What's the difference between gitpython's diff and git log's diff?

Question

I don't understand the difference between the output of git log -p and that of gitpython in terms of diffs.

For example, for some merge commit f534e1..., git log -p gives:

commit f534e1...
Merge: ....
Author: ....
Date: ...
Merge pull request ... from ...

with no diff, which I assume to be correct (the diff comes with the next log entry - one of the parents of f534e1...).

I would like to achieve the same effect with gitpython. I'm trying:

repo = Repo("...")
for c in repo.iter_commits():
 print c.hexsha
 print c.summary
 print c.diff()

I get:

f534e1...
Merge pull request ... from ...
[<git.diff.Diff object at 0x102cd3490>]

with some diff.

What diff is this? Why do I get it here? How can I mimic the behavior of git log -p?

• user3159253: What are the actual properties of the git.diff.Diff object?
• Bogdan Vasilescu: gitpython.readthedocs.org/en/latest/…
• user3159253: Huh, I'd read the docs :) I asked what are the values of those properties? Have you compared the resulted diff (if it actually contains any significant difference, not just an object of git.diff.Diff class), with git diff ... with each of the merge parents.

Answer 1

According to the documentation, c.diff() will compare the commit against the index, i.e. staged changes.

It appears that git log -p will produce a special-format diff against all of the given commit's parents. Something roughly comparable can be achieved with the following code (based on your example).

repo = Repo("...")
for c in repo.iter_commits():
 print c.hexsha
 print c.summary
 for p in c.parents:
 handle_diff(c.diff(p))

The latter would yield a Diff object with all relevant information.

If what you really want is the exact format produced by git log -p, you may also call repo.git.log(p=True) and parse the output yourself.