How to diff PDF files with Git

how-to-diff-pdf-files-with-git-and-diffpdf.md

One can use MD5 or plain text diff to see differences in PDF files. If that's not enough, here's how to use diff-pdf which knows how to diff based on appearance or words:

• brew install diff-pdf
• edit your ~/.gitconfig to add this:

[difftool "diffpdf"]
  cmd = diff-pdf --view \"$LOCAL\" \"$REMOTE\"

• then use with:

git difftool --tool=diffpdf your_pdf_file.pdf

I suspect there's maybe some way to force always using a specific difftool for files with a specific extension; in the mean time I'm just using a bash function for this.

Also useful in .bash_profile:

function git_diff_pdf {
  yes | git difftool --tool=diffpdf $1
}

then you can use it like:

git_diff_pdf your_pdf_file.pdf

This is great, thank you!

As of 2016, it seems that the brew formular has changed its name to diff-pdf. So I tried the following:

brew install diff-pdf

which gives the diff-pdf binary. In OS X, it is convenient to use --view option for viewing the differences in a window.

[difftool "diffpdf"]
  cmd = diff-pdf --view \"$LOCAL\" \"$REMOTE\"

and it worked nice to me.

@wookayin great, thanks for the feedback! I've updated the gist.

@thbar thanks for the tool

However, in order to have it run, I changed the command to

diff-pdf --view \"$LOCAL\" \"$REMOTE\"

@ggrrll thanks! I've updated the gist.

Thanks for this gist. Note: diffpdf under Linux doesn't seem to be have a --view option. So in this case the following is sufficient (the escaping of the double quotes is also not necessary in my point of view):

[difftool "diffpdf"]
    cmd = diffpdf "$LOCAL" "$REMOTE"

Note: brew install diff-pdf installs https://vslavik.github.io/diff-pdf/ (GPL licensed, sources: https://github.com/vslavik/diff-pdf), not the "commercial Windows graphical user interface" DiffPDF as mentioned in this gist. See https://formulae.brew.sh/formula/diff-pdf.

@seppestas thanks for the notice! I just updated the link.

@grubersjoe I believe this is (as pointed out by @seppestas) because it's not the correct program (although named almost exactly the same). Check out this version, which installs on Linux and has the --view option: https://github.com/vslavik/diff-pdf

Ah... I see 😄. I've mistakenly used http://www.qtrac.eu/diffpdf-foss.html. Still works though.

From https://nbdime.readthedocs.io/en/latest/vcs.html#manual-registration :

To associate the diff driver with a file type, add the following entry to the appropriate .gitattributes file:

*.ipynb diff=jupyternotebook

Would be great to have this support in git core.

Note: brew install diff-pdf installs https://vslavik.github.io/diff-pdf/ (GPL licensed, sources: https://github.com/vslavik/diff-pdf), not the "commercial Windows graphical user interface" DiffPDF as mentioned in this gist. See https://formulae.brew.sh/formula/diff-pdf.

FWIW, an earlier version of DiffPDF is still available under the GPL v2 from http://www.qtrac.eu/diffpdf-foss.html (and packaged in Debian).

If you use git-annex, you can diff PDFs via

[difftool "diffpdf"]
    cmd = diffpdf \"$(cat $LOCAL)\" \"$(cat $REMOTE)\"

Rationale: The $LOCAL and $REMOTE variables point to temporary files containing the two versions. With git-annex, they contain only a path to the actual file in the .git/annex/objects store. That's why we need to cat these files to get the actual path and make them viewed in diffpdf.

Edit:

For some more ninja-coolness, you can use

[difftool "diffpdf"]
       cmd = diffpdf \"$(git rev-parse --show-toplevel)/$(cat $LOCAL)\" \"$(git rev-parse --show-toplevel)/$(cat $REMOTE)\"

which makes them paths absolute, so you can execute this from a sub-directory of your repository, too!

With a recent version of git (I'm using 2.40.1) you can do this pretty easily using diff in both the git config and .gitattributes (official docs):

In your project's .gitattributes:

*.pdf diff=diff-pdf

In your ~/.gitconfig or your project's .git/config:

[diff "diff-pdf"]
        command = f() { diff-pdf --view "$2 $1;" } ";" f