cascremers / pdfdiff

Command-line tool to inspect the difference between (the text in) two PDF files

pdfdiff

Command-line tool to inspect the difference between (the text in) two PDF files.

Purpose and function

pdfdiff takes two arguments, each being the filename of a PDF file, and generates a textual diff between the two. It visualises this diff using the first diff-viewer it finds on the system.

pdfdiff relies on pdftotext to extract the plaintext from a PDF file. However, small changes in the text between two PDF files can make a huge difference in the resulting extracted text. More often than not, the difference is so large that doing a diff on the output does not yield a sensible result.

The main function of this program, pdfdiff, is to normalize the output of pdftotext, such that the result is suitable for diff viewing. To achieve this, it attempts to detect sentence endings to reformat paragraphs and lines. Along the way, it removes some ligature encodings to give diff viewers an easier time. After this normalisation procedure, diff viewers commonly yield a substantially better comparison between the contents of the files.

Note that if a single file is provided as input, pdfdiff will directly output the normalised text, enabling its use as a preprocessor for other tools.

Running pdfdiff.py

After downloading, either run it through python or make it executable (chmod +x pdfdiff.py) to use it directly from the commandline.

Requirements

• pdftotext, which is part of the xpdf package.

• Python

• A diff viewer, preferably one that supports unicode, like kdiff3 or meld. If these don't work for you, you can use xxdiff, tkdiff, opendiff, vimdiff, or even good old diff. You only need one of these to use pdfdiff.py.

Note that for most Linux distributions, installing xpdf is usually sufficient to get it working. (Afterwards one might want to upgrade to a better diff viewer though).

Caveats

• pdfdiff ignores many elements of PDF files, such as figures. As a result, if the (textual) difference between two files is empty, there is no guarantee that the PDF files are identical.

• Some PDF files do not contain embedded text. In this case pdftotext will not work correctly, and will return empty diffs. In this case you would need to resort to OCR (Optical Character Recognition) to extract the text. This is outside the scope of this program.

Use cases

• (Scientific) Reviews: you reviewed version A of a paper, and receive version B, and wonder what the changes are.

License

Currently the pdfdiff sources are licensed under the GPL 2, as indicated in the source code. Contact Cas Cremers if you have any questions.