FR: git diff output compatibility, function label for hunks

[botovq]

git's diff has a heuristic to determine what function a patch hunk applies to: it prints what it determines to be the function name after the second @@. Some diff implementations have variants of this under the -p option (or --show-c-function in GNU diff). This is tremendously helpful for reviewing patches in text files. I noticed I use this all the time when composing commit messages and it's seemingly missing from jj:

https://github.com/martinvonz/jj/blob/e55d03a2ee611d08f09a38e01e0cb47c1d6cd315/cli/src/diff_util.rs#L1406-L1413

Compare git's output containing @@ -123,7 +123,7 @@ tls_get_peer_cert_common_name(struct tls *ctx, char **common_name)

$ git show 67f33188 | cat
commit 67f33188bdca6d7612732a59795cf65c184f00ab
Author: Theo Buehler <tb@openbsd.org>
Date:   Thu Nov 7 15:00:24 2024 +0100

diff --git a/lib/libtls/tls_conninfo.c b/lib/libtls/tls_conninfo.c
index 30a4e9403025..1cffba0c3da6 100644
--- a/lib/libtls/tls_conninfo.c
+++ b/lib/libtls/tls_conninfo.c
@@ -123,7 +123,7 @@ tls_get_peer_cert_common_name(struct tls *ctx, char **common_name)
 {
 	if (ctx->ssl_peer_cert == NULL)
 		return (-1);
-	return tls_get_common_name(ctx, ctx->ssl_peer_cert, common_name);
+	return tls_get_common_name(ctx, ctx->ssl_peer_cert, NULL, common_name);
 }

 static int

with jujutsu's output where the function name is missing and only @@ -123,7 +123,7 @@ is shown:

$ jj show --git --no-pager
Commit ID: 67f33188bdca6d7612732a59795cf65c184f00ab
Change ID: klzopryvxnuswttvuynokmlxkrzpsrxy
Author: Theo Buehler <tb@openbsd.org> (2024-11-07 15:00:24)
Committer: Theo Buehler <tb@openbsd.org> (2024-11-08 01:38:34)

    (no description set)

diff --git a/lib/libtls/tls_conninfo.c b/lib/libtls/tls_conninfo.c
index 30a4e94030..1cffba0c3d 100644
--- a/lib/libtls/tls_conninfo.c
+++ b/lib/libtls/tls_conninfo.c
@@ -123,7 +123,7 @@
 {
 	if (ctx->ssl_peer_cert == NULL)
 		return (-1);
-	return tls_get_common_name(ctx, ctx->ssl_peer_cert, common_name);
+	return tls_get_common_name(ctx, ctx->ssl_peer_cert, NULL, common_name);
 }

 static int

The OpenBSD diff manual gives a hint how this could be implemented. I haven't looked at GNU diff and git-diff code - I expect their heuristics to be very similar:

With unified and context diffs, show with each change the first 40 characters of the last line before the context beginning with a letter, an underscore or a dollar sign. For C source code following standard layout conventions, this will show the prototype of the function the change applies to.

Of course this isn't perfect. One well-known problem with this heuristic is that C goto labels tend to take up the spot of the function name, which in turn leads to a common workaround/style convention to add a space in front of the label to get the function name again.

While I'm not sure this is helpful in all languages, and may be considered visual clutter by some, I find this invaluable and a way of having this available would be super helpful.

[ilyagr]

I think modern-day Git has heuristics specific to various languages for this. At some point, I thought of stealing them, but they are GPL-licensed.

This would also make sense as a part of https://github.com/GitoxideLabs/gitoxide. If they have this feature, we could more or less easily borrow it, probably.