<![CDATA[I've run into an interesting problem that I believe doesn't have a well-known solution.]]>I've run into an interesting problem that I believe doesn't have a well-known solution.

Given a diff file, compute a derived diff file which only removes/adds lines that match a regular expression, leaving non-matching lines unchanged.

I would appreciate suggestions/advice!

If I solve it myself, I'll write a blog post.

]]>https://board.circlewithadot.net/topic/141e2863-abb2-4d7c-a470-6094b83d4004/i-ve-run-into-an-interesting-problem-that-i-believe-doesn-t-have-a-well-known-solution.RSS for NodeFri, 01 May 2026 13:11:14 GMTSun, 19 Apr 2026 22:47:04 GMT60<![CDATA[Reply to I've run into an interesting problem that I believe doesn't have a well-known solution. on Mon, 20 Apr 2026 07:50:10 GMT]]>@krans Thinking of your example, surely that's not automatically decideable?

context1
-irrelevant1
-old
+new
+irrelevant2
context2

Should the output become
context1
irrelevant1
new
irrelevant2
context2

or

context1
new
irrelevant1
irrelevant2
context2
?

or any other potential resultant output?

]]>https://board.circlewithadot.net/post/https://infosec.exchange/users/dascandy/statuses/116435937539045119https://board.circlewithadot.net/post/https://infosec.exchange/users/dascandy/statuses/116435937539045119Mon, 20 Apr 2026 07:50:10 GMT<![CDATA[Reply to I've run into an interesting problem that I believe doesn't have a well-known solution. on Mon, 20 Apr 2026 07:11:10 GMT]]>@krans Right, that's not a thing regexes can do. I was expecting something to be wrong still but couldn't quite find what. I'm also somewhat sure the header lines would be wrong in counts, which is also not fixable in regex.

Best of luck!

]]>https://board.circlewithadot.net/post/https://infosec.exchange/users/dascandy/statuses/116435784206842504https://board.circlewithadot.net/post/https://infosec.exchange/users/dascandy/statuses/116435784206842504Mon, 20 Apr 2026 07:11:10 GMT<![CDATA[Reply to I've run into an interesting problem that I believe doesn't have a well-known solution. on Mon, 20 Apr 2026 06:29:28 GMT]]>@dascandy Context lines need to be reordered. Consider:

context1
-old
-irrelevant1
+new
+irrelevant2
context2

The positions of the "irrelevant1" and "new" lines in the diff have to be swapped.

]]>https://board.circlewithadot.net/post/https://mastodon.me.uk/users/krans/statuses/116435620235305457https://board.circlewithadot.net/post/https://mastodon.me.uk/users/krans/statuses/116435620235305457Mon, 20 Apr 2026 06:29:28 GMT<![CDATA[Reply to I've run into an interesting problem that I believe doesn't have a well-known solution. on Mon, 20 Apr 2026 06:27:22 GMT]]>@krans 2nd regex does do that? That's why I prefix the !KeEp to those you want to keep, so I can then remove the +/- from the rest. Should replace with space though, lemme fix that

]]>https://board.circlewithadot.net/post/https://infosec.exchange/users/dascandy/statuses/116435611996731284https://board.circlewithadot.net/post/https://infosec.exchange/users/dascandy/statuses/116435611996731284Mon, 20 Apr 2026 06:27:22 GMT<![CDATA[Reply to I've run into an interesting problem that I believe doesn't have a well-known solution. on Mon, 20 Apr 2026 06:26:54 GMT]]>@krans I did kinda expect that and yeah, that context is unregexable

]]>https://board.circlewithadot.net/post/https://infosec.exchange/users/dascandy/statuses/116435610157171985https://board.circlewithadot.net/post/https://infosec.exchange/users/dascandy/statuses/116435610157171985Mon, 20 Apr 2026 06:26:54 GMT<![CDATA[Reply to I've run into an interesting problem that I believe doesn't have a well-known solution. on Mon, 20 Apr 2026 06:24:32 GMT]]>@dascandy (I tried this already)

]]>https://board.circlewithadot.net/post/https://mastodon.me.uk/users/krans/statuses/116435600810296098https://board.circlewithadot.net/post/https://mastodon.me.uk/users/krans/statuses/116435600810296098Mon, 20 Apr 2026 06:24:32 GMT<![CDATA[Reply to I've run into an interesting problem that I believe doesn't have a well-known solution. on Mon, 20 Apr 2026 06:24:01 GMT]]>@dascandy This creates patches that don't apply, because non-matching deletion lines must be converted into context lines.

]]>https://board.circlewithadot.net/post/https://mastodon.me.uk/users/krans/statuses/116435598806045387https://board.circlewithadot.net/post/https://mastodon.me.uk/users/krans/statuses/116435598806045387Mon, 20 Apr 2026 06:24:01 GMT<![CDATA[Reply to I've run into an interesting problem that I believe doesn't have a well-known solution. on Mon, 20 Apr 2026 06:22:49 GMT]]>@krans

s/^[+-].*regex/!KeEp&/
s/^[+-]/ /
s/!KeEp//

?

]]>https://board.circlewithadot.net/post/https://infosec.exchange/users/dascandy/statuses/116435594093690945https://board.circlewithadot.net/post/https://infosec.exchange/users/dascandy/statuses/116435594093690945Mon, 20 Apr 2026 06:22:49 GMT<![CDATA[Reply to I've run into an interesting problem that I believe doesn't have a well-known solution. on Mon, 20 Apr 2026 05:59:11 GMT]]>@mag I'm changing the wording in a log message and there are about 1500 tests affected.

Ooh, that is potentially *exactly* what I need — if the files I needed to update were in a git repository.

]]>https://board.circlewithadot.net/post/https://mastodon.me.uk/users/krans/statuses/116435501125944985https://board.circlewithadot.net/post/https://mastodon.me.uk/users/krans/statuses/116435501125944985Mon, 20 Apr 2026 05:59:11 GMT<![CDATA[Reply to I've run into an interesting problem that I believe doesn't have a well-known solution. on Sun, 19 Apr 2026 23:03:38 GMT]]>@krans what's your use case for this? The closest I've come across in the past is using `EDITOR=<some sed command> git add -e <file>` to prevent +/- lines that don't match a regex from being committed.

]]>https://board.circlewithadot.net/post/https://mastodon.me.uk/users/mag/statuses/116433867155107161https://board.circlewithadot.net/post/https://mastodon.me.uk/users/mag/statuses/116433867155107161Sun, 19 Apr 2026 23:03:38 GMT