mirror of
https://github.com/lotabout/skim.git
synced 2026-09-10 07:16:23 -04:00
29ca6f22bd
* fix: preserve input order with --no-sort Workers grab 4096-item chunks from a shared queue, so the order in which worker results are concatenated is nondeterministic. With --no-sort the merged list was left in that arrival order, which scrambles the display order once the item count exceeds num_workers * chunk_size (~24k items on a 10-core machine) and makes --filter output nondeterministic for large inputs. Sort matched items by rank.index (the original input position) when no_sort is set: each worker sorts its accumulator in prepare, and the final merge exploits the k sorted runs, mirroring the sorted path. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * avoid extra sorts by exploiting the thread_pool stability --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: Loric ANDRE <loric.andre@pm.me> |
||
|---|---|---|
| .. | ||
| common | ||
| snapshots | ||
| ansi.rs | ||
| binds.rs | ||
| case.rs | ||
| cli.rs | ||
| defaults.rs | ||
| highlighting.rs | ||
| issues.rs | ||
| keys.rs | ||
| keys_interactive.rs | ||
| layout.rs | ||
| listen.rs | ||
| matcher.rs | ||
| mouse.rs | ||
| multiline.rs | ||
| normalize.rs | ||
| options.rs | ||
| popup.rs | ||
| preview.rs | ||
| split_match.rs | ||
| tiebreak.rs | ||
| unix.rs | ||