Ghostcript 9.21 seems to have a regression related to Unicode metadata

Fix issue #147 : unpaper loses DPI information, affects —pdf-renderer tess4
Make —pdf-renderer tess4 more informative, less FUD
2017-03-24 15:15:46 -07:00 · 2017-03-24 13:23:03 -07:00 · 2017-03-24 12:58:40 -07:00 · 2017-03-23 08:13:28 -07:00 · 2017-03-16 00:18:07 -07:00 · 2017-03-13 22:25:25 -07:00
60 changed files with 3938 additions and 2301 deletions
@@ -11,11 +11,14 @@
 */*/*/*/*/*/*.pyc
 */*/*/*/*/*/*/*.pyc
 .cache/
+.eggs/
+.github/
 .git/
 .ipynb_checkpoints/
 .ruffus_history.sqlite
 bin/
 build/
+docs/
 dist/
 htmlcov/
 include/
@@ -27,5 +30,5 @@ tests/cache/
 tests/output/
 tests/resources/private/
 tmp/
-venv-3.4/
-venv-3.5/
+venv*/
+wheelhouse/
@@ -1,9 +1,11 @@
 # Development environment
 *.pyc
 *.sublime-*
-venv-*/
+venv*/
 pyvenv.cfg
 tasks.py
+.bash_history
+.ruffus_history.sqlite

 # Package building
 *.egg-info/
@@ -27,8 +29,11 @@ htmlcov/
 # Testing
 log/
 /*.pdf
+/*.qdf
+*.ipynb
 .ipynb_checkpoints/
 tests/cache/
 tests/output/
 tests/resources/private/
 tmp/
+pdfbox-app*.jar
@@ -1,45 +1,60 @@
-sudo: required
 dist: trusty
 language: python
-
-cache: pip
-
 cache:
-    directories:
-        - $HOME/.ccache
-        - packages
-        - tests/cache
+  pip: true
+  directories:
+  - packages
+  - tests/cache
+  - $HOME/Library/Caches/Homebrew

-python:
-    - 3.4
-    - 3.5
+matrix:
+  include:
+    - os: linux
+      sudo: required
+      language: python
+      python: 3.4
+    - os: linux
+      sudo: required
+      language: python
+      python: 3.5
+    - os: linux
+      sudo: required
+      language: python
+      python: 3.6
+    - os: osx
+      osx_image: xcode8
+      language: generic

 before_cache:
-    - rm -f $HOME/.cache/pip/log/debug.log
+- rm -f $HOME/.cache/pip/log/debug.log

 before_install:
-    # Ubuntu packages
-    - sudo add-apt-repository ppa:vshn/ghostscript -y  # for ghostscript 9.16 (trusty has 9.10)
-    - sudo add-apt-repository ppa:heyarje/libav-11 -y  # for libav11, which is for unpaper
-    - sudo apt-get update -qq  # must go after all add-apt-repo
-    - sudo apt-get install -y ghostscript tesseract-ocr tesseract-ocr-deu tesseract-ocr-eng tesseract-ocr-fra qpdf poppler-utils libavformat56 libavcodec56 libavutil54 libffi-dev
-
-    # pip
-    - pip install --upgrade pip
-
-    # Download, make and install unpaper (using ccache)
-    - mkdir -p packages
-    - "[ -f packages/unpaper_6.1-1.deb ] || wget -q https://dl.dropboxusercontent.com/u/28971240/unpaper_6.1-1.deb -O packages/unpaper_6.1-1.deb"
-    - sudo dpkg -i packages/unpaper_6.1-1.deb
+- if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then bash .travis/linux_before_install.sh ; fi
+- if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then bash .travis/osx_before_install.sh ; fi

 install:
-    - pip install -r requirements.txt
-    - pip install -r test_requirements.txt
+- pip3 install .
+- pip3 install -r requirements.txt -r test_requirements.txt

 script:
-    - python setup.py clean
-    - python setup.py install
-    - python setup.py test
+- mv ocrmypdf dont_import_this_ocrmypdf
+- pytest
+- mv dont_import_this_ocrmypdf ocrmypdf

-os:
-    - linux
+after_success:
+- if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then bash .travis/osx_brew.sh ; fi
+
+# See https://www.appneta.com/blog/pypi-deployment-with-travis-ci/ for
+# steps to set up testpypi deploy for untagged builds if desired
+
+deploy:
+  provider: pypi
+  user: ocrmypdf-travis
+  password:
+    secure: DTFOmmNL6olA0+yXvp4u9jXZlZeqrJsJ0526jzqf4a3gZ6jnGTq5UI6WzRsslSyoMMfXKtHQebqHM6ogSgCZinyZ3ufHJo8fn9brxbEc2gsiWkbj5o3bGwdWMT1vNNE7XW0VCpw87rZ1EEwjl4FJHFudMlPR1yfU5+uq0k0PACo=
+  distributions: "sdist bdist_wheel"
+  on:
+    branch: master
+    tags: true
+    condition: $TRAVIS_PYTHON_VERSION == "3.6" && $TRAVIS_OS_NAME == "linux"
+  skip_upload_docs: true
@@ -0,0 +1,75 @@
+from string import Template
+from subprocess import run, PIPE
+import re
+
+recipe_template = Template("""
+class Ocrmypdf < Formula
+  include Language::Python::Virtualenv
+
+  desc "Adds an OCR text layer to scanned PDF files"
+  homepage "https://github.com/jbarlow83/OCRmyPDF"
+  ${ocrmypdf_url}
+  ${ocrmypdf_sha256}
+
+  depends_on :python3
+  depends_on "pkg-config" => :build
+  depends_on "zlib"
+  depends_on "libffi"
+  depends_on "tesseract"
+  depends_on "ghostscript"
+  depends_on "unpaper"
+  depends_on "qpdf"
+
+  # For Pillow source install
+  depends_on "openjpeg"
+  depends_on "freetype"
+  depends_on "libpng"
+
+${resources}
+
+  def install
+    ENV.append ["SETUPTOOLS_SCM_PRETEND_VERSION"], "v${ocrmypdf_version}"
+    ENV.each do |key, value|
+      puts "#{key}:#{value}"
+    end
+    virtualenv_install_with_resources
+  end
+
+  test do
+    # `test do` will create, run in and delete a temporary directory.
+    #
+    # The installed folder is not in the path, so use the entire path to any
+    # executables being tested: `system "#{bin}/program", "do", "something"`.
+    system "#{bin}/ocrmypdf", "--version"
+  end
+end
+""")
+
+def main():
+    p = run(['poet', '--single', 'ocrmypdf'],
+            encoding='utf-8', stdout=PIPE, check=True)
+
+    ocrmypdf_lines = p.stdout.splitlines()
+    ocrmypdf_url = ocrmypdf_lines[1].strip()
+    ocrmypdf_sha256 = ocrmypdf_lines[2].strip()
+
+    ocrmypdf_version = re.search(
+        r'ocrmypdf-(.+)\.tar.*', ocrmypdf_url).group(1)
+    print(f"Autobrewing {ocrmypdf_version}")
+
+    p = run(['poet', '--resources', 'ocrmypdf'],
+            encoding='utf-8', stdout=PIPE, check=True)
+
+    poet_resources = p.stdout
+
+    # Remove the duplicate "ocrmypdf" resource block
+    all_resources = poet_resources.split('resource')
+    kept_resources = [block for block in all_resources if 'ocrmypdf' not in block]
+    resources = 'resource'.join(kept_resources)
+
+    with open('ocrmypdf.rb', 'w') as out:
+        out.write(recipe_template.substitute(**locals()))
+
+
+if __name__ == '__main__':
+    main()
@@ -0,0 +1,24 @@
+#!/bin/bash
+set -euo pipefail
+set -x
+
+sudo add-apt-repository ppa:vshn/ghostscript -y
+sudo add-apt-repository ppa:heyarje/libav-11 -y
+sudo apt-get update -qq
+sudo apt-get install -y \
+	ghostscript \
+	tesseract-ocr \
+	tesseract-ocr-deu \
+	tesseract-ocr-eng \
+	tesseract-ocr-fra \
+	qpdf \
+	poppler-utils \
+	libavformat56 \
+	libavcodec56 \
+	libavutil54 \
+	libffi-dev
+
+pip install --upgrade pip
+mkdir -p packages
+[ -f packages/unpaper_6.1-1.deb ] || wget -q https://dl.dropboxusercontent.com/u/28971240/unpaper_6.1-1.deb -O packages/unpaper_6.1-1.deb
+sudo dpkg -i packages/unpaper_6.1-1.deb
@@ -0,0 +1,16 @@
+#!/bin/bash
+set -euo pipefail
+set -x
+
+brew update
+
+brew install openjpeg jbig2dec libtiff     # image libraries
+brew install qpdf
+brew install ghostscript
+brew install python3
+brew install libxml2 libffi leptonica
+brew install unpaper   # optional
+brew install tesseract
+
+pip3 install --upgrade pip
+pip3 install wheel
@@ -0,0 +1,8 @@
+#!/bin/bash
+set -euo pipefail
+set -x
+
+pip3 install homebrew-pypi-poet
+python3 .travis/autobrew.py
+brew audit ocrmypdf.rb
+cat ocrmypdf.rb
@@ -1,91 +1,54 @@
 # OCRmyPDF
 #
-# VERSION               3.2
-FROM      debian:stretch
+# VERSION               4.4.2
+FROM      ubuntu:16.10
 MAINTAINER James R. Barlow <jim@purplerock.ca>

-# Add unprivileged user
+RUN apt-get update && apt-get install -y --no-install-recommends \
+  software-properties-common python-software-properties \
+  python3-wheel \
+  python3-reportlab \
+  python3-venv \
+  ghostscript \
+  qpdf \
+  poppler-utils \
+  unpaper \
+  libffi-dev \ 
+  tesseract-ocr \
+  tesseract-ocr-eng \
+  tesseract-ocr-fra \
+  tesseract-ocr-spa \
+  tesseract-ocr-deu
+
+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 
+# (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
+
+# 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
+
+# Remove the junk, including the source version of application since it was
+# already installed
+RUN rm -rf /tmp/* /var/tmp/* /root/* /application/ocrmypdf \
+  && apt-get autoremove -y \
+  && apt-get autoclean -y
+
 RUN useradd docker \
  && mkdir /home/docker \
  && chown docker:docker /home/docker

-# Update system and install our dependencies
-# If this command takes too Docker hub's automated build will timeout,
-# so try it in portions
-RUN apt-get update && apt-get install -y --no-install-recommends \
-  locales \
-  python3 \
-  python3-pip \
-  python3-venv \
-  python3-reportlab \
-  python3-pil \
-  python3-wheel
-
-RUN apt-get install -y --no-install-recommends \
-  unpaper \
-  qpdf \
-  poppler-utils \
-  tesseract-ocr \
-  tesseract-ocr-deu tesseract-ocr-spa tesseract-ocr-eng tesseract-ocr-fra
-
-RUN apt-get install -qy --no-install-recommends \
-  libffi-dev \
-  libpython3-dev \
-  gcc
-
-# Install Ghostscript from Debian sid to work around JPEG 2000 issue in
-# Debian stretch libgs9 or gs 9.16~dfsg-2.1
-
-COPY ./share/etc-apt-sources.list /etc/apt/sources.list
-
-RUN apt-get update && apt-get install -y ghostscript/sid
-
-
-# Enforce UTF-8
-# Borrowed from https://index.docker.io/u/crosbymichael/python/ 
-RUN dpkg-reconfigure locales && \
-  locale-gen C.UTF-8 && \
-  /usr/sbin/update-locale LANG=C.UTF-8
-ENV LC_ALL C.UTF-8
-
-
-# Set up a Python virtualenv and take all of the system packages, so we can
-# rely on the platform packages rather than importing GCC and compiling them
-RUN pyvenv /appenv \
-  && pyvenv --system-site-packages /appenv
-
-COPY . /application/
-
-# Replace stock Tesseract 3.04.00 font with improved sharp2.ttf that resolves
-# issues in many PDF viewers.
-# Discussion is in https://github.com/tesseract-ocr/tesseract/issues/182
-COPY ./share/sharp2.ttf /usr/share/tesseract-ocr/tessdata/pdf.ttf
-RUN chmod 644 /usr/share/tesseract-ocr/tessdata/pdf.ttf
-
-# Set this here to force a docker version, allowing non-tagged versions to
-# be built
-# ENV SETUPTOOLS_SCM_PRETEND_VERSION=v3.3.0
-
-# Install application and dependencies
-# In this arrangement Pillow and reportlab will be provided by the system
-# Even though ocrmypdf is locally present, pull from PyPI because
-# Dockerhub and setuptools_scm clash
-RUN . /appenv/bin/activate; \
-  pip install --upgrade pip \
-  && pip install ocrmypdf \
-  && pip install --no-cache-dir -r /application/test_requirements.txt
-
-# Remove the junk
-RUN apt-get remove -qy gcc
-RUN apt-get autoremove -y && apt-get clean -y
-RUN rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/* /root/*
-
 USER docker
 WORKDIR /home/docker

-ENV OCRMYPDF_TEST_OUTPUT=/tmp/test-output
-ENV OCRMYPDF_SHARP_TTF=1
-
 # Must use array form of ENTRYPOINT
 # Non-array form does not append other arguments, because that is "intuitive"
-ENTRYPOINT ["/application/docker-wrapper.sh"]
+ENTRYPOINT ["/application/docker-wrapper.sh"]
@@ -1,14 +1,17 @@
 # OCRmyPDF polyglot
 #
-# VERSION               3.2
+# VERSION               4.4.2
 FROM      jbarlow83/ocrmypdf:latest
 MAINTAINER James R. Barlow <jim@purplerock.ca>

-# Update system and install our dependencies
 USER root
+
+# Update system and install our dependencies
 RUN apt-get update && apt-get install -y --no-install-recommends \
  tesseract-ocr-all 

+RUN apt-get autoremove -y && apt-get clean -y
+
 USER docker

 # Must use array form of ENTRYPOINT
@@ -0,0 +1,64 @@
+# OCRmyPDF
+#
+# VERSION               4.4.2
+FROM 	ubuntu:16.10
+MAINTAINER James R. Barlow <jim@purplerock.ca>
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+  software-properties-common python-software-properties \
+  python3-wheel \
+  python3-reportlab \
+  python3-venv \
+  ghostscript \
+  qpdf \
+  poppler-utils \
+  unpaper \
+  libffi-dev
+
+RUN add-apt-repository ppa:alex-p/tesseract-ocr
+
+RUN apt-get update \
+	&& apt-get autoremove -y \
+	&& apt-get install -y --no-install-recommends \
+		tesseract-ocr \
+		tesseract-ocr-eng \
+		tesseract-ocr-fra \
+		tesseract-ocr-deu \
+		tesseract-ocr-spa \
+		tesseract-ocr-por \
+		tesseract-ocr-ara \
+		tesseract-ocr-rus \
+		tesseract-ocr-chi-sim 
+
+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 
+# (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
+
+# 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
+
+# Remove the junk, including the source version of application since it was
+# already installed
+RUN rm -rf /tmp/* /var/tmp/* /root/* /application/ocrmypdf \
+  && apt-get autoremove -y \
+  && apt-get autoclean -y
+
+RUN useradd docker \
+  && mkdir /home/docker \
+  && chown docker:docker /home/docker
+
+USER docker
+WORKDIR /home/docker
+
+# Must use array form of ENTRYPOINT
+# Non-array form does not append other arguments, because that is "intuitive"
+ENTRYPOINT ["/application/docker-wrapper.sh"]
@@ -14,6 +14,7 @@ include .dockerignore
 # tests
 include pytest.ini
 recursive-include tests *.jpg
+recursive-include tests *.png
 recursive-include tests *.pdf
 recursive-include tests *.py
 recursive-include tests *.rst
@@ -35,7 +36,6 @@ recursive-exclude docs/_build *

 # support files
 recursive-include ocrmypdf/data *
-recursive-include share *
 include *.py
 exclude tasks.py

@@ -60,7 +60,17 @@ 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
-``apt-get install ocrmypdf``.
+
+.. code-block:: bash
+
+   apt-get install ocrmypdf
+
+and macOS users may simply
+
+.. code-block:: bash
+
+   brew tap jbarlow83/ocrmypdf
+   brew install ocrmypdf

 For everyone else, `see our documentation <https://ocrmypdf.readthedocs.io/en/latest/installation.html>`_ for installation steps.

@@ -3,8 +3,94 @@ RELEASE NOTES

 OCRmyPDF uses `semantic versioning <http://semver.org/>`_.

-v4.3.1:
-=======
+
+v4.5.2
+======
+
+-  Fix issue #147. ``--pdf-renderer tess4 --clean`` will produce an oversized page containing the original image in the bottom left corner, due to loss DPI information.
+-  Make "using Tesseract 4.0" warning less ominous
+-  Set up machinery for homebrew OCRmyPDF tap
+
+
+v4.5.1
+======
+
+-  Fix issue #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; 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
+======
+
+-  The Docker images (ocrmypdf, ocrmypdf-polyglot, ocrmypdf-tess4) are now based on Ubuntu 16.10 instead of Debian stretch
+
+   + This makes supporting the Tesseract 4 image easier
+   + This could be a disruptive change for any Docker users who built customized these images with their own changes, and made those changes in a way that depends on Debian and not Ubuntu
+
+-  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
+======
+
+-  To prevent a `TIFF output error <https://github.com/python-pillow/Pillow/issues/2206>`_ caused by img2pdf >= 0.2.1 and Pillow <= 3.4.2, dependencies have been tightened
+-  The Tesseract 4.00 simultaneous process limit was increased from 1 to 2, since it was observed that 1 lowers performance
+-  Documentation improvements to describe the ``--tesseract-config`` feature 
+-  Added test cases and fixed error handling for ``--tesseract-config``
+-  Tweaks to setup.py to deal with issues in the v4.4 release
+
+v4.4
+====
+
+-  Tesseract 4.00 is now supported on an experimental basis.
+
+   +  A new rendering option ``--pdf-renderer tess4`` exploits Tesseract 4's new text-only output PDF mode. See the documentation on PDF Renderers for details.
+   +  The ``--tesseract-oem`` argument allows control over the Tesseract 4 OCR engine mode (tesseract's ``--oem``). Use ``--tesseract-oem 2`` to enforce the new LSTM mode.
+   +  Fixed poor performance with Tesseract 4.00 on Linux
+
+-  Fixed an issue that caused corruption of output to stdout in some cases
+-  Removed test for Pillow JPEG and PNG support, as the minimum supported version of Pillow now enforces this
+-  OCRmyPDF now tests that the intended destination file is writable before proceeding
+-  The test suite now requires ``pytest-helpers-namespace`` to run (but not install)
+-  Significant code reorganization to make OCRmyPDF re-entrant and improve performance. All changes should be backward compatible for the v4.x series.
+
+   + 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
+======

 -  Fixed an issue where pages produced by the "hocr" renderer after a Tesseract timeout would be rotated incorrectly if the input page was rotated with a /Rotate marker
 -  Fixed a file handle leak in LeptonicaErrorTrap that would cause a "too many open files" error for files around hundred pages of pages long when ``--deskew`` or ``--remove-background`` or other Leptonica based image processing features were in use, depending on the system value of ``ulimit -n``
@@ -14,8 +100,8 @@ v4.3.1:
 -  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:
-=====
+v4.3
+====

 -  New feature ``--remove-background`` to detect and erase the background of color and grayscale images
 -  Better documentation
@@ -25,21 +111,21 @@ v4.3:
   +  This does not improve performance since temporary files are still used for buffering
   +  Some output validation is disabled in this mode

-v4.2.5:
-=======
+v4.2.5
+======

 -  Fixed an issue (#100) with PDFs that omit the optional /BitsPerComponent parameter on images
 -  Removed non-free file milk.pdf


-v4.2.4:
-=======
+v4.2.4
+======

 -  Fixed an error (#90) caused by PDFs that use stencil masks properly
 -  Fixed handling of PDFs that try to draw images or stencil masks without properly setting up the graphics state (such images are now ignored for the purposes of calculating DPI)

-v4.2.3:
-=======
+v4.2.3
+======

 -  Fixed an issue with PDFs that store page rotation (/Rotate) in an indirect object
 -  Integrated a few fixes to simplify downstream packaging (Debian)
@@ -52,21 +138,21 @@ v4.2.3:
 -  Deprecated the OCRmyPDF.sh shell script


-v4.2.2:
-=======
+v4.2.2
+======

 -  Improvements to documentation


-v4.2.1:
-=======
+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:
-=====
+v4.2
+====

 -  ocrmypdf will now try to convert single image files to PDFs if they are provided as input (#15)

@@ -98,14 +184,14 @@ v4.2:

 -  Ghostscript now runs in "safer" mode where possible

-v4.1.4:
-=======
+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:
-=======
+v4.1.3
+======

 -  More helpful error message for PDFs with version 4 security handler
 -  Update usage instructions for Windows/Docker users
@@ -113,50 +199,50 @@ v4.1.3:
 -  Add a few leptonica wrapper functions (no effect on most users)


-v4.1.2:
-=======
+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:
-=====
+v4.1
+====

 -  ``--rotate-pages`` now only rotates pages when reasonably confidence in the orientation. This behavior can be adjusted with the new argument ``--rotate-pages-threshold``
 -  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:
-=======
+v4.0.7
+======

 -  Minor correction to Ghostscript output settings


-v4.0.6:
-=======
+v4.0.6
+======

 -  Update install instructions
 -  Provide a sRGB profile instead of using Ghostscript's


-v4.0.5:
-=======
+v4.0.5
+======

 -  Remove some verbose debug messages from v4.0.4
 -  Fixed temporary that wasn't being deleted
 -  DPI is now calculated correctly for cropped images, along with other image transformations
 -  Inline images are now checked during DPI calculation instead of rejecting the image

-v4.0.4:
-=======
+v4.0.4
+======

 Released with verbose debug message turned on. Do not use. Skip to v4.0.5.


-v4.0.3:
-=======
+v4.0.3
+======

 New features
 ------------
@@ -173,8 +259,8 @@ Fixes
 -  Docker: fix blank JPEG2000 issue by insisting on Ghostscript versions that have this fixed


-v4.0.2:
-=======
+v4.0.2
+======

 Fixes
 -----
@@ -185,8 +271,8 @@ Fixes
 -  Fixed use of chmod on Docker that broke most test cases


-v4.0.1:
-=======
+v4.0.1
+======

 Fixes
 -----
@@ -194,8 +280,8 @@ Fixes
 -  Fixed a KeyError if tesseract fails to find page orientation information


-v4.0:
-=====
+v4.0
+====

 New features
 ------------
@@ -229,8 +315,8 @@ Changes
   to correct the problem.


-v3.2.1:
-=======
+v3.2.1
+======

 Changes
 -------
@@ -239,8 +325,8 @@ Changes
 -  Tweaked the Dockerfiles


-v3.2:
-=====
+v3.2
+====

 New features
 ------------
@@ -261,22 +347,22 @@ Changes



-v3.1.1:
-=======
+v3.1.1
+======

 Changes
 -------

 -  Fixed bug that caused incorrect page size and DPI calculations on documents with mixed page sizes

-v3.1:
-=====
+v3.1
+====

 Changes
 -------

 -  Default output format is now PDF/A-2b instead of PDF/A-1b
-  Python 3.5 and OS X El Capitan are now supported platforms - no changes were
+-  Python 3.5 and macOS El Capitan are now supported platforms - no changes were
   needed to implement support
 -  Improved some error messages related to missing input files
 -  Fixed issue #20 - uppercase .PDF extension not accepted
@@ -287,8 +373,8 @@ Changes
   Currently it always chooses the 'hocrtransform' renderer but that behavior may change.
 -  Set up Travis CI automatic integration testing

-v3.0:
-=====
+v3.0
+====

 New features
 ------------
@@ -437,9 +523,9 @@ Notes and known issues
   images almost never contain inline images.


-v2.2-stable (2014-09-29):
-=========================
+v2.2-stable (2014-09-29)
+========================

 OCRmyPDF versions 1 and 2 were implemented as shell scripts. OCRmyPDF 3.0+ is a fork that gradually replaced all shell scripts with Python while maintaining the existing command line arguments. No one is maintaining old versions.

-For details on older versions, see the `final version of its release notes <https://github.com/fritz-hh/OCRmyPDF/blob/7fd3dbdf42ca53a619412ce8add7532c5e81a9d1/RELEASE_NOTES.md>`_.
+For details on older versions, see the `final version of its release notes <https://github.com/fritz-hh/OCRmyPDF/blob/7fd3dbdf42ca53a619412ce8add7532c5e81a9d1/RELEASE_NOTES.md>`_.
@@ -1,4 +1,10 @@
-check-manifest>=0.33
-setuptools-scm>=1.11.1
-twine>=1.8.1
-coverage>=4.2
+check-manifest >= 0.34
+twine >= 1.8.1
+coverage >= 4.3.1
+pytest-xdist >= 1.15.0
+GitPython == 2.1.0
+
+# Known good versions: 1.11.1
+# Known broken versions: 1.15.0
+setuptools-scm == 1.11.1
+setuptools-scm-git-archive == 1.0
@@ -0,0 +1,127 @@
+Advanced features
+=================
+
+Control of OCR options
+----------------------
+
+OCRmyPDF provides many features to control the behavior of the OCR engine, Tesseract.
+
+When OCR is skipped
+"""""""""""""""""""
+
+If a page in a PDF seems to have text, by default OCRmyPDF will exit without modifying the PDF. This is to ensure that PDFs that were previously OCRed or were "born digital" rather than scanned are not processed. 
+
+If ``--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.
+
+
+Time and image size limits
+""""""""""""""""""""""""""
+
+By default, OCRmyPDF permits tesseract to run for only three minutes (180 seconds) per page. This is usually more than enough time to find all text on a reasonably sized page with modern hardware. 
+
+If a page is skipped, it will be inserted without OCR. If preprocessing was requested, the preprocessed image layer will be inserted.
+
+If you want to adjust the amount of time spent on OCR, change ``--tesseract-timeout``.  You can also automatically skip images that exceed a certain number of megapixels with ``--skip-big``. (A 300 DPI, 8.5×11" page is 8.4 megapixels.)
+
+.. code-block:: bash
+
+	# 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 environment variable ``OCRMYPDF_TESSERACT`` for the full path *to the tesseract executable* first. 
+
+For example, if you are testing tesseract 4.00 and don't wish to disturb your tesseract 3.04 installation, you can launch OCRmyPDF as follows:
+
+.. code-block:: bash
+
+	env \
+		OCRMYPDF_TESSERACT=/home/user/src/tesseract4/api/tesseract \
+		TESSDATA_PREFIX=/home/user/src/tesseract4 \
+		ocrmypdf --pdf-renderer tess4 --tesseract-oem 2 input.pdf output.pdf
+
+* ``TESSDATA_PREFIX`` directs tesseract 4.0 to use LSTM training data. This is a tesseract environment variable.
+* ``--pdf-renderer tess4`` takes advantage of new tesseract 4.0 PDF renderer in OCRmyPDF. (Tesseract 4.0 only.)
+* ``--tesseract-oem 1`` requests tesseract 4.0's new LSTM engine. (Tesseract 4.0 only.)
+
+Overriding other support programs
+"""""""""""""""""""""""""""""""""
+
+In addition to tesseract, OCRmyPDF uses the following external binaries:
+
+* ``gs`` (Ghostscript)
+* ``unpaper``
+* ``qpdf``
+
+In each case OCRmyPDF will check the environment variable ``OCRMYPDF_{program}`` before asking the system to find ``{program}`` on the PATH. For example, you could redirect OCRmyPDF to ``OCRMYPDF_GS`` to override Ghostscript.
+
+Changing tesseract configuration variables
+""""""""""""""""""""""""""""""""""""""""""
+
+You can override tesseract's default `control parameters <https://github.com/tesseract-ocr/tesseract/wiki/ControlParams>`_ with a configuration file.
+
+As an example, this configuration will disable Tesseract's dictionary for current language. Normally the dictionary is helpful for interpolating words that are unclear, but it may interfere with OCR if the document does not contain many words (for example, a list of part numbers).
+
+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
+
+then run ocrmypdf as follows (along with any other desired arguments):
+
+.. code-block:: bash
+
+	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.
+
+
+Changing the PDF renderer
+-------------------------
+
+rasterizing
+  Converting a PDF to an image for display.
+
+rendering
+  Creating a new PDF from other data (such as an existing PDF).
+
+
+OCRmyPDF has three PDF renderers: ``hocr``, ``tesseract`` and ``tess4``. The renderer may be selected using ``--pdf-renderer``. The default is ``auto`` which lets OCRmyPDF select the renderer to use. Currently, ``auto`` always selects ``hocr``. 
+
+The ``hocr`` renderer
+"""""""""""""""""""""
+
+The ``hocr`` renderer is the default because it works in most cases. In this mode the whole PDF is rasterized, the raster image is run through OCR to generate a .hocr file, which is an HTML-like file that specifies the location of all identified words.
+
+The .hocr file is then rendered as a PDF and merged with the image layer.
+
+The image layer is copied from the original PDF page if possible, avoiding potentially lossy transcoding or loss of other PDF information. If preprocessing is specified, then the image layer is a new PDF.
+
+This is the only option for tesseract 3.02 and older.
+
+
+The ``tesseract`` renderer
+""""""""""""""""""""""""""
+
+The tesseract renderer uses tesseract's capability to produce a PDF directly. In version 3, tesseract automatically combined the image layer and text, meaning that this mode *always* transcodes and loses potentially loses image quality and other PDF information.
+
+It does a much better job on non-Latin text.
+
+In a future release this will become the "tess3" renderer and ultimately will be dropped.
+
+
+The ``tess4`` renderer
+""""""""""""""""""""""
+
+The tess4 renderer uses tesseract 4.00 alpha's text-only PDF feature added in January 2017. This combines the advantages of the tesseract and hocr renderers, transcoding the image layer only if required by preprocessing options.
+
+Ghostscript PDF/A still sometimes inserts spaces between words when the tess4 renderer is used, affecting search quality.  ``--output-pdf pdf`` may be used to avoid this issue.
@@ -0,0 +1,142 @@
+Batch processing
+================
+
+This article provides information about running OCRmyPDF on multiple files or configuring it as a service triggered by file system events.
+
+Batch jobs
+----------
+
+Consider using the excellent `GNU Parallel <https://www.gnu.org/software/parallel/>`_ to apply OCRmyPDF to multiple files at once.
+
+Both ``parallel`` and ``ocrmypdf`` will try to use all available processors. To maximize parallelism without overloading your system with processes, consider using ``parallel -j 2`` to limit parallel to running two jobs at once.
+
+This command will run all ocrmypdf all files named ``*.pdf`` in the current directory and write them to the previous created ``output/`` folder. It will not search subdirectories.
+
+The ``--tag`` argument tells parallel to print the filename as a prefix whenever a message is printed, so that one can trace any errors to the file that produced them.
+
+.. code-block:: bash
+
+	parallel --tag -j 2 ocrmypdf '{}' 'output/{}' ::: *.pdf
+
+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 
+
+.. code-block:: bash
+
+	find . --printf '%p' -name '*.pdf' -exec ocrmypdf '{}' '{}' \;
+
+This only runs one ``ocrmypdf`` process at a time. This variation uses ``find`` to create a directory list and ``parallel`` to parallelize runs of ``ocrmypdf``, again updating files in place.
+
+.. code-block:: bash
+
+	find . -name '*.pdf' | parallel --tag -j 2 ocrmypdf '{}' '{}'
+
+
+Sample script
+"""""""""""""
+
+This user contributed script also provides an example of batch processing.
+
+.. code-block:: python
+
+	#!/usr/bin/env python3
+	# Walk through directory tree, replacing all files with OCR'd version
+	# Contributed by DeliciousPickle@github
+
+	import logging
+	import os
+	import subprocess
+	import sys
+
+	script_dir = os.path.dirname(os.path.realpath(__file__))
+	print(script_dir + '/ocr-tree.py: Start')
+
+	if len(sys.argv) > 1:
+	    start_dir = sys.argv[1]
+	else:
+	    start_dir = '.'
+
+	if len(sys.argv) > 2:
+	    log_file = sys.argv[2]
+	else:
+	    log_file = script_dir + '/ocr-tree.log'
+
+	logging.basicConfig(
+			level=logging.INFO, format='%(asctime)s %(message)s', 
+			filename=log_file, filemode='w')
+
+	for dir_name, subdirs, file_list in os.walk(start_dir):
+	    logging.info('\n')
+	    logging.info(dir_name + '\n')
+	    os.chdir(dir_name)
+	    for filename in file_list:
+	        file_ext = os.path.splitext(filename)[1]
+	        if file_ext == '.pdf':
+	            full_path = dir_name + '/' + filename
+	            print(full_path)
+	            cmd = ["ocrmypdf",  "--deskew", filename, filename]
+	            logging.info(cmd)
+	            proc = subprocess.Popen(
+	            	cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
+	            result = proc.stdout.read()
+	            if proc.returncode == 6:
+	                print("Skipped document because it already contained text")
+	            elif proc.returncode == 0:
+	                print("OCR complete")
+	            logging.info(result)
+
+API
+"""
+
+OCRmyPDF is currently supported as a command line interface. Due to limitations in one of the libraries OCRmyPDF depends on, it is not yet usable as an API.
+
+
+Huge batch jobs
+"""""""""""""""
+
+If you have thousands of files to work with, contact the author.
+
+
+Hot (watched) folders
+---------------------
+
+To set up a "hot folder" that will trigger OCR for every file inserted, use a program like Python `watchdog <https://pypi.python.org/pypi/watchdog>`_ (supports all major OS).
+
+One could then configure a scanner to automatically place scanned files in a hot folder, so that they will be queued for OCR and copied to the destination.
+
+.. code-block:: bash
+
+	pip install watchdog
+
+watchdog installs the command line program ``watchmedo``, which can be told to run ``ocrmypdf`` on any .pdf added to the current directory (``.``) and place the result in the previously created ``out/`` folder.
+
+.. code-block:: bash
+
+	cd hot-folder
+	mkdir out
+	watchmedo shell-command \
+		--patterns="*.pdf" \
+		--ignore-directories \
+		--command='ocrmypdf "${watch_src_path}" "out/${watch_src_path}" ' \
+		.  # don't forget the final dot
+
+For more complex behavior you can write a Python script around to use the watchdog API.
+
+On file servers, you could configure watchmedo as a system service so it will run all the time.
+
+Caveats
+"""""""
+
+* ``watchmedo`` may not work properly on a networked file system, depending on the capabilities of the file system client and server.
+* This simple recipe does not filter for the type of file system event, so file copies, deletes and moves, and directory operations, will all be sent to ocrmypdf, producing errors in several cases. Disable your watched folder if you are doing anything other than copying files to it.
+* If the source and destination directory are the same, watchmedo may create an infinite loop.
+* On BSD, FreeBSD and older versions of macOS, you may need to increase the number of file descriptors to monitor more files, using ``ulimit -n 1024`` to watch a folder of up to 1024 files.
+
+Alternatives
+""""""""""""
+
+* `Watchman <https://facebook.github.io/watchman/>`_ is a more powerful alternative to ``watchmedo``.
+
+
@@ -52,7 +52,7 @@ master_doc = 'index'

 # General information about the project.
 project = 'ocrmypdf'
-copyright = '2016, James R. Barlow'
+copyright = '2017, James R. Barlow'
 author = 'James R. Barlow'

 # The version info for the project you're documenting, acts as replacement for
@@ -62,6 +62,23 @@ 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>`.


+Produce PDF and text file containing OCR text
+"""""""""""""""""""""""""""""""""""""""""""""
+
+This produces a file named "output.pdf" and a companion text file named "output.txt". The ``pdftotext`` program from 
+`Poppler <https://poppler.freedesktop.org/>`_ is used to extract text from the finished PDF.
+
+
+.. code-block:: bash
+
+	ocrmypdf input.pdf - | tee output.pdf | pdftotext - output.txt
+
+.. note::
+
+	To get pdftotext, Debian/Ubuntu users may ``apt-get install poppler-utils`` 
+	and macOS users may ``brew install poppler`` respectively.
+
+
 OCR images, not PDFs
 --------------------

@@ -71,7 +88,13 @@ Use a program like `img2pdf <https://gitlab.mister-muffin.de/josch/img2pdf>`_ to

 	img2pdf my-images*.jpg | ocrmypdf - myfile.pdf

-If given a single image as input, OCRmyPDF will try converting it to a PDF on its own.  This feature may be removed at some point, because OCRmyPDF does not specialize in converting images to PDFs.
+If given a single image as input, OCRmyPDF will try converting it to a PDF on its own.  If the DPI specified in the image is incorrect, it can be overridden with ``--image-dpi``:
+
+.. code-block:: bash
+
+	ocrmypdf --image-dpi 300 image.png myfile.pdf
+
+This feature may be removed at some point, because OCRmyPDF does not specialize in converting images to PDFs.

 You can also use Tesseract 3.04+ directly to convert single page images or multi-page TIFFs to PDF:

@@ -94,60 +117,28 @@ OCRmyPDF perform some image processing on each page of a PDF, if desired.  The s

 * ``--clean-final`` uses unpaper to clean up pages before OCR and inserts the page into the final output.  You will want to review each page to ensure that unpaper did not remove something important.

+.. note::
+
+	In many cases image processing will rasterize PDF pages as images, potentially losing quality.
+
+.. warning::
+
+	``--clean-final`` and ``-remove-background`` may leave undesirable visual artifacts in some images where their algorithms have shortcomings. Files should be visually reviewed after using these options.
+

 OCR and correct document skew (crooked scan)
 """"""""""""""""""""""""""""""""""""""""""""

+Deskew:
+
 .. code-block:: bash

 	ocrmypdf --deskew input.pdf output.pdf

-
-Hot (watched) folders
---------------------
-
-To set up a "hot folder" that will trigger an OCR operation for every file inserted, use a program like Python `watchdog <https://pypi.python.org/pypi/watchdog>`_ (supports all major OS).
+Image processing commands can be combined. The order in which options are given does not matter. OCRmyPDF always applies the steps of the image processing pipeline in the same order (rotate, remove background, deskew, clean).

 .. code-block:: bash

-	pip install watchdog
-
-watchdog installs the command line program ``watchmedo``, which can be told to run ``ocrmypdf`` on any .pdf added to the current directory (``.``) and place the result in the previously created ``out/`` folder.
-
-.. code-block:: bash
-
-	cd hot-folder
-	mkdir out
-	watchmedo shell-command \
-		--patterns="*.pdf" \
-		--ignore-directories \
-		--command='ocrmypdf "${watch_src_path}" "out/${watch_src_path}" ' \
-		.  # don't forget the final dot
-
-For more complex behavior you can write a Python script around to use the watchdog API.
-
-On file servers, you could configure watchmedo as a system service so it will run all the time.
-
-Caveats
-"""""""
-
-* ``watchmedo`` may not work properly on a networked file system, depending on the capabilities of the file system client and server.
-* This simple recipe does not filter for the type of file system event, so file copies, deletes and moves, and directory operations, will all be sent to ocrmypdf, producing errors in several cases. Disable your watched folder if you are doing anything other than copying files to it.
-* If the source and destination directory are the same, watchmedo may create an infinite loop.
+	ocrmypdf --deskew --clean --rotate-pages input.pdf output.pdf


-Batch jobs
----------
-
-Consider using the excellent `GNU Parallel <https://www.gnu.org/software/parallel/>`_ to apply OCRmyPDF to multiple files at once.
-
-Both ``parallel`` and ``ocrmypdf`` will try to use all available processors. To maximize parallelism without overloading your system with processes, consider using ``parallel -j 2`` to limit parallel to running two jobs at once.
-
-This command will run all ocrmypdf all files named ``*.pdf`` in the current directory and write them to the previous created ``output/`` folder.
-
-.. code-block:: bash
-
-	parallel -j 2 ocrmypdf '{}' 'output/{}' ::: *.pdf
-
-If you have thousands of files to work with, contact the author.
- 
@@ -20,6 +20,9 @@ Contents:
   installation
   languages
   cookbook
+   advanced
+   batch
+   security
   errors


@@ -10,6 +10,18 @@ Users of Debian 9 ("stretch") or later or Ubuntu 16.10 or later may simply

 	apt-get install ocrmypdf

+
+Installing on macOS
+-------------------
+
+.. code-block:: bash
+
+   brew tap jbarlow83/ocrmypdf
+   brew install ocrmypdf
+
+
+Users who previously installed OCRmyPDF on macOS may need to remove the ``pip`` based installation (``pip3 uninstall ocrmypdf``).
+
 .. _Docker:

 Installing the Docker image
@@ -27,7 +39,7 @@ successfully, your system is ready to download and execute the image:

   docker run hello-world
   
-OCRmyPDF will use all available CPU cores.  By default, the VirtualBox machine instance on Windows and OS X 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:
+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

@@ -37,28 +49,34 @@ OCRmyPDF will use all available CPU cores.  By default, the VirtualBox machine i
   docker-machine start "yourVM"
   eval $(docker-machine env "yourVM")

-Assuming you have a Docker engine running somewhere, you can run these commands to download
-the image:
+Assuming you have a Docker engine running, you can download one of the three available images:
+
+-----------------------------+---------------------------------------------+---------------------------------------------------------------------------------+
+| Image name                  | Download command                            | Notes                                                                           |
+-----------------------------+---------------------------------------------+---------------------------------------------------------------------------------+
+| ocrmypdf                    | ``docker pull jbarlow83/ocrmypdf``          | Latest ocrmypdf with Tesseract 3.04. Includes English, French, German, Spanish. |
+-----------------------------+---------------------------------------------+---------------------------------------------------------------------------------+
+| ocrmypdf-polyglot           | ``docker pull jbarlow83/ocrmypdf-polyglot`` | As above, with all available language packs.                                    |
+-----------------------------+---------------------------------------------+---------------------------------------------------------------------------------+
+| ocrmypdf-tess4              | ``docker pull jbarlow83/ocrmypdf-tess4``    | Latest ocrmypdf with Tesseract 4.00.00alpha and English, French, German,        |
+|                             |                                             | Spanish, Portuguese, Chinese Simplified, Arabic and Russian (the top 8).        |
+-----------------------------+---------------------------------------------+---------------------------------------------------------------------------------+
+
+For example:

 .. code-block:: bash

-   docker pull jbarlow83/ocrmypdf
+   docker pull jbarlow83/ocrmypdf-tess4

 Then tag it to give a more convenient name, just ocrmypdf:

 .. code-block:: bash

-   docker tag jbarlow83/ocrmypdf ocrmypdf
+   docker tag jbarlow83/ocrmypdf-tess4 ocrmypdf

 .. _docker-polyglot:

-This image contains language packs for English, French, Spanish and German. The alternative "polyglot" image provides `all available language packs <https://github.com/tesseract-ocr/tesseract/blob/master/doc/tesseract.1.asc#languages>`_:
-
-.. code-block:: bash
-
-   # Alternative step: If you need all language packs
-   docker pull jbarlow83/ocrmypdf-polyglot
-   docker tag jbarlow83/ocrmypdf-polyglot ocrmypdf
+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:

@@ -78,13 +96,29 @@ In this worked example, the current working directory contains an input file cal

   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.

+For convenience, a shell alias can hide the docker command:

-Installing on macOS (formerly Mac OS X)
---------------------------------------
+.. code-block:: bash

-These instructions probably work on all macOS supported by Homebrew. OCRmyPDF is known to work on Yosemite and El Capitan, and regularly tested on El Capitan.
+   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
+
+
+Manual installation on macOS
+----------------------------
+
+These instructions probably work on all macOS supported by Homebrew.

 If it's not already present, `install Homebrew <http://brew.sh/>`_.

@@ -105,6 +139,8 @@ Install or upgrade the required Homebrew packages, if any are missing:
   brew install libxml2 libffi leptonica
   brew install unpaper   # optional
   
+Python 3.4, 3.5 and 3.6 are supported.
+
 Install the required Tesseract OCR engine with the language packs you plan to use:
   
 .. code-block:: bash
@@ -136,6 +172,40 @@ The command line program should now be available:

   ocrmypdf --help

+
+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 to the system Python, then install as follows (note this installs new packages
+into your system Python, which could interfere with other programs):
+
+.. 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
 ------------------------------

@@ -153,7 +223,6 @@ Update apt-get:
 .. code-block:: bash

   sudo apt-get update
-   sudo apt-get upgrade
   
 Install system dependencies:

@@ -221,7 +290,7 @@ where /c/Users/sampleuser is a Unix representation of the Windows path C:\\Users
 Installing HEAD revision from sources
 -------------------------------------

-If you have ``git`` and ``python3.4`` or ``python3.5`` installed, you can install from source. When the ``pip`` installer runs,
+If you have ``git`` and Python 3.4 or newer installed, you can install from source. When the ``pip`` installer runs,
 it will alert you if dependencies are missing.

 To install the HEAD revision from sources in the current Python 3 environment:
@@ -23,8 +23,8 @@ Debian and Ubuntu users
 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``.

-Mac OS X (macOS) users
----------------------
+macOS users
+-----------

 You can install additional language packs by :ref:`installing Tesseract using Homebrew with all language packs <macos-all-languages>`.

@@ -0,0 +1,392 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+ "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<!-- Generated by graphviz version 2.38.0 (20140413.2041)
+ -->
+<!-- Title: Pipeline: Pages: 1 -->
+<svg width="1544pt" height="1089pt"
+ viewBox="0.00 0.00 1544.00 1089.14" 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 1085.14)">
+<title>Pipeline:</title>
+<polygon fill="white" stroke="none" points="-4,4 -4,-1085.14 1540,-1085.14 1540,4 -4,4"/>
+<g id="clust1" class="cluster"><title>clustertasks</title>
+<polygon fill="none" stroke="black" points="8,-8 8,-1073.14 1528,-1073.14 1528,-8 8,-8"/>
+<text text-anchor="middle" x="768" y="-1045.14" 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="black" points="750.705,-1027.14 541.295,-1027.14 537.295,-1023.14 537.295,-991.141 746.705,-991.141 750.705,-995.141 750.705,-1027.14"/>
+<polyline fill="none" stroke="black" points="746.705,-1023.14 537.295,-1023.14 "/>
+<polyline fill="none" stroke="black" points="746.705,-1023.14 746.705,-991.141 "/>
+<polyline fill="none" stroke="black" points="746.705,-1023.14 750.705,-1027.14 "/>
+<text text-anchor="middle" x="644" y="-1003.14" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.triage</text>
+</g>
+<!-- t1 -->
+<g id="node2" class="node"><title>t1</title>
+<polygon fill="#efa03b" stroke="black" points="769.969,-969.141 522.031,-969.141 518.031,-965.141 518.031,-933.141 765.969,-933.141 769.969,-937.141 769.969,-969.141"/>
+<polyline fill="none" stroke="black" points="765.969,-965.141 518.031,-965.141 "/>
+<polyline fill="none" stroke="black" points="765.969,-965.141 765.969,-933.141 "/>
+<polyline fill="none" stroke="black" points="765.969,-965.141 769.969,-969.141 "/>
+<text text-anchor="middle" x="644" y="-945.141" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.repair_pdf</text>
+</g>
+<!-- t0&#45;&gt;t1 -->
+<g id="edge1" class="edge"><title>t0&#45;&gt;t1</title>
+<path fill="none" stroke="#0044a0" d="M644,-991.035C644,-987.344 644,-983.354 644,-979.384"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="647.5,-979.142 644,-969.142 640.5,-979.142 647.5,-979.142"/>
+</g>
+<!-- t2 -->
+<g id="node3" class="node"><title>t2</title>
+<polygon fill="#efa03b" stroke="black" points="869.82,-890.215 644,-907.116 418.18,-890.215 418.391,-862.869 869.609,-862.869 869.82,-890.215"/>
+<polygon fill="none" stroke="black" points="873.852,-893.922 644,-911.125 414.148,-893.922 414.419,-858.871 873.581,-858.871 873.852,-893.922"/>
+<text text-anchor="middle" x="644" y="-876.657" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.split_pages</text>
+</g>
+<!-- t1&#45;&gt;t2 -->
+<g id="edge2" class="edge"><title>t1&#45;&gt;t2</title>
+<path fill="none" stroke="#0044a0" d="M644,-933.019C644,-929.4 644,-925.445 644,-921.399"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="647.5,-921.295 644,-911.295 640.5,-921.295 647.5,-921.295"/>
+</g>
+<!-- t18 -->
+<g id="node20" class="node"><title>t18</title>
+<polygon fill="#efa03b" stroke="black" points="383.516,-832.173 20.4844,-832.173 16.4844,-828.173 16.4844,-796.173 379.516,-796.173 383.516,-800.173 383.516,-832.173"/>
+<polyline fill="none" stroke="black" points="379.516,-828.173 16.4844,-828.173 "/>
+<polyline fill="none" stroke="black" points="379.516,-828.173 379.516,-796.173 "/>
+<polyline fill="none" stroke="black" points="379.516,-828.173 383.516,-832.173 "/>
+<text text-anchor="middle" x="200" y="-808.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.generate_postscript_stub</text>
+</g>
+<!-- t1&#45;&gt;t18 -->
+<g id="edge29" class="edge"><title>t1&#45;&gt;t18</title>
+<path fill="none" stroke="#0044a0" d="M518.14,-936.259C481.198,-930.408 441.05,-922.284 405,-911.141 344.394,-892.407 278.709,-859.14 238.448,-837.118"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="239.963,-833.956 229.517,-832.188 236.581,-840.085 239.963,-833.956"/>
+</g>
+<!-- t21 -->
+<g id="node22" class="node"><title>t21</title>
+<polygon fill="#efa03b" stroke="black" points="751.49,-156.424 963,-129.109 1174.51,-156.424 1174.31,-200.62 751.688,-200.62 751.49,-156.424"/>
+<polygon fill="none" stroke="black" points="747.474,-152.906 963,-125.073 1178.53,-152.906 1178.29,-204.623 747.705,-204.623 747.474,-152.906"/>
+<text text-anchor="middle" x="963" y="-162.639" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.merge_pages_qpdf</text>
+</g>
+<!-- t1&#45;&gt;t21 -->
+<g id="edge39" class="edge"><title>t1&#45;&gt;t21</title>
+<path fill="none" stroke="#0044a0" d="M769.832,-943.539C1011.96,-929.143 1519,-890.056 1519,-815.173 1519,-815.173 1519,-815.173 1519,-309.173 1519,-237.474 1342.87,-202.283 1188.87,-185.216"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1189.08,-181.719 1178.76,-184.118 1188.32,-188.678 1189.08,-181.719"/>
+</g>
+<!-- t3 -->
+<g id="node4" class="node"><title>t3</title>
+<polygon fill="#efa03b" stroke="black" points="712.317,-832.173 405.683,-832.173 401.683,-828.173 401.683,-796.173 708.317,-796.173 712.317,-800.173 712.317,-832.173"/>
+<polyline fill="none" stroke="black" points="708.317,-828.173 401.683,-828.173 "/>
+<polyline fill="none" stroke="black" points="708.317,-828.173 708.317,-796.173 "/>
+<polyline fill="none" stroke="black" points="708.317,-828.173 712.317,-832.173 "/>
+<text text-anchor="middle" x="557" y="-808.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.rasterize_preview</text>
+</g>
+<!-- t2&#45;&gt;t3 -->
+<g id="edge3" class="edge"><title>t2&#45;&gt;t3</title>
+<path fill="none" stroke="#0044a0" d="M614.06,-858.777C605.418,-852.173 595.983,-844.963 587.319,-838.342"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="589.37,-835.504 579.299,-832.213 585.119,-841.066 589.37,-835.504"/>
+</g>
+<!-- t4 -->
+<g id="node5" class="node"><title>t4</title>
+<polygon fill="#efa03b" stroke="black" points="688.082,-774.173 429.918,-774.173 425.918,-770.173 425.918,-738.173 684.082,-738.173 688.082,-742.173 688.082,-774.173"/>
+<polyline fill="none" stroke="black" points="684.082,-770.173 425.918,-770.173 "/>
+<polyline fill="none" stroke="black" points="684.082,-770.173 684.082,-738.173 "/>
+<polyline fill="none" stroke="black" points="684.082,-770.173 688.082,-774.173 "/>
+<text text-anchor="middle" x="557" y="-750.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.orient_page</text>
+</g>
+<!-- t2&#45;&gt;t4 -->
+<g id="edge5" class="edge"><title>t2&#45;&gt;t4</title>
+<path fill="none" stroke="#0044a0" d="M695.691,-858.661C705.833,-851.667 715.136,-842.912 721,-832.173 728.668,-818.13 730.919,-808.727 721,-796.173 714.915,-788.471 700.759,-782.127 683.18,-776.959"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="683.659,-773.462 673.091,-774.208 681.817,-780.216 683.659,-773.462"/>
+</g>
+<!-- t3&#45;&gt;t4 -->
+<g id="edge4" class="edge"><title>t3&#45;&gt;t4</title>
+<path fill="none" stroke="#0044a0" d="M557,-796.066C557,-792.375 557,-788.385 557,-784.415"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="560.5,-784.174 557,-774.174 553.5,-784.174 560.5,-784.174"/>
+</g>
+<!-- t5 -->
+<g id="node6" class="node"><title>t5</title>
+<polygon fill="#efa03b" stroke="black" points="1064.73,-716.173 689.268,-716.173 685.268,-712.173 685.268,-680.173 1060.73,-680.173 1064.73,-684.173 1064.73,-716.173"/>
+<polyline fill="none" stroke="black" points="1060.73,-712.173 685.268,-712.173 "/>
+<polyline fill="none" stroke="black" points="1060.73,-712.173 1060.73,-680.173 "/>
+<polyline fill="none" stroke="black" points="1060.73,-712.173 1064.73,-716.173 "/>
+<text text-anchor="middle" x="875" y="-692.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.rasterize_with_ghostscript</text>
+</g>
+<!-- t4&#45;&gt;t5 -->
+<g id="edge6" class="edge"><title>t4&#45;&gt;t5</title>
+<path fill="none" stroke="#0044a0" d="M653.706,-738.143C689.783,-731.789 730.943,-724.541 767.902,-718.033"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="768.965,-721.399 778.206,-716.218 767.751,-714.505 768.965,-721.399"/>
+</g>
+<!-- t15 -->
+<g id="node13" class="node"><title>t15</title>
+<polygon fill="#ff69b4" stroke="black" points="440.27,-396.173 59.7304,-396.173 55.7304,-392.173 55.7304,-360.173 436.27,-360.173 440.27,-364.173 440.27,-396.173"/>
+<polyline fill="none" stroke="black" points="436.27,-392.173 55.7304,-392.173 "/>
+<polyline fill="none" stroke="black" points="436.27,-392.173 436.27,-360.173 "/>
+<polyline fill="none" stroke="black" points="436.27,-392.173 440.27,-396.173 "/>
+<text text-anchor="middle" x="248" y="-372.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.ocr_tesseract_textonly_pdf</text>
+</g>
+<!-- t4&#45;&gt;t15 -->
+<g id="edge14" class="edge"><title>t4&#45;&gt;t15</title>
+<path fill="none" stroke="#0044a0" d="M425.857,-745.263C361.241,-732.425 297,-703.891 297,-641.173 297,-641.173 297,-641.173 297,-523.173 297,-480.18 277.182,-433.63 262.754,-405.436"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="265.677,-403.477 257.916,-396.265 259.486,-406.743 265.677,-403.477"/>
+</g>
+<!-- t12 -->
+<g id="node15" class="node"><title>t12</title>
+<polygon fill="#00cc66" stroke="black" points="781,-406.173 458.414,-378.173 781,-350.173 1103.59,-378.173 781,-406.173"/>
+<text text-anchor="middle" x="781" y="-372.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.select_image_layer</text>
+</g>
+<!-- t4&#45;&gt;t12 -->
+<g id="edge20" class="edge"><title>t4&#45;&gt;t12</title>
+<path fill="none" stroke="#0044a0" d="M471.449,-738.167C421.018,-722.605 367,-693.681 367,-641.173 367,-641.173 367,-641.173 367,-523.173 367,-480.717 347.218,-458.43 377,-428.173 392.917,-412.001 494.592,-399.721 591.948,-391.499"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="592.284,-394.983 601.96,-390.666 591.704,-388.007 592.284,-394.983"/>
+</g>
+<!-- t19 -->
+<g id="node18" class="node"><title>t19</title>
+<polygon fill="#efa03b" stroke="black" points="323.772,-270.173 78.2284,-270.173 74.2284,-266.173 74.2284,-234.173 319.772,-234.173 323.772,-238.173 323.772,-270.173"/>
+<polyline fill="none" stroke="black" points="319.772,-266.173 74.2284,-266.173 "/>
+<polyline fill="none" stroke="black" points="319.772,-266.173 319.772,-234.173 "/>
+<polyline fill="none" stroke="black" points="319.772,-266.173 323.772,-270.173 "/>
+<text text-anchor="middle" x="199" y="-246.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.skip_page</text>
+</g>
+<!-- t4&#45;&gt;t19 -->
+<g id="edge26" class="edge"><title>t4&#45;&gt;t19</title>
+<path fill="none" stroke="#0044a0" d="M425.782,-752.078C271.736,-744.795 37,-720.655 37,-641.173 37,-641.173 37,-641.173 37,-377.173 37,-325.92 88.7624,-292.989 133.769,-274.043"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="135.289,-277.203 143.249,-270.211 132.666,-270.713 135.289,-277.203"/>
+</g>
+<!-- t17 -->
+<g id="node19" class="node"><title>t17</title>
+<polygon fill="#66ccff" stroke="black" points="1499.67,-328.173 1094.33,-328.173 1090.33,-324.173 1090.33,-292.173 1495.67,-292.173 1499.67,-296.173 1499.67,-328.173"/>
+<polyline fill="none" stroke="black" points="1495.67,-324.173 1090.33,-324.173 "/>
+<polyline fill="none" stroke="black" points="1495.67,-324.173 1495.67,-292.173 "/>
+<polyline fill="none" stroke="black" points="1495.67,-324.173 1499.67,-328.173 "/>
+<text text-anchor="middle" x="1295" y="-304.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.ocr_tesseract_and_render_pdf</text>
+</g>
+<!-- t4&#45;&gt;t17 -->
+<g id="edge28" class="edge"><title>t4&#45;&gt;t17</title>
+<path fill="none" stroke="#0044a0" d="M688.477,-752.611C947.521,-745.804 1499,-721.841 1499,-641.173 1499,-641.173 1499,-641.173 1499,-455.173 1499,-407.201 1508.25,-383.766 1474,-350.173 1465.98,-342.302 1448.74,-335.804 1427.99,-330.522"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1428.77,-327.11 1418.23,-328.18 1427.14,-333.917 1428.77,-327.11"/>
+</g>
+<!-- t6 -->
+<g id="node7" class="node"><title>t6</title>
+<polygon fill="#efa03b" stroke="black" points="1089.08,-658.173 664.92,-658.173 660.92,-654.173 660.92,-622.173 1085.08,-622.173 1089.08,-626.173 1089.08,-658.173"/>
+<polyline fill="none" stroke="black" points="1085.08,-654.173 660.92,-654.173 "/>
+<polyline fill="none" stroke="black" points="1085.08,-654.173 1085.08,-622.173 "/>
+<polyline fill="none" stroke="black" points="1085.08,-654.173 1089.08,-658.173 "/>
+<text text-anchor="middle" x="875" y="-634.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.preprocess_remove_background</text>
+</g>
+<!-- t5&#45;&gt;t6 -->
+<g id="edge7" class="edge"><title>t5&#45;&gt;t6</title>
+<path fill="none" stroke="#0044a0" d="M875,-680.066C875,-676.375 875,-672.385 875,-668.415"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="878.5,-668.174 875,-658.174 871.5,-668.174 878.5,-668.174"/>
+</g>
+<!-- t11 -->
+<g id="node14" class="node"><title>t11</title>
+<polygon fill="#efa03b" stroke="black" points="1096,-484.173 712.187,-456.173 1096,-428.173 1479.81,-456.173 1096,-484.173"/>
+<text text-anchor="middle" x="1096" y="-450.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.select_visible_page_image</text>
+</g>
+<!-- t5&#45;&gt;t11 -->
+<g id="edge18" class="edge"><title>t5&#45;&gt;t11</title>
+<path fill="none" stroke="#0044a0" d="M1051.07,-680.0px8C1072.76,-674.712 1090.1,-667.576 1098,-658.173 1136.53,-612.317 1122.5,-537.602 1109.06,-493.335"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1112.33,-492.043 1105.96,-483.573 1105.65,-494.164 1112.33,-492.043"/>
+</g>
+<!-- t7 -->
+<g id="node8" class="node"><title>t7</title>
+<polygon fill="#efa03b" stroke="black" points="1039.68,-600.173 720.324,-600.173 716.324,-596.173 716.324,-564.173 1035.68,-564.173 1039.68,-568.173 1039.68,-600.173"/>
+<polyline fill="none" stroke="black" points="1035.68,-596.173 716.324,-596.173 "/>
+<polyline fill="none" stroke="black" points="1035.68,-596.173 1035.68,-564.173 "/>
+<polyline fill="none" stroke="black" points="1035.68,-596.173 1039.68,-600.173 "/>
+<text text-anchor="middle" x="878" y="-576.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.preprocess_deskew</text>
+</g>
+<!-- t6&#45;&gt;t7 -->
+<g id="edge8" class="edge"><title>t6&#45;&gt;t7</title>
+<path fill="none" stroke="#0044a0" d="M875.916,-622.066C876.114,-618.375 876.328,-614.385 876.541,-610.415"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="880.049,-610.347 877.089,-600.174 873.059,-609.972 880.049,-610.347"/>
+</g>
+<!-- t6&#45;&gt;t11 -->
+<g id="edge17" class="edge"><title>t6&#45;&gt;t11</title>
+<path fill="none" stroke="#0044a0" d="M990.695,-622.152C1016.54,-616.369 1039.2,-609.066 1049,-600.173 1078.78,-573.156 1089.74,-526.964 1093.74,-494.403"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1097.24,-494.666 1094.82,-484.351 1090.28,-493.924 1097.24,-494.666"/>
+</g>
+<!-- t8 -->
+<g id="node9" class="node"><title>t8</title>
+<polygon fill="#efa03b" stroke="black" points="1030.88,-542.173 729.117,-542.173 725.117,-538.173 725.117,-506.173 1026.88,-506.173 1030.88,-510.173 1030.88,-542.173"/>
+<polyline fill="none" stroke="black" points="1026.88,-538.173 725.117,-538.173 "/>
+<polyline fill="none" stroke="black" points="1026.88,-538.173 1026.88,-506.173 "/>
+<polyline fill="none" stroke="black" points="1026.88,-538.173 1030.88,-542.173 "/>
+<text text-anchor="middle" x="878" y="-518.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.preprocess_clean</text>
+</g>
+<!-- t7&#45;&gt;t8 -->
+<g id="edge9" class="edge"><title>t7&#45;&gt;t8</title>
+<path fill="none" stroke="#0044a0" d="M878,-564.066C878,-560.375 878,-556.385 878,-552.415"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="881.5,-552.174 878,-542.174 874.5,-552.174 881.5,-552.174"/>
+</g>
+<!-- t7&#45;&gt;t11 -->
+<g id="edge16" class="edge"><title>t7&#45;&gt;t11</title>
+<path fill="none" stroke="#0044a0" d="M995.441,-564.147C1011.26,-558.713 1026.62,-551.566 1040,-542.173 1057.45,-529.919 1070.89,-510.334 1080.22,-493.051"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1083.55,-494.243 1084.99,-483.747 1077.31,-491.053 1083.55,-494.243"/>
+</g>
+<!-- t9 -->
+<g id="node10" class="node"><title>t9</title>
+<polygon fill="#efa03b" stroke="black" points="694.109,-474.173 389.891,-474.173 385.891,-470.173 385.891,-438.173 690.109,-438.173 694.109,-442.173 694.109,-474.173"/>
+<polyline fill="none" stroke="black" points="690.109,-470.173 385.891,-470.173 "/>
+<polyline fill="none" stroke="black" points="690.109,-470.173 690.109,-438.173 "/>
+<polyline fill="none" stroke="black" points="690.109,-470.173 694.109,-474.173 "/>
+<text text-anchor="middle" x="540" y="-450.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.select_ocr_image</text>
+</g>
+<!-- t8&#45;&gt;t9 -->
+<g id="edge10" class="edge"><title>t8&#45;&gt;t9</title>
+<path fill="none" stroke="#0044a0" d="M790.517,-506.09C743.487,-496.907 685.383,-485.561 636.976,-476.109"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="637.602,-472.665 627.116,-474.183 636.26,-479.535 637.602,-472.665"/>
+</g>
+<!-- t8&#45;&gt;t11 -->
+<g id="edge15" class="edge"><title>t8&#45;&gt;t11</title>
+<path fill="none" stroke="#0044a0" d="M934.424,-506.09C958.693,-498.742 987.536,-490.01 1014.09,-481.971"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1015.2,-485.293 1023.75,-479.045 1013.17,-478.593 1015.2,-485.293"/>
+</g>
+<!-- t10 -->
+<g id="node11" class="node"><title>t10</title>
+<polygon fill="#00cc66" stroke="black" points="1440.1,-396.173 1125.9,-396.173 1121.9,-392.173 1121.9,-360.173 1436.1,-360.173 1440.1,-364.173 1440.1,-396.173"/>
+<polyline fill="none" stroke="black" points="1436.1,-392.173 1121.9,-392.173 "/>
+<polyline fill="none" stroke="black" points="1436.1,-392.173 1436.1,-360.173 "/>
+<polyline fill="none" stroke="black" points="1436.1,-392.173 1440.1,-396.173 "/>
+<text text-anchor="middle" x="1281" y="-372.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.ocr_tesseract_hocr</text>
+</g>
+<!-- t9&#45;&gt;t10 -->
+<g id="edge11" class="edge"><title>t9&#45;&gt;t10</title>
+<path fill="none" stroke="#0044a0" d="M631.992,-438.111C655.104,-434.327 679.903,-430.697 703,-428.173 884.404,-408.343 931.549,-425.567 1113,-406.173 1133.45,-403.987 1155.2,-400.994 1176.03,-397.794"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1176.83,-401.211 1186.17,-396.209 1175.75,-394.295 1176.83,-401.211"/>
+</g>
+<!-- t9&#45;&gt;t15 -->
+<g id="edge13" class="edge"><title>t9&#45;&gt;t15</title>
+<path fill="none" stroke="#0044a0" d="M474.47,-438.117C429.651,-426.452 370.0px1,-410.957 323.342,-398.782"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="324.123,-395.369 313.564,-396.237 322.36,-402.143 324.123,-395.369"/>
+</g>
+<!-- t13 -->
+<g id="node12" class="node"><title>t13</title>
+<polygon fill="#00cc66" stroke="black" points="683.043,-328.173 374.957,-328.173 370.957,-324.173 370.957,-292.173 679.043,-292.173 683.043,-296.173 683.043,-328.173"/>
+<polyline fill="none" stroke="black" points="679.043,-324.173 370.957,-324.173 "/>
+<polyline fill="none" stroke="black" points="679.043,-324.173 679.043,-292.173 "/>
+<polyline fill="none" stroke="black" points="679.043,-324.173 683.043,-328.173 "/>
+<text text-anchor="middle" x="527" y="-304.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.render_hocr_page</text>
+</g>
+<!-- t10&#45;&gt;t13 -->
+<g id="edge12" class="edge"><title>t10&#45;&gt;t13</title>
+<path fill="none" stroke="#0044a0" d="M1186.17,-360.113C1162.35,-356.33 1136.79,-352.699 1113,-350.173 929.956,-330.738 880.531,-342.246 693.485,-328.255"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="693.548,-324.75 683.311,-327.48 693.016,-331.73 693.548,-324.75"/>
+</g>
+<!-- t14 -->
+<g id="node17" class="node"><title>t14</title>
+<polygon fill="#00cc66" stroke="black" points="1072.92,-328.173 705.08,-328.173 701.08,-324.173 701.08,-292.173 1068.92,-292.173 1072.92,-296.173 1072.92,-328.173"/>
+<polyline fill="none" stroke="black" points="1068.92,-324.173 701.08,-324.173 "/>
+<polyline fill="none" stroke="black" points="1068.92,-324.173 1068.92,-292.173 "/>
+<polyline fill="none" stroke="black" points="1068.92,-324.173 1072.92,-328.173 "/>
+<text text-anchor="middle" x="887" y="-304.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.render_hocr_debug_page</text>
+</g>
+<!-- t10&#45;&gt;t14 -->
+<g id="edge25" class="edge"><title>t10&#45;&gt;t14</title>
+<path fill="none" stroke="#0044a0" d="M1179.28,-360.133C1124,-350.873 1055.54,-339.405 998.77,-329.896"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="999.118,-326.405 988.677,-328.205 997.962,-333.309 999.118,-326.405"/>
+</g>
+<!-- t16 -->
+<g id="node16" class="node"><title>t16</title>
+<polygon fill="#00cc66" stroke="black" points="636.299,-270.173 345.701,-270.173 341.701,-266.173 341.701,-234.173 632.299,-234.173 636.299,-238.173 636.299,-270.173"/>
+<polyline fill="none" stroke="black" points="632.299,-266.173 341.701,-266.173 "/>
+<polyline fill="none" stroke="black" points="632.299,-266.173 632.299,-234.173 "/>
+<polyline fill="none" stroke="black" points="632.299,-266.173 636.299,-270.173 "/>
+<text text-anchor="middle" x="489" y="-246.173" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.combine_layers</text>
+</g>
+<!-- t13&#45;&gt;t16 -->
+<g id="edge22" class="edge"><title>t13&#45;&gt;t16</title>
+<path fill="none" stroke="#0044a0" d="M515.392,-292.066C512.519,-287.833 509.379,-283.205 506.299,-278.666"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="509.048,-276.483 500.537,-270.174 503.256,-280.414 509.048,-276.483"/>
+</g>
+<!-- t15&#45;&gt;t16 -->
+<g id="edge21" class="edge"><title>t15&#45;&gt;t16</title>
+<path fill="none" stroke="#0044a0" d="M263.466,-359.936C281.645,-340.695 313.598,-309.935 347,-292.173 361.3,-284.568 377.219,-278.339 393.079,-273.27"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="394.389,-276.53 402.932,-270.262 392.345,-269.835 394.389,-276.53"/>
+</g>
+<!-- t11&#45;&gt;t12 -->
+<g id="edge19" class="edge"><title>t11&#45;&gt;t12</title>
+<path fill="none" stroke="#0044a0" d="M1009.92,-434.405C967.591,-424.191 916.524,-411.87 873.373,-401.459"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="873.983,-398.006 863.441,-399.063 872.341,-404.811 873.983,-398.006"/>
+</g>
+<!-- t11&#45;&gt;t14 -->
+<g id="edge24" class="edge"><title>t11&#45;&gt;t14</title>
+<path fill="none" stroke="#0044a0" d="M1334.06,-445.464C1386.49,-438.209 1431.63,-426.12 1449,-406.173 1465.34,-387.403 1466.07,-368.289 1449,-350.173 1422.45,-321.987 1153.75,-330.0px 1083.05,-328.037"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1083,-324.531 1072.85,-327.564 1082.68,-331.524 1083,-324.531"/>
+</g>
+<!-- t11&#45;&gt;t17 -->
+<g id="edge27" class="edge"><title>t11&#45;&gt;t17</title>
+<path fill="none" stroke="#0044a0" d="M1278.64,-441.463C1364.44,-433.122 1450.86,-421.174 1464,-406.173 1480.4,-387.448 1479.49,-369.65 1464,-350.173 1457.83,-342.418 1443.67,-336.074 1426.02,-330.926"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1426.46,-327.418 1415.89,-328.188 1424.63,-334.176 1426.46,-327.418"/>
+</g>
+<!-- t12&#45;&gt;t16 -->
+<g id="edge23" class="edge"><title>t12&#45;&gt;t16</title>
+<path fill="none" stroke="#0044a0" d="M624.426,-363.718C509.174,-353.04 370.992,-338.277 362,-328.173 351.364,-316.22 352.491,-305.04 362,-292.173 366.819,-285.651 372.71,-280.186 379.287,-275.609"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="381.274,-278.496 387.956,-270.274 377.605,-272.534 381.274,-278.496"/>
+</g>
+<!-- t20 -->
+<g id="node21" class="node"><title>t20</title>
+<polygon fill="#efa03b" stroke="black" points="234.918,-156.426 480,-129.117 725.082,-156.426 724.853,-200.614 235.147,-200.614 234.918,-156.426"/>
+<polygon fill="none" stroke="black" points="230.903,-152.847 480,-125.09 729.097,-152.847 728.829,-204.616 231.171,-204.616 230.903,-152.847"/>
+<text text-anchor="middle" x="480" y="-162.639" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.merge_pages_ghostscript</text>
+</g>
+<!-- t16&#45;&gt;t20 -->
+<g id="edge33" class="edge"><title>t16&#45;&gt;t20</title>
+<path fill="none" stroke="#0044a0" d="M487.091,-233.881C486.46,-228.165 485.726,-221.515 484.974,-214.703"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="488.446,-214.253 483.87,-204.697 481.488,-215.021 488.446,-214.253"/>
+</g>
+<!-- t16&#45;&gt;t21 -->
+<g id="edge37" class="edge"><title>t16&#45;&gt;t21</title>
+<path fill="none" stroke="#0044a0" d="M602.543,-234.159C645.082,-227.582 693.805,-219.809 738,-212.173 748.572,-210.346 759.416,-208.431 770.357,-206.468"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="771.206,-209.872 780.427,-204.653 769.964,-202.983 771.206,-209.872"/>
+</g>
+<!-- t14&#45;&gt;t20 -->
+<g id="edge34" class="edge"><title>t14&#45;&gt;t20</title>
+<path fill="none" stroke="#0044a0" d="M837.094,-292.063C776.411,-271.259 672.165,-235.52 591.922,-208.01"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="592.855,-204.63 582.26,-204.697 590.584,-211.251 592.855,-204.63"/>
+</g>
+<!-- t14&#45;&gt;t21 -->
+<g id="edge38" class="edge"><title>t14&#45;&gt;t21</title>
+<path fill="none" stroke="#0044a0" d="M896.319,-292.063C906.85,-272.729 924.406,-240.497 938.861,-213.958"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="942.118,-215.294 943.828,-204.838 935.971,-211.946 942.118,-215.294"/>
+</g>
+<!-- t19&#45;&gt;t20 -->
+<g id="edge30" class="edge"><title>t19&#45;&gt;t20</title>
+<path fill="none" stroke="#0044a0" d="M257.907,-234.081C284.597,-226.336 317.298,-216.848 349.316,-207.558"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="350.595,-210.831 359.224,-204.683 348.645,-204.108 350.595,-210.831"/>
+</g>
+<!-- t19&#45;&gt;t21 -->
+<g id="edge35" class="edge"><title>t19&#45;&gt;t21</title>
+<path fill="none" stroke="#0044a0" d="M323.792,-235.042C326.891,-234.738 329.964,-234.447 333,-234.173 512.532,-217.929 559.043,-233.851 738,-212.173 751.869,-210.493 766.169,-208.472 780.522,-206.237"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="781.136,-209.683 790.462,-204.657 780.037,-202.77 781.136,-209.683"/>
+</g>
+<!-- t17&#45;&gt;t20 -->
+<g id="edge32" class="edge"><title>t17&#45;&gt;t20</title>
+<path fill="none" stroke="#0044a0" d="M1195.47,-292.133C1071.63,-270.93 856.728,-234.138 694.688,-206.395"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="695.028,-202.903 684.581,-204.665 693.846,-209.802 695.028,-202.903"/>
+</g>
+<!-- t17&#45;&gt;t21 -->
+<g id="edge36" class="edge"><title>t17&#45;&gt;t21</title>
+<path fill="none" stroke="#0044a0" d="M1254.29,-292.063C1205.19,-271.428 1121.13,-236.1 1055.9,-208.683"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="1056.99,-205.345 1046.42,-204.697 1054.28,-211.799 1056.99,-205.345"/>
+</g>
+<!-- t18&#45;&gt;t20 -->
+<g id="edge31" class="edge"><title>t18&#45;&gt;t20</title>
+<path fill="none" stroke="#0044a0" d="M120.266,-796.105C70.9512,-780.34 17,-751.202 17,-699.173 17,-699.173 17,-699.173 17,-309.173 17,-269.597 31.8505,-255.79 65,-234.173 92.9782,-215.927 154.085,-202.475 221.024,-192.724"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="221.533,-196.187 230.939,-191.311 220.546,-189.256 221.533,-196.187"/>
+</g>
+<!-- t22 -->
+<g id="node23" class="node"><title>t22</title>
+<polygon fill="#efa03b" stroke="black" points="554.065,-47.332 721,-20.0049 887.935,-47.332 887.779,-91.5483 554.221,-91.5483 554.065,-47.332"/>
+<polygon fill="none" stroke="black" points="550.054,-43.9306 721,-15.9468 891.946,-43.9306 891.764,-95.5532 550.236,-95.5532 550.054,-43.9306"/>
+<text text-anchor="middle" x="721" y="-53.5531" font-family="Times,serif" font-size="20.00">ocrmypdf.pipeline.copy_final</text>
+</g>
+<!-- t20&#45;&gt;t22 -->
+<g id="edge41" class="edge"><title>t20&#45;&gt;t22</title>
+<path fill="none" stroke="#0044a0" d="M556.585,-133.609C580.784,-122.857 607.72,-110.888 632.67,-99.8018"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="634.408,-102.859 642.125,-95.6004 631.566,-96.4625 634.408,-102.859"/>
+</g>
+<!-- t21&#45;&gt;t22 -->
+<g id="edge40" class="edge"><title>t21&#45;&gt;t22</title>
+<path fill="none" stroke="#0044a0" d="M888.417,-134.636C863.556,-123.635 835.613,-111.27 809.783,-99.8401"/>
+<polygon fill="#0044a0" stroke="#0044a0" points="810.933,-96.5216 800.372,-95.6756 808.1,-102.923 810.933,-96.5216"/>
+</g>
+</g>
+</svg>
@@ -0,0 +1,44 @@
+PDF Security Issues
+===================
+
+	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.
+
+The disclaimer applies: this software has no warranties of any kind.
+
+PDFs may contain malware
+------------------------
+
+PDF is a rich, complex file format. The official PDF 1.7 specification, ISO 32000:2008, is hundreds of packages long and references several annexes each of which are similar in length. PDFs can contain video, audio, JavaScript and other programming, and forms. In some cases, they can open internet connections to pre-selected URLs. All of these possible attack vectors.
+
+In short, PDFs `may contain viruses <https://security.stackexchange.com/questions/64052/can-a-pdf-file-contain-a-virus>`_.
+
+This `article <https://theinvisiblethings.blogspot.ca/2013/02/converting-untrusted-pdfs-into-trusted.html>`_ describes a high-paranoia method which allows potentially hostile PDFs to be viewed and rasterized safely in a disposable virtual machine. A trusted PDF created in this manner is converted to images and loses all information making it searchable and losing all compression. OCRmyPDF could be used restore searchability.
+
+How OCRmyPDF processes PDFs
+---------------------------
+
+OCRmyPDF must open and interpret your PDF in order to insert an OCR layer. First, it runs all PDFs through `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.
+
+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.
+
+Using OCRmyPDF online
+---------------------
+
+OCRmyPDF is not designed to be deployed "as a service", in a setting where a user/attacker could upload a file for OCR processing online. It is not designed to be secure in this case.
+
+Abbyy Cloud OCR is a viable commercial alternative with a web services API. The author also provides professional services that include OCR and building databases around PDFs, and is happy to provide consultation.
+
+Password protection, digital signatures and certification
+---------------------------------------------------------
+
+OCRmyPDF cannot remove password protection from a PDF. ``qpdf``, one of its dependencies, has this capability. After OCR is applied, password protection is not permitted on PDF/A documents but the file can be converted to regular PDF.
+
+Many programs exist which are capable of inserting an image of someone's signature. On its own, this offers no security guarantees. It is trivial to remove the signature image and apply it to other files. This practice offers no real security.
+
+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. 
@@ -1,34 +1,15 @@
-from enum import IntEnum
-import os
-from collections.abc import Iterable
 import pkg_resources

+PROGRAM_NAME = 'ocrmypdf'
+
 VERSION = pkg_resources.get_distribution('ocrmypdf').version


-class ExitCode(IntEnum):
-    ok = 0
-    bad_args = 1
-    input_file = 2
-    missing_dependency = 3
-    invalid_output_pdf = 4
-    file_access_error = 5
-    already_done_ocr = 6
-    child_process_error = 7
-    encrypted_pdf = 8
-    other_error = 15
-    ctrl_c = 130
+# These imports are for v4.x backward compatibility for consumers of ocrmypdf
+# (if any). They are deprecated and will be removed in v5.x.
+from .exec import ghostscript, qpdf, tesseract, unpaper, get_program
+from .exceptions import ExitCode
+from .helpers import page_number, is_iterable_notstr


-def get_program(name):
-    envvar = 'OCRMYPDF_' + name.upper()
-    return os.environ.get(envvar, name)
-
-
-def page_number(input_file):
-    return int(os.path.basename(input_file)[0:6])
-
-
-def is_iterable_notstr(thing):
-    return isinstance(thing, Iterable) and not isinstance(thing, str)

@@ -0,0 +1,59 @@
+#!/usr/bin/env python3
+# © 2016 James R. Barlow: github.com/jbarlow83
+
+
+from enum import IntEnum
+
+class ExitCode(IntEnum):
+    ok = 0
+    bad_args = 1
+    input_file = 2
+    missing_dependency = 3
+    invalid_output_pdf = 4
+    file_access_error = 5
+    already_done_ocr = 6
+    child_process_error = 7
+    encrypted_pdf = 8
+    invalid_config = 9
+    other_error = 15
+    ctrl_c = 130
+
+
+class ExitCodeException(Exception):
+    pass
+
+
+class PdfMergeFailedError(ExitCodeException):
+    exit_code = ExitCode.input_file
+
+
+class MissingDependencyError(ExitCodeException):
+    exit_code = ExitCode.missing_dependency
+
+
+class UnsupportedImageFormatError(ExitCodeException):
+    exit_code = ExitCode.input_file
+
+
+class DpiError(ExitCodeException):
+    exit_code = ExitCode.input_file
+
+
+class PriorOcrFoundError(ExitCodeException):
+    exit_code = ExitCode.already_done_ocr
+
+
+class InputFileError(ExitCodeException):
+    exit_code = ExitCode.input_file
+
+
+class SubprocessOutputError(ExitCodeException):
+    exit_code = ExitCode.other_error
+
+
+class EncryptedPdfError(ExitCodeException):
+    exit_code = ExitCode.encrypted_pdf
+
+
+class TesseractConfigError(ExitCodeException):
+    exit_code = ExitCode.invalid_config
@@ -0,0 +1,11 @@
+#!/usr/bin/env python3
+# © 2016 James R. Barlow: github.com/jbarlow83
+
+"""Wrappers to manage subprocess calls"""
+
+import os
+
+
+def get_program(name):
+    envvar = 'OCRMYPDF_' + name.upper()
+    return os.environ.get(envvar, name)
@@ -0,0 +1,102 @@
+#!/usr/bin/env python3
+# © 2015 James R. Barlow: github.com/jbarlow83
+
+from tempfile import NamedTemporaryFile
+from subprocess import Popen, PIPE, STDOUT, check_call, CalledProcessError, \
+    check_output
+from shutil import copy
+from functools import lru_cache
+from . import get_program
+from ..pdfa import SRGB_ICC_PROFILE
+
+
+@lru_cache(maxsize=1)
+def version():
+    args_gs = [
+        get_program('gs'),
+        '--version'
+    ]
+    try:
+        version = check_output(
+                args_gs, close_fds=True, universal_newlines=True,
+                stderr=STDOUT)
+    except CalledProcessError as e:
+        print("Could not find Ghostscript executable on system PATH.",
+              file=sys.stderr)
+        raise MissingDependencyError from e
+
+    return version.strip()
+
+
+def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
+                  pageno=1):
+    with NamedTemporaryFile(delete=True) as tmp:
+        args_gs = [
+            get_program('gs'),
+            '-dQUIET',
+            '-dSAFER',
+            '-dBATCH',
+            '-dNOPAUSE',
+            '-sDEVICE=%s' % raster_device,
+            '-dFirstPage=%i' % pageno,
+            '-dLastPage=%i' % pageno,
+            '-o', tmp.name,
+            '-r{0}x{1}'.format(str(round(xres)), str(round(yres))),
+            input_file
+        ]
+
+        p = Popen(args_gs, close_fds=True, stdout=PIPE, stderr=STDOUT,
+                  universal_newlines=True)
+        stdout, _ = p.communicate()
+        if 'error' in stdout:
+            log.error(stdout)  # Ghostscript puts errors in stdout
+        else:
+            log.debug(stdout)
+
+        if p.returncode == 0:
+            copy(tmp.name, output_file)
+        else:
+            log.error('Ghostscript rendering failed')
+
+
+def generate_pdfa(pdf_pages, output_file, log, threads=1):
+    with NamedTemporaryFile(delete=True) as gs_pdf:
+        args_gs = [
+            get_program("gs"),
+            "-dQUIET",
+            "-dBATCH",
+            "-dNOPAUSE",
+            '-dNumRenderingThreads=' + str(threads),
+            "-sDEVICE=pdfwrite",
+            "-dAutoRotatePages=/None",
+            "-sColorConversionStrategy=/RGB",
+            "-sProcessColorModel=DeviceRGB",
+            "-dJPEGQ=95",
+            "-dPDFA=2",
+            "-dPDFACompatibilityPolicy=1",
+            "-sOutputFile=" + gs_pdf.name,
+        ]
+        args_gs.extend(pdf_pages)
+        p = Popen(args_gs, close_fds=True, stdout=PIPE, stderr=STDOUT,
+                  universal_newlines=True)
+        stdout, _ = p.communicate()
+
+        if 'error' in stdout:
+            log.error(stdout)
+        elif 'overprint mode not set' in stdout:
+            # Unless someone is going to print PDF/A documents on a
+            # magical sRGB printer I can't see the removal of overprinting
+            # being a problem....
+            log.debug(
+                "Ghostscript had to remove PDF 'overprinting' from the "
+                "input file to complete PDF/A conversion. "
+                )
+        else:
+            log.debug(stdout)
+
+        if p.returncode == 0:
+            # Ghostscript does not change return code when it fails to create
+            # PDF/A - check PDF/A status elsewhere
+            copy(gs_pdf.name, output_file)
+        else:
+            log.error('Ghostscript PDF/A failed')
@@ -7,7 +7,9 @@ import sys
 import os
 import re

-from . import ExitCode, get_program
+from ..exceptions import InputFileError, SubprocessOutputError, \
+    MissingDependencyError, EncryptedPdfError
+from . import get_program


@lru_cache(maxsize=1)
@@ -20,21 +22,25 @@ def version():
        versions = check_output(
            args_qpdf, close_fds=True, universal_newlines=True,
            stderr=STDOUT)
-    except CalledProcessError:
-        print("Could not find qpdf executable on system PATH.")
-        sys.exit(ExitCode.missing_dependency)
+    except CalledProcessError as e:
+        print("Could not find qpdf executable on system PATH.",
+              file=sys.stderr)
+        raise MissingDependencyError() from e

    qpdf_version = re.match(r'qpdf version (.+)', versions).group(1)
    return qpdf_version


-def check(input_file, log):
+def check(input_file, log=None):
    args_qpdf = [
        get_program('qpdf'),
        '--check',
        input_file
    ]

+    if log is None:
+        import logging as log
+
    try:
        check_output(args_qpdf, stderr=STDOUT, universal_newlines=True)
    except CalledProcessError as e:
@@ -52,6 +58,14 @@ def check(input_file, log):
    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 = [
        get_program('qpdf'), input_file, output_file
@@ -64,20 +78,20 @@ def repair(input_file, output_file, log):
            log.debug(e.output)
            return

-        if e.returncode == 2 and e.output.find("invalid password"):
+        if _probably_encrypted(e):
            log.error("{0}: this PDF is password-protected - password must "
                      "be removed for OCR".format(input_file))
-            sys.exit(ExitCode.input_file)
+            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)
-            sys.exit(ExitCode.input_file)
+            raise InputFileError() from e
        else:
            log.error("{0}: unknown error".format(
                      input_file))
            log.error(e.output)
-            sys.exit(ExitCode.unknown)
+            raise SubprocessOutputError() from e


 def get_npages(input_file, log):
@@ -88,7 +102,7 @@ def get_npages(input_file, log):
    except CalledProcessError as e:
        if e.returncode == 2 and e.output.find('No such file'):
            log.error(e.output)
-            sys.exit(ExitCode.input_file)
+            raise InputFileError() from e
    return int(pages)


@@ -6,8 +6,11 @@ import os
 import re
 import shutil
 from functools import lru_cache
-from . import ExitCode, get_program, page_number
+from ..exceptions import MissingDependencyError, TesseractConfigError
+from ..helpers import page_number
+from . import get_program
 from collections import namedtuple
+from textwrap import dedent

 from subprocess import Popen, PIPE, CalledProcessError, \
    TimeoutExpired, check_output, STDOUT, DEVNULL
@@ -50,14 +53,47 @@ def version():
        versions = check_output(
                args_tess, close_fds=True, universal_newlines=True,
                stderr=STDOUT)
-    except CalledProcessError:
-        print("Could not find Tesseract executable on system PATH.")
-        sys.exit(ExitCode.missing_dependency)
+    except CalledProcessError as e:
+        print("Could not find Tesseract executable on system PATH.",
+              file=sys.stderr)
+        raise MissingDependencyError from e

    tesseract_version = re.match(r'tesseract\s(.+)', versions).group(1)
    return tesseract_version


+def v4():
+    "Is this Tesseract v4.0?"
+    return (version() >= '4')
+
+
+def has_textonly_pdf():
+    if version() == '4.00.00alpha':
+        # textonly_pdf added during the 4.00.00alpha cycle, so we must test
+        # more carefully to see if it is present
+        args_tess = [
+            get_program('tesseract'),
+            '--print-parameters'
+        ]
+        try:
+            params = check_output(
+                    args_tess, close_fds=True, universal_newlines=True,
+                    stderr=STDOUT)
+        except CalledProcessError as e:
+            print("Could not --print-parameters from tesseract",
+                  file=sys.stderr)
+            raise MissingDependencyError from e
+        if 'textonly_pdf' in params:
+            return True
+    else:
+        return v4()
+
+
+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 = [
@@ -69,19 +105,31 @@ def languages():
                args_tess, close_fds=True, universal_newlines=True,
                stderr=STDOUT)
    except CalledProcessError as e:
-        print("Tesseract failed to report available languages.")
-        print("Output from Tesseract:")
-        print("-" * 40)
-        print(e.output)
-        sys.exit(ExitCode.missing_dependency)
+        msg = dedent("""Tesseract failed to report available languages.
+        Output from Tesseract:
+        -----------
+        """)
+        msg += e.output
+        print(msg, file=sys.stderr)
+        raise MissingDependencyError from e
    return set(lang.strip() for lang in langs.splitlines()[1:])


-def get_orientation(input_file, language: list, timeout: float, log):
-    args_tesseract = [
+def tess_base_args(languages, engine_mode):
+    args = [
        get_program('tesseract'),
-        '-l', '+'.join(language),
-        '-psm', '0',
+    ]
+    if languages:
+        args.extend(['-l', '+'.join(languages)])
+    if engine_mode is not None and v4():
+        args.extend(['--oem', str(engine_mode)])
+    return args
+
+
+def get_orientation(input_file, language: list, engine_mode, timeout: float,
+                    log):
+    args_tesseract = tess_base_args(language, engine_mode) + [
+        psm(), '0',
        input_file,
        'stdout'
    ]
@@ -138,6 +186,8 @@ def tesseract_log_output(log, stdout, input_file):
            log.warning(prefix + "unsure about page orientation")
        elif 'error' in line.lower() or 'exception' in line.lower():
            log.error(prefix + line.strip())
+        elif 'read_params_file' in line.lower():
+            log.error(prefix + line.strip())
        else:
            log.info(prefix + line.strip())

@@ -159,18 +209,16 @@ def _generate_null_hocr(output_hocr, image):
        f.write(HOCR_TEMPLATE.format(w, h))


-def generate_hocr(input_file, output_hocr, language: list, tessconfig: list,
-                  timeout: float, pageinfo_getter, pagesegmode: int, log):
+def generate_hocr(input_file, output_hocr, language: list, engine_mode,
+                  tessconfig: list,
+                  timeout: float, pagesegmode: int, log):

    badxml = os.path.splitext(output_hocr)[0] + '.badxml'

-    args_tesseract = [
-        get_program('tesseract'),
-        '-l', '+'.join(language)
-    ]
+    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)])

    args_tesseract.extend([
        input_file,
@@ -178,6 +226,7 @@ def generate_hocr(input_file, output_hocr, language: list, tessconfig: list,
        'hocr'
    ] + tessconfig)
    try:
+        log.debug(args_tesseract)
        stdout = check_output(
            args_tesseract, close_fds=True, stderr=STDOUT,
            universal_newlines=True, timeout=timeout)
@@ -189,6 +238,8 @@ def generate_hocr(input_file, output_hocr, language: list, tessconfig: list,
        _generate_null_hocr(output_hocr, input_file)
    except CalledProcessError as e:
        tesseract_log_output(log, e.output, input_file)
+        if 'read_params_file: parameter not found' in e.output:
+            raise TesseractConfigError() from e
        if 'Image too large' in e.output:
            _generate_null_hocr(output_hocr, input_file)
            return
@@ -196,6 +247,7 @@ def generate_hocr(input_file, output_hocr, language: list, tessconfig: list,
        raise e from e
    else:
        tesseract_log_output(log, stdout, input_file)
+
        if os.path.exists(badxml + '.html'):
            # Tesseract 3.02 appends suffix ".html" on its own (.badxml.html)
            shutil.move(badxml + '.html', badxml)
@@ -219,24 +271,28 @@ def generate_hocr(input_file, output_hocr, language: list, tessconfig: list,


 def generate_pdf(input_image, skip_pdf, output_pdf, language: list,
+                 engine_mode, text_only: bool,
                 tessconfig: list, timeout: float, pagesegmode: int, log):
    '''Use Tesseract to render a PDF.

    input_image -- image to analyze
    skip_pdf -- if we time out, use this file as output
+    output_pdf -- file to generate
    language -- list of languages to consider
+    engine_mode -- engine mode argument for tess v4
+    text_only -- enable tesseract text only mode?
    tessconfig -- tesseract configuration
    timeout -- timeout (seconds)
    log -- logger object
    '''

-    args_tesseract = [
-        get_program('tesseract'),
-        '-l', '+'.join(language)
-    ]
+    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:
+        args_tesseract.extend(['-c', 'textonly_pdf=1'])

    args_tesseract.extend([
        input_image,
@@ -245,6 +301,7 @@ def generate_pdf(input_image, skip_pdf, output_pdf, language: list,
    ] + tessconfig)

    try:
+        log.debug(args_tesseract)
        stdout = check_output(
            args_tesseract, close_fds=True, stderr=STDOUT,
            universal_newlines=True, timeout=timeout)
@@ -253,6 +310,9 @@ def generate_pdf(input_image, skip_pdf, output_pdf, language: list,
        shutil.copy(skip_pdf, output_pdf)
    except CalledProcessError as e:
        tesseract_log_output(log, e.output, input_image)
+        if 'read_params_file: parameter not found' in e.output:
+            raise TesseractConfigError() from e
+
        if 'Image too large' in e.output:
            shutil.copy(skip_pdf, output_pdf)
            return
@@ -8,7 +8,8 @@ from tempfile import NamedTemporaryFile
 import sys
 import os
 from functools import lru_cache
-from . import ExitCode, get_program
+from ..exceptions import MissingDependencyError
+from . import get_program


@lru_cache(maxsize=1)
@@ -47,17 +48,17 @@ def run(input_file, output_file, dpi, log, mode_args):
                im = im.convert(mode='1')
            else:
                im = im.convert(mode='RGB')
-        except IOError:
+        except IOError as e:
            log.error(
                    "Could not convert image with type " + im.mode)
-            sys.exit(ExitCode.missing_dependency)
+            raise MissingDependencyError() from e

    try:
        suffix = SUFFIXES[im.mode]
    except KeyError:
        log.error(
                "Failed to convert image to a supported format.")
-        sys.exit(ExitCode.missing_dependency)
+        raise MissingDependencyError() from e

    with NamedTemporaryFile(suffix=suffix) as input_pnm, \
            NamedTemporaryFile(suffix=suffix, mode="r+b") as output_pnm:
@@ -77,7 +78,8 @@ def run(input_file, output_file, dpi, log, mode_args):
            raise e from e
        else:
            log.debug(stdout)
-            Image.open(output_pnm.name).save(output_file)
+            # unpaper sets dpi to 72
+            Image.open(output_pnm.name).save(output_file, dpi=(dpi, dpi))


 def deskew(input_file, output_file, dpi, log):
@@ -1,64 +0,0 @@
-#!/usr/bin/env python3
-# © 2015 James R. Barlow: github.com/jbarlow83
-
-from tempfile import NamedTemporaryFile
-from subprocess import Popen, PIPE, check_call
-from shutil import copy
-from . import get_program
-from .pdfa import SRGB_ICC_PROFILE
-
-
-def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
-                  pageno=1):
-    with NamedTemporaryFile(delete=True) as tmp:
-        args_gs = [
-            get_program('gs'),
-            '-dQUIET',
-            '-dSAFER',
-            '-dBATCH',
-            '-dNOPAUSE',
-            '-sDEVICE=%s' % raster_device,
-            '-dFirstPage=%i' % pageno,
-            '-dLastPage=%i' % pageno,
-            '-o', tmp.name,
-            '-r{0}x{1}'.format(str(round(xres)), str(round(yres))),
-            input_file
-        ]
-
-        p = Popen(args_gs, close_fds=True, stdout=PIPE, stderr=PIPE,
-                  universal_newlines=True)
-        stdout, stderr = p.communicate()
-        if stdout:
-            if 'error' in stdout:
-                log.error(stdout)  # Ghostscript puts errors in stdout
-            else:
-                log.debug(stdout)
-        if stderr:
-            log.error(stderr)
-
-        if p.returncode == 0:
-            copy(tmp.name, output_file)
-        else:
-            log.error('Ghostscript rendering failed')
-
-
-def generate_pdfa(pdf_pages, output_file, threads=1):
-    with NamedTemporaryFile(delete=True) as gs_pdf:
-        args_gs = [
-            get_program("gs"),
-            "-dQUIET",
-            "-dBATCH",
-            "-dNOPAUSE",
-            '-dNumRenderingThreads=' + str(threads),
-            "-sDEVICE=pdfwrite",
-            "-dAutoRotatePages=/None",
-            "-sColorConversionStrategy=/RGB",
-            "-sProcessColorModel=DeviceRGB",
-            "-dJPEGQ=95",
-            "-dPDFA=2",
-            "-sPDFACompatibilityPolicy=2",
-            "-sOutputFile=" + gs_pdf.name,
-        ]
-        args_gs.extend(pdf_pages)
-        check_call(args_gs)
-        copy(gs_pdf.name, output_file)
@@ -0,0 +1,79 @@
+#!/usr/bin/env python3
+# © 2016 James R. Barlow: github.com/jbarlow83
+
+from functools import partial
+from collections.abc import Iterable
+from contextlib import suppress
+import sys
+import os
+
+
+def re_symlink(input_file, soft_link_name, log=None):
+    """
+    Helper function: relinks soft symbolic link if necessary
+    """
+
+    if log is None:
+        prdebug = partial(print, file=sys.stderr)
+    else:
+        prdebug = log.debug
+
+    # 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.")
+        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)
+        try:
+            os.unlink(soft_link_name)
+        except:
+            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)
+
+    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
+    )
+
+
+def is_iterable_notstr(thing):
+    return isinstance(thing, Iterable) and not isinstance(thing, str)
+
+
+def page_number(input_file):
+    return int(os.path.basename(input_file)[0:6])
+
+
+def is_file_writable(test_file):
+    """Intentionally racy test if target is writable.
+
+    We intend to write to the output file if and only if we succeed and
+    can replace it atomically. Before doing the OCR work, make sure
+    the location is writable.
+    """
+    if os.path.exists(test_file):
+        return os.access(
+            test_file, os.W_OK,
+            effective_ids=(os.access in os.supports_effective_ids))
+    else:
+        try:
+            fp = open(test_file, 'wb')
+        except OSError as e:
+            return False
+        else:
+            fp.close()
+            with suppress(OSError):
+                os.unlink(test_file)
+        return True
@@ -0,0 +1 @@
+"""Bindings to external libraries"""
@@ -1,8 +1,8 @@
 from cffi import FFI

-ffi = FFI()
-ffi.set_source("ocrmypdf.lib._leptonica", None)
-ffi.cdef("""
+ffibuilder = FFI()
+ffibuilder.set_source("ocrmypdf.lib._leptonica", None)
+ffibuilder.cdef("""
 typedef signed char             l_int8;
 typedef unsigned char           l_uint8;
 typedef short                   l_int16;
@@ -56,7 +56,7 @@ typedef struct Box    BOX;

 """)

-ffi.cdef("""
+ffibuilder.cdef("""
 PIX * pixRead ( const char *filename );
 PIX * pixScale ( PIX *pixs, l_float32 scalex, l_float32 scaley );
 l_int32 pixFindSkew ( PIX *pixs, l_float32 *pangle, l_float32 *pconf );
@@ -189,4 +189,4 @@ void lept_free(void *ptr);


 if __name__ == '__main__':
-    ffi.compile()
+    ffibuilder.compile(verbose=True)
@@ -2,7 +2,7 @@
 # © 2015 James R. Barlow: github.com/jbarlow83

 from subprocess import Popen, PIPE
-from decimal import Decimal, getcontext
+from decimal import Decimal
 from math import hypot
 import re
 import sys
@@ -89,16 +89,32 @@ def _is_unit_square(shorthand):
    pairwise = zip(values, UNIT_SQUARE)
    return all([isclose(a, b, rel_tol=1e-3) for a, b in pairwise])

-RasterSettings = namedtuple('RasterSettings',
+XobjectSettings = namedtuple('XobjectSettings',
    ['name', 'shorthand', 'stack_depth'])

 InlineSettings = namedtuple('InlineSettings',
    ['settings', 'shorthand', 'stack_depth'])

-ContentsInfo = namedtuple('ContentsInfo', ['raster_settings', 'inline_images'])
+ContentsInfo = namedtuple('ContentsInfo', ['xobject_settings', 'inline_images'])


-def _interpret_contents(contentstream):
+def _normalize_stack(operations):
+    """Fix runs of qQ's in the stack
+
+    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
+        else:
+            yield (operands, command)
+
+
+def _interpret_contents(contentstream, initial_shorthand=UNIT_SQUARE):
    """Interpret the PDF content stream

    The stack represents the state of the PDF graphics stack.  We are only
@@ -123,27 +139,32 @@ def _interpret_contents(contentstream):

    operations = contentstream.operations
    stack = []
-    ctm = _matrix_from_shorthand(UNIT_SQUARE)
-    image_raster_settings = []
+    ctm = _matrix_from_shorthand(initial_shorthand)
+    xobject_settings = []
    inline_images = []

-    for op in operations:
+    for n, op in enumerate(_normalize_stack(operations)):
        operands, command = op
        if command == b'q':
            stack.append(ctm)
            if len(stack) > 32:
-                raise RuntimeError("PDF graphics stack overflow")
+                raise RuntimeError(
+                    "PDF graphics stack overflow, command %i" % n)
        elif command == b'Q':
-            ctm = stack.pop()
+            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':
            image_name = operands[0]
-            raster = RasterSettings(
+            settings = XobjectSettings(
                name=image_name, shorthand=_shorthand_from_matrix(ctm),
                stack_depth=len(stack))
-            image_raster_settings.append(raster)
+            xobject_settings.append(settings)
        elif command == b'INLINE IMAGE':
            settings = operands['settings']
            inline = InlineSettings(
@@ -152,7 +173,7 @@ def _interpret_contents(contentstream):
            inline_images.append(inline)

    return ContentsInfo(
-        raster_settings=image_raster_settings,
+        xobject_settings=xobject_settings,
        inline_images=inline_images)


@@ -220,8 +241,8 @@ def _get_dpi(ctm_shorthand, image_size):
    return (dpi_w, dpi_h)


-def _find_page_inline_images(page, pageinfo, contentsinfo):
-    "Find inline images on the page"
+def _find_inline_images(contentsinfo):
+    "Find inline images in the contentstream"

    for n, inline in enumerate(contentsinfo.inline_images):
        image = {}
@@ -232,7 +253,10 @@ def _find_page_inline_images(page, pageinfo, contentsinfo):
            image['bpc'] = inline.settings['/BPC']
        else:
            image['bpc'] = 8
-        image['color'] = FRIENDLY_COLORSPACE.get(inline.settings['/CS'], '-')
+        if '/CS' in inline.settings:
+            image['color'] = FRIENDLY_COLORSPACE.get(inline.settings['/CS'], '-')
+        else:
+            image['color'] = '-'
        image['comp'] = FRIENDLY_COMP.get(image['color'], '?')
        if '/F' in inline.settings:
            filter_ = inline.settings['/F']
@@ -248,20 +272,44 @@ def _find_page_inline_images(page, pageinfo, contentsinfo):
        yield image


-def _find_page_regular_images(page, pageinfo, contentsinfo):
-    "Find images stored in XObject resources"
+def _image_xobjects(container):
+    """Search for all XObject-based images in the container

-    try:
-        page['/Resources']['/XObject']
-    except KeyError:
+    Usually the container is a page, but it could also be a Form XObject
+    that contains images. Filter out the Form XObjects which are dealt with
+    elsewhere.
+
+    Generate a sequence of tuples (image, xobj container), where container,
+    where xobj is the name of the object and image is the object itself,
+    since the object does not know its own name.
+
+    """
+
+    if '/Resources' not in container:
        return
-    for xobj in page['/Resources']['/XObject']:
-        # PyPDF2 returns the keys as an iterator
-        pdfimage = page['/Resources']['/XObject'][xobj]
-        if pdfimage['/Subtype'] != '/Image':
-            continue
+    resources = container['/Resources']
+    if '/XObject' not in resources:
+        return
+    for xobj in resources['/XObject']:
+        candidate = resources['/XObject'][xobj]
+        if candidate['/Subtype'] == '/Image':
+            image = candidate
+            yield (image, xobj)
+
+
+def _find_regular_images(container, contentsinfo):
+    """Find images stored in the container's /Resources /XObject
+
+    Usually the container is a page, but it could also be a Form XObject
+    that contains images.
+
+    Generates images with their DPI at time of drawing.
+
+    """
+
+    for pdfimage, xobj in _image_xobjects(container):
        image = {}
-        image['name'] = str(xobj)
+        image['name'] = xobj
        image['width'] = pdfimage['/Width']
        image['height'] = pdfimage['/Height']
        if '/BitsPerComponent' in pdfimage:
@@ -305,12 +353,12 @@ def _find_page_regular_images(page, pageinfo, contentsinfo):

        image['dpi_w'] = image['dpi_h'] = 0

-        for raster in contentsinfo.raster_settings:
+        for xobj in contentsinfo.xobject_settings:
            # Loop in case the same image is display multiple times on a page
-            if raster.name != image['name']:
+            if xobj.name != image['name']:
                continue

-            if raster.stack_depth == 0 and _is_unit_square(raster.shorthand):
+            if xobj.stack_depth == 0 and _is_unit_square(xobj.shorthand):
                # At least one PDF in the wild (and test suite) draws an image
                # when the graphics stack depth is 0, meaning that the image
                # gets drawn into a square of 1x1 PDF units (or 1/72",
@@ -319,7 +367,7 @@ def _find_page_regular_images(page, pageinfo, contentsinfo):
                continue

            dpi_w, dpi_h = _get_dpi(
-                raster.shorthand, (image['width'], image['height']))
+                xobj.shorthand, (image['width'], image['height']))

            # When image is used multiple times take the highest DPI it is
            # rendered at
@@ -327,16 +375,92 @@ def _find_page_regular_images(page, pageinfo, contentsinfo):
            image['dpi_h'] = max(dpi_h, image.get('dpi_h', 0))

        DPI_PREC = Decimal('1.000')
+        dpi = Decimal(image['dpi_w'] * image['dpi_h']).sqrt()
        image['dpi_w'] = Decimal(image['dpi_w']).quantize(DPI_PREC)
        image['dpi_h'] = Decimal(image['dpi_h']).quantize(DPI_PREC)
-        dpi = Decimal(image['dpi_w'] * image['dpi_h']).sqrt()
        image['dpi'] = dpi.quantize(DPI_PREC)
        yield image


-def _find_page_images(page, pageinfo, contentsinfo):
-    yield from _find_page_inline_images(page, pageinfo, contentsinfo)
-    yield from _find_page_regular_images(page, pageinfo, contentsinfo)
+def _find_form_xobject_images(pdf, container, contentsinfo):
+    """Find any images that are in Form XObjects in the container
+
+    The container may be a page, or a parent Form XObject.
+
+    """
+    if '/Resources' not in container:
+        return
+    resources = container['/Resources']
+    if '/XObject' not in resources:
+        return
+    for xobj in resources['/XObject']:
+        candidate = resources['/XObject'][xobj]
+        if candidate['/Subtype'] != '/Form':
+            continue
+
+        form_xobject = candidate
+        for settings in contentsinfo.xobject_settings:
+            if settings.name != xobj:
+                continue
+
+            # Find images once for each time this Form XObject is drawn.
+            # This could be optimized to cache the multiple drawing events
+            # 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, form_xobject, ctm_shorthand)
+
+
+def _find_images(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.
+
+    On a typical page images are stored inline or as regular images
+    in an XObject.
+
+    Form XObjects may include inline images, XObject images,
+    and recursively, other Form XObjects; and also vector drawing commands.
+
+    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
+    on one page it may be drawn at differing resolutions, and our objective
+    is to find the resolution at which the page can be rastered without
+    downsampling.
+
+    """
+
+    if container.get('/Type') == '/Page':
+        # 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)
+
+        # 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)
+
+        # 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))
+
+        # 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)
+    else:
+        return
+
+    contentsinfo = _interpret_contents(contentstream, initial_shorthand)
+
+    yield from _find_inline_images(contentsinfo)
+    yield from _find_regular_images(container, contentsinfo)
+    yield from _find_form_xobject_images(pdf, container, contentsinfo)


 def _page_has_text(pdf, page):
@@ -381,15 +505,8 @@ def _pdf_get_pageinfo(infile, pageno: int):
    except KeyError:
        pageinfo['rotate'] = 0

-    try:
-        contentstream = pypdf.pdf.ContentStream(page.getContents(), pdf)
-    except AttributeError as e:
-        return pageinfo
-
-    contentsinfo = _interpret_contents(contentstream)
-    pageinfo['images'] = [im for im in _find_page_images(
-                                page, pageinfo, contentsinfo)]
-
+    pageinfo['images'] = [im for im in
+                          _find_images(pdf, page)]
    if pageinfo['images']:
        xres = max(image['dpi_w'] for image in pageinfo['images'])
        yres = max(image['dpi_h'] for image in pageinfo['images'])
@@ -404,7 +521,6 @@ def _pdf_get_pageinfo(infile, pageno: int):

 def pdf_get_all_pageinfo(infile):
    pdf = pypdf.PdfFileReader(infile)
-    getcontext().prec = 6
    return [_pdf_get_pageinfo(infile, n) for n in range(pdf.numPages)]


@@ -3,7 +3,6 @@
 #
 # Generate a PDFA_def.ps file for Ghostscript >= 9.14

-from __future__ import print_function, absolute_import, division
 from string import Template
 import codecs
 import pkg_resources
@@ -19,10 +18,9 @@ SRGB_ICC_PROFILE = pkg_resources.resource_filename(
 # files, from the Ghostscript documentation. Lines beginning with % are
 # comments. Python substitution variables have a '$' prefix.
 pdfa_def_template = u"""%!
-% This is a sample prefix file for creating a PDF/A document.
-% Feel free to modify entries marked with "Customize".
-% This assumes an ICC profile to reside in the file (ISO Coated sb.icc),
-% unless the user modifies the corresponding line below.
+% 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)
@@ -71,7 +69,7 @@ def encode_text_string(s: str) -> str:
    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 fall could be encoded as PdfDocEncoding literals provided
+    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.
    '''
@@ -140,21 +138,20 @@ def file_claims_pdfa(filename):
                'conformance': 'No XMP metadata'}

    pdfa_dict = {attr.localName: attr.value for attr in pdfa_nodes}
-    pdfa_dict['pass'] = False
-    pdfa_dict['output'] = 'pdf'
-    if pdfa_dict:
-        part_conformance = pdfa_dict['part'] + pdfa_dict['conformance']
-        valid_part_conforms = {'1A', '1B', '2A', '2B', '2U', '3A', '3B', '3U'}
+    if not pdfa_dict:
+        return {'pass': False, 'output': 'pdf',
+                'conformance': 'No XMP metadata'}

-        conformance = 'PDF/A-{}'.format(
-            part_conformance)
+    part_conformance = pdfa_dict['part'] + pdfa_dict['conformance']
+    valid_part_conforms = {'1A', '1B', '2A', '2B', '2U', '3A', '3B', '3U'}

-        if part_conformance in valid_part_conforms:
-            pdfa_dict['pass'] = True
-            pdfa_dict['output'] = 'pdfa'
-        pdfa_dict['conformance'] = conformance
-    else:
-        pdfa_dict['conformance'] = 'PDF'
+    conformance = 'PDF/A-{}'.format(
+        part_conformance)
+
+    if part_conformance in valid_part_conforms:
+        pdfa_dict['pass'] = True
+        pdfa_dict['output'] = 'pdfa'
+    pdfa_dict['conformance'] = conformance

    return pdfa_dict

@@ -1,338 +0,0 @@
-<?xml version="1.0" encoding="UTF-8" standalone="no"?>
-<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
- "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
-<!-- Generated by graphviz version 2.38.0 (20140413.2041)
- -->
-<!-- Title: Pipeline: Pages: 1 -->
-<svg width="1444pt" height="973pt"
- viewBox="0.00 0.00 1444.00 973.18" 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(4 969.181)">
-<title>Pipeline:</title>
-<polygon fill="white" stroke="none" points="-4,4 -4,-969.181 1440,-969.181 1440,4 -4,4"/>
-<g id="clust1" class="cluster"><title>clustertasks</title>
-<polygon fill="none" stroke="black" points="8,-8 8,-957.181 1428,-957.181 1428,-8 8,-8"/>
-<text text-anchor="middle" x="718" y="-929.181" 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="black" points="1219.77,-911.181 1034.23,-911.181 1030.23,-907.181 1030.23,-875.181 1215.77,-875.181 1219.77,-879.181 1219.77,-911.181"/>
-<polyline fill="none" stroke="black" points="1215.77,-907.181 1030.23,-907.181 "/>
-<polyline fill="none" stroke="black" points="1215.77,-907.181 1215.77,-875.181 "/>
-<polyline fill="none" stroke="black" points="1215.77,-907.181 1219.77,-911.181 "/>
-<text text-anchor="middle" x="1125" y="-887.181" font-family="Times,serif" font-size="20.00">triage</text>
-</g>
-<!-- t1 -->
-<g id="node2" class="node"><title>t1</title>
-<polygon fill="#efa03b" stroke="black" points="1238.54,-853.181 1015.46,-853.181 1011.46,-849.181 1011.46,-817.181 1234.54,-817.181 1238.54,-821.181 1238.54,-853.181"/>
-<polyline fill="none" stroke="black" points="1234.54,-849.181 1011.46,-849.181 "/>
-<polyline fill="none" stroke="black" points="1234.54,-849.181 1234.54,-817.181 "/>
-<polyline fill="none" stroke="black" points="1234.54,-849.181 1238.54,-853.181 "/>
-<text text-anchor="middle" x="1125" y="-829.181" font-family="Times,serif" font-size="20.00">repair_pdf</text>
-</g>
-<!-- t0&#45;&gt;t1 -->
-<g id="edge1" class="edge"><title>t0&#45;&gt;t1</title>
-<path fill="none" stroke="#0044a0" d="M1125,-875.075C1125,-871.384 1125,-867.394 1125,-863.424"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1128.5,-863.182 1125,-853.182 1121.5,-863.182 1128.5,-863.182"/>
-</g>
-<!-- t2 -->
-<g id="node3" class="node"><title>t2</title>
-<polygon fill="#efa03b" stroke="black" points="1329.11,-774.253 1125,-791.156 920.888,-774.253 921.078,-746.905 1328.92,-746.905 1329.11,-774.253"/>
-<polygon fill="none" stroke="black" points="1333.13,-777.932 1125,-795.167 916.866,-777.932 917.11,-742.906 1332.89,-742.906 1333.13,-777.932"/>
-<text text-anchor="middle" x="1125" y="-760.694" font-family="Times,serif" font-size="20.00">split_pages</text>
-</g>
-<!-- t1&#45;&gt;t2 -->
-<g id="edge2" class="edge"><title>t1&#45;&gt;t2</title>
-<path fill="none" stroke="#0044a0" d="M1125,-817.059C1125,-813.44 1125,-809.484 1125,-805.438"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1128.5,-805.334 1125,-795.334 1121.5,-805.334 1128.5,-805.334"/>
-</g>
-<!-- t15 -->
-<g id="node17" class="node"><title>t15</title>
-<polygon fill="#efa03b" stroke="black" points="632.082,-716.208 293.918,-716.208 289.918,-712.208 289.918,-680.208 628.082,-680.208 632.082,-684.208 632.082,-716.208"/>
-<polyline fill="none" stroke="black" points="628.082,-712.208 289.918,-712.208 "/>
-<polyline fill="none" stroke="black" points="628.082,-712.208 628.082,-680.208 "/>
-<polyline fill="none" stroke="black" points="628.082,-712.208 632.082,-716.208 "/>
-<text text-anchor="middle" x="461" y="-692.208" font-family="Times,serif" font-size="20.00">generate_postscript_stub</text>
-</g>
-<!-- t1&#45;&gt;t15 -->
-<g id="edge23" class="edge"><title>t1&#45;&gt;t15</title>
-<path fill="none" stroke="#0044a0" d="M1026.3,-817.165C989.157,-810.572 946.587,-802.788 908,-795.181 781.899,-770.319 636.479,-738.51 546.633,-718.49"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="547.289,-715.05 536.767,-716.289 545.765,-721.882 547.289,-715.05"/>
-</g>
-<!-- t18 -->
-<g id="node19" class="node"><title>t18</title>
-<polygon fill="#efa03b" stroke="black" points="1025.79,-156.452 1221,-129.134 1416.21,-156.452 1416.03,-200.654 1025.97,-200.654 1025.79,-156.452"/>
-<polygon fill="none" stroke="black" points="1021.78,-152.971 1221,-125.091 1420.22,-152.971 1420.01,-204.657 1021.99,-204.657 1021.78,-152.971"/>
-<text text-anchor="middle" x="1221" y="-162.669" font-family="Times,serif" font-size="20.00">merge_pages_qpdf</text>
-</g>
-<!-- t1&#45;&gt;t18 -->
-<g id="edge33" class="edge"><title>t1&#45;&gt;t18</title>
-<path fill="none" stroke="#0044a0" d="M1238.67,-826.639C1321.4,-814.236 1419,-782.049 1419,-699.208 1419,-699.208 1419,-699.208 1419,-319.208 1419,-280.294 1424,-263.161 1398,-234.208 1389.64,-224.894 1379.82,-216.923 1369.17,-210.101"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1370.83,-207.019 1360.46,-204.872 1367.23,-213.02 1370.83,-207.019"/>
-</g>
-<!-- t3 -->
-<g id="node4" class="node"><title>t3</title>
-<polygon fill="#efa03b" stroke="black" points="1186.88,-716.208 905.117,-716.208 901.117,-712.208 901.117,-680.208 1182.88,-680.208 1186.88,-684.208 1186.88,-716.208"/>
-<polyline fill="none" stroke="black" points="1182.88,-712.208 901.117,-712.208 "/>
-<polyline fill="none" stroke="black" points="1182.88,-712.208 1182.88,-680.208 "/>
-<polyline fill="none" stroke="black" points="1182.88,-712.208 1186.88,-716.208 "/>
-<text text-anchor="middle" x="1044" y="-692.208" font-family="Times,serif" font-size="20.00">rasterize_preview</text>
-</g>
-<!-- t2&#45;&gt;t3 -->
-<g id="edge3" class="edge"><title>t2&#45;&gt;t3</title>
-<path fill="none" stroke="#0044a0" d="M1097.12,-742.813C1089.24,-736.345 1080.66,-729.296 1072.73,-722.788"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1074.71,-719.888 1064.76,-716.249 1070.27,-725.299 1074.71,-719.888"/>
-</g>
-<!-- t4 -->
-<g id="node5" class="node"><title>t4</title>
-<polygon fill="#efa03b" stroke="black" points="1258.15,-658.208 1023.85,-658.208 1019.85,-654.208 1019.85,-622.208 1254.15,-622.208 1258.15,-626.208 1258.15,-658.208"/>
-<polyline fill="none" stroke="black" points="1254.15,-654.208 1019.85,-654.208 "/>
-<polyline fill="none" stroke="black" points="1254.15,-654.208 1254.15,-622.208 "/>
-<polyline fill="none" stroke="black" points="1254.15,-654.208 1258.15,-658.208 "/>
-<text text-anchor="middle" x="1139" y="-634.208" font-family="Times,serif" font-size="20.00">orient_page</text>
-</g>
-<!-- t2&#45;&gt;t4 -->
-<g id="edge5" class="edge"><title>t2&#45;&gt;t4</title>
-<path fill="none" stroke="#0044a0" d="M1171.65,-742.795C1181.42,-735.692 1190.44,-726.864 1196,-716.208 1203.4,-702.023 1203.46,-694.36 1196,-680.208 1192.84,-674.222 1188.31,-668.98 1183.15,-664.445"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1185.3,-661.676 1175.25,-658.302 1181,-667.203 1185.3,-661.676"/>
-</g>
-<!-- t3&#45;&gt;t4 -->
-<g id="edge4" class="edge"><title>t3&#45;&gt;t4</title>
-<path fill="none" stroke="#0044a0" d="M1073.02,-680.102C1081.95,-674.836 1091.92,-668.963 1101.34,-663.407"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1103.32,-666.302 1110.16,-658.209 1099.77,-660.272 1103.32,-666.302"/>
-</g>
-<!-- t5 -->
-<g id="node6" class="node"><title>t5</title>
-<polygon fill="#efa03b" stroke="black" points="1271.3,-600.208 920.701,-600.208 916.701,-596.208 916.701,-564.208 1267.3,-564.208 1271.3,-568.208 1271.3,-600.208"/>
-<polyline fill="none" stroke="black" points="1267.3,-596.208 916.701,-596.208 "/>
-<polyline fill="none" stroke="black" points="1267.3,-596.208 1267.3,-564.208 "/>
-<polyline fill="none" stroke="black" points="1267.3,-596.208 1271.3,-600.208 "/>
-<text text-anchor="middle" x="1094" y="-576.208" font-family="Times,serif" font-size="20.00">rasterize_with_ghostscript</text>
-</g>
-<!-- t4&#45;&gt;t5 -->
-<g id="edge6" class="edge"><title>t4&#45;&gt;t5</title>
-<path fill="none" stroke="#0044a0" d="M1125.25,-622.102C1121.76,-617.759 1117.94,-613.003 1114.21,-608.353"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1116.65,-605.812 1107.66,-600.209 1111.2,-610.197 1116.65,-605.812"/>
-</g>
-<!-- t10 -->
-<g id="node12" class="node"><title>t10</title>
-<polygon fill="#00cc66" stroke="black" points="334,-348.208 35.7812,-320.208 334,-292.208 632.219,-320.208 334,-348.208"/>
-<text text-anchor="middle" x="334" y="-314.208" font-family="Times,serif" font-size="20.00">select_image_layer</text>
-</g>
-<!-- t4&#45;&gt;t10 -->
-<g id="edge15" class="edge"><title>t4&#45;&gt;t10</title>
-<path fill="none" stroke="#0044a0" d="M1019.87,-634.085C799.18,-622.991 349,-592.156 349,-525.208 349,-525.208 349,-525.208 349,-465.208 349,-428.606 344.147,-387.052 339.961,-358.022"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="343.393,-357.303 338.46,-347.926 336.469,-358.332 343.393,-357.303"/>
-</g>
-<!-- t16 -->
-<g id="node15" class="node"><title>t16</title>
-<polygon fill="#efa03b" stroke="black" points="1380.34,-270.208 1159.66,-270.208 1155.66,-266.208 1155.66,-234.208 1376.34,-234.208 1380.34,-238.208 1380.34,-270.208"/>
-<polyline fill="none" stroke="black" points="1376.34,-266.208 1155.66,-266.208 "/>
-<polyline fill="none" stroke="black" points="1376.34,-266.208 1376.34,-234.208 "/>
-<polyline fill="none" stroke="black" points="1376.34,-266.208 1380.34,-270.208 "/>
-<text text-anchor="middle" x="1268" y="-246.208" font-family="Times,serif" font-size="20.00">skip_page</text>
-</g>
-<!-- t4&#45;&gt;t16 -->
-<g id="edge20" class="edge"><title>t4&#45;&gt;t16</title>
-<path fill="none" stroke="#0044a0" d="M1258.35,-631.473C1326.7,-619.809 1399,-591.745 1399,-525.208 1399,-525.208 1399,-525.208 1399,-397.208 1399,-350.33 1418.37,-328.748 1389,-292.208 1383.58,-285.469 1376.99,-279.879 1369.71,-275.244"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1371.3,-272.123 1360.88,-270.231 1367.85,-278.211 1371.3,-272.123"/>
-</g>
-<!-- t14 -->
-<g id="node16" class="node"><title>t14</title>
-<polygon fill="#66ccff" stroke="black" points="1380.24,-338.208 999.76,-338.208 995.76,-334.208 995.76,-302.208 1376.24,-302.208 1380.24,-306.208 1380.24,-338.208"/>
-<polyline fill="none" stroke="black" points="1376.24,-334.208 995.76,-334.208 "/>
-<polyline fill="none" stroke="black" points="1376.24,-334.208 1376.24,-302.208 "/>
-<polyline fill="none" stroke="black" points="1376.24,-334.208 1380.24,-338.208 "/>
-<text text-anchor="middle" x="1188" y="-314.208" font-family="Times,serif" font-size="20.00">tesseract_ocr_and_render_pdf</text>
-</g>
-<!-- t4&#45;&gt;t14 -->
-<g id="edge22" class="edge"><title>t4&#45;&gt;t14</title>
-<path fill="none" stroke="#0044a0" d="M1244.5,-622.179C1297.25,-607.003 1350,-578.503 1350,-525.208 1350,-525.208 1350,-525.208 1350,-465.208 1350,-422.752 1366.76,-403.169 1340,-370.208 1330.4,-358.377 1317.91,-349.368 1304.22,-342.513"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1305.62,-339.304 1295.07,-338.313 1302.7,-345.666 1305.62,-339.304"/>
-</g>
-<!-- t6 -->
-<g id="node7" class="node"><title>t6</title>
-<polygon fill="#efa03b" stroke="black" points="1074.74,-542.208 779.258,-542.208 775.258,-538.208 775.258,-506.208 1070.74,-506.208 1074.74,-510.208 1074.74,-542.208"/>
-<polyline fill="none" stroke="black" points="1070.74,-538.208 775.258,-538.208 "/>
-<polyline fill="none" stroke="black" points="1070.74,-538.208 1070.74,-506.208 "/>
-<polyline fill="none" stroke="black" points="1070.74,-538.208 1074.74,-542.208 "/>
-<text text-anchor="middle" x="925" y="-518.208" font-family="Times,serif" font-size="20.00">preprocess_deskew</text>
-</g>
-<!-- t5&#45;&gt;t6 -->
-<g id="edge7" class="edge"><title>t5&#45;&gt;t6</title>
-<path fill="none" stroke="#0044a0" d="M1042.38,-564.102C1024.52,-558.185 1004.35,-551.5 985.823,-545.362"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="986.901,-542.032 976.308,-542.209 984.699,-548.677 986.901,-542.032"/>
-</g>
-<!-- t9 -->
-<g id="node11" class="node"><title>t9</title>
-<polygon fill="#efa03b" stroke="black" points="1013,-426.208 694.762,-398.208 1013,-370.208 1331.24,-398.208 1013,-426.208"/>
-<text text-anchor="middle" x="1013" y="-392.208" font-family="Times,serif" font-size="20.00">select_image_for_pdf</text>
-</g>
-<!-- t5&#45;&gt;t9 -->
-<g id="edge13" class="edge"><title>t5&#45;&gt;t9</title>
-<path fill="none" stroke="#0044a0" d="M1093.89,-563.996C1093.29,-548.44 1091.11,-525.146 1084,-506.208 1073.94,-479.426 1056.13,-452.609 1040.86,-432.521"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1043.62,-430.369 1034.72,-424.619 1038.09,-434.663 1043.62,-430.369"/>
-</g>
-<!-- t7 -->
-<g id="node8" class="node"><title>t7</title>
-<polygon fill="#efa03b" stroke="black" points="993.95,-484.208 716.05,-484.208 712.05,-480.208 712.05,-448.208 989.95,-448.208 993.95,-452.208 993.95,-484.208"/>
-<polyline fill="none" stroke="black" points="989.95,-480.208 712.05,-480.208 "/>
-<polyline fill="none" stroke="black" points="989.95,-480.208 989.95,-448.208 "/>
-<polyline fill="none" stroke="black" points="989.95,-480.208 993.95,-484.208 "/>
-<text text-anchor="middle" x="853" y="-460.208" font-family="Times,serif" font-size="20.00">preprocess_clean</text>
-</g>
-<!-- t6&#45;&gt;t7 -->
-<g id="edge8" class="edge"><title>t6&#45;&gt;t7</title>
-<path fill="none" stroke="#0044a0" d="M903.006,-506.102C896.726,-501.216 889.771,-495.807 883.095,-490.615"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="884.901,-487.586 874.859,-484.209 880.604,-493.111 884.901,-487.586"/>
-</g>
-<!-- t6&#45;&gt;t9 -->
-<g id="edge12" class="edge"><title>t6&#45;&gt;t9</title>
-<path fill="none" stroke="#0044a0" d="M977.454,-506.003C987.329,-500.504 996.562,-493.377 1003,-484.208 1012.67,-470.441 1015.74,-452.229 1016.16,-436.195"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1019.66,-436.018 1016.09,-426.041 1012.66,-436.063 1019.66,-436.018"/>
-</g>
-<!-- t8 -->
-<g id="node9" class="node"><title>t8</title>
-<polygon fill="#00cc66" stroke="black" points="676.666,-416.208 387.334,-416.208 383.334,-412.208 383.334,-380.208 672.666,-380.208 676.666,-384.208 676.666,-416.208"/>
-<polyline fill="none" stroke="black" points="672.666,-412.208 383.334,-412.208 "/>
-<polyline fill="none" stroke="black" points="672.666,-412.208 672.666,-380.208 "/>
-<polyline fill="none" stroke="black" points="672.666,-412.208 676.666,-416.208 "/>
-<text text-anchor="middle" x="530" y="-392.208" font-family="Times,serif" font-size="20.00">ocr_tesseract_hocr</text>
-</g>
-<!-- t7&#45;&gt;t8 -->
-<g id="edge9" class="edge"><title>t7&#45;&gt;t8</title>
-<path fill="none" stroke="#0044a0" d="M769.399,-448.125C724.645,-438.98 669.394,-427.691 623.252,-418.262"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="623.748,-414.791 613.25,-416.219 622.347,-421.65 623.748,-414.791"/>
-</g>
-<!-- t7&#45;&gt;t9 -->
-<g id="edge11" class="edge"><title>t7&#45;&gt;t9</title>
-<path fill="none" stroke="#0044a0" d="M894.62,-448.039C911.413,-441.112 931.162,-432.966 949.666,-425.333"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="951.009,-428.565 958.918,-421.516 948.339,-422.094 951.009,-428.565"/>
-</g>
-<!-- t11 -->
-<g id="node10" class="node"><title>t11</title>
-<polygon fill="#00cc66" stroke="black" points="938.109,-338.208 653.891,-338.208 649.891,-334.208 649.891,-302.208 934.109,-302.208 938.109,-306.208 938.109,-338.208"/>
-<polyline fill="none" stroke="black" points="934.109,-334.208 649.891,-334.208 "/>
-<polyline fill="none" stroke="black" points="934.109,-334.208 934.109,-302.208 "/>
-<polyline fill="none" stroke="black" points="934.109,-334.208 938.109,-338.208 "/>
-<text text-anchor="middle" x="794" y="-314.208" font-family="Times,serif" font-size="20.00">render_hocr_page</text>
-</g>
-<!-- t8&#45;&gt;t11 -->
-<g id="edge10" class="edge"><title>t8&#45;&gt;t11</title>
-<path fill="none" stroke="#0044a0" d="M589.247,-380.152C629.442,-368.58 682.727,-353.241 724.863,-341.111"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="725.853,-344.468 734.495,-338.338 723.917,-337.741 725.853,-344.468"/>
-</g>
-<!-- t12 -->
-<g id="node14" class="node"><title>t12</title>
-<polygon fill="#00cc66" stroke="black" points="1118.49,-270.208 775.514,-270.208 771.514,-266.208 771.514,-234.208 1114.49,-234.208 1118.49,-238.208 1118.49,-270.208"/>
-<polyline fill="none" stroke="black" points="1114.49,-266.208 771.514,-266.208 "/>
-<polyline fill="none" stroke="black" points="1114.49,-266.208 1114.49,-234.208 "/>
-<polyline fill="none" stroke="black" points="1114.49,-266.208 1118.49,-270.208 "/>
-<text text-anchor="middle" x="945" y="-246.208" font-family="Times,serif" font-size="20.00">render_hocr_debug_page</text>
-</g>
-<!-- t8&#45;&gt;t12 -->
-<g id="edge19" class="edge"><title>t8&#45;&gt;t12</title>
-<path fill="none" stroke="#0044a0" d="M620.537,-380.161C641.947,-376.512 664.736,-372.939 686,-370.208 714.866,-366.5 927.304,-369.633 947,-348.208 963.502,-330.256 960.281,-301.363 954.611,-280.149"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="957.925,-279.012 951.707,-270.433 951.218,-281.017 957.925,-279.012"/>
-</g>
-<!-- t13 -->
-<g id="node13" class="node"><title>t13</title>
-<polygon fill="#00cc66" stroke="black" points="753.082,-270.208 494.918,-270.208 490.918,-266.208 490.918,-234.208 749.082,-234.208 753.082,-238.208 753.082,-270.208"/>
-<polyline fill="none" stroke="black" points="749.082,-266.208 490.918,-266.208 "/>
-<polyline fill="none" stroke="black" points="749.082,-266.208 749.082,-234.208 "/>
-<polyline fill="none" stroke="black" points="749.082,-266.208 753.082,-270.208 "/>
-<text text-anchor="middle" x="622" y="-246.208" font-family="Times,serif" font-size="20.00">add_text_layer</text>
-</g>
-<!-- t11&#45;&gt;t13 -->
-<g id="edge16" class="edge"><title>t11&#45;&gt;t13</title>
-<path fill="none" stroke="#0044a0" d="M749.258,-302.039C726.862,-293.446 699.574,-282.975 676.019,-273.936"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="677.108,-270.605 666.518,-270.29 674.6,-277.14 677.108,-270.605"/>
-</g>
-<!-- t9&#45;&gt;t10 -->
-<g id="edge14" class="edge"><title>t9&#45;&gt;t10</title>
-<path fill="none" stroke="#0044a0" d="M877.633,-382.056C761.19,-369.023 593.788,-350.286 475.578,-337.054"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="475.697,-333.546 465.37,-335.912 474.919,-340.503 475.697,-333.546"/>
-</g>
-<!-- t9&#45;&gt;t12 -->
-<g id="edge18" class="edge"><title>t9&#45;&gt;t12</title>
-<path fill="none" stroke="#0044a0" d="M998.887,-371.447C994.931,-364.003 990.702,-355.831 987,-348.208 983.822,-341.663 967.96,-305.616 956.669,-279.865"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="959.82,-278.337 952.601,-270.582 953.409,-281.146 959.82,-278.337"/>
-</g>
-<!-- t9&#45;&gt;t14 -->
-<g id="edge21" class="edge"><title>t9&#45;&gt;t14</title>
-<path fill="none" stroke="#0044a0" d="M1065.03,-374.611C1088.59,-364.381 1116.27,-352.359 1139.4,-342.314"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1140.9,-345.479 1148.68,-338.285 1138.11,-339.058 1140.9,-345.479"/>
-</g>
-<!-- t10&#45;&gt;t13 -->
-<g id="edge17" class="edge"><title>t10&#45;&gt;t13</title>
-<path fill="none" stroke="#0044a0" d="M417.688,-300.029C455.345,-291.399 499.718,-281.231 537.473,-272.578"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="538.461,-275.943 547.427,-270.297 536.898,-269.12 538.461,-275.943"/>
-</g>
-<!-- t17 -->
-<g id="node18" class="node"><title>t17</title>
-<polygon fill="#efa03b" stroke="black" points="540.724,-156.455 770,-129.143 999.276,-156.455 999.062,-200.646 540.938,-200.646 540.724,-156.455"/>
-<polygon fill="none" stroke="black" points="536.704,-152.903 770,-125.112 1003.3,-152.903 1003.05,-204.648 536.954,-204.648 536.704,-152.903"/>
-<text text-anchor="middle" x="770" y="-162.669" font-family="Times,serif" font-size="20.00">merge_pages_ghostscript</text>
-</g>
-<!-- t13&#45;&gt;t17 -->
-<g id="edge27" class="edge"><title>t13&#45;&gt;t17</title>
-<path fill="none" stroke="#0044a0" d="M653.026,-234.114C665.992,-226.971 681.653,-218.343 697.24,-209.755"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="699.319,-212.606 706.388,-204.715 695.941,-206.475 699.319,-212.606"/>
-</g>
-<!-- t13&#45;&gt;t18 -->
-<g id="edge31" class="edge"><title>t13&#45;&gt;t18</title>
-<path fill="none" stroke="#0044a0" d="M753.329,-235.235C756.584,-234.884 759.812,-234.541 763,-234.208 873.497,-222.668 902.059,-228.201 1012,-212.208 1023.74,-210.5 1035.81,-208.554 1047.95,-206.456"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1048.91,-209.84 1058.16,-204.659 1047.7,-202.946 1048.91,-209.84"/>
-</g>
-<!-- t12&#45;&gt;t17 -->
-<g id="edge28" class="edge"><title>t12&#45;&gt;t17</title>
-<path fill="none" stroke="#0044a0" d="M908.314,-234.114C892.552,-226.77 873.424,-217.858 854.484,-209.033"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="855.759,-205.766 845.216,-204.715 852.803,-212.111 855.759,-205.766"/>
-</g>
-<!-- t12&#45;&gt;t18 -->
-<g id="edge32" class="edge"><title>t12&#45;&gt;t18</title>
-<path fill="none" stroke="#0044a0" d="M1002.86,-234.114C1029.07,-226.37 1061.19,-216.881 1092.64,-207.59"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1093.77,-210.905 1102.37,-204.715 1091.79,-204.192 1093.77,-210.905"/>
-</g>
-<!-- t16&#45;&gt;t17 -->
-<g id="edge24" class="edge"><title>t16&#45;&gt;t17</title>
-<path fill="none" stroke="#0044a0" d="M1155.57,-234.546C1111.13,-227.821 1059.63,-219.853 1013,-212.208 1001.53,-210.326 989.752,-208.365 977.872,-206.363"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="978.316,-202.888 967.872,-204.672 977.149,-209.79 978.316,-202.888"/>
-</g>
-<!-- t16&#45;&gt;t18 -->
-<g id="edge29" class="edge"><title>t16&#45;&gt;t18</title>
-<path fill="none" stroke="#0044a0" d="M1258.03,-233.915C1254.58,-227.933 1250.55,-220.928 1246.43,-213.785"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1249.24,-211.645 1241.21,-204.729 1243.17,-215.141 1249.24,-211.645"/>
-</g>
-<!-- t14&#45;&gt;t17 -->
-<g id="edge26" class="edge"><title>t14&#45;&gt;t17</title>
-<path fill="none" stroke="#0044a0" d="M1172.36,-302.154C1164.33,-293.013 1154.66,-281.362 1147,-270.208 1136.64,-255.121 1141.88,-244.862 1127,-234.208 1125.19,-232.915 1055.67,-220.232 978.45,-206.446"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="978.964,-202.982 968.505,-204.672 977.735,-209.874 978.964,-202.982"/>
-</g>
-<!-- t14&#45;&gt;t18 -->
-<g id="edge30" class="edge"><title>t14&#45;&gt;t18</title>
-<path fill="none" stroke="#0044a0" d="M1310.22,-302.184C1345.8,-294.579 1378.07,-284.119 1389,-270.208 1408.78,-245.028 1391.25,-224.765 1361.2,-209.211"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1362.48,-205.944 1351.96,-204.742 1359.44,-212.246 1362.48,-205.944"/>
-</g>
-<!-- t15&#45;&gt;t17 -->
-<g id="edge25" class="edge"><title>t15&#45;&gt;t17</title>
-<path fill="none" stroke="#0044a0" d="M289.656,-684.797C164.646,-670.853 17,-641.95 17,-583.208 17,-583.208 17,-583.208 17,-319.208 17,-213.417 306.554,-181.327 526.553,-172.213"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="526.794,-175.706 536.646,-171.808 526.514,-168.711 526.794,-175.706"/>
-</g>
-<!-- t19 -->
-<g id="node20" class="node"><title>t19</title>
-<polygon fill="#efa03b" stroke="black" points="844.361,-47.3406 995,-20.006 1145.64,-47.3406 1145.5,-91.5689 844.501,-91.5689 844.361,-47.3406"/>
-<polygon fill="none" stroke="black" points="840.355,-43.9965 995,-15.9349 1149.65,-43.9965 1149.48,-95.5749 840.519,-95.5749 840.355,-43.9965"/>
-<text text-anchor="middle" x="995" y="-53.565" font-family="Times,serif" font-size="20.00">copy_final</text>
-</g>
-<!-- t17&#45;&gt;t19 -->
-<g id="edge35" class="edge"><title>t17&#45;&gt;t19</title>
-<path fill="none" stroke="#0044a0" d="M841.81,-133.486C864.177,-122.839 889.024,-111.011 912.085,-100.034"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="913.64,-103.17 921.165,-95.7118 910.631,-96.8497 913.64,-103.17"/>
-</g>
-<!-- t18&#45;&gt;t19 -->
-<g id="edge34" class="edge"><title>t18&#45;&gt;t19</title>
-<path fill="none" stroke="#0044a0" d="M1151.35,-134.66C1128.23,-123.705 1102.26,-111.399 1078.23,-100.009"/>
-<polygon fill="#0044a0" stroke="#0044a0" points="1079.66,-96.8131 1069.12,-95.6933 1076.66,-103.139 1079.66,-96.8131"/>
-</g>
-</g>
-</svg>
@@ -1,2 +0,0 @@
-[pytest]
-norecursedirs = lib .pc .git
@@ -1,9 +1,9 @@
 # requirements.txt can be used to replicate the developer's build environment
 # setup.py lists a separate set of requirements that are looser to simplify
 # installation
-ruffus==2.6.3
-Pillow==3.3.0
-reportlab==3.2.0
-PyPDF2==1.26
-img2pdf==0.2.1
-cffi==1.5.2
+ruffus == 2.6.3
+Pillow == 4.0.0
+reportlab == 3.3.0
+PyPDF2 == 1.26
+img2pdf == 0.2.3
+cffi == 1.9.1
@@ -6,4 +6,8 @@ test=pytest

 [check-manifest]
 ignore =
-	.github
+	.github
+
+[tool:pytest]
+norecursedirs = lib .pc .git output cache resources
+testpaths = tests
@@ -9,7 +9,7 @@ if sys.version_info < (3, 4):
    print("Python 3.4 or newer is required", file=sys.stderr)
    sys.exit(1)

-from setuptools import setup  # nopep8
+from setuptools import setup, find_packages  # nopep8
 from subprocess import STDOUT, check_output, CalledProcessError  # nopep8
 from collections.abc import Mapping  # nopep8
 import re  # nopep8
@@ -155,12 +155,12 @@ if not forced and command.startswith('install') or \
        command in ['check', 'test', 'nosetests', 'easy_install']:
    check_external_program(
        program='tesseract',
-        need_version='3.03',  # limited by Travis CI / Ubuntu 12.04 backports
+        need_version='3.03',  # limited by Travis CI / Ubuntu 14.04 backports
        package={'darwin': 'tesseract', 'linux': 'tesseract-ocr'}
    )
    check_external_program(
        program='gs',
-        need_version='9.15',  # limited by Travis CI / Ubuntu 12.04 backports
+        need_version='9.15',  # limited by Travis CI / Ubuntu 14.04 backports
        package='ghostscript'
    )
    check_external_program(
@@ -171,7 +171,7 @@ if not forced and command.startswith('install') or \
    )
    check_external_program(
        program='qpdf',
-        need_version='5.0.0',   # limited by Travis CI / Ubuntu 12.04 backports
+        need_version='5.1.1',   # limited by Travis CI / Ubuntu 14.04 backports
        package='qpdf',
        version_check_args=['--version']
    )
@@ -191,11 +191,14 @@ setup(
    url='https://github.com/jbarlow83/OCRmyPDF',
    author='James R. Barlow',
    author_email='jim@purplerock.ca',
-    license='Public Domain',
-    packages=['ocrmypdf'],
+    license='MIT',
+    packages=find_packages(exclude=["tests", "tests.*"]),
    keywords=['PDF', 'OCR', 'optical character recognition', 'PDF/A', 'scanning'],
    classifiers=[
        "Programming Language :: Python :: 3",
+        "Programming Language :: Python :: 3.4",
+        "Programming Language :: Python :: 3.5",
+        "Programming Language :: Python :: 3.6",
        "Development Status :: 5 - Production/Stable",
        "Environment :: Console",
        "Intended Audience :: End Users/Desktop",
@@ -211,21 +214,20 @@ setup(
        "Topic :: Text Processing :: Linguistic",
        ],
    setup_requires=[
-        'setuptools_scm',
-        'cffi>=1.5.0',
-        'pytest-runner'
+        'setuptools_scm',       # so that version will work
+        'cffi>=1.9.1'           # to build the leptonica module
    ],
    use_scm_version={'version_scheme': 'post-release'},
    cffi_modules=[
-        'ocrmypdf/lib/compile_leptonica.py:ffi'
+        'ocrmypdf/lib/compile_leptonica.py:ffibuilder'
    ],
    install_requires=[
        'ruffus==2.6.3',        # pinned - ocrmypdf implements a 2.6.3 workaround
-        'Pillow>=3.1.0',        # Pillow is pretty stable
+        'Pillow>=4.0.0',        # Pillow < 4 has BytesIO/TIFF bug w/img2pdf 0.2.3
        'reportlab>=3.2.0',     # oldest released version with sane image handling
        'PyPDF2>=1.26',         # pure Python, so track HEAD closely
-        'img2pdf>=0.2.1',       # pure Python, so track HEAD closely
-        'cffi>=1.5.0'           # oldest version ever tested
+        'img2pdf>=0.2.3',       # pure Python, so track HEAD closely
+        'cffi>=1.9.1'           # must be a setup and install requirement
    ],
    tests_require=tests_require,
    entry_points={
@@ -1,4 +0,0 @@
-deb http://httpredir.debian.org/debian stretch main
-deb http://httpredir.debian.org/debian stretch-updates main
-deb http://security.debian.org stretch/updates main
-deb http://ftp.de.debian.org/debian sid main contrib non-free
@@ -1 +1,2 @@
-pytest>=2.7.2
+pytest >= 3.0
+pytest-helpers-namespace
@@ -0,0 +1,135 @@
+#!/usr/bin/env python3
+# © 2017 James R. Barlow: github.com/jbarlow83
+
+import sys
+import os
+import platform
+
+pytest_plugins = ['helpers_namespace']
+
+import pytest
+from pathlib import Path
+from subprocess import Popen, PIPE
+
+
+if sys.version_info.major < 3:
+    print("Requires Python 3.4+")
+    sys.exit(1)
+
+
+@pytest.helpers.register
+def is_linux():
+    return platform.system() == 'Linux'
+
+
+@pytest.helpers.register
+def is_macos():
+    return platform.system() == 'Darwin'
+
+
+@pytest.helpers.register
+def running_in_docker():
+    # Docker creates a file named /.dockerenv (newer versions) or
+    # /.dockerinit (older) -- this is undocumented, not an offical test
+    return os.path.exists('/.dockerenv') or os.path.exists('/.dockerinit')
+
+
+@pytest.helpers.register
+def running_in_travis():
+    return os.environ.get('TRAVIS') == 'true'
+
+
+TESTS_ROOT = os.path.abspath(os.path.dirname(__file__))
+SPOOF_PATH = os.path.join(TESTS_ROOT, 'spoof')
+PROJECT_ROOT = os.path.dirname(TESTS_ROOT)
+OCRMYPDF = [sys.executable, '-m', 'ocrmypdf']
+
+
+@pytest.helpers.register
+def spoof(**kwargs):
+    """Modify environment variables to override subprocess executables
+
+    spoof(program1='replacement', ...)
+
+    Before running any executable, ocrmypdf checks the environment variable
+    OCRMYPDF_PROGRAMNAME to override default program name/location, e.g.
+    OCRMYPDF_GS redirects from the system path Ghostscript ("gs") to elsewhere.
+
+    """
+    env = os.environ.copy()
+
+    for replace_program, with_spoof in kwargs.items():
+        spoofer = os.path.join(SPOOF_PATH, with_spoof)
+        if not os.access(spoofer, os.X_OK):
+            os.chmod(spoofer, 0o755)
+        env['OCRMYPDF_' + replace_program.upper()] = spoofer
+    return env
+
+
+@pytest.fixture
+def resources():
+    return Path(TESTS_ROOT) / 'resources'
+
+
+@pytest.fixture
+def ocrmypdf_exec():
+    return OCRMYPDF
+
+
+@pytest.fixture(scope="function")
+def outdir(tmpdir):
+    return Path(str(tmpdir))
+
+
+@pytest.fixture(scope="function")
+def outpdf(tmpdir):
+    return str(Path(str(tmpdir)) / 'out.pdf')
+
+
+@pytest.fixture(scope="function")
+def no_outpdf(tmpdir):
+    """This just documents the fact that a test is not expected to produce
+    output. Unfortunately an assertion failure inside a test fixture produces
+    an error rather than a test failure, so no testing is done. It's up to
+    the test to confirm that no output file was created."""
+    return str(Path(str(tmpdir)) / 'no_output.pdf')
+
+
+@pytest.helpers.register
+def check_ocrmypdf(input_file, output_file, *args, env=None):
+    "Run ocrmypdf and confirmed that a valid file was created"
+
+    p, out, err = run_ocrmypdf(input_file, output_file, *args, env=env)
+    #print(err)  # ensure py.test collects the output, use -s to view
+    assert p.returncode == 0, "<stderr>\n" + err + "\n</stderr>"
+    assert os.path.exists(str(output_file)), "Output file not created"
+    assert os.stat(str(output_file)).st_size > 100, "PDF too small or empty"
+    assert out == "", \
+        "The following was written to stdout and should not have been: \n" + \
+        "<stdout>\n" + out + "\n</stdout>"
+    return output_file
+
+
+@pytest.helpers.register
+def run_ocrmypdf(input_file, output_file, *args, env=None):
+    "Run ocrmypdf and let caller deal with results"
+
+    if env is None:
+        env = os.environ
+
+    p_args = OCRMYPDF + list(args) + [str(input_file), str(output_file)]
+    p = Popen(
+        p_args, close_fds=True, stdout=PIPE, stderr=PIPE,
+        universal_newlines=True, env=env)
+    out, err = p.communicate()
+    #print(err)
+
+    return p, out, err
+
+
+@pytest.helpers.register
+def first_page_dimensions(pdf):
+    from ocrmypdf import pageinfo
+    info = pageinfo.pdf_get_all_pageinfo(str(pdf))
+    page0 = info[0]
+    return (page0['width_inches'], page0['height_inches'])
@@ -31,6 +31,9 @@ In some cases they were converted from one image format to another without other
    *   - LinnSequencer.jpg, linn.pdf, linn.txt
        - `Wikimedia: LinnSequencer`_
        - Creative Commons BY-SA 3.0
+    *   - typewriter.png, 2400dpi.pdf
+        - `Wikimedia: Triumph typewrtier text Linzensoep`_
+        * Creative Commons BY-SA 2.5


 Files generated for this project
@@ -61,6 +64,9 @@ under the terms of the license in LICENSE.rst.
    *   - epson.pdf
        - @lowesjam
        - a linearized PDF containing some unusual indirect objects, created by an Epson printer; printout of a Wikipedia article (CC BY-SA)
+    *   - formxobject.pdf
+        - @jbarlow83
+        - hand-crafted exotic PDF containing an image inside a Form XObject
    *   - francais.pdf
        - @jbarlow83
        - a page containing French accents (diacritics)  
@@ -76,6 +82,9 @@ under the terms of the license in LICENSE.rst.
    *   - missing_docinfo.pdf
        - @jbarlow83
        - PDF file with no /DocumentInfo section 
+    *   - overlay.pdf
+        - @maxandersen
+        - PDF file generated by PDFPen pro that triggered content stream parse errors

 Assemblies
 ==========
@@ -104,4 +113,6 @@ These test resources are assemblies from other previously mentioned files, relea

 .. _`Wikimedia: JPEG2000 Lichtenstein`: https://en.wikipedia.org/wiki/JPEG_2000#/media/File:Jpeg2000_2-level_wavelet_transform-lichtenstein.png

-.. _`Linux (Wikipedia Article)`: https://de.wikipedia.org/wiki/Linux 
+.. _`Linux (Wikipedia Article)`: https://de.wikipedia.org/wiki/Linux 
+
+.. _`Wikimedia: Triumph typewrtier text Linzensoep`: https://commons.wikimedia.org/wiki/File:Triumph.typewriter_text_Linzensoep.gif
@@ -0,0 +1,35 @@
+#!/usr/bin/env python3
+import sys
+import os
+from subprocess import check_call
+
+"""Replicate one type of Ghostscript feature elision warning during
+PDF/A creation."""
+
+
+def real_ghostscript(argv):
+    gs_args = ['gs'] + argv[1:]
+    os.execvp("gs", gs_args)
+    return  # Not reachable
+
+
+elision_warning = """GPL Ghostscript 9.20: Setting Overprint Mode to 1
+not permitted in PDF/A-2, overprint mode not set"""
+
+
+def main():
+    if '--version' in sys.argv:
+        print('9.20')
+        print('SPOOFED: ' + os.path.basename(__filename__))
+        sys.exit(0)
+
+    gs_args = ['gs'] + sys.argv[1:]
+    check_call(gs_args)
+
+    if '-sDEVICE=pdfwrite' in sys.argv[1:]:
+        print(elision_warning)
+
+    sys.exit(0)
+
+if __name__ == '__main__':
+    main()
@@ -0,0 +1,42 @@
+#!/usr/bin/env python3
+import sys
+import os
+
+"""Replicate Ghostscript PDF/A conversion failure by suppressing some
+arguments"""
+
+
+def real_ghostscript(argv):
+    gs_args = ['gs'] + argv[1:]
+    os.execvp("gs", gs_args)
+    return  # Not reachable
+
+
+def main():
+    if '--version' in sys.argv:
+        print('9.20')
+        print('SPOOFED: ' + os.path.basename(__filename__))
+        sys.exit(0)
+
+    # Unless some argument is calling for PDFA generation, forward to
+    # real ghostscript
+    if not any(arg.startswith('-dPDFA') for arg in sys.argv):
+        real_ghostscript(sys.argv)
+        return
+
+    # Remove the two arguments that tell ghostscript to create a PDF/A
+    # Does not remove the Postscript definition file - not necessary
+    # to cause PDF/A creation failure
+    argv = []
+    for arg in sys.argv:
+        if arg.startswith('-dPDFA'):
+            continue
+        elif arg.startswith('-dPDFACompatibilityPolicy'):
+            continue
+        argv.append(arg)
+
+    real_ghostscript(argv)
+
+
+if __name__ == '__main__':
+    main()
@@ -37,6 +37,7 @@ def real_tesseract():
 def main():
    operation = sys.argv[-1]
    # For anything unexpected operation, defer to real tesseract binary
+    # Currently this includes all use of "--tesseract-config"
    if operation != 'hocr' and operation != 'pdf' and operation != 'stdout':
        real_tesseract()
        return  # Not reachable
@@ -52,6 +53,11 @@ def main():
        ['tesseract', '--version'],
        stderr=subprocess.STDOUT)

+    if b'4.00.00alpha' in tess_version:
+        # Tesseract 4.x alpha is a moving target, don't cache it
+        real_tesseract()
+        return
+
    m.update(tess_version)

    # Insert this source file into the hash function, to ensure that any
@@ -65,12 +71,21 @@ def main():
        lang = sys.argv[sys.argv.index('-l') + 1]
        m.update(lang.encode())
    except ValueError:
-        pass
-    try:
-        psm = sys.argv[sys.argv.index('-psm') + 1]
-        m.update(psm.encode())
-    except ValueError:
-        pass
+        m.update(b'default-lang')
+
+    psm_arg = ''
+    if '--psm' in sys.argv:
+        psm_arg = '--psm'
+    elif '-psm' in sys.argv:
+        psm_arg = '-psm'
+    if psm_arg:
+        try:
+            psm = sys.argv[sys.argv.index(psm_arg) + 1]
+            m.update(psm.encode())
+        except ValueError:
+            m.update(b'default-psm')
+    else:
+        m.update(b'default-psm')

    if operation == 'stdout' and psm != '0':
        real_tesseract()
@@ -2,11 +2,13 @@
 # © 2015 James R. Barlow: github.com/jbarlow83

 from ocrmypdf import hocrtransform
-from ocrmypdf.tesseract import HOCR_TEMPLATE
+from ocrmypdf.exec.tesseract import HOCR_TEMPLATE
+from ocrmypdf.exec import qpdf
 from reportlab.pdfgen.canvas import Canvas
 from PIL import Image
 from tempfile import NamedTemporaryFile
 from contextlib import suppress
+from pathlib import Path
 import os
 import shutil
 import pytest
@@ -15,47 +17,25 @@ import pytest
 import sys


-if sys.version_info.major < 3:
-    print("Requires Python 3.4+")
-    sys.exit(1)
-
-TESTS_ROOT = os.path.abspath(os.path.dirname(__file__))
-SPOOF_PATH = os.path.join(TESTS_ROOT, 'spoof')
-PROJECT_ROOT = os.path.dirname(TESTS_ROOT)
-OCRMYPDF = os.path.join(PROJECT_ROOT, 'OCRmyPDF.sh')
-TEST_RESOURCES = os.path.join(PROJECT_ROOT, 'tests', 'resources')
-TEST_OUTPUT = os.environ.get(
-    'OCRMYPDF_TEST_OUTPUT',
-    default=os.path.join(PROJECT_ROOT, 'tests', 'output', 'hocrtransform'))
-
-
-def setup_module():
-    with suppress(FileNotFoundError):
-        shutil.rmtree(TEST_OUTPUT)
-    with suppress(FileExistsError):
-        os.makedirs(TEST_OUTPUT)
-    with open(_make_output('blank.hocr'), 'w') as f:
+@pytest.fixture
+def blank_hocr(tmpdir):
+    filename = Path(str(tmpdir)) / "blank.hocr"
+    with open(str(filename), 'w') as f:
        f.write(HOCR_TEMPLATE)
+    return filename


-def _make_input(input_basename):
-    return os.path.join(TEST_RESOURCES, input_basename)
-
-
-def _make_output(output_basename):
-    return os.path.join(TEST_OUTPUT, output_basename)
-
-
-def test_mono_image():
+def test_mono_image(blank_hocr, outdir):
    im = Image.new('1', (8, 8), 0)
    for n in range(8):
        im.putpixel((n, n), 1)
-    im.save(_make_output('mono.tif'), format='TIFF')
-
-    hocr = hocrtransform.HocrTransform(_make_output('blank.hocr'), 300)
-    hocr.to_pdf(_make_output('mono.pdf'), imageFileName=_make_output('mono.tif'))
-
+    im.save(outdir / 'mono.tif', format='TIFF')

+    hocr = hocrtransform.HocrTransform(str(blank_hocr), 300)
+    hocr.to_pdf(
+        str(outdir / 'mono.pdf'), imageFileName=str(outdir / 'mono.tif'))
+
+    qpdf.check(str(outdir / 'mono.pdf'))



@@ -14,38 +14,9 @@ import pytest
 import sys


-if sys.version_info.major < 3:
-    print("Requires Python 3.4+")
-    sys.exit(1)
-
-TESTS_ROOT = os.path.abspath(os.path.dirname(__file__))
-SPOOF_PATH = os.path.join(TESTS_ROOT, 'spoof')
-PROJECT_ROOT = os.path.dirname(TESTS_ROOT)
-OCRMYPDF = os.path.join(PROJECT_ROOT, 'OCRmyPDF.sh')
-TEST_RESOURCES = os.path.join(PROJECT_ROOT, 'tests', 'resources')
-TEST_OUTPUT = os.environ.get(
-    'OCRMYPDF_TEST_OUTPUT',
-    default=os.path.join(PROJECT_ROOT, 'tests', 'output', 'pageinfo'))
-
-
-def setup_module():
-    with suppress(FileNotFoundError):
-        shutil.rmtree(TEST_OUTPUT)
-    with suppress(FileExistsError):
-        os.makedirs(TEST_OUTPUT)
-
-
-def _make_input(input_basename):
-    return os.path.join(TEST_RESOURCES, input_basename)
-
-
-def _make_output(output_basename):
-    return os.path.join(TEST_OUTPUT, output_basename)
-
-
-def test_single_page_text():
-    filename = os.path.join(TEST_OUTPUT, 'text.pdf')
-    pdf = Canvas(filename, pagesize=(8*72, 6*72))
+def test_single_page_text(outdir):
+    filename = outdir / 'text.pdf'
+    pdf = Canvas(str(filename), pagesize=(8*72, 6*72))
    text = pdf.beginText()
    text.setFont('Helvetica', 12)
    text.setTextOrigin(1*72, 3*72)
@@ -55,7 +26,7 @@ def test_single_page_text():
    pdf.showPage()
    pdf.save()

-    pdfinfo = pageinfo.pdf_get_all_pageinfo(filename)
+    pdfinfo = pageinfo.pdf_get_all_pageinfo(str(filename))

    assert len(pdfinfo) == 1
    page = pdfinfo[0]
@@ -64,28 +35,26 @@ def test_single_page_text():
    assert len(page['images']) == 0


-def test_single_page_image():
-    filename = os.path.join(TEST_OUTPUT, 'image-mono.pdf')
+@pytest.mark.skipif(sys.version_info < (3, 5), reason="needs Path.read_bytes")
+def test_single_page_image(outdir):
+    filename = outdir / 'image-mono.pdf'

-    with NamedTemporaryFile(mode='wb+', suffix='.png') as im_tmp:
-        im = Image.new('1', (8, 8), 0)
-        for n in range(8):
-            im.putpixel((n, n), 1)
-        im.save(im_tmp.name, format='PNG')
+    im_tmp = outdir / 'tmp.png'
+    im = Image.new('1', (8, 8), 0)
+    for n in range(8):
+        im.putpixel((n, n), 1)
+    im.save(str(im_tmp), format='PNG')

-        imgsize = ((img2pdf.ImgSize.dpi, 8), (img2pdf.ImgSize.dpi, 8))
-        layout_fun = img2pdf.get_layout_fun(None, imgsize, None, None, None)
+    imgsize = ((img2pdf.ImgSize.dpi, 8), (img2pdf.ImgSize.dpi, 8))
+    layout_fun = img2pdf.get_layout_fun(None, imgsize, None, None, None)

-        im_tmp.seek(0)
-        im_bytes = im_tmp.read()
-        pdf_bytes = img2pdf.convert(
-                im_bytes, producer="img2pdf", with_pdfrw=False,
-                layout_fun=layout_fun)
+    im_bytes = im_tmp.read_bytes()
+    pdf_bytes = img2pdf.convert(
+            im_bytes, producer="img2pdf", with_pdfrw=False,
+            layout_fun=layout_fun)
+    filename.write_bytes(pdf_bytes)

-        with open(filename, 'wb') as pdf:
-            pdf.write(pdf_bytes)
-
-    pdfinfo = pageinfo.pdf_get_all_pageinfo(filename)
+    pdfinfo = pageinfo.pdf_get_all_pageinfo(str(filename))

    assert len(pdfinfo) == 1
    page = pdfinfo[0]
@@ -102,9 +71,9 @@ def test_single_page_image():
    assert abs(pdfimage['dpi_h'] - 8) < 1e-5


-def test_single_page_inline_image():
-    filename = os.path.join(TEST_OUTPUT, 'image-mono-inline.pdf')
-    pdf = Canvas(filename, pagesize=(8*72, 6*72))
+def test_single_page_inline_image(outdir):
+    filename = outdir / 'image-mono-inline.pdf'
+    pdf = Canvas(str(filename), pagesize=(8*72, 6*72))
    with NamedTemporaryFile() as im_tmp:
        im = Image.new('1', (8, 8), 0)
        for n in range(8):
@@ -115,7 +84,7 @@ def test_single_page_inline_image():
        pdf.showPage()
        pdf.save()

-    pdfinfo = pageinfo.pdf_get_all_pageinfo(filename)
+    pdfinfo = pageinfo.pdf_get_all_pageinfo(str(filename))
    print(pdfinfo)
    pdfimage = pdfinfo[0]['images'][0]
    assert (pdfimage['dpi_w'] - 8) < 1e-5
@@ -123,12 +92,19 @@ def test_single_page_inline_image():
    assert pdfimage['width'] == 8


-def test_jpeg():
-    filename = _make_input('c02-22.pdf')
+def test_jpeg(resources, outdir):
+    filename = resources / 'c02-22.pdf'

-    pdfinfo = pageinfo.pdf_get_all_pageinfo(filename)
+    pdfinfo = pageinfo.pdf_get_all_pageinfo(str(filename))

    pdfimage = pdfinfo[0]['images'][0]
    assert pdfimage['enc'] == 'jpeg'
    assert (pdfimage['dpi_w'] - 150) < 1e-5

+
+def test_form_xobject(resources):
+    filename = resources / 'formxobject.pdf'
+
+    pdfinfo = pageinfo.pdf_get_all_pageinfo(str(filename))
+    pdfimage = pdfinfo[0]['images'][0]
+    assert pdfimage['width'] == 50
@@ -0,0 +1,28 @@
+#!/usr/bin/env python3
+# © 2017 James R. Barlow: github.com/jbarlow83
+
+import pytest
+from ocrmypdf.exceptions import ExitCode
+from ocrmypdf.exec import tesseract
+
+
+# Skip all tests in this file if not tesseract 3
+pytestmark = pytest.mark.skipif(tesseract.v4(),
+                                reason="tesseract 3.x required")
+
+
+def test_textonly_pdf_on_tess3(resources, no_outpdf):
+    p, _, _ = pytest.helpers.run_ocrmypdf(
+        resources / 'linn.pdf',
+        no_outpdf, '--pdf-renderer', 'tess4')
+
+    assert p.returncode == ExitCode.missing_dependency
+
+
+def test_oem_on_tess3(resources, no_outpdf):
+    p, _, err = pytest.helpers.run_ocrmypdf(
+        resources / 'aspect.pdf',
+        no_outpdf, '--tesseract-oem', '1')
+
+    assert p.returncode == ExitCode.ok
+    assert 'argument ignored' in err
@@ -0,0 +1,43 @@
+#!/usr/bin/env python3
+# © 2017 James R. Barlow: github.com/jbarlow83
+
+import pytest
+from ocrmypdf.exceptions import ExitCode
+from ocrmypdf.exec import tesseract
+from ocrmypdf import pageinfo
+import sys
+
+
+# Skip all tests in this file if not tesseract 4
+pytestmark = pytest.mark.skipif(
+    not (tesseract.v4() and tesseract.has_textonly_pdf()),
+    reason="tesseract 4.0 with textonly_pdf feature required")
+
+check_ocrmypdf = pytest.helpers.check_ocrmypdf
+run_ocrmypdf = pytest.helpers.run_ocrmypdf
+spoof = pytest.helpers.spoof
+
+
+def test_textonly_pdf(resources, outdir):
+    check_ocrmypdf(
+        resources / 'linn.pdf',
+        outdir / 'linn_textonly.pdf', '--pdf-renderer', 'tess4')
+
+
+@pytest.mark.skipif(sys.version_info < (3, 5), reason="needs math.isclose")
+def test_pagesize_consistency_tess4(resources, outpdf):
+    from math import isclose
+
+    infile = resources / 'linn.pdf'
+
+    before_dims = pytest.helpers.first_page_dimensions(infile)
+
+    check_ocrmypdf(
+        infile,
+        outpdf, '--pdf-renderer', 'tess4',
+        '--clean', '--deskew', '--remove-background', '--clean-final')
+
+    after_dims = pytest.helpers.first_page_dimensions(outpdf)
+
+    assert isclose(before_dims[0], after_dims[0])
+    assert isclose(before_dims[1], after_dims[1])
Author	SHA1	Message	Date
James R. Barlow	199de96cff	Ghostcript 9.21 seems to have a regression related to Unicode metadata	2017-03-24 15:15:46 -07:00
James R. Barlow	8ddbe81513	Fix issue #147 : unpaper loses DPI information, affects —pdf-renderer tess4	2017-03-24 13:23:03 -07:00
James R. Barlow	a3e26e0498	Make —pdf-renderer tess4 more informative, less FUD	2017-03-24 12:58:40 -07:00
James R. Barlow	4ad129d8d8	docs: Don't recommend system-site-packages anymore (Ubuntu 16.04) Not needed since reportlab 3.4 comes with a wheel, and that was the main difficulty.	2017-03-23 08:13:28 -07:00
James R. Barlow	dfb9fa0736	autobrew: missing deps	2017-03-16 00:18:07 -07:00
James R. Barlow	eb036898e9	Update documentation with macOS homebrew tap [ci skip]	2017-03-13 22:25:25 -07:00
James R. Barlow	7c6aa76a2a	Fix brew audit failure	2017-03-13 17:58:31 -07:00
James R. Barlow	f035cb1088	Fixed issue #142 — closed streams raise an exception on fork attempt	2017-03-13 15:52:57 -07:00
James R. Barlow	35162166c5	Fix UnboundLocalError in autobrew.py	2017-03-13 15:51:39 -07:00
James R. Barlow	107f6abcb1	Fix autobrew build issues - missing deps	2017-03-12 17:07:38 -07:00
James R. Barlow	760a939e7d	Further autobrew tweaks	2017-03-11 17:05:26 -08:00
James R. Barlow	72660d0dec	MacOS skip the one test that needs poppler, to save installing poppler	2017-03-11 17:03:26 -08:00
James R. Barlow	8444a8f211	before_deploy doesn’t run unless something is going to be deployed	2017-03-11 17:03:20 -08:00
James R. Barlow	4a1fec8328	Improvements to macOS test and work on homebrew tap autobrew Squashed commits: [3f06c1e] Try setting up homebrew tap autobuilding [01532f1] Strict mode error in brew	2017-03-11 17:00:54 -08:00
James R. Barlow	42547f6017	Revert "Finalize Dockerfile move; unfortunately not supported by Docker Hub" Unfortunately because of this issue https://github.com/docker/hub-feedback/issues/292 Docker Hub effectively automatically changes directory to the location of Dockerfile, meaning needed files are not in the right places, so this has to be reverted. This reverts commit `65c9a07dde`. [ci skip]	2017-03-10 22:49:08 -08:00
James R. Barlow	0ccf564f03	Revert "Move Dockerfiles out of root" This reverts commit `3d3b3abc1b`.	2017-03-10 22:46:03 -08:00
James R. Barlow	65c9a07dde	Finalize Dockerfile move; unfortunately not supported by Docker Hub Unfortunately because of this issue https://github.com/docker/hub-feedback/issues/292 Docker Hub effectively automatically changes directory to the location of Dockerfile, meaning needed files are not in the right places, so this has to be reverted.	2017-03-10 22:45:31 -08:00
James R. Barlow	4700a19322	Move pipeline.svg out of root	2017-03-10 17:18:38 -08:00
James R. Barlow	3d3b3abc1b	Move Dockerfiles out of root	2017-03-10 17:17:36 -08:00
James R. Barlow	7cd2770a13	Fix issue #137 - proportions of non-square resolution distorted Distortion mainly affected —force-ocr	2017-02-26 17:13:16 -08:00
James R. Barlow	7b94129d9e	v4.5 notes	2017-02-14 13:03:48 -08:00
James R. Barlow	d1a0065ef8	Create test case for Form XObjects	2017-02-14 12:51:15 -08:00
James R. Barlow	5a817370fd	Warn more strongly about —pdf-renderer tesseract until fix is widely propagated	2017-02-14 11:33:07 -08:00
James R. Barlow	ab0a210763	Update dockerfile.tess4 yet again Installing Tess4 PPA over Tess3 proved too much pain, so sever the link between this and the jbarlow83/ocrmypdf image, starting each from scratch. Also the complete set of language packs proves too much - the build seems likely to fail when trying to install so many.	2017-02-14 11:32:05 -08:00
James R. Barlow	9f800736bc	Fix running_in_docker() check failing on newer Docker This test has to work to ensure spoof/tesseract_cache.py has a writable directory to put cache into. Otherwise those tests fail.	2017-02-13 02:16:06 -08:00
James R. Barlow	c9a83afad6	Improve batch processing examples	2017-02-13 02:14:32 -08:00
James R. Barlow	5e14274f10	pageinfo: learn to extract image information from Form XObjects	2017-02-11 16:48:59 -08:00
James R. Barlow	167470b4bd	Re-fix Dockerfile.tess4 [ci skip]	2017-02-10 15:43:49 -08:00
James R. Barlow	f06d3c2ec2	Fix tesseract 3.04 on tesseract 4 on image [skip ci]	2017-02-10 08:38:22 -08:00
James R. Barlow	74c99a8a77	v4.4.2 release notes	2017-02-06 21:56:55 -08:00
James R. Barlow	0e4d312ee2	Adjust Travis deploy to PyPI settings -only on master branch -only Python 3.6 build uploads, so the others don’t compete -don’t upload docs to PyPI	2017-02-06 21:27:59 -08:00
James R. Barlow	589f19559d	Rewrite Dockerfiles to use ubuntu 16.10 base system Debian now has a few disadvantages: -there is no convenient PPA for Debian tesseract 4.0, but there is for Ubuntu -Ubuntu sets locale to UTF-8 automatically removing the need to do this All three ocrmypdf docker images are now based on a common Ubuntu 16.10 image, derived from the one used to build ocrmypdf-tess4. -polyglot now differs from -tess4 only by opting into the tess4 PPA. Both Ubuntu 16.10 and Debian stretch use tesseract 3.04.01 now making the sharp.ttf patch unnecessary. /etc/apt/sources has been unused for a while now both have newer Ghostscripts.	2017-02-06 14:39:29 -08:00
James R. Barlow	f28bc25dc0	Configure travis to handle deployment to PyPI; also lint .travis.yml	2017-02-06 13:50:53 -08:00
James R. Barlow	a0657ad937	Prevent use of —pdf-renderer tess4 on tesseract 3	2017-02-06 13:49:43 -08:00
James R. Barlow	5b8d88af4c	Suggest use of aliases to hide docker run	2017-01-30 15:08:02 -08:00
James R. Barlow	fa82b50340	Adding missing file Dockerfile.tess4	2017-01-29 18:34:01 -08:00
James R. Barlow	005216bc57	Support ocrmypdf-tess4	2017-01-29 18:26:52 -08:00
James R. Barlow	e748fdcf6f	v4.4.1 release notes	2017-01-28 22:23:35 -08:00
James R. Barlow	8c17c9918e	Add documentation and test cases for —tesseract-config This parameter has existed for along time but never really got any attention.	2017-01-28 22:06:51 -08:00
James R. Barlow	ea0dd99d0b	More documentation updates	2017-01-28 15:35:59 -08:00
James R. Barlow	e0cc67afae	docs: suggest —oem 1	2017-01-28 14:58:25 -08:00
James R. Barlow	04f9cbe364	Describe how to use tesseract 4.0 while 3.04 is installed	2017-01-27 18:13:59 -08:00
James R. Barlow	99afebd033	tesseract jobs_limit(2) At least on macOS with my quadcore performance improves with two tesseracts in parallel (20% gain). Hard to say how this will affect Linux, but stepping up to 2 jobs seems justifiable.	2017-01-27 18:13:12 -08:00
James R. Barlow	a6feacc810	travis: fix ‘pip install’ by moving working code out of the way	2017-01-27 14:33:23 -08:00
James R. Barlow	65e4b1672f	cffi: verbose=True	2017-01-27 14:17:13 -08:00
James R. Barlow	46cc0dd190	Revert "Do we need to exclude ocrmypdf.lib?" This reverts commit `678b9fb603`.	2017-01-27 13:51:30 -08:00
James R. Barlow	678b9fb603	Do we need to exclude ocrmypdf.lib?	2017-01-27 13:49:11 -08:00
James R. Barlow	49ab0c1f0b	setup.py: cffi is definitely needed in setup_requires	2017-01-27 13:43:47 -08:00
James R. Barlow	ab490a7736	Experiment: update *requirements.txt, use more current travis build steps Perhaps this works around the pip/setup.py asymmetry that broke the 4.4 release.	2017-01-27 13:13:14 -08:00
James R. Barlow	e4ce1dae35	setup.py: for some reason, subpackages must be explicitly specified	2017-01-27 00:37:05 -08:00
James R. Barlow	179b812acb	Fix readthedocs build error	2017-01-26 23:57:51 -08:00
jbarlow83andGitHub	7f170517ec	Note about pytest-helpers-namespace	2017-01-26 23:15:32 -08:00
James R. Barlow	5480da4f04	Additional docs updates for v4.4	2017-01-26 23:02:44 -08:00
James R. Barlow	9a15a4db10	Ensure specified destination is writable before starting pipeline process	2017-01-26 22:08:24 -08:00
James R. Barlow	55aeaec293	Autorotation check: Replace duplicated tests with parameterized test	2017-01-26 18:07:59 -08:00
James R. Barlow	f6df1fb40c	Fix test suite regression: output files dumped in tests/resources	2017-01-26 18:07:09 -08:00
James R. Barlow	b889a89c36	Fix remaining 3.4/3.5 regressions	2017-01-26 17:53:27 -08:00
James R. Barlow	1976dc6f30	Fix issue #121 “pop from empty list” (content stream parsing error)	2017-01-26 17:24:40 -08:00
James R. Barlow	e864c65d26	(Hopefully) Fix Path <-> py.path conversion on Py3.4/3.5	2017-01-26 17:19:15 -08:00
James R. Barlow	02fba02d31	Refactor test suite to use fixtures to manage paths	2017-01-26 16:38:59 -08:00
James R. Barlow	fb9e7c82f6	Move duplicate test code into common namespace	2017-01-26 13:36:52 -08:00
James R. Barlow	77d31bf646	Add renderers page (missed from previous)	2017-01-26 13:20:44 -08:00
James R. Barlow	29ca799bcf	Move pytest.ini into setup.cfg	2017-01-26 12:45:38 -08:00
James R. Barlow	467b7f0163	Update docs for eventual v4.4 release	2017-01-26 12:29:11 -08:00
James R. Barlow	bad67c6dc5	Rename ‘tesstop’ to ‘tess4’ There’s no reason text-only PDF shouldn’t become the default for tesseract 4.	2017-01-26 12:28:51 -08:00
James R. Barlow	ac40426971	Implement “tesstop” (tesseract v4 text-only pages - working name)	2017-01-20 17:16:01 -08:00
James R. Barlow	7acfaf6d34	pipeline: rename some of the stages, for clarity	2017-01-20 17:15:00 -08:00
James R. Barlow	99e47c9c04	tesseract: add support for using v4 textonly_pdf feature	2017-01-20 17:06:23 -08:00
James R. Barlow	d7904e2251	Travis now has Python 3.6, test against it	2017-01-20 14:26:17 -08:00
James R. Barlow	68aef489de	Merge branch 'master' (4.3.5, Python 3.6 support) into develop # Conflicts: # dev_requirements.txt # requirements.txt	2017-01-20 14:25:28 -08:00
James R. Barlow	3f9adcd5e0	Document idea for producing companion text files	2017-01-19 16:48:05 -08:00
James R. Barlow	6cc5135d2d	Output to stdout: ensure stdout is flushed to prevent truncation errors	2017-01-19 16:41:10 -08:00
James R. Barlow	d4c72b371f	Forward --oem argument to tesseract 4	2017-01-18 21:37:50 -08:00
James R. Barlow	18b6f05657	Resolve issue #124 - poor performance with Tesseract v4 It seems that Tesseract v4 on a platform with OpenMP working correctly while perform poorly with ocrmypdf because each will also soak up all available CPUs. Running N^2 processes/threads on a N-core CPU where each wants 100% of CPU turns out to be detrimental. So, we restrict ocrmypdf w/tessv4 to a single Tesseract process at a time, for now. Alternative may be to limit OpenMP threads if throughput is higher.	2017-01-18 17:52:12 -08:00
James R. Barlow	c42d9baa26	tesseract: for v4, use --psm while keeping -psm for v3 At the moment v4 accepts both but who knows if this will get dropped, so do as document for each version.	2017-01-18 17:43:47 -08:00
James R. Barlow	6e27ecd2b9	Finalize ‘exec’ migration and make it backward compatibility for now	2017-01-18 17:40:50 -08:00
James R. Barlow	482692396e	Add installation instructions for Ubuntu 16.04	2017-01-17 08:17:56 -08:00
James R. Barlow	c48acf165a	v4.3.5: Python 3.6 compatibility	2017-01-03 00:45:33 -08:00
James R. Barlow	9e004c3ec0	Another attempt at py 3.4/3.5 Revert to exactly what the previous passing build specified.	2017-01-03 00:34:26 -08:00
James R. Barlow	7be4e9c919	fix setuptools-scm for py 3.4, 3.5	2017-01-03 00:25:57 -08:00
James R. Barlow	5ec38a4bed	Update requirements files and documentation for Python 3.6 - no code changes	2017-01-03 00:11:34 -08:00
James R. Barlow	f246779b8e	pdfa: documentation, remove from __future__	2016-12-12 15:10:10 -08:00
James R. Barlow	a7d8cdf061	Don’t copy pageinfo - job manager already provides a copy of real pdfinfo	2016-12-12 15:09:41 -08:00
James R. Barlow	620745c812	pipeline: don’t use qpdf to check page count again We already know the number of pages at this stage.	2016-12-12 15:09:11 -08:00
James R. Barlow	b8767e5ba9	Rename exe -> exec, more Unix-y and suggestive	2016-12-10 15:34:00 -08:00
James R. Barlow	d33a50660d	Replace most sys.exit() with raising exceptions Because ruffus doesn’t handle exceptions well I tended to call sys.exit to make sure we got out of dodge when needed. However, sys.exit is not ideal for the Python API this is moving towards, so this introduces proper exceptions for the various cases that retain suggested error codes. Only __main__.py should call sys.exit now, everyone else has to throw an exception. For now the worker raising a fatal exception is logging messages rather than passing an exception object with the fatal error message, mainly because ruffus doesn’t properly marshall the exception object so we just check “what is the name of the exception class that caused ruffus to thrown an RethrownJobError”? Also fixed along the way was the wrong return code being shown for encrypted PDF checking, and incorrect use of str.find (e.output.find) in boolean logic (str.find returns -1 on failure to find, which is True).	2016-12-10 15:24:24 -08:00
James R. Barlow	4ee9658e97	Move external program wrappers to ocrmypdf.exe package	2016-12-09 16:54:24 -08:00
James R. Barlow	dd1b84e7ba	More refactoring - helpers.py	2016-12-09 16:31:08 -08:00
James R. Barlow	4c677e6c47	Extract pipeline out of __main__.py and into pipeline.py This leaves __main__.py to handle command line arguments while pipeline.py runs the pipeline - mostly. They are still somewhat intertwined, with __main__.py doing essential things for pipeline.py, etc., and some helper functions that could go in their own module. All tests pass after this major refactor.	2016-12-09 16:17:12 -08:00
James R. Barlow	f0f889440b	Merge branch 'master' into feature/ooruffus	2016-12-08 16:36:03 -08:00
James R. Barlow	cc9ceaeb74	v4.3.4: release notes	2016-12-08 16:34:09 -08:00
James R. Barlow	ad2fa8d1d7	Fix MANIFEST for .png	2016-12-08 16:25:04 -08:00
James R. Barlow	adc1580742	Help py.test collect output in more cases	2016-12-08 16:21:07 -08:00
James R. Barlow	4d3b44d6df	ghostscript: cleanup harmless error message printed for overprint Redirect stderr->stdout to hopefully make GS output easier to work with overall, since the previous code didn’t seem to account for mixed used properly.	2016-12-08 16:19:15 -08:00
James R. Barlow	e57aa0eee2	pageinfo: fix “decimal.InvalidOperation: quantize result has too many digits” And add new test case for this.	2016-12-08 16:06:53 -08:00
James R. Barlow	1ae1d116c7	Make setup.py license internally consistent	2016-12-08 16:06:31 -08:00
James R. Barlow	097a69d07f	pageinfo: fix “decimal.InvalidOperation: quantize result has too many digits” And add new test case for this.	2016-12-08 16:04:14 -08:00
James R. Barlow	a81ce87a50	Remove non-reentrant options checking and logging setup	2016-12-05 14:13:36 -08:00
James R. Barlow	88be0d43a0	Make setup.py license internally consistent	2016-12-03 21:37:24 -08:00
James R. Barlow	ff16a00a3d	Remove test for Pillow JPEG and PNG As of 3.1.1, our minimum version, these codecs are now required by default for a successful installation, effectively solving the problem of Pillow installed without libjpeg/libpng.	2016-12-03 14:25:46 -08:00
James R. Barlow	8982b3e1e2	Update requirements -update requirements.txt and dev_requirements.txt to more recent version -setup.py updated to Ubuntu 14.04 rather than 12.04 backports -request at least Pillow 3.1.1 now (since this makes jpeg/png mandatory)	2016-12-03 14:14:07 -08:00
James R. Barlow	be0fa35d14	Merge branch 'master' into feature/ooruffus	2016-12-03 14:02:43 -08:00
James R. Barlow	9f51ed9d01	Finalize v4.3.3 release notes	2016-12-03 00:39:24 -08:00
James R. Barlow	731e6792c7	Add test cases for Ghostscript PDF/A warnings	2016-12-03 00:32:09 -08:00
James R. Barlow	c35ec0b4aa	ghostscript: more effort at error logging	2016-12-03 00:22:03 -08:00
James R. Barlow	03aaf575dc	v4.3.3 release notes, fix more gs 9.20 issues	2016-12-02 16:26:34 -08:00
James R. Barlow	9a060579ba	Move work_folder into multiprocessing manager	2016-12-02 01:39:17 -08:00
James R. Barlow	d40a5c4f7a	Remove all remaining traces of ‘options’ global state from task runners	2016-12-02 01:31:57 -08:00
James R. Barlow	21f7dc3377	Distribute ‘options’ to worker processes via the multiprocessing manager	2016-12-02 01:06:11 -08:00
James R. Barlow	43c13a1ed9	Replace pdfinfo, pdfinfo_lock with multiprocessing manager Using a context manager to guard the pdfinfo list makes the lock unnecessary. (Although it was probably unnecessary in the first place anyway.)	2016-12-01 23:36:30 -08:00
James R. Barlow	6bc3f189e1	Remove “WrappedLogger” - does not do anything useful Never really investigated the reason why ruffus returns a mutex to go along with its logger. It seems that the mutex is only needed if one wanted to make multiple successive calls to a log function and have them appear appear atomically. It is not needed to protect the logger proxy because accessing the proxy triggers IPC in the child process that handles the multiprocessing.Manager() object. The logging wrapper only logs one line at a time, so the mutex does not actually protect logging sequence. Cut it. Also manager.Lock() returns a threading.Lock object so the purpose of it is actually to help processes share a thread-level lock. It would be more appropriate to use a semaphore based multiprocessing.Lock.	2016-12-01 15:27:07 -08:00
James R. Barlow	2c5437135c	Remove temporary re_symlink logging shim	2016-12-01 00:31:42 -08:00
James R. Barlow	444da02523	Fix mistake made in converting pipeline; incredibly, all tests pass now	2016-12-01 00:30:19 -08:00
James R. Barlow	00e8af2381	Reactivate the pipeline; surprisingly works in quick test	2016-12-01 00:03:03 -08:00
James R. Barlow	401b21864f	Convert to object oriented ruffus syntax (does not run) I experimented with the idea of using asyncio-based processing but realized that that does not solve the import time binding problem that is the real issue. Therefore the simpler refactoring is to convert to ruffus-oo syntax and get things working again. build_pipeline() is really ugly at the moment. The old syntax had its advantages. This test reproduces the complete pipeline graph but does not work otherwise.	2016-11-30 23:58:26 -08:00
James R. Barlow	de939951d4	Record version in debug log	2016-11-29 15:30:50 -08:00
James R. Barlow	7725d16a26	Fix exception on inline stencil masks with no /CS attribute	2016-11-24 22:37:00 -08:00
James R. Barlow	8a74408d83	Add security suggestions	2016-11-21 20:58:31 -08:00
James R. Barlow	3d0dc95a06	Moved venvs	2016-11-21 20:40:22 -08:00
James R. Barlow	04a57a3cc2	OS X -> macOS	2016-11-21 20:40:06 -08:00
James R. Barlow	d0c22ce01d	v4.3.2 release notes	2016-11-10 23:16:08 -08:00
James R. Barlow	23c95e9660	ghostscript: elide overprinting to fix PDF/A errors in GS 9.20 It looks like GS 9.19 can incorrectly set overprinting for the text layer even though this makes no sense in PDF/A, or at least someone produced PDFs that have this after a Tesseract PDF -> GS PDF/A conversion. GS 9.20 complains about this. Instead of aborting, elide the feature. See http://git.ghostscript.com/?p=ghostpdl.git;a=commitdiff;h=094d5a1880f1cb9ed320ca9353eb69436e09b594 and issue #107. It looks like it is better to elide features and warn about elision rather than abort with an error.	2016-11-10 14:48:02 -08:00
James R. Barlow	eecab9b95d	pdfa: fix KeyError on pdfa_dict if document has some xmp metadata but not exactly what we’re looking for	2016-11-09 05:41:12 -08:00