grepdiff: invert match #36

Activity

1. twaugh commented on Dec 6, 2019

   

   twaugh

   on Dec 6, 2019

   Owner

   More actions

   --invert-match sounds like the right parameter name for this. Probably needs an invert_regexecs() similar to match regexecs(), and a way of switching between them appropriately (maybe a function pointer).

   I don't have time to code this myself right now. Want to have a go at it?
2. camsteffen commented on Dec 7, 2019

   

   camsteffen

   on Dec 7, 2019

   Author

   More actions

   Thanks for your response. I'm also a little too busy to claim this right now.
3. jansmets commented on Apr 11, 2021

   

   jansmets

   on Apr 11, 2021

   More actions

   Would be useful indeed.

   👍React with 👍1Reacted by Profpatsch
4. twaugh commented on Aug 27, 2025

   

   twaugh

   on Aug 27, 2025

   Owner

   More actions

   Should --invert-match and --output-matching=hunk show individual hunks that don't match, or entire files that have at least one non-matching hunk?

   How should --invert-match interact with --only-match=...? Should it invert the pattern matching or the line-type filtering? For example, with a hunk like this:

   @@ -1,3 +1,3 @@
    unchanged
   -removed_line_with_foo
   +added_line_with_bar
   +added_line_with_foo
   

   should grepdiff --invert-match --only-match=add foo not show this file because the addition lines match 'foo'?

   I'm particularly interested in which workflows an --invert-match option enables. I'm thinking something like "using git, unstage hunks that do not match", but what others are there?
5. camsteffen commented on Aug 29, 2025

   

   camsteffen

   on Aug 29, 2025

   Author

   More actions

   Should --invert-match and --output-matching=hunk show individual hunks that don't match, or entire files that have at least one non-matching hunk?

   The former. It fits the use case. It's the corollary of grep --invert-match. The latter could be derived from it if needed but that seems like a pretty unlikely use case.

   How should --invert-match interact with --only-match=...? Should it invert the pattern matching or the line-type filtering?

   This seems like a question of order of operations. I think the sensible thing is to apply --only-match and then apply --invert-match. That way the --only-match is part of what gets inverted and so you get more functionality with the combined semantics. If you wanted to apply the --only-match filtering at the end, I think you could use a subsequent invocation of grepdiff.

   For example, ...should grepdiff --invert-match --only-match=add foo not show this file because the addition lines match 'foo'?

   Right, except I'd say "should not show this hunk".

   I'm particularly interested in which workflows an --invert-match option enables. I'm thinking something like "using git, unstage hunks that do not match", but what others are there?

   Yeah that is the use case I thought of. I can't think of a very different use case. A slightly different use case is to create a patch that partially reverts a commit.