зеркало из https://github.com/golang/tools.git
0ba9c8439e
Following the edge case discovered in golang/go#60201, take a more scientific approach to improving symbol match scoring: - Add a conformance test that compares Matcher with SymbolMatcher, querying all identifiers in x/tools. The two are not expected to agree in all cases, but this test helped find interesting ranking edge cases, which are added to the ranking test. - Don't count a capital letter in the middle of a sequence of capital letters (e.g. the M in YAML) as a word start. This was the inconsistency that led to golang/go#60201. - Compute the sequence bonus before role score; role score should take precedent. - Simplify the sequence scoring logic: a sequential character gets the same score as a word start, unless it is the final character in the pattern in which case we also adjust for whether it completes a word or segment. This feels like a reasonable heuristic. - Fix a bug in final-rune adjustment where we were checking the next input rune for a segment start, not a separator. Notably, the scoring improvements above were all derived from first principles, and happened to also improve the conformance rate in the new test. Additionally, make the following cleanup: - s/character/rune throughout, since that's what we mean - add debugging support for more easily understanding the match algorithm - add additional commentary - add benchmarks Fixes golang/go#60201 Change-Id: I838898c49cbb69af083a8cc837612da047778c40 Reviewed-on: https://go-review.googlesource.com/c/tools/+/531697 Reviewed-by: Alan Donovan <adonovan@google.com> Auto-Submit: Robert Findley <rfindley@google.com> LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> |
||
|---|---|---|
| .. | ||
| input.go | ||
| input_test.go | ||
| matcher.go | ||
| matcher_test.go | ||
| self_test.go | ||
| symbol.go | ||
| symbol_test.go | ||