| Summary: | Find action algorithm leads to unexpected results | ||
|---|---|---|---|
| Product: | [Applications] kate | Reporter: | tomashnyk |
| Component: | application | Assignee: | KWrite Developers <kwrite-bugs-null> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | waqar.17a |
| Priority: | NOR | ||
| Version First Reported In: | 25.08.1 | ||
| Target Milestone: | --- | ||
| Platform: | Other | ||
| OS: | Linux | ||
| Latest Commit: | https://invent.kde.org/frameworks/kcoreaddons/-/commit/d445210768f14140443ed0a8256868f93067803c | Version Fixed/Implemented In: | |
| Sentry Crash Report: | |||
| Attachments: |
bad order (suggested order drawn)
good order (also shows hte highlight bug) |
||
Created attachment 186514 [details]
good order (also shows hte highlight bug)
We need to upstream some changes from Kate to the KFuzzyMatcher library. A possibly relevant merge request was started @ https://invent.kde.org/frameworks/kcoreaddons/-/merge_requests/541 Git commit d445210768f14140443ed0a8256868f93067803c by Christoph Cullmann, on behalf of Waqar Ahmed. Committed on 17/11/2025 at 16:50. Pushed by cullmann into branch 'master'. Update fuzzy matching logic changes from Kate The algorithm has been updated in Kate, this patch upstreams all the improvements. Most notably, it now tries to prefer sequential matches and matches that happen at the start of a word M +3 -3 autotests/kfuzzymatchertest.cpp M +84 -48 src/lib/text/kfuzzymatcher.cpp https://invent.kde.org/frameworks/kcoreaddons/-/commit/d445210768f14140443ed0a8256868f93067803c |
Created attachment 186513 [details] bad order (suggested order drawn) I came accross this when I wanted to change sessions. I press F2 and start typing "session". What happens: I only get good results when I get to "sessio" What I would expect: I get good results already when I type "se" or "ses" The thing is that the algorithm seem to be looking for finds *s*e*s*i*o*m, so for "se", it finds for example "kate upward Split viEw" (capitals emphasize what matches). It also finds "SourcEs: elm" [I think that is what is being matched - when part of the match is in the prefix, that is in "sources" in this case, the matched letters are not highlighted, which also sounds like a bug, should report it separately? - it is visible on the second screenshot, where I made a red note about it]. That is all great and good, but I think some ordering of the results is in order. What I would suggest that the results be sorted by whether they contain "se", "ses", "sess" and so on as it is typed. I assume most people type a word they look for. If they were looking for "sources: elm", they could get that by typing "s e". Probably looking for "^ se" and then " se" first would also make sense, so we first match commands that either start with the string or that have a word that starts with the string. And while at it, per each such match, my orderly me might like it if they were sorted alphabetically (now the order seems random, I guess they are written as they are ordered in the menu?)