The diff-highlight script works on heuristics, so it can be
wrong. Let's document some of the wrong-ness in case
somebody feels like working on it.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently we only bother highlighting single-line hunks. The
rationale was that the purpose of highlighting is to point
out small changes between two similar lines that are
otherwise hard to see. However, that meant we missed similar
cases where two lines were changed together, like:
-foo(buf);
-bar(buf);
+foo(obj->buf);
+bar(obj->buf);
Each of those changes is simple, and would benefit from
highlighting (the "obj->" parts in this case).
This patch considers whole hunks at a time. For now, we
consider only the case where the hunk has the same number of
removed and added lines, and assume that the lines from each
segment correspond one-to-one. While this is just a
heuristic, in practice it seems to generate sensible
results (especially because we now omit highlighting on
completely-changed lines, so when our heuristic is wrong, we
tend to avoid highlighting at all).
Based on an original idea and implementation by Michał
Kiedrowicz.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The current code structure assumes that we will only look at
a pair of lines at any given time, and that the end result
should always be to output that pair. However, we want to
eventually handle multi-line hunks, which will involve
collating pairs of removed/added lines. Let's refactor the
code to return highlighted pairs instead of printing them.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If you have a change like:
-foo
+bar
we end up highlighting the entirety of both lines (since the
whole thing is changed). But the point of diff highlighting
is to pinpoint the specific change in a pair of lines that
are mostly identical. In this case, the highlighting is just
noise, since there is nothing to pinpoint, and we are better
off doing nothing.
The implementation looks for "interesting" pairs by checking
to see whether they actually have a matching prefix or
suffix that does not simply consist of colorization and
whitespace. However, the implementation makes it easy to
plug in other heuristics, too, like:
1. Depending on the source material, the set of "boring"
characters could be tweaked to include language-specific
stuff (like braces or semicolons for C).
2. Instead of saying "an interesting line has at least one
character of prefix or suffix", we could require that
less than N percent of the line be highlighted.
The simple "ignore whitespace, and highlight if there are
any matched characters" implemented by this patch seems to
give good results on git.git. I'll leave experimentation
with other heuristics to somebody who has a dataset that
does not look good with the current code.
Based on an original idea and implementation by Michał
Kiedrowicz.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
These perl features can catch bugs, and we shouldn't be
violating any of the strict rules or creating any warnings,
so let's turn them on.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is a simple and stupid script for highlighting
differing parts of lines in a unified diff. See the README
for a discussion of the limitations.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
