pdf-stuff

Scripts for working with PDFs-- comparing PDFs, extracting glyphs, that sort of thing.

Provides:

• pdfdiff.py
  • A script for identifying differences in PDFs by comparing them page by page as images.
• pdf2text.py
  • A script for extracting the text from a PDF.

Installation

I haven't yet uploaded this to PyPI, but you can still clone the repo and install it thus:

# Install like a regular Python package
pip install .

# For development/hacking
pip install --editable .

You can also just run the scripts in a standalone fashion, i.e. without installing the package:

# Standalone
python pdf_stuff/pdfdiff.py one.pdf two.pdf --output="one-vs-two.pdf"

Usage

pdfdiff.py

Compare two PDFs to look for differences.

# Compare two PDFs
pdfdiff.py one.pdf another.pdf

# See help message for more options
pdfdiff.py --help

Example

I created a few PDFs using my makedown project, which is basically a Makefile that wraps pandoc. The default is example/makedown-1.pdf and the one with a bit of customization is example/makedown-2.pdf.

We can see the difference between my slightly customized template and the default by comparing them via:

pdfdiff.py example/makedown-1.pdf example/makedown-2.pdf

The results are mostly identical, except for the lists, which use a consistent bullet type and less spacing when generated via my template.

A	B	Diff

pdf2text.py

Extract the text from a PDF.

Note that this does not (currently) perform any OCR nor attempt to handle any weird glyph shenanigans; it is essentially equivalent to copy-and-pasting text from a PDF that you've opened up in a PDF viewer.

# Extract the text from `file.pdf` and save as `file.pdf.txt`
pdf2text.py file.pdf --output=file.pdf.txt

# If output unspecified, it will print to stdout
pdf2text.py file.pdf

# Use the `pdfplumber` backend
pdf2text.py file.pdf --backend=pdfplumber

pdfmeta.py

Print metadata from the PDF. May eventually support modifying the metadata as well.

# Default, with date parsing and colored output
pdfmeta.py file.pdf

# Verbatim, no colored output
pdfmeta.py --verbatim file.pdf

TODO

Move the following from scattered notebooks into this repo:

• Metadata extraction & pretty-printing
• Image extraction
• Text extraction
• OCR
• HTML image w/ text overlay

Improvements

• More options for pdfdiff.py and pdf2txt.py
  • Handle page ranges
• Options specific to pdfdiff.py
  • Omit pages that are identical
  • Handle pages of different sizes but same aspect ratio
  • Implement more diff algorithms
• Add debug logging to pdfdiff.py
• Image (rather than PDF) output for pdfdiff.py
• Allow for custom page separator in pdf2txt.py?
• pdfmeta.py
  • Allow setting metadata
  • Are there additional things that count as metadata we should look for?