page from the old we
+        invalidated to its new one.
+        """
+        try:
+            pageref = dest_node[0]
+            if pageref['/Type'] == '/Page' and pageref.objgen in pageref_remap:
+                new_objgen = pageref_remap[pageref.objgen]
+                dest_node[0] = pdf_base.get_object(new_objgen)
+        except (IndexError, TypeError) as e:
+            log.warning("This file may contain invalid table of contents entries")
+            log.debug(e)
+
+    def visit_remap_dest(pdf_base, node, log):
+        """
+        Visitor function to fix ToC entries
+
+        Test for the two types of references to pages that can occur in ToCs.
+        Both types have the same final format (an indirect reference to the
+        target page).
+        """
+        if '/Dest' in node:
+            # /Dest reference to another page (old method)
+            remap_dest(node['/Dest'])
+        elif '/A' in node:
+            # /A (action) command set to "GoTo" (newer method)
+            if '/S' in node['/A'] and node['/A']['/S'] == '/GoTo':
+                remap_dest(node['/A']['/D'])
+
+    _traverse_toc(pdf_base, visit_remap_dest, log)
+
+
+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 = {}
+
+    # Walk the table of contents first, to trigger pikepdf/qpdf to resolve all
+    # page references in the table of contents. Some PDF generators put invalid
+    # references in the ToC, so we want to resolve them to null before we
+    # create any references, or the ToC will be corrupted
+    _traverse_toc(pdf_base, None, log)
+
+    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
+            strip_old = context.get_options().redo_ocr
+            _weave_layers_graft(
+                pdf_base=pdf_base,
+                page_num=page_num,
+                text=text,
+                font=font,
+                font_key=font_key,
+                rotation=text_misaligned,
+                procset=procset,
+                strip_old_text=strip_old,
+                log=log,
+            )
+
+        # Correct the rotation if applicable
+        pdf_base.pages[page_num - 1].Rotate = (
+            content_rotation - autorotate_correction
+        ) % 360
+
+        if len(keep_open) > MAX_OPEN_PAGE_PDFS:
+            # 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 + f'_working{page_num}.pdf'
+            pdf_base.save(interim)
+            del pdf_base
+            keep_open = []
+
+            pdf_base = pikepdf.open(interim)
+            procset = pdf_base.pages[0].Resources.ProcSet
+            font, font_key = None, None  # Reacquire this information
+
+    _fix_toc(pdf_base, pagerefs, log)
+    pdf_base.save(output_file)

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

@@ -21,40 +21,153 @@ import os
 import re
 import sys
 from subprocess import run, STDOUT, PIPE, CalledProcessError
-from ..exceptions import MissingDependencyError
+from ..exceptions import MissingDependencyError, ExitCode
+from collections.abc import Mapping
 
 
-
-def get_version(program, *,
-        version_arg='--version', regex=r'(\d+(\.\d+)*)'):
+def get_version(program, *, version_arg='--version', regex=r'(\d+(\.\d+)*)'):
     "Get the version of the specified program"
-    args_prog = [
-        program,
-        version_arg
-    ]
+    args_prog = [program, version_arg]
     try:
         proc = run(
-            args_prog, close_fds=True, universal_newlines=True,
-            stdout=PIPE, stderr=STDOUT, check=True)
+            args_prog,
+            close_fds=True,
+            universal_newlines=True,
+            stdout=PIPE,
+            stderr=STDOUT,
+            check=True,
+        )
         output = proc.stdout
     except FileNotFoundError as e:
         raise MissingDependencyError(
-            "Could not find program '{}' on the PATH".format(
-                program)) from e
+            f"Could not find program '{program}' on the PATH"
+        ) from e
     except CalledProcessError as e:
-        if e.returncode < 0:
+        if e.returncode != 0:
             raise MissingDependencyError(
-                "Ran program '{}' but it exited with an error:\n{}".format(
-                    program, e.output)) from e
+                f"Ran program '{program}' but it exited with an error:\n{e.output}"
+            ) from e
         raise MissingDependencyError(
-            "Could not find program '{}' on the PATH".format(
-                program)) from e
+            f"Could not find program '{program}' on the PATH"
+        ) from e
     try:
         version = re.match(regex, output.strip()).group(1)
     except AttributeError as e:
         raise MissingDependencyError(
-            ("The program '{}' did not report its version. "
-            "Message was:\n{}").format(program, output)
+            f"The program '{program}' did not report its version. "
+            f"Message was:\n{output}"
         )
 
     return version
+
+
+missing_program = '''
+The program '{program}' could not be executed or was not found on your
+system PATH.
+'''
+
+missing_optional_program = '''
+The program '{program}' could not be executed or was not found on your
+system PATH.  This program is required when you use the
+{required_for} arguments.  You could try omitting these arguments, or install
+the package.
+'''
+
+missing_recommend_program = '''
+The program '{program}' could not be executed or was not found on your
+system PATH.  This program is recommended when using the {required_for} arguments,
+but not required, so we will proceed.  For best results, install the program.
+'''
+
+old_version = '''
+OCRmyPDF requires '{program}' {need_version} or higher.  Your system appears
+to have {found_version}.  Please update this program.
+'''
+
+old_version_required_for = '''
+OCRmyPDF requires '{program}' {need_version} or higher when run with the
+{required_for} arguments.  If you omit these arguments, OCRmyPDF may be able to
+proceed.  For best results, install the program.
+'''
+
+osx_install_advice = '''
+If you have homebrew installed, try these command to install the missing
+package:
+    brew install {package}
+'''
+
+linux_install_advice = '''
+On systems with the aptitude package manager (Debian, Ubuntu), try these
+commands:
+    sudo apt-get update
+    sudo apt-get install {package}
+
+On RPM-based systems (Red Hat, Fedora), search for instructions on
+installing the RPM for {program}.
+'''
+
+
+def _get_platform():
+    if sys.platform.startswith('freebsd'):
+        return 'freebsd'
+    elif sys.platform.startswith('linux'):
+        return 'linux'
+    return sys.platform
+
+
+def _error_trailer(log, program, package, **kwargs):
+    if isinstance(package, Mapping):
+        package = package[_get_platform()]
+
+    if _get_platform() == 'darwin':
+        log.info(osx_install_advice.format(**locals()))
+    elif _get_platform() == 'linux':
+        log.info(linux_install_advice.format(**locals()))
+
+
+def _error_missing_program(log, 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(
+    log, 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(
+    *,
+    log,
+    program,
+    package,
+    version_checker,
+    need_version,
+    required_for=None,
+    recommended=False,
+):
+    try:
+        found_version = version_checker()
+    except (CalledProcessError, FileNotFoundError, MissingDependencyError):
+        _error_missing_program(log, program, package, required_for, recommended)
+        if not recommended:
+            sys.exit(ExitCode.missing_dependency)
+        return
+
+    if found_version < need_version:
+        _error_old_version(
+            log, program, package, need_version, found_version, required_for
+        )
+        if not recommended:
+            sys.exit(ExitCode.missing_dependency)
+
+    log.debug(f'Found {program} {found_version}')

src/ocrmypdf/exec/ghostscript.py

@@ -15,16 +15,17 @@
 # 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
 import re
-import sys
+from functools import lru_cache
+from os import fspath
+from shutil import copy
+from subprocess import PIPE, STDOUT, run
+from tempfile import NamedTemporaryFile
+
 from PIL import Image
+
 from . import get_version
-from ..exceptions import SubprocessOutputError, MissingDependencyError
-from ..helpers import fspath
+from ..exceptions import SubprocessOutputError
 
 
 @lru_cache(maxsize=1)
@@ -33,8 +34,7 @@ def version():
 
 
 def jpeg_passthrough_available():
-    """
-    Returns True if the installed version of Ghostscript supports JPEG passthru
+    """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
@@ -46,20 +46,72 @@ def jpeg_passthrough_available():
     the feature available. (However, we don't use 9.24 at all, so the first
     version that allows JPEG passthrough is 9.25.
 
-    Regardless, in ocrmypdf 6.x we are ignoring this new feature entirely to
-    avoid new behavior.
     """
-    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):
+def extract_text(input_file, pageno=1):
+    """Use the txtwrite device to get text layout information out
+
+    For details on options of -dTextFormat see
+    https://www.ghostscript.com/doc/current/VectorDevices.htm#TXT
+
+    Format is like
+    <page>
+    <line>
+    <span bbox="left top right bottom" font="..." size="...">
+    <char bbox="...." c="X"/>
+
+    :param pageno: number of page to extract, or all pages if None
+    :return: XML-ish text representation in bytes
     """
-    Rasterize one page of a PDF at resolution (xres, yres) in canvas units.
+
+    if pageno is not None:
+        pages = ['-dFirstPage=%i' % pageno, '-dLastPage=%i' % pageno]
+    else:
+        pages = []
+
+    args_gs = (
+        [
+            'gs',
+            '-dQUIET',
+            '-dSAFER',
+            '-dBATCH',
+            '-dNOPAUSE',
+            '-sDEVICE=txtwrite',
+            '-dTextFormat=0',
+        ]
+        + pages
+        + ['-o', '-', fspath(input_file)]
+    )
+
+    p = run(args_gs, stdout=PIPE, stderr=PIPE)
+    if p.returncode != 0:
+        raise SubprocessOutputError(
+            'Ghostscript text extraction failed\n%s\n%s\n%s'
+            % (input_file, p.stdout.decode(), p.stderr.decode())
+        )
+
+    return p.stdout
+
+
+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
@@ -73,29 +125,40 @@ def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
     :param log:
     :param pageno: page number to rasterize (beginning at page 1)
     :param page_dpi: resolution tuple (x, y) overriding output image DPI
+    :param rotation: 0, 90, 180, 270: clockwise angle to rotate page
+    :param filter_vector: if True, remove vector graphics objects
     :return:
     """
     res = xres, yres
     int_res = round(xres), round(yres)
     if not page_dpi:
         page_dpi = res
-    with NamedTemporaryFile(delete=True) as tmp:
-        args_gs = [
-            'gs',
-            '-dQUIET',
-            '-dSAFER',
-            '-dBATCH',
-            '-dNOPAUSE',
-            '-sDEVICE=%s' % raster_device,
-            '-dFirstPage=%i' % pageno,
-            '-dLastPage=%i' % pageno,
-            '-r{0}x{1}'.format(str(int_res[0]), str(int_res[1])),
-            '-o', tmp.name,
-            fspath(input_file)
-        ]
 
-        p = run(args_gs, stdout=PIPE, stderr=STDOUT,
-                universal_newlines=True)
+    with NamedTemporaryFile(delete=True) as tmp:
+        args_gs = (
+            [
+                'gs',
+                '-dQUIET',
+                '-dSAFER',
+                '-dBATCH',
+                '-dNOPAUSE',
+                f'-sDEVICE={raster_device}',
+                f'-dFirstPage={pageno}',
+                f'-dLastPage={pageno}',
+                f'-r{str(int_res[0])}x{str(int_res[1])}',
+            ]
+            + (['-dFILTERVECTOR'] if filter_vector else [])
+            + [
+                '-o',
+                tmp.name,
+                '-dAutoRotatePages=/None',  # Probably has no effect on raster
+                '-f',
+                fspath(input_file),
+            ]
+        )
+
+        log.debug(args_gs)
+        p = run(args_gs, stdout=PIPE, stderr=STDOUT, universal_newlines=True)
         if _gs_error_reported(p.stdout):
             log.error(p.stdout)
         else:
@@ -112,20 +175,57 @@ def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
 
         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])
+            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.resize(expected_size).save(
-                    fspath(output_file), dpi=page_dpi)
-            else:
-                copy(tmp.name, fspath(output_file))
+                    f"Ghostscript: resize output image {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)
 
 
-def generate_pdfa(pdf_pages, output_file, compression, log,
-                  threads=1, pdf_version='1.5', pdfa_part='2'):
+def generate_pdfa(
+    pdf_pages,
+    output_file,
+    compression,
+    log,
+    threads=1,
+    pdf_version='1.5',
+    pdfa_part='2',
+):
+    """Generate a PDF/A.
+
+    The pdf_pages, a list files, will be merged into output_file. One or more
+    PDF files may be merged. One of the files in this list must be a pdfmark
+    file that provides Ghostscript with details on how to perform the PDF/A
+    conversion. By default with we pick PDF/A-2b, but this works for 1 or 3.
+
+    compression can be 'jpeg', 'lossless', or an empty string. In 'jpeg',
+    Ghostscript is instructed to convert color and grayscale images to DCT
+    (JPEG encoding). In 'lossless' Ghostscript is told to convert images to
+    Flate (lossless/PNG). If the parameter is omitted Ghostscript is left to
+    make its own decisions about how to encode images; it appears to use a
+    heuristic to decide how to encode images. As of Ghostscript 9.25, we
+    support passthrough JPEG which allows Ghostscript to avoid transcoding
+    images entirely. (The feature was added in 9.23 but broken, and the 9.24
+    release of Ghostscript had regressions, so we don't support it until 9.25.)
+    """
     compression_args = []
     if compression == 'jpeg':
         compression_args = [
@@ -152,35 +252,39 @@ def generate_pdfa(pdf_pages, output_file, compression, log,
     # git commit fe1c025d.
     strategy = 'RGB' if version() >= '9.19' else '/RGB'
 
-    if version() >= '9.23':
+    if version() == '9.23':
         # 9.23: new feature JPEG passthrough is broken in some cases, best to
         # disable it always
         # https://bugs.ghostscript.com/show_bug.cgi?id=699216
-        # fixed in 9.24, but to avoid changing expected behavior we disable it
-        # for ocrmypdf 6.x
         compression_args.append('-dPassThroughJPEGImages=false')
 
     with NamedTemporaryFile(delete=True) as gs_pdf:
-        args_gs = [
-            "gs",
-            "-dQUIET",
-            "-dBATCH",
-            "-dNOPAUSE",
-            "-dCompatibilityLevel=" + str(pdf_version),
-            "-dNumRenderingThreads=" + str(threads),
-            "-sDEVICE=pdfwrite",
-            "-dAutoRotatePages=/None",
-            "-sColorConversionStrategy=" + strategy,
-            "-sProcessColorModel=DeviceRGB"
-        ] + compression_args + [
-            "-dJPEGQ=95",
-            "-dPDFA=" + pdfa_part,
-            "-dPDFACompatibilityPolicy=1",
-            "-sOutputFile=" + gs_pdf.name,
-        ]
-        args_gs.extend(pdf_pages)
-        p = run(args_gs, stdout=PIPE, stderr=STDOUT,
-                universal_newlines=True)
+        # nb no need to specify ProcessColorModel when ColorConversionStrategy
+        # is set; see:
+        # https://bugs.ghostscript.com/show_bug.cgi?id=699392
+        args_gs = (
+            [
+                "gs",
+                "-dQUIET",
+                "-dBATCH",
+                "-dNOPAUSE",
+                "-dCompatibilityLevel=" + str(pdf_version),
+                "-dNumRenderingThreads=" + str(threads),
+                "-sDEVICE=pdfwrite",
+                "-dAutoRotatePages=/None",
+                "-sColorConversionStrategy=" + strategy,
+            ]
+            + compression_args
+            + [
+                "-dJPEGQ=95",
+                "-dPDFA=" + pdfa_part,
+                "-dPDFACompatibilityPolicy=1",
+                "-sOutputFile=" + gs_pdf.name,
+            ]
+        )
+        args_gs.extend(fspath(s) for s in pdf_pages)  # Stringify Path objs
+        log.debug(args_gs)
+        p = run(args_gs, stdout=PIPE, stderr=STDOUT, universal_newlines=True)
 
         if _gs_error_reported(p.stdout):
             log.error(p.stdout)
@@ -191,14 +295,14 @@ def generate_pdfa(pdf_pages, output_file, compression, log,
             log.debug(
                 "Ghostscript had to remove PDF 'overprinting' from the "
                 "input file to complete PDF/A conversion. "
-                )
+            )
         else:
             log.debug(p.stdout)
 
         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)
+            copy(gs_pdf.name, fspath(output_file))
         else:
             log.error('Ghostscript PDF/A rendering failed')
             raise SubprocessOutputError()

src/ocrmypdf/exec/jbig2enc.py

@@ -0,0 +1,59 @@
+# © 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 functools import lru_cache
+from subprocess import PIPE, run
+
+from . import get_version
+from ..exceptions import MissingDependencyError
+
+
+@lru_cache(maxsize=1)
+def version():
+    return get_version('jbig2', regex=r'jbig2enc (\d+(\.\d+)*).*')
+
+
+def available():
+    try:
+        version()
+    except MissingDependencyError:
+        return False
+    return True
+
+
+def convert_group(*, cwd, infiles, out_prefix):
+    args = [
+        'jbig2',
+        '-b',
+        out_prefix,
+        '-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
+
+
+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

@@ -0,0 +1,70 @@
+# © 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 functools import lru_cache
+from subprocess import run
+from tempfile import NamedTemporaryFile
+
+from PIL import Image
+
+from . import get_version
+from ..exceptions import MissingDependencyError
+
+
+@lru_cache(maxsize=1)
+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):
+    if input_file.endswith('.jpg'):
+        im = Image.open(input_file)
+        with NamedTemporaryFile(suffix='.png') as tmp:
+            im.save(tmp)
+            args = [
+                'pngquant',
+                '--force',
+                '--skip-if-larger',
+                '--output',
+                output_file,
+                '--quality',
+                f'{quality_min}-{quality_max}',
+                '--',
+                tmp.name,
+            ]
+            run(args)
+    else:
+        args = [
+            'pngquant',
+            '--force',
+            '--skip-if-larger',
+            '--output',
+            output_file,
+            '--quality',
+            f'{quality_min}-{quality_max}',
+            '--',
+            input_file,
+        ]
+        run(args)

src/ocrmypdf/exec/qpdf.py

@@ -15,41 +15,29 @@
 # 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
+from os import fspath
+from subprocess import PIPE, STDOUT, CalledProcessError, run
 
-from ..exceptions import InputFileError, SubprocessOutputError, \
-    MissingDependencyError, EncryptedPdfError
-from . import  get_version
-from ..helpers import re_symlink
+from . import get_version
 
 
 @lru_cache(maxsize=1)
 def version():
-    return get_version('qpdf', regex=r'qpdf version (.+)')    
+    return get_version('qpdf', regex=r'qpdf version (.+)')
 
 
 def check(input_file, log=None):
-    args_qpdf = [
-        'qpdf',
-        '--check',
-        input_file
-    ]
+    args_qpdf = ['qpdf', '--check', fspath(input_file)]
 
     if log is None:
         import logging as log
 
     try:
-        run(args_qpdf, stderr=STDOUT, stdout=PIPE, universal_newlines=True, 
-            check=True)
+        run(args_qpdf, stderr=STDOUT, stdout=PIPE, universal_newlines=True, check=True)
     except CalledProcessError as e:
         if e.returncode == 2:
-            log.error("{0}: not a valid PDF, and could not repair it.".format(
-                input_file))
+            log.error("%s: not a valid PDF, and could not repair it.", input_file)
             log.error("Details:")
             log.error(e.output)
         elif e.returncode == 3:
@@ -59,169 +47,3 @@ def check(input_file, log=None):
             log.warning(e.output)
         return False
     return True
-
-
-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
-        else:
-            log.error("{0}: unknown error".format(
-                      input_file))
-            log.error(e.output)
-            raise SubprocessOutputError() from e
-
-
-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)
-
-

src/ocrmypdf/exec/tesseract.py

@@ -15,26 +15,28 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-import sys
 import os
-import re
 import shutil
-from functools import lru_cache
+import sys
 from collections import namedtuple
-from textwrap import dedent
-from subprocess import PIPE, CalledProcessError, \
-    TimeoutExpired, check_output, STDOUT
 from contextlib import suppress
+from functools import lru_cache
+from os import fspath
+from subprocess import (
+    PIPE,
+    STDOUT,
+    CalledProcessError,
+    TimeoutExpired,
+    check_output,
+    run,
+)
+from textwrap import dedent
 
-import PyPDF2 as pypdf
-
+from . import get_version
 from ..exceptions import MissingDependencyError, TesseractConfigError
 from ..helpers import page_number
-from . import get_version
 
-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"
@@ -43,8 +45,8 @@ 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'>
@@ -67,57 +69,53 @@ def v4():
 @lru_cache(maxsize=1)
 def has_textonly_pdf():
     """Does Tesseract have textonly_pdf capability?
-    
-    Available in 3.05.01, and v4.00.00alpha since January 2017. Best to 
+
+    Available in v4.00.00alpha since January 2017. Best to
     parse the parameter list
     """
-    args_tess = [
-        'tesseract',
-        '--print-parameters',
-        'pdf'
-    ]
+    args_tess = ['tesseract', '--print-parameters', 'pdf']
     params = ''
     try:
-        params = check_output(
-                args_tess, universal_newlines=True, stderr=STDOUT)
+        params = check_output(args_tess, universal_newlines=True, stderr=STDOUT)
     except CalledProcessError as e:
-        print("Could not --print-parameters from tesseract",
-              file=sys.stderr)
+        print("Could not --print-parameters from tesseract", file=sys.stderr)
         raise MissingDependencyError 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'
-
-
 @lru_cache(maxsize=1)
 def languages():
-    args_tess = [
-        'tesseract',
-        '--list-langs'
-    ]
-    try:
-        langs = check_output(
-                args_tess, universal_newlines=True, stderr=STDOUT)
-    except CalledProcessError as e:
-        msg = dedent("""Tesseract failed to report available languages.
+    def lang_error(output):
+        msg = dedent(
+            """Tesseract failed to report available languages.
         Output from Tesseract:
         -----------
-        """)
-        msg += e.output
+        """
+        )
+        msg += output
         print(msg, file=sys.stderr)
+
+    args_tess = ['tesseract', '--list-langs']
+    try:
+        proc = run(
+            args_tess, universal_newlines=True, stdout=PIPE, stderr=STDOUT, check=True
+        )
+        output = proc.stdout
+    except CalledProcessError as e:
+        lang_error(e.output)
         raise MissingDependencyError from e
-    return set(lang.strip() for lang in langs.splitlines()[1:])
+
+    header, *rest = output.splitlines()
+    if not header.startswith('List of available languages'):
+        lang_error(output)
+        raise MissingDependencyError
+    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():
@@ -125,23 +123,24 @@ def tess_base_args(langs, 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):
     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)
+        stdout = check_output(args_tesseract, stderr=STDOUT, timeout=timeout)
     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
     else:
@@ -153,32 +152,24 @@ 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))
+    prefix = f"{(page_number(input_file)):4d}: [tesseract] "
 
     try:
         text = stdout.decode()
     except UnicodeDecodeError:
-        log.error(prefix + "command line output was not utf-8. " +
-            "This usually means Tesseract's language packs do not match "
-            "the installed version of Tesseract.")
+        log.error(
+            prefix
+            + "command line output was not utf-8. "
+            + "This usually means Tesseract's language packs do not match "
+            "the installed version of Tesseract."
+        )
         text = stdout.decode('utf-8', 'backslashreplace')
 
     lines = text.splitlines()
@@ -191,6 +182,14 @@ def tesseract_log_output(log, stdout, input_file):
             log.warning(prefix + "lots of diacritics - possibly poor OCR")
         elif line.startswith('OSD: Weak margin'):
             log.warning(prefix + "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(prefix + line.strip())
+            problem = line.split('found: ')[1]
+            raise TesseractConfigError(problem)
         elif 'error' in line.lower() or 'exception' in line.lower():
             log.error(prefix + line.strip())
         elif 'warning' in line.lower():
@@ -201,8 +200,10 @@ def tesseract_log_output(log, stdout, input_file):
             log.info(prefix + line.strip())
 
 
-def page_timedout(log, input_file):
-    prefix = "{0:4d}: [tesseract] ".format(page_number(input_file))
+def page_timedout(log, input_file, timeout):
+    if timeout == 0:
+        return
+    prefix = f"{(page_number(input_file)):4d}: [tesseract] "
     log.warning(prefix + " took too long to OCR - skipping")
 
 
@@ -220,10 +221,18 @@ 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,
+    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'))
@@ -232,7 +241,7 @@ def generate_hocr(input_file, output_files, language: list, engine_mode,
     args_tesseract = tess_base_args(language, engine_mode)
 
     if pagesegmode is not None:
-        args_tesseract.extend([psm(), str(pagesegmode)])
+        args_tesseract.extend(['--psm', str(pagesegmode)])
 
     if user_words:
         args_tesseract.extend(['--user-words', user_words])
@@ -242,30 +251,18 @@ 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)
+        stdout = check_output(args_tesseract, stderr=STDOUT, timeout=timeout)
     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
@@ -283,33 +280,34 @@ def use_skip_page(text_only, skip_pdf, output_pdf, output_text):
     with open(output_text, 'w') as f:
         f.write('[skipped page]')
 
-    if not text_only:
+    if skip_pdf and not text_only:
+        # Substitute a "skipped page"
         with suppress(FileNotFoundError):
             os.remove(output_pdf)  # In case it was partially created
         os.symlink(skip_pdf, output_pdf)
         return
 
-    # For text only we must create a blank page with dimensions identical
-    # to the skip page because this is equivalent to a page with no text
-
-    pdf_in = pypdf.PdfFileReader(skip_pdf)
-    page0 = pdf_in.pages[0]
-
+    # Or normally, just write a 0 byte file to the output to indicate a skip
     with open(output_pdf, 'wb') as out:
-        pdf_out = pypdf.PdfFileWriter()
-        w, h = page0.mediaBox.getWidth(), page0.mediaBox.getHeight()
-        # If skip page has a /Rotate key, replicate the rotation
-        rotation = int(page0.get('/Rotate', 0))
-        if rotation % 180 == 90:
-            w, h = h, w
-        pdf_out.addBlankPage(w, h)
-        pdf_out.write(out)
+        out.write(b'')
 
 
-def generate_pdf(*, input_image, skip_pdf, output_pdf, output_text,
-                 language: list, engine_mode, text_only: bool,
-                 tessconfig: list, timeout: float, pagesegmode: int,
-                 user_words, user_patterns, log):
+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.
 
     input_image -- image to analyze
@@ -327,9 +325,9 @@ def generate_pdf(*, input_image, skip_pdf, output_pdf, output_text,
     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:
+    if text_only and has_textonly_pdf():
         args_tesseract.extend(['-c', 'textonly_pdf=1'])
 
     if user_words:
@@ -343,28 +341,18 @@ def generate_pdf(*, input_image, skip_pdf, 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)
+        stdout = check_output(args_tesseract, stderr=STDOUT, timeout=timeout)
         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

src/ocrmypdf/exec/unpaper.py

@@ -18,14 +18,16 @@
 # unpaper documentation:
 # https://github.com/Flameeyes/unpaper/blob/master/doc/basic-concepts.md
 
-from subprocess import CalledProcessError, STDOUT, check_output
-from tempfile import NamedTemporaryFile
-import sys
 import os
+import shlex
+import subprocess
+import sys
 from functools import lru_cache
-from ..exceptions import MissingDependencyError
-from . import get_version
+from subprocess import PIPE, STDOUT, CalledProcessError
+from tempfile import TemporaryDirectory
 
+from . import get_version
+from ..exceptions import MissingDependencyError, SubprocessOutputError
 
 try:
     from PIL import Image
@@ -40,11 +42,7 @@ def version():
 
 
 def run(input_file, output_file, dpi, log, mode_args):
-    args_unpaper = [
-        'unpaper',
-        '-v',
-        '--dpi', str(dpi)
-    ] + mode_args
+    args_unpaper = ['unpaper', '-v', '--dpi', str(dpi)] + mode_args
 
     SUFFIXES = {'1': '.pbm', 'L': '.pgm', 'RGB': '.ppm'}
 
@@ -68,35 +66,64 @@ def run(input_file, output_file, dpi, log, mode_args):
         im.close()
         raise MissingDependencyError() from e
 
-    with NamedTemporaryFile(suffix=suffix) as input_pnm, \
-            NamedTemporaryFile(suffix=suffix, mode="r+b") as output_pnm:
+    with TemporaryDirectory() as tmpdir:
+        input_pnm = os.path.join(tmpdir, f'input{suffix}')
+        output_pnm = os.path.join(tmpdir, f'output{suffix}')
         im.save(input_pnm, format='PPM')
         im.close()
 
-        os.unlink(output_pnm.name)
-
-        args_unpaper.extend([input_pnm.name, output_pnm.name])
+        # To prevent any shenanigans from accepting arbitrary parameters in
+        # --unpaper-args, we:
+        # 1) run with cwd set to a tmpdir with only unpaper's files
+        # 2) forbid the use of '/' in arguments, to prevent changing paths
+        # 3) append absolute paths for the input and output file
+        # This should ensure that a user cannot clobber some other file with
+        # their unpaper arguments (whether intentionally or otherwise)
+        args_unpaper.extend([input_pnm, output_pnm])
         try:
-            stdout = check_output(
-                args_unpaper, close_fds=True,
-                universal_newlines=True, stderr=STDOUT,
-                )
+            proc = subprocess.run(
+                args_unpaper,
+                check=True,
+                close_fds=True,
+                universal_newlines=True,
+                stderr=STDOUT,
+                cwd=tmpdir,
+                stdout=PIPE,
+            )
         except CalledProcessError as e:
             log.debug(e.output)
             raise e from e
         else:
-            log.debug(stdout)
-            # unpaper sets dpi to 72
-            Image.open(output_pnm.name).save(output_file, dpi=(dpi, dpi))
+            log.debug(proc.stdout)
+            # unpaper sets dpi to 72; fix this
+            try:
+                Image.open(output_pnm).save(output_file, dpi=(dpi, dpi))
+            except (FileNotFoundError, OSError):
+                raise SubprocessOutputError(
+                    "unpaper: failed to produce the expected output file. Called with: "
+                    + str(args_unpaper)
+                ) from None
 
 
-def clean(input_file, output_file, dpi, log):
-    run(input_file, output_file, dpi, log, [
-        '--layout', 'none',
-        '--mask-scan-size', '100',  # don't blank out narrow columns
+def validate_custom_args(args: str):
+    unpaper_args = shlex.split(args)
+    if any('/' in arg for arg in unpaper_args):
+        raise ValueError('No filenames allowed in --unpaper-args')
+    return unpaper_args
+
+
+def clean(input_file, output_file, dpi, log, unpaper_args=None):
+    default_args = [
+        '--layout',
+        'none',
+        '--mask-scan-size',
+        '100',  # don't blank out narrow columns
         '--no-border-align',  # don't align visible content to borders
-        '--no-mask-center',   # don't center visible content within page
-        '--no-grayfilter',    # don't remove light gray areas
-        '--no-blackfilter',   # don't remove solid black areas
-        '--no-deskew',        # don't deskew
-    ])
+        '--no-mask-center',  # don't center visible content within page
+        '--no-grayfilter',  # don't remove light gray areas
+        '--no-blackfilter',  # don't remove solid black areas
+        '--no-deskew',  # don't deskew
+    ]
+    if not unpaper_args:
+        unpaper_args = default_args
+    run(input_file, output_file, dpi, log, unpaper_args)

src/ocrmypdf/helpers.py

@@ -15,23 +15,22 @@
 # You should have received a copy of the GNU General Public License
 # along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
 
-from PIL import Image
-
-from functools import partial
-from collections.abc import Iterable
-from contextlib import suppress, contextmanager
-from pathlib import Path
-import sys
-import os
 import multiprocessing
+import os
+import sys
 import warnings
+from collections.abc import Iterable
+from contextlib import suppress
+from functools import partial, wraps
+from pathlib import Path
 
 
 def re_symlink(input_file, soft_link_name, log=None):
     """
     Helper function: relinks soft symbolic link if necessary
     """
-
+    input_file = os.fspath(input_file)
+    soft_link_name = os.fspath(soft_link_name)
     if log is None:
         prdebug = partial(print, file=sys.stderr)
     else:
@@ -39,32 +38,29 @@ def re_symlink(input_file, soft_link_name, log=None):
 
     # 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.")
+        prdebug(
+            "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("%s exists and is not a link" % soft_link_name)
         try:
             os.unlink(soft_link_name)
         except OSError:
             prdebug("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("trying to create a broken symlink to %s" % input_file)
 
     prdebug("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):
@@ -72,8 +68,8 @@ def is_iterable_notstr(thing):
 
 
 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])
+    """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():
@@ -84,13 +80,14 @@ def available_cpu_count():
 
     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
 
 
@@ -104,20 +101,15 @@ def is_file_writable(test_file):
     p = Path(test_file)
 
     if p.is_symlink():
-        # Python 3.5 does not accept parameters for Path.resolve() and behaves
-        # as if strict=True (throws an exception on failure). Python 3.6
-        # defaults to strict=False. This implements strict=False like behavior
-        # for Python 3.5.
-        if sys.version_info[0:2] <= (3, 5):
-            p = Path(os.path.realpath(str(p)))
-        else:
-            p = p.resolve(strict=False)
+        p = p.resolve(strict=False)
 
     # p.is_file() throws an exception in some cases
     if p.exists() and p.is_file():
         return os.access(
-            str(p), os.W_OK,
-            effective_ids=(os.access in os.supports_effective_ids))
+            os.fspath(p),
+            os.W_OK,
+            effective_ids=(os.access in os.supports_effective_ids),
+        )
     else:
         try:
             fp = p.open('wb')
@@ -130,43 +122,26 @@ def is_file_writable(test_file):
         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
+def flatten_groups(groups):
+    for obj in groups:
+        if is_iterable_notstr(obj):
+            yield from obj
         else:
-            if isinstance(path, (str, bytes)):
-                return path
-            else:
-                raise TypeError("expected __fspath__() to return str or bytes, "
-                                "not " + type(path).__name__)
+            yield obj
 
-        raise TypeError(
-            "expected str, bytes, pathlib.Path or os.PathLike object, not "
-            + path_type.__name__)
 
-else:
-    universal_open = open
-    fspath = os.fspath
+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
@@ -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:
@@ -191,38 +192,58 @@ class HocrTransform():
             # draw the bbox border
             if showBoundingboxes:
                 pdf.rect(
-                    pt.x1, self.height - pt.y2, pt.x2 - pt.x1, pt.y2 - pt.y1,
-                    fill=1)
-                    
+                    pt.x1, self.height - pt.y2, pt.x2 - pt.x1, pt.y2 - pt.y1, fill=1
+                )
+
         found_lines = False
         for line in self.hocr.findall(
-                ".//%sspan[@class='%s']" % (self.xmlns, "ocr_line")):
+            ".//%sspan[@class='%s']" % (self.xmlns, "ocr_line")
+        ):
             found_lines = True
-            self._do_line(pdf, line, "ocrx_word", fontname, invisibleText, 
-                          interwordSpaces, showBoundingboxes)
+            self._do_line(
+                pdf,
+                line,
+                "ocrx_word",
+                fontname,
+                invisibleText,
+                interwordSpaces,
+                showBoundingboxes,
+            )
 
         if not found_lines:
             # Tesseract did not report any lines (just words)
             root = self.hocr.find(".//%sdiv[@class='%s']" % (self.xmlns, "ocr_page"))
-            self._do_line(pdf, root, "ocrx_word", fontname, invisibleText,
-                          interwordSpaces, showBoundingboxes)
+            self._do_line(
+                pdf,
+                root,
+                "ocrx_word",
+                fontname,
+                invisibleText,
+                interwordSpaces,
+                showBoundingboxes,
+            )
         # put the image on the page, scaled to fill the page
         if imageFileName is not None:
-            pdf.drawImage(imageFileName, 0, 0,
-                          width=self.width, height=self.height)
+            pdf.drawImage(imageFileName, 0, 0, width=self.width, height=self.height)
 
         # finish up the page and save it
         pdf.showPage()
         pdf.save()
 
-
     @classmethod
     def polyval(cls, poly, x):
         return x * poly[0] + poly[1]
 
-
-    def _do_line(self, pdf, line, elemclass, fontname, invisibleText, 
-                 interwordSpaces, showBoundingboxes):
+    def _do_line(
+        self,
+        pdf,
+        line,
+        elemclass,
+        fontname,
+        invisibleText,
+        interwordSpaces,
+        showBoundingboxes,
+    ):
         pxl_line_coords = self.element_coordinates(line)
         line_box = self.pt_from_pixel(pxl_line_coords)
         line_height = line_box.y2 - line_box.y1
@@ -255,23 +276,20 @@ class HocrTransform():
             pdf.setLineWidth(0.5)
             # negate slope because it is defined as a rise/run in pixel
             # coordinates and page coordinates have the y axis flipped
-            pdf.line(line_box.x1, 
-                     baseline_y2,
-                     line_box.x2,
-                     self.polyval((-slope, baseline_y2), 
-                                  line_box.x2 - line_box.x1))
+            pdf.line(
+                line_box.x1,
+                baseline_y2,
+                line_box.x2,
+                self.polyval((-slope, baseline_y2), line_box.x2 - line_box.x1),
+            )
             # light green for bounding box of word/line
             pdf.setDash(6, 3)
             pdf.setStrokeColorRGB(1, 0, 0)
 
-        text.setTextTransform(
-            cos_a, -sin_a, sin_a, cos_a,
-            line_box.x1, baseline_y2
-        )
+        text.setTextTransform(cos_a, -sin_a, sin_a, cos_a, line_box.x1, baseline_y2)
         pdf.setFillColorRGB(0, 0, 0)  # text in black
 
-        elements = line.findall(
-                ".//%sspan[@class='%s']" % (self.xmlns, elemclass))
+        elements = line.findall(".//%sspan[@class='%s']" % (self.xmlns, elemclass))
         for elem in elements:
             elemtxt = self._get_element_text(elem).strip()
             elemtxt = self.replace_unsupported_chars(elemtxt)
@@ -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:
                 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

@@ -19,8 +19,8 @@
 from cffi import FFI
 
 ffibuilder = FFI()
-ffibuilder.set_source("ocrmypdf.lib._leptonica", None)
-ffibuilder.cdef("""
+ffibuilder.cdef(
+    """
 typedef signed char             l_int8;
 typedef unsigned char           l_uint8;
 typedef short                   l_int16;
@@ -32,6 +32,8 @@ typedef double                  l_float64;
 typedef long long               l_int64;
 typedef unsigned long long      l_uint64;
 
+typedef int l_ok; /*!< return type 0 if OK, 1 on error */
+
 struct Pix
 {
     l_uint32             w;           /* width in pixels                   */
@@ -61,6 +63,17 @@ struct PixColormap
 };
 typedef struct PixColormap  PIXCMAP;
 
+/*! Array of pix */
+struct Pixa
+{
+    l_int32             n;          /*!< number of Pix in ptr array        */
+    l_int32             nalloc;     /*!< number of Pix ptrs allocated      */
+    l_uint32            refcount;   /*!< reference count (1 if no clones)  */
+    struct Pix        **pix;        /*!< the array of ptrs to pix          */
+    struct Boxa        *boxa;       /*!< array of boxes                    */
+};
+typedef struct Pixa PIXA;
+
 struct Box
 {
     l_int32            x;
@@ -72,6 +85,69 @@ 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 {
+    L_DEFAULT_ENCODE  = 0,  /*!< use default encoding based on image        */
+    L_JPEG_ENCODE     = 1,  /*!< use dct encoding: 8 and 32 bpp, no cmap    */
+    L_G4_ENCODE       = 2,  /*!< use ccitt g4 fax encoding: 1 bpp           */
+    L_FLATE_ENCODE    = 3,  /*!< use flate encoding: any depth, cmap ok     */
+    L_JP2K_ENCODE     = 4   /*!< use jp2k encoding: 8 and 32 bpp, no cmap   */
+};
+
+/*! Compressed image data */
+struct L_Compressed_Data
+{
+    l_int32            type;         /*!< encoding type: L_JPEG_ENCODE, etc   */
+    l_uint8           *datacomp;     /*!< gzipped raster data                 */
+    size_t             nbytescomp;   /*!< number of compressed bytes          */
+    char              *data85;       /*!< ascii85-encoded gzipped raster data */
+    size_t             nbytes85;     /*!< number of ascii85 encoded bytes     */
+    char              *cmapdata85;   /*!< ascii85-encoded uncompressed cmap   */
+    char              *cmapdatahex;  /*!< hex pdf array for the cmap          */
+    l_int32            ncolors;      /*!< number of colors in cmap            */
+    l_int32            w;            /*!< image width                         */
+    l_int32            h;            /*!< image height                        */
+    l_int32            bps;          /*!< bits/sample; typ. 1, 2, 4 or 8      */
+    l_int32            spp;          /*!< samples/pixel; typ. 1 or 3          */
+    l_int32            minisblack;   /*!< tiff g4 photometry                  */
+    l_int32            predictor;    /*!< flate data has PNG predictors       */
+    size_t             nbytes;       /*!< number of uncompressed raster bytes */
+    l_int32            res;          /*!< resolution (ppi)                    */
+};
+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  */
@@ -90,10 +166,50 @@ enum {
                       /*!< the array with clones (e.g., pix)              */
 };
 
-""")
+/*! Flags for method of extracting barcode widths */
+enum {
+    L_USE_WIDTHS = 1,     /*!< use histogram of barcode widths           */
+    L_USE_WINDOWS = 2     /*!< find best window for decoding transitions */
+};
 
-ffibuilder.cdef("""
+/*! Flags for barcode formats */
+enum {
+    L_BF_UNKNOWN = 0,     /*!< unknown format                            */
+    L_BF_ANY = 1,         /*!< try decoding with all known formats       */
+    L_BF_CODE128 = 2,     /*!< decode with Code128 format                */
+    L_BF_EAN8 = 3,        /*!< decode with EAN8 format                   */
+    L_BF_EAN13 = 4,       /*!< decode with EAN13 format                  */
+    L_BF_CODE2OF5 = 5,    /*!< decode with Code 2 of 5 format            */
+    L_BF_CODEI2OF5 = 6,   /*!< decode with Interleaved 2 of 5 format     */
+    L_BF_CODE39 = 7,      /*!< decode with Code39 format                 */
+    L_BF_CODE93 = 8,      /*!< decode with Code93 format                 */
+    L_BF_CODABAR = 9,     /*!< decode with Code93 format                 */
+    L_BF_UPCA = 10        /*!< decode with UPC A format                  */
+};
+
+enum {
+    L_SEVERITY_EXTERNAL = 0,   /* Get the severity from the environment   */
+    L_SEVERITY_ALL      = 1,   /* Lowest severity: print all messages     */
+    L_SEVERITY_DEBUG    = 2,   /* Print debugging and higher messages     */
+    L_SEVERITY_INFO     = 3,   /* Print informational and higher messages */
+    L_SEVERITY_WARNING  = 4,   /* Print warning and higher messages       */
+    L_SEVERITY_ERROR    = 5,   /* Print error and higher messages         */
+    L_SEVERITY_NONE     = 6    /* Highest severity: print no messages     */
+};
+
+enum {
+    SEL_DONT_CARE  = 0,
+    SEL_HIT        = 1,
+    SEL_MISS       = 2
+};
+
+"""
+)
+
+ffibuilder.cdef(
+    """
 PIX * pixRead ( const char *filename );
+PIX * pixReadMem ( const l_uint8 *data, size_t size );
 PIX * pixScale ( PIX *pixs, l_float32 scalex, l_float32 scaley );
 l_int32 pixFindSkew ( PIX *pixs, l_float32 *pangle, l_float32 *pconf );
 l_int32 pixWriteImpliedFormat ( const char *filename, PIX *pix, l_int32 quality, l_int32 progressive );
@@ -105,6 +221,11 @@ pixWriteMemPng(l_uint8  **pdata,
 
 void pixDestroy ( PIX **ppix );
 
+l_ok
+pixEqual(PIX      *pix1,
+         PIX      *pix2,
+         l_int32 *psame);
+
 PIX *
 pixEndianByteSwapNew(PIX  *pixs);
 
@@ -125,6 +246,8 @@ l_int32 pixSerializeToMemory ( PIX *pixs, l_uint32 **pdata, size_t *pnbytes );
 
 PIX * pixConvertRGBToLuminance(PIX *pixs);
 
+PIX * pixConvertTo8(PIX     *pixs, l_int32  cmapflag);
+
 PIX * pixRemoveColormap(PIX *pixs, l_int32  type);
 
 l_int32
@@ -150,6 +273,18 @@ pixOtsuThreshOnBackgroundNorm(PIX       *pixs,
                               l_float32  scorefract,
                               l_int32   *pthresh);
 
+PIX *
+pixMaskedThreshOnBackgroundNorm(PIX       *pixs,
+                                PIX       *pixim,
+                                l_int32    sx,
+                                l_int32    sy,
+                                l_int32    thresh,
+                                l_int32    mincount,
+                                l_int32    smoothx,
+                                l_int32    smoothy,
+                                l_float32  scorefract,
+                                l_int32   *pthresh);
+
 PIX *
 pixCleanBackgroundToWhite(PIX       *pixs,
                           PIX       *pixim,
@@ -233,13 +368,13 @@ pixGetAverageMaskedRGB(PIX        *pixs,
                        l_float32  *pgval,
                        l_float32  *pbval);
 
-PIX * 
+PIX *
 pixGlobalNormRGB(PIX * 	pixd,
                  PIX * 	pixs,
                  l_int32 	rval,
                  l_int32 	gval,
                  l_int32 	bval,
-                 l_int32 	mapval); 
+                 l_int32 	mapval);
 
 PIX *
 pixInvert(PIX * pixd,
@@ -250,13 +385,109 @@ pixRemoveColormapGeneral(PIX     *pixs,
                          l_int32  type,
                          l_int32  ifnocmap);
 
-void                 
+l_int32
+pixGenerateCIData(PIX           *pixs,
+                  l_int32        type,
+                  l_int32        quality,
+                  l_int32        ascii85,
+                  L_COMP_DATA **pcid);
+
+SARRAY *
+pixProcessBarcodes(PIX      *pixs,
+                   l_int32   format,
+                   l_int32   method,
+                   SARRAY  **psaw,
+                   l_int32 debugflag);
+
+PIX *
+pixaGetPix(PIXA    *pixa,
+           l_int32  index,
+           l_int32 accesstype);
+
+BOX*
+pixaGetBox 	(PIXA *  	pixa,
+		    l_int32  	index,
+		    l_int32  	accesstype );
+
+PIXA *
+pixExtractBarcodes(PIX     *pixs,
+                   l_int32 debugflag);
+
+BOXA *
+pixLocateBarcodes ( PIX *pixs, l_int32 thresh, PIX **ppixb, PIX **ppixm );
+
+SARRAY *
+pixReadBarcodes(PIXA     *pixa,
+                l_int32   format,
+                l_int32   method,
+                SARRAY  **psaw,
+                l_int32 debugflag);
+
+l_int32
+l_generateCIDataForPdf(const char *fname,
+                       PIX *pix,
+                       l_int32 quality,
+                       L_COMP_DATA **pcid);
+
+
+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
-lept_free(void *ptr);  
-""")
+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
+sarrayDestroy(SARRAY  **psa);
+
+void
+lept_free(void *ptr);
+
+void selDestroy ( SEL **psel );
+
+l_int32
+setMsgSeverity(l_int32 newsev);
+
+"""
+)
+
+
+ffibuilder.set_source("ocrmypdf.lib._leptonica", None)
 
 if __name__ == '__main__':
     ffibuilder.compile(verbose=True)

src/ocrmypdf/optimize.py

@@ -0,0 +1,515 @@
+# © 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 logging
+import sys
+from collections import defaultdict
+from os import fspath
+from pathlib import Path
+
+from PIL import Image
+
+import pikepdf
+from pikepdf import Name, Dictionary, Array
+
+from . import leptonica
+from ._jobcontext import JobContext
+from .exec import jbig2enc, pngquant
+from .helpers import re_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
+
+    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
+
+    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'
+
+    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)
+        for future in concurrent.futures.as_completed(futures):
+            proc = future.result()
+            log.debug(proc.stderr.decode())
+
+
+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 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(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):
+    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:
+            for xref in images:
+                log.debug(image_name_fn(root, xref))
+                executor.submit(
+                    pngquant.quantize,
+                    image_name_fn(root, xref),
+                    png_name(root, xref),
+                    png_quality[0],
+                    png_quality[1],
+                )
+
+    for xref in images:
+        im_obj = pike.get_object(xref, 0)
+        try:
+            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
+
+        # When a PNG is inserted into a PDF, we more or less copy the IDAT section from
+        # the PDF and transfer the rest of the PNG headers to PDF image metadata.
+        # One thing we have to do is tell the PDF reader whether a predictor was used
+        # on the image before Flate encoding. (Typically one is.)
+        # According to Leptonica source, PDF readers don't actually need us
+        # to specify the correct predictor, they just need a value of either:
+        #   1 - no predictor
+        #   10-14 - there is a predictor
+        # Leptonica's compdata->predictor only tells TRUE or FALSE
+        # From there the PNG decoder can infer the rest from the file.
+        # In practice the predictor should be Paeth, 14, so we'll use that.
+        # See:
+        #   - PDF RM 7.4.4.4 Table 10
+        #   - https://github.com/DanBloomberg/leptonica/blob/master/src/pdfio2.c#L757
+        predictor = 14 if compdata.predictor > 0 else 1
+        dparms = Dictionary(Predictor=predictor)
+        if predictor > 1:
+            dparms.BitsPerComponent = compdata.bps  # Yes, this is redundant
+            dparms.Colors = compdata.spp
+            dparms.Columns = compdata.w
+
+        im_obj.BitsPerComponent = compdata.bps
+        im_obj.Width = compdata.w
+        im_obj.Height = compdata.h
+
+        if compdata.ncolors > 0:
+            # .ncolors is the number of colors in the palette, not the number of
+            # colors used in a true color image
+            palette_pdf_string = compdata.get_palette_pdf_string()
+            palette_data = pikepdf.Object.parse(palette_pdf_string)
+            palette_stream = pikepdf.Stream(pike, bytes(palette_data))
+            palette = [
+                Name.Indexed,
+                Name.DeviceRGB,
+                compdata.ncolors - 1,
+                palette_stream,
+            ]
+            cs = palette
+        else:
+            if compdata.spp == 1:
+                # PDF interprets binary-1 as black in 1bpp, but PNG sets
+                # black to 0 for 1bpp. Create a palette that informs the PDF
+                # of the mapping - seems cleaner to go this way but pikepdf
+                # needs to be patched to support it.
+                # palette = [Name.Indexed, Name.DeviceGray, 1, b"\xff\x00"]
+                # cs = palette
+                cs = Name.DeviceGray
+            elif compdata.spp == 3:
+                cs = Name.DeviceRGB
+            elif compdata.spp == 4:
+                cs = Name.DeviceCMYK
+        if compdata.bps == 1:
+            im_obj.Decode = [1, 0]  # Bit of a kludge but this inverts photometric too
+        im_obj.ColorSpace = cs
+        im_obj.write(compdata.read(), filter=Name.FlateDecode, decode_parms=dparms)
+
+
+def optimize(input_file, output_file, log, context):
+
+    options = context.get_options()
+    if options.optimize == 0:
+        re_symlink(input_file, output_file, log)
+        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
+
+    pike = pikepdf.Pdf.open(input_file)
+
+    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,
+        preserve_pdfa=True,
+        object_stream_mode=pikepdf.ObjectStreamMode.generate,
+    )
+
+    input_size = Path(input_file).stat().st_size
+    output_size = Path(target_file).stat().st_size
+    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("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):
+    from tempfile import TemporaryDirectory
+    from shutil import copy
+
+    class OptimizeOptions:
+        """Emulate ocrmypdf's options"""
+
+        def __init__(self, jobs, optimize, jpeg_quality, png_quality, jb2lossy):
+            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
+
+    logging.basicConfig(level=logging.DEBUG)
+    log = logging.getLogger()
+
+    ctx = JobContext()
+    options = OptimizeOptions(
+        jobs=jobs,
+        optimize=int(level),
+        jpeg_quality=0,  # Use default
+        png_quality=0,
+        jb2lossy=False,
+    )
+    ctx.set_options(options)
+
+    with TemporaryDirectory() as td:
+        tmpout = Path(td) / 'out.pdf'
+        optimize(infile, tmpout, log, ctx)
+        copy(fspath(tmpout), fspath(outfile))
+
+
+if __name__ == '__main__':
+    main(sys.argv[1], sys.argv[2], sys.argv[3])

src/ocrmypdf/pdfa.py

@@ -31,36 +31,28 @@ Ghostscript's handling of pdfmark.
 
 """
 
-from string import Template
+import os
 from binascii import hexlify
-from datetime import datetime
-from xml.parsers.expat import ExpatError
+from pathlib import Path
+from string import Template
+
 import pkg_resources
-import PyPDF2 as pypdf
-from defusedxml.minidom import parseString as defused_parseString
-from unittest.mock import patch
+
+import pikepdf
 
 ICC_PROFILE_RELPATH = 'data/sRGB.icc'
 
-SRGB_ICC_PROFILE = pkg_resources.resource_filename(
-    'ocrmypdf', ICC_PROFILE_RELPATH)
+SRGB_ICC_PROFILE = pkg_resources.resource_filename('ocrmypdf', ICC_PROFILE_RELPATH)
 
 
 # This is a template written in PostScript which is needed to create PDF/A
 # files, from the Ghostscript documentation. Lines beginning with % are
 # comments. Python substitution variables have a '$' prefix.
 pdfa_def_template = u"""%!
-% This is derived from Ghostscript's template for creating a PDF/A document.
-% This is a small PostScript program that includes some necessary information
-% to create a PDF/A compliant file.
-
 % Define entries in the document Info dictionary :
-/ICCProfile ($icc_profile)
+/ICCProfile $icc_profile
 def
 
-[$docinfo
-  /DOCINFO pdfmark
-
 % Define an ICC profile :
 
 [/_objdef {icc_PDFA} /type /stream /OBJ pdfmark
@@ -91,234 +83,67 @@ 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_ascii(s: str) -> str:
-    """
-    Aggressively strip non-ASCII and PDF escape sequences
-
-    Ghostscript 9.24+ lost support for UTF-16BE in pdfmark files for reasons
-    given in GhostPDL commit e997c683. Our temporary workaround is use ASCII
-    and drop all non-ASCII characters. A slightly improved alternative would
-    be to implement PdfDocEncoding in pikepdf and encode to that, or handle
-    metadata there.
-    """
-    trans = str.maketrans({
-        '(': '',
-        ')': '',
-        '\\': '',
-        '\0': ''
-    })
-    return s.translate(trans).encode('ascii', errors='replace').decode()
-
-
-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.
-    """
-
-    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:
-    """
-    Decode a pdfmark date to a Python datetime object
-
-    A pdfmark date is a string in a paritcular format. See the pdfmark
-    Reference for the specification.
-
-    """
-    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, ascii_docinfo=False):
-    """Create a Postscript pdfmark file for Ghostscript.
-
-    pdfmark contains the various objects as strings; these must be encoded in
-    ASCII, and dates have a special format.
-
-    :param icc_profile: filename of the ICC profile to include in pdfmark
-    :param icc_identifier: ICC identifier such as 'sRGB'
-    :param pdfmark: a dictionary containing keys to include the pdfmark
-    :param ascii_docinfo: if True, the docinfo block must be encoded in pure
-        ASCII and may not contain UTF-16BE-BOM-hex encoded strings, as
-        required for Ghostscript 9.24+
+    :param target_filename: filename to save
+    :param icc: ICC identifier such as 'sRGB'
 
     :returns: a string containing the entire pdfmark
-
     """
-
-    # 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')
-
-    def docinfo_gen():
-        if not ascii_docinfo:
-            docinfo_line_template = '  {key} <{value}>'
-            encode = encode_text_string
-        else:
-            docinfo_line_template = '  {key} ({value})'
-            encode = _encode_ascii
-        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(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', ascii_docinfo=False):
     if icc == 'sRGB':
         icc_profile = SRGB_ICC_PROFILE
     else:
         raise NotImplementedError("Only supporting sRGB")
 
-    ps = _get_pdfa_def(icc_profile, icc, pdfmark, ascii_docinfo=ascii_docinfo)
+    # pdfmark must contain the full path to the ICC profile, and pdfmark must be
+    # also encoded in ASCII. ocrmypdf can be installed anywhere, including to
+    # paths that have a non-ASCII character in the filename. Ghostscript
+    # accepts hex-encoded strings and converts them to byte strings, so
+    # we encode the path with fsencode() and use the hex representation.
+    # UTF-16 not accepted here. (Even though ASCII encodable is the usual case,
+    # do this always to avoid making it a rare conditional.)
+    bytes_icc_profile = os.fsencode(icc_profile)
+    hex_icc_profile = hexlify(bytes_icc_profile)
+    icc_profile = '<' + hex_icc_profile.decode('ascii') + '>'
+
+    t = Template(pdfa_def_template)
+    ps = t.substitute(icc_profile=icc_profile, icc_identifier=icc)
 
     # We should have encoded everything to pure ASCII by this point, and
     # to be safe, only allow ASCII in PostScript
-    with open(target_filename, 'w', encoding='ascii') as f:
-        f.write(ps)
+    Path(target_filename).write_text(ps, encoding='ascii')
 
 
 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.
     """
-    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']
+    pdf = pikepdf.open(filename)
+    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 = 'PDF/A-{}'.format(
-        part_conformance)
-
-    if part_conformance in valid_part_conforms:
+    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

@@ -16,33 +16,29 @@
 # 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 re
-import sys
 from collections import namedtuple
-from collections.abc import MutableMapping, Mapping
-import warnings
-from pathlib import Path
+from decimal import Decimal
 from enum import Enum
-from contextlib import contextmanager
+from math import hypot, isclose
+from os import fspath
+from pathlib import Path
+from unittest.mock import Mock
+from warnings import warn
+import re
 
-import PyPDF2 as pypdf
-from .lib import fitz
-from .helpers import universal_open, fspath
+from pikepdf import PdfMatrix
+import pikepdf
+
+from . import ghosttext
+
+from ..exceptions import EncryptedPdfError, MissingDependencyError
 
 
+Colorspace = Enum('Colorspace', 'gray rgb cmyk lab icc index sep devn pattern jpeg2000')
 
-matrix_mult = pypdf.pdf.utils.matrixMultiply
-
-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')
-
+Encoding = Enum(
+    'Encoding', 'ccitt jpeg jpeg2000 jbig2 asciihex ascii85 lzw flate ' + 'runlength'
+)
 
 FRIENDLY_COLORSPACE = {
     '/DeviceGray': Colorspace.gray,
@@ -73,7 +69,7 @@ FRIENDLY_ENCODING = {
     '/A85': Encoding.ascii85,
     '/LZW': Encoding.lzw,
     '/Fl': Encoding.flate,
-    '/RL': Encoding.runlength
+    '/RL': Encoding.runlength,
 }
 
 FRIENDLY_COMP = {
@@ -81,67 +77,48 @@ 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',
-    ['settings', 'shorthand', 'stack_depth'])
+XobjectSettings = namedtuple('XobjectSettings', ['name', 'shorthand', 'stack_depth'])
 
-ContentsInfo = namedtuple('ContentsInfo', 
-    ['xobject_settings', 'inline_images', 'found_text'])
+InlineSettings = namedtuple('InlineSettings', ['iimage', 'shorthand', 'stack_depth'])
+
+ContentsInfo = namedtuple(
+    'ContentsInfo', ['xobject_settings', 'inline_images', 'found_vector']
+)
+
+TextboxInfo = namedtuple('TextboxInfo', ['bbox', 'is_visible', 'is_corrupt'])
 
 
-def _normalize_stack(operations):
-    """Fix runs of qQ's in the stack
+class VectorInfo:
+    def __init__(self):
+        pass
 
-    For some reason PyPDF2 converts runs of qqq, QQ, QQQq, etc. into single
-    operations.  Break this silliness up and issue each stack operation
-    individually so we don't lose count.
 
-    """
-    for operands, command in operations:
-        if re.match(br'Q*q+$', command):   # Zero or more Q, one or more q
-            for char in command:           # Split into individual bytes
-                yield ([], bytes([char]))  # Yield individual bytes
+def _normalize_stack(graphobjs):
+    """Convert runs of qQ's in the stack into single graphobjs"""
+    for operands, operator in graphobjs:
+        operator = str(operator)
+        if re.match(r'Q*q+$', 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
@@ -159,53 +136,66 @@ def _interpret_contents(contentstream, initial_shorthand=UNIT_SQUARE):
 
     PDF units suit our needs so we initialize ctm to the identity matrix.
 
-    PyPDF2 replaces inline images with a fake "INLINE IMAGE" operator.
-
+    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.
     """
 
-    operations = contentstream.operations
     stack = []
-    ctm = _matrix_from_shorthand(initial_shorthand)
+    ctm = PdfMatrix(initial_shorthand)
     xobject_settings = []
     inline_images = []
-    found_text = False
+    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(operations)):
-        operands, command = op
-        if command == b'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 == b'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 == b'cm':
-            ctm = matrix_mult(
-                _matrix_from_shorthand(operands), ctm)
-        elif command == b'Do':
+                # 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 operator == 'Do':
             image_name = operands[0]
             settings = XobjectSettings(
-                name=image_name, shorthand=_shorthand_from_matrix(ctm),
-                stack_depth=len(stack))
+                name=image_name, shorthand=ctm.shorthand, stack_depth=len(stack)
+            )
             xobject_settings.append(settings)
-        elif command == b'INLINE IMAGE':
-            settings = operands['settings']
+        elif operator == 'INLINE IMAGE':  # BI/ID/EI are grouped into this
+            iimage = operands[0]
             inline = InlineSettings(
-                settings=settings, shorthand=_shorthand_from_matrix(ctm),
-                stack_depth=len(stack))
+                iimage=iimage, shorthand=ctm.shorthand, stack_depth=len(stack)
+            )
             inline_images.append(inline)
-        elif command in (b'Tj', b'TJ', b'"', b"'"):
-            found_text = True
-
+        elif operator in vector_ops:
+            found_vector = True
 
     return ContentsInfo(
         xobject_settings=xobject_settings,
         inline_images=inline_images,
-        found_text=found_text)
+        found_vector=found_vector,
+    )
 
 
 def _get_dpi(ctm_shorthand, image_size):
@@ -275,73 +265,49 @@ 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 = name
+        self._name = str(name)
         self._shorthand = shorthand
-        if inline:
-            # Fixme does not work for inline images with non abbreviated
-            # fields
+
+        if inline is not None:
             self._origin = 'inline'
-            self._width = inline.settings['/W']
-            self._height = inline.settings['/H']
-            self._bpc = inline.settings.get('/BPC', 8)
-            try:
-                self._color = FRIENDLY_COLORSPACE[inline.settings['/CS']]
-            except Exception:
-                self._color = '-'
-            self._comp = FRIENDLY_COMP.get(self._color, '?')
-            if '/F' in inline.settings:
-                filter_ = inline.settings['/F']
-                if isinstance(filter_, pypdf.generic.ArrayObject):
-                    filter_ = filter_[0]
-                self._enc = FRIENDLY_ENCODING.get(filter_, 'image')
-            else:
-                self._enc = 'image'
-        elif pdfimage:
+            pim = inline.iimage
+        elif pdfimage is not None:
             self._origin = 'xobject'
-            self._width = pdfimage['/Width']
-            self._height = pdfimage['/Height']
-            if '/BitsPerComponent' in pdfimage:
-                self._bpc = pdfimage['/BitsPerComponent']
-            else:
-                self._bpc = 8
+            pim = pikepdf.PdfImage(pdfimage)
+        self._width = pim.width
+        self._height = pim.height
 
-            # Fixme: this is incorrectly treats explicit masks as stencil masks,
-            # but good enough for now. Explicit masks have /ImageMask true but are
-            # never called for in content stream, instead are drawn as a /Mask on
-            # other images. For our purposes finding out the details of /Mask
-            # will seldom matter.
-            if '/ImageMask' in pdfimage:
-                self._type = 'stencil' if pdfimage['/ImageMask'].value \
-                    else 'image'
-            else:
-                self._type = 'image'
-            if '/Filter' in pdfimage:
-                filter_ = pdfimage['/Filter']
-                if isinstance(filter_, pypdf.generic.ArrayObject):
-                    filter_ = filter_[0]
-                self._enc = FRIENDLY_ENCODING.get(filter_, 'image')
-            else:
-                self._enc = 'image'
-            if '/ColorSpace' in pdfimage:
-                cs = pdfimage['/ColorSpace']
-                if isinstance(cs, pypdf.generic.ArrayObject):
-                    cs = cs[0]
-                self._color = FRIENDLY_COLORSPACE.get(cs, '-')
-            else:
-                self._color = FRIENDLY_COLORSPACE[Colorspace.jpeg2000] \
-                    if self._enc == Encoding.jpeg2000 else '?'
+        # If /ImageMask is true, then this image is a stencil mask
+        # (Images that draw with this stencil mask will have a reference to
+        # it in their /Mask, but we don't actually need that information)
+        if pim.image_mask:
+            self._type = 'stencil'
+        else:
+            self._type = 'image'
 
-            self._comp = FRIENDLY_COMP.get(self._color, '?')
+        self._bpc = int(pim.bits_per_component)
+        try:
+            self._enc = FRIENDLY_ENCODING.get(pim.filters[0], 'image')
+        except IndexError:
+            self._enc = '?'
 
-            # 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]
+        try:
+            self._color = FRIENDLY_COLORSPACE.get(pim.colorspace, '?')
+        except NotImplementedError:
+            self._color = '?'
+        if self._enc == Encoding.jpeg2000:
+            self._color = Colorspace.jpeg2000
+
+        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]
 
     @property
     def name(self):
@@ -383,34 +349,25 @@ class ImageInfo:
     def yres(self):
         return _get_dpi(self._shorthand, (self._width, self._height))[1]
 
-    def __getitem__(self, item):
-        warnings.warn("ImageInfo.__getitem__", DeprecationWarning)
-        if item in ('name', 'width', 'height', 'bpc', 'color', 'comp', 'enc'):
-            return getattr(self, item)
-        elif item == 'dpi_w':
-            return Decimal(self.xres).quantize(self.DPI_PREC)
-        elif item == 'dpi_h':
-            return Decimal(self.yres).quantize(self.DPI_PREC)
-        elif item == 'dpi':
-            return Decimal(self.xres * self.yres).sqrt().quantize(
-                self.DPI_PREC)
-        else:
-            raise KeyError(item)
-
     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):
@@ -431,8 +388,11 @@ def _image_xobjects(container):
     resources = container['/Resources']
     if '/XObject' not in resources:
         return
-    for xobj in resources['/XObject']:
-        candidate = resources['/XObject'][xobj]
+    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)
@@ -445,7 +405,6 @@ def _find_regular_images(container, contentsinfo):
     that contains images.
 
     Generates images with their DPI at time of drawing.
-
     """
 
     for pdfimage, xobj in _image_xobjects(container):
@@ -464,8 +423,7 @@ 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):
@@ -479,8 +437,9 @@ def _find_form_xobject_images(pdf, container, contentsinfo):
     resources = container['/Resources']
     if '/XObject' not in resources:
         return
-    for xobj in resources['/XObject']:
-        candidate = resources['/XObject'][xobj]
+    xobjs = resources['/XObject'].as_dict()
+    for xobj in xobjs:
+        candidate = xobjs[xobj]
         if candidate['/Subtype'] != '/Form':
             continue
 
@@ -494,11 +453,12 @@ 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.
@@ -507,7 +467,7 @@ 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
@@ -518,99 +478,110 @@ def _find_images(*, pdf, container, shorthand=None):
     """
 
     if container.get('/Type') == '/Page' and '/Contents' in container:
-        # For a /Page the content stream is attached to the page's /Contents
-        page = container
-        contentstream = pypdf.pdf.ContentStream(page.getContents(), pdf)
         initial_shorthand = shorthand or UNIT_SQUARE
-    elif container.get('/Type') == '/XObject' and \
-            container['/Subtype'] == '/Form':
-        # For a Form XObject that content stream is attached to the XObject
-        contentstream = pypdf.pdf.ContentStream(container, pdf)
-
+    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 = _matrix_from_shorthand(shorthand or UNIT_SQUARE)
+        ctm = PdfMatrix(shorthand) if shorthand else PdfMatrix.identity()
 
         # A Form XObject may provide its own matrix to map form space into
         # user space. Get this if one exists
-        form_matrix = _matrix_from_shorthand(
-                container.get('/Matrix', UNIT_SQUARE))
+        form_shorthand = container.get('/Matrix', PdfMatrix.identity())
+        form_matrix = PdfMatrix(form_shorthand)
 
         # Concatenate form matrix with CTM to ensure CTM is correct for
         # drawing this instance of the XObject
-        ctm = matrix_mult(form_matrix, ctm)
-        initial_shorthand = _shorthand_from_matrix(ctm)
+        ctm = form_matrix @ ctm
+        initial_shorthand = ctm.shorthand
     else:
         return
 
-    contentsinfo = _interpret_contents(contentstream, initial_shorthand)
+    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    
-    contentstream = pypdf.pdf.ContentStream(page.getContents(), pdf)
-    contentsinfo = _interpret_contents(contentstream, UNIT_SQUARE)
-    if contentsinfo.found_text:
-        return True
+    pw, ph = float(page_width), float(page_height)
 
-    # 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:    
-            for xobj in resources['/XObject']:
-                candidate = resources['/XObject'][xobj]
-                if candidate['/Subtype'] != '/Form':
-                    continue
-                form_xobject = candidate                
-                # Content stream is attached to Form XObject dictionary
-                contentstream = pypdf.pdf.ContentStream(form_xobject, pdf)
-                sub_contentsinfo = _interpret_contents(
-                    contentstream, UNIT_SQUARE)
-                if sub_contentsinfo.found_text:
-                    return True
-    return False
+    margin_ratio = 0.125
+    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 bbox in text_blocks:
+        if rects_intersect(bbox, interior_bbox):
+            has_text = True
+            break
+    return has_text
 
 
-def _page_has_text(infile, pageno):
-    doc = fitz.Document(infile)
-    text = doc.getPageText(pageno)
-    if text.strip() != '':
-        return True
-    return False
+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):
+def _pdf_get_pageinfo(pdf, pageno: int, infile, xmltext):
     pageinfo = {}
     pageinfo['pageno'] = pageno
     pageinfo['images'] = []
 
-    if isinstance(pdf, Path):
-        pdf = pypdf.PdfFileReader(str(pdf))
-    elif isinstance(pdf, str):
-        pdf = pypdf.PdfFileReader(pdf)
-
     page = pdf.pages[pageno]
+    mediabox = [Decimal(d) for d in page.MediaBox.as_list()]
+    width_pt = mediabox[2] - mediabox[0]
+    height_pt = mediabox[3] - mediabox[1]
 
-    if fitz:
-        pageinfo['has_text'] = _page_has_text(str(infile), pageno)
+    if xmltext is not None:
+        bboxes = ghosttext.page_get_textblocks(
+            fspath(infile), pageno, xmltext=xmltext, height=height_pt
+        )
+        pageinfo['bboxes'] = bboxes
     else:
-        pageinfo['has_text'] = _naive_find_text(pdf=pdf, page=page)
+        # pdfminer required for this section
+        try:
+            from .layout import get_page_analysis, get_text_boxes
+        except ImportError:
+            raise MissingDependencyError(
+                "pdfminer is required for this feature. Your distribution "
+                "may not have installed it."
+            )
+        pscript5_mode = str(pdf.docinfo.get('/Creator')).startswith('PScript5')
+        miner = get_page_analysis(infile, pageno, pscript5_mode)
+        pageinfo['textboxes'] = list(simplify_textboxes(miner, get_text_boxes))
+        bboxes = (box.bbox for box in pageinfo['textboxes'])
 
-    width_pt = page.mediaBox.getWidth()
-    height_pt = page.mediaBox.getHeight()
+    pageinfo['has_text'] = _page_has_text(bboxes, width_pt, height_pt)
 
     userunit = page.get('/UserUnit', Decimal(1.0))
+    if not isinstance(userunit, Decimal):
+        userunit = Decimal(userunit)
     pageinfo['userunit'] = userunit
     pageinfo['width_inches'] = width_pt * userunit / Decimal(72.0)
     pageinfo['height_inches'] = height_pt * userunit / Decimal(72.0)
@@ -621,32 +592,55 @@ 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 = max(image['dpi_w'] for image in pageinfo['images'])
-        yres = max(image['dpi_h'] for image in 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):
-    with universal_open(infile, 'rb') as f:
-        pdf = pypdf.PdfFileReader(f)
-        return [PageInfo(pdf, n, infile) for n in range(pdf.numPages)]
+def _pdf_get_all_pageinfo(infile, detailed_analysis=False, log=None):
+    if not log:
+        log = Mock()
+
+    pdf = pikepdf.open(infile)
+    if pdf.is_encrypted:
+        raise EncryptedPdfError()  # Triggered by encryption with empty passwd
+    if detailed_analysis:
+        pages_xml = None
+    else:
+        pages_xml = ghosttext.extract_text_xml(infile, pdf, pageno=None, log=log)
+
+    pages = []
+    for n in range(len(pdf.pages)):
+        page_xml = pages_xml[n] if pages_xml else None
+        page = PageInfo(pdf, n, infile, page_xml, detailed_analysis)
+        pages.append(page)
+
+    return pages, pdf
 
 
 class PageInfo:
-    def __init__(self, pdf, pageno, infile):
+    def __init__(self, pdf, pageno, infile, xmltext, detailed_analysis=False):
         self._pageno = pageno
         self._infile = infile
-        self._pageinfo = _pdf_get_pageinfo(pdf, pageno, infile)
+        self._pageinfo = _pdf_get_pageinfo(pdf, pageno, infile, xmltext)
+        self._detailed_analysis = detailed_analysis
 
     @property
     def pageno(self):
@@ -656,6 +650,16 @@ class PageInfo:
     def has_text(self):
         return self._pageinfo['has_text']
 
+    @property
+    def has_corrupt_text(self):
+        if not self._detailed_analysis:
+            raise NotImplementedError('Did not do detailed analysis')
+        return any(tbox.is_corrupt for tbox in self._pageinfo['textboxes'])
+
+    @property
+    def has_vector(self):
+        return self._pageinfo['has_vector']
+
     @property
     def width_inches(self):
         return self._pageinfo['width_inches']
@@ -687,6 +691,28 @@ class PageInfo:
     def images(self):
         return self._pageinfo['images']
 
+    def get_textareas(self, visible=None, corrupt=None):
+        def predicate(obj, want_visible, want_corrupt):
+            result = True
+            if want_visible is not None:
+                if obj.is_visible != want_visible:
+                    result = False
+            if want_corrupt is not None:
+                if obj.is_corrupt != want_corrupt:
+                    result = False
+            return result
+
+        if 'textboxes' not in self._pageinfo:
+            if visible is not None and corrupt is not None:
+                raise NotImplementedError('Ghostscript textboxes cannot be classified')
+            return self._pageinfo['bboxes']
+
+        return (
+            obj.bbox
+            for obj in self._pageinfo['textboxes']
+            if predicate(obj, visible, corrupt)
+        )
+
     @property
     def xres(self):
         return self._pageinfo.get('xres', None)
@@ -708,25 +734,28 @@ 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
-    
-    """
-    def __init__(self, infile):
+    """Get summary information about a PDF"""
+
+    def __init__(self, infile, detailed_page_analysis=False, log=None):
         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, detailed_page_analysis, log=log
+        )
+        self._needs_rendering = pdf.root.get('/NeedsRendering', False)
+        self._has_acroform = '/AcroForm' in pdf.root
 
     @property
     def pages(self):
@@ -741,6 +770,10 @@ 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)):
@@ -748,8 +781,8 @@ 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]
@@ -758,24 +791,18 @@ class PdfInfo:
         return len(self._pages)
 
     def __repr__(self):
-        return "<PdfInfo('...'), page count={}>".format(len(self))
-
-    # def __getstate__(self):
-    #     state = {'_infile': self._infile}
-    #     return state
-    #
-    # def __setstate__(self, state):
-    #     self._infile = state['_infile']
-    #     self._pages = _pdf_get_all_pageinfo(self._infile)
+        return f"<PdfInfo('...'), page count={len(self)}>"
 
 
 def main():
     import argparse
+
     parser = argparse.ArgumentParser()
     parser.add_argument('infile')
     args = parser.parse_args()
     info = _pdf_get_all_pageinfo(args.infile)
     from pprint import pprint
+
     pprint(info)
 
 

src/ocrmypdf/pdfinfo/ghosttext.py

@@ -0,0 +1,98 @@
+# © 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
+import xml.etree.ElementTree as ET
+
+from ..exec import ghostscript
+
+# Forgive me for I have sinned
+# I am using regular expressions to parse XML. However the XML in this case,
+# generated by Ghostscript, is self-consistent enough to be parseable.
+regex_remove_char_tags = re.compile(
+    br"""
+    <char\b
+    (?:   [^>]   # anything single character but >
+        | \">\"  # special case: trap ">"
+    )*
+    />           # terminate with '/>'
+""",
+    re.VERBOSE,
+)
+
+
+def page_get_textblocks(infile, pageno, xmltext, height):
+    """Get text boxes out of Ghostscript txtwrite xml"""
+
+    root = xmltext
+    if not hasattr(xmltext, 'findall'):
+        return []
+
+    def blocks():
+        for span in root.findall('.//span'):
+            bbox_str = span.attrib['bbox']
+            font_size = span.attrib['size']
+            pts = [int(pt) for pt in bbox_str.split()]
+            pts[1] = pts[1] - int(float(font_size) + 0.5)
+            bbox_topdown = tuple(pts)
+            bb = bbox_topdown
+            bbox_bottomup = (bb[0], height - bb[3], bb[2], height - bb[1])
+            yield bbox_bottomup
+
+    def joined_blocks():
+        prev = None
+        for bbox in blocks():
+            if prev is None:
+                prev = bbox
+            if bbox[1] == prev[1] and bbox[3] == prev[3]:
+                gap = prev[2] - bbox[0]
+                height = abs(bbox[3] - bbox[1])
+                if gap < height:
+                    # Join boxes
+                    prev = (prev[0], prev[1], bbox[2], bbox[3])
+                    continue
+            # yield previously joined bboxes and start anew
+            yield prev
+            prev = bbox
+        if prev is not None:
+            yield prev
+
+    return [block for block in joined_blocks()]
+
+
+def extract_text_xml(infile, pdf, pageno=None, log=None):
+    existing_text = ghostscript.extract_text(infile, pageno=None)
+    existing_text = regex_remove_char_tags.sub(b' ', existing_text)
+
+    try:
+        root = ET.fromstringlist([b'<document>\n', existing_text, b'</document>\n'])
+        page_xml = root.findall('page')
+    except ET.ParseError as e:
+        log.error(
+            "An error occurred while attempting to retrieve existing text in "
+            "the input file. Will attempt to continue assuming that there is "
+            "no existing text in the file. The error was:"
+        )
+        log.error(e)
+        page_xml = [None] * len(pdf.pages)
+
+    page_count_difference = len(pdf.pages) - len(page_xml)
+    if page_count_difference != 0:
+        log.error("The number of pages in the input file is inconsistent.")
+        if page_count_difference > 0:
+            page_xml.extend([None] * page_count_difference)
+    return page_xml

src/ocrmypdf/pdfinfo/layout.py

@@ -0,0 +1,299 @@
+# © 2018 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+import re
+from math import copysign
+from pathlib import Path
+from unittest.mock import patch
+
+import pdfminer.encodingdb
+import pdfminer.pdfdevice
+import pdfminer.pdfinterp
+from pdfminer.converter import PDFLayoutAnalyzer
+from pdfminer.glyphlist import glyphname2unicode
+from pdfminer.layout import LAParams, LTChar, LTPage, LTTextBox
+from pdfminer.pdfdocument import PDFTextExtractionNotAllowed
+from pdfminer.pdffont import PDFFont, PDFSimpleFont, PDFUnicodeNotDefined
+from pdfminer.pdfpage import PDFPage
+from pdfminer.utils import bbox2str, matrix2str
+
+from ..exceptions import EncryptedPdfError
+
+STRIP_NAME = re.compile(r'[0-9]+')
+
+#
+# Unconditional pdfminer patches
+#
+
+
+def name2unicode(name):
+    """Fix pdfminer's name2unicode function
+
+    Font cids that are mapped to names of the form /g123 seem to be, by convention
+    characters with no corresponding Unicode entry. These can be subsetted fonts
+    or symbolic fonts. There seems to be no way to map /g123 fonts to Unicode,
+    barring a ToUnicode data structure.
+    """
+    if name in glyphname2unicode:
+        return glyphname2unicode[name]
+    if name.startswith('g') or name.startswith('a'):
+        raise KeyError(name)
+    if name.startswith('uni'):
+        try:
+            return chr(int(name[3:], 16))
+        except ValueError:  # Not hexadecimal
+            raise KeyError(name)
+    m = STRIP_NAME.search(name)
+    if not m:
+        raise KeyError(name)
+    return chr(int(m.group(0)))
+
+
+pdfminer.encodingdb.name2unicode = name2unicode
+
+original_PDFFont_init = PDFFont.__init__
+
+
+def PDFFont__init__(self, descriptor, widths, default_width=None):
+    original_PDFFont_init(self, descriptor, widths, default_width)
+    # PDF spec says descent should be negative
+    # A font with a positive descent implies it floats entirely above the
+    # baseline, i.e. it's not really a baseline anymore. I have fonts that
+    # claim a positive descent, but treating descent as positive always seems
+    # to misposition text.
+    if self.descent > 0:
+        self.descent = -self.descent
+
+
+PDFFont.__init__ = PDFFont__init__
+
+original_PDFSimpleFont_init = PDFSimpleFont.__init__
+
+
+def PDFSimpleFont__init__(self, descriptor, widths, spec):
+    # Font encoding is specified either by a name of
+    # built-in encoding or a dictionary that describes
+    # the differences.
+    original_PDFSimpleFont_init(self, descriptor, widths, spec)
+    # pdfminer is incorrect. If there is no ToUnicode and no Encoding, do not
+    # assume Unicode conversion is possible. RM 9.10.2
+    if not self.unicode_map and 'Encoding' not in spec:
+        self.cid2unicode = {}
+    return
+
+
+PDFSimpleFont.__init__ = PDFSimpleFont__init__
+#
+# pdfminer patches when creator is PScript5.dll
+#
+
+
+def PDFType3Font__PScript5_get_height(self):
+    h = self.bbox[3] - self.bbox[1]
+    if h == 0:
+        h = self.ascent - self.descent
+    return h * copysign(1.0, self.vscale)
+
+
+def PDFType3Font__PScript5_get_descent(self):
+    return self.descent * copysign(1.0, self.vscale)
+
+
+def PDFType3Font__PScript5_get_ascent(self):
+    return self.ascent * copysign(1.0, self.vscale)
+
+
+class LTStateAwareChar(LTChar):
+    """A subclass of LTChar that tracks text render mode at time of drawing"""
+
+    __slots__ = (
+        'rendermode',
+        '_text',
+        'matrix',
+        'fontname',
+        'adv',
+        'upright',
+        'size',
+        'width',
+        'height',
+        'bbox',
+        'x0',
+        'x1',
+        'y0',
+        'y1',
+    )
+
+    def __init__(
+        self,
+        matrix,
+        font,
+        fontsize,
+        scaling,
+        rise,
+        text,
+        textwidth,
+        textdisp,
+        ncs,
+        graphicstate,
+        textstate,
+    ):
+        super().__init__(
+            matrix,
+            font,
+            fontsize,
+            scaling,
+            rise,
+            text,
+            textwidth,
+            textdisp,
+            ncs,
+            graphicstate,
+        )
+        self.rendermode = textstate.render
+
+    def is_compatible(self, obj):
+        """Check if characters can be combined into a textline
+
+        We consider characters compatible if:
+            - the Unicode mapping is known, and both have the same render mode
+            - the Unicode mapping is unknown but both are part of the same font
+        """
+        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 '�'
+        return self._text
+
+    def __repr__(self):
+        return '<%s %s matrix=%s rendermode=%r font=%r adv=%s text=%r>' % (
+            self.__class__.__name__,
+            bbox2str(self.bbox),
+            matrix2str(self.matrix),
+            self.rendermode,
+            self.fontname,
+            self.adv,
+            self.get_text(),
+        )
+
+
+class TextPositionTracker(PDFLayoutAnalyzer):
+    """A page layout analyzer that pays attention to text visibility"""
+
+    def __init__(self, rsrcmgr, pageno=1, laparams=None):
+        super().__init__(rsrcmgr, pageno, laparams)
+        self.textstate = None
+        self.result = None
+
+    def begin_page(self, page, ctm):
+        super().begin_page(page, ctm)
+        self.cur_item = LTPage(self.pageno, page.mediabox)
+
+    def end_page(self, page):
+        assert not self._stack, str(len(self._stack))
+        assert isinstance(self.cur_item, LTPage), str(type(self.cur_item))
+        if self.laparams is not None:
+            self.cur_item.analyze(self.laparams)
+        self.pageno += 1
+        self.receive_layout(self.cur_item)
+
+    def render_string(self, textstate, seq, ncs, graphicstate):
+        self.textstate = textstate.copy()
+        super().render_string(self.textstate, seq, ncs, graphicstate)
+
+    def render_char(
+        self, matrix, font, fontsize, scaling, rise, cid, ncs, graphicstate
+    ):
+        try:
+            text = font.to_unichr(cid)
+            assert isinstance(text, str), str(type(text))
+        except PDFUnicodeNotDefined:
+            text = self.handle_undefined_char(font, cid)
+        textwidth = font.char_width(cid)
+        textdisp = font.char_disp(cid)
+        item = LTStateAwareChar(
+            matrix,
+            font,
+            fontsize,
+            scaling,
+            rise,
+            text,
+            textwidth,
+            textdisp,
+            ncs,
+            graphicstate,
+            self.textstate,
+        )
+        self.cur_item.add(item)
+        return item.adv
+
+    def handle_undefined_char(self, font, cid):
+        # log.info('undefined: %r, %r', font, cid)
+        return (font.fontname, cid)
+
+    def receive_layout(self, ltpage):
+        self.result = ltpage
+
+    def get_result(self):
+        return self.result
+
+
+def get_page_analysis(infile, pageno, pscript5_mode):
+    rman = pdfminer.pdfinterp.PDFResourceManager(caching=True)
+    dev = TextPositionTracker(rman, laparams=LAParams())
+    interp = pdfminer.pdfinterp.PDFPageInterpreter(rman, dev)
+
+    if pscript5_mode:
+        patcher = patch.multiple(
+            'pdfminer.pdffont.PDFType3Font',
+            spec=True,
+            get_ascent=PDFType3Font__PScript5_get_ascent,
+            get_descent=PDFType3Font__PScript5_get_descent,
+            get_height=PDFType3Font__PScript5_get_height,
+        )
+        patcher.start()
+
+    try:
+        with Path(infile).open('rb') as f:
+            page = PDFPage.get_pages(f, pagenos=[pageno], maxpages=0)
+            interp.process_page(next(page))
+    except PDFTextExtractionNotAllowed:
+        raise EncryptedPdfError()
+    finally:
+        if pscript5_mode:
+            patcher.stop()
+
+    return dev.get_result()
+
+
+def get_text_boxes(obj):
+    for child in obj:
+        if isinstance(child, (LTTextBox)):
+            yield child
+        else:
+            try:
+                yield from get_text_boxes(child)
+            except TypeError:
+                continue

test_requirements.txt

@@ -1,5 +0,0 @@
-pytest >= 3.2
-pytest-helpers-namespace
-pytest-xdist
-pytest-cov
-pytest-timeout == 1.2.1

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.text__pdf__txt/stderr.bin

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

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

@@ -1,3 +0,0 @@
-Tesseract Open Source OCR Engine v4.0.0-beta.1 with Leptonica
-Too few characters. Skipping this page
-OSD: Weak margin (0.00) for 24 blob text block, but using orientation anyway: 0

tests/cache/aspect/__-l__eng__000001.image__000001.rendered__pdf__txt/stderr.bin

@@ -1,3 +0,0 @@
-Tesseract Open Source OCR Engine v4.0.0-beta.1 with Leptonica
-Too few characters. Skipping this page
-OSD: Weak margin (0.00) for 24 blob text block, but using orientation anyway: 0

tests/cache/aspect/__-l__eng__000001.image__000001.rendered__pdf__txt/txt.bin

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

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

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

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

@@ -5,20 +5,27 @@
  <head>
   <title></title>
 <meta http-equiv="Content-Type" content="text/html;charset=utf-8" />
-  <meta name='ocr-system' content='tesseract 4.0.0-beta.1' />
-  <meta name='ocr-capabilities' content='ocr_page ocr_carea ocr_par ocr_line ocrx_word'/>
+  <meta name='ocr-system' content='tesseract 4.0.0' />
+  <meta name='ocr-capabilities' content='ocr_page ocr_carea ocr_par ocr_line ocrx_word ocrp_wconf'/>
 </head>
 <body>
-  <div class='ocr_page' id='page_1' title='image "/var/folders/37/78_114p552q16vv6vmgm5kr00000gn/T/com.github.ocrmypdf.4nwu81i8/000001.ocr.png"; bbox 0 0 1000 800; ppageno 0'>
+  <div class='ocr_page' id='page_1' title='image "/var/folders/37/78_114p552q16vv6vmgm5kr00000gn/T/com.github.ocrmypdf.jz4b2s0s/000001.ocr.png"; bbox 0 0 1000 800; ppageno 0'>
    <div class='ocr_carea' id='block_1_1' title="bbox 296 96 704 504">
     <p class='ocr_par' id='par_1_1' lang='eng' title="bbox 296 96 704 504">
-     <span class='ocr_line' id='line_1_1' title="bbox 296 96 704 504; baseline 0 296; x_size 169.33333; x_descenders 42.333332; x_ascenders 42.333336"><span class='ocrx_word' id='word_1_1' title='bbox 296 96 704 504; x_wconf 95'><strong> </strong></span> 
+     <span class='ocr_line' id='line_1_1' title="bbox 296 96 704 504; baseline 0 296; x_size 169.33333; x_descenders 42.333332; x_ascenders 42.333336">
+      <span class='ocrx_word' id='word_1_1' title='bbox 296 96 704 504; x_wconf 95'><strong><em> </em></strong></span>
      </span>
     </p>
    </div>
    <div class='ocr_carea' id='block_1_2' title="bbox 150 592 841 622">
     <p class='ocr_par' id='par_1_2' lang='eng' title="bbox 150 592 841 622">
-     <span class='ocr_line' id='line_1_2' title="bbox 150 592 841 622; baseline 0 -6; x_size 30; x_descenders 6; x_ascenders 8"><span class='ocrx_word' id='word_1_2' title='bbox 150 592 230 616; x_wconf 96'><strong>This</strong></span> <span class='ocrx_word' id='word_1_3' title='bbox 260 592 384 616; x_wconf 95'><strong>should</strong></span> <span class='ocrx_word' id='word_1_4' title='bbox 413 592 449 616; x_wconf 95'><strong>be</strong></span> <span class='ocrx_word' id='word_1_5' title='bbox 479 600 493 616; x_wconf 95'><strong>a</strong></span> <span class='ocrx_word' id='word_1_6' title='bbox 523 592 668 622; x_wconf 95'><strong>perfect</strong></span> <span class='ocrx_word' id='word_1_7' title='bbox 698 592 841 616; x_wconf 55'><strong>circle:</strong></span> 
+     <span class='ocr_line' id='line_1_2' title="bbox 150 592 841 622; baseline 0 -6; x_size 30; x_descenders 6; x_ascenders 8">
+      <span class='ocrx_word' id='word_1_2' title='bbox 150 592 230 616; x_wconf 96'><strong><em>This</em></strong></span>
+      <span class='ocrx_word' id='word_1_3' title='bbox 260 592 384 616; x_wconf 95'><strong><em>should</em></strong></span>
+      <span class='ocrx_word' id='word_1_4' title='bbox 413 592 449 616; x_wconf 95'><strong><em>be</em></strong></span>
+      <span class='ocrx_word' id='word_1_5' title='bbox 479 600 493 616; x_wconf 95'><strong><em>a</em></strong></span>
+      <span class='ocrx_word' id='word_1_6' title='bbox 523 592 668 622; x_wconf 95'><strong><em>perfect</em></strong></span>
+      <span class='ocrx_word' id='word_1_7' title='bbox 698 592 841 616; x_wconf 55'><strong><em>circle:</em></strong></span>
      </span>
     </p>
    </div>

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

@@ -1,3 +1 @@
-Tesseract Open Source OCR Engine v4.0.0-beta.1 with Leptonica
-Too few characters. Skipping this page
-OSD: Weak margin (0.00) for 24 blob text block, but using orientation anyway: 0
+Tesseract Open Source OCR Engine v4.0.0 with Leptonica

tests/cache/cardinal/__-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/cardinal/__-l__eng__-c__textonly_pdf=1__000002.ocr.png__000002.text__pdf__txt/stderr.bin

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

tests/cache/cardinal/__-l__eng__-c__textonly_pdf=1__000003.ocr.png__000003.text__pdf__txt/stderr.bin

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

tests/cache/cardinal/__-l__eng__-c__textonly_pdf=1__000004.ocr.png__000004.text__pdf__txt/stderr.bin

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

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

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

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

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

tests/cache/cardinal/__-l__eng__000002.ocr.png__000002.text__pdf__txt/stderr.bin

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

tests/cache/cardinal/__-l__eng__000002.ocr.png__000002__hocr__txt/stderr.bin

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