fallintoplace opened a new pull request, #2148: URL: https://github.com/apache/iceberg-go/pull/2148
**What** - Index equality-delete metric ranges during scan planning. **Why** - The current matcher checks every sequence-applicable delete even when nearly all ranges are disjoint. **Implementation** - Add immutable interval indexes for large global and partition buckets. - Index one usable equality field per entry. Keep uncertain entries on the fallback path. - Keep sequence filtering and the existing metrics matcher as the final check. - Use the linear path for small sequence suffixes, unusable data bounds or unselective queries. - Add comparisons against the linear reference and a lookup benchmark. **Benchmark** Existing `BenchmarkEqualityDeleteIndexMetrics/metrics`: index construction plus 10,000 data-file lookups against 10,000 deletes in 100 disjoint range groups. Both versions attach 100 deletes per data file. Apple M1 Pro, darwin/arm64, Go 1.26.3. Median of 3 runs, one iteration each, `-cpu=1`. | | Before | After | | --- | ---: | ---: | | Time | 895.9 ms/op | 48.1 ms/op | | Bytes | 52,477,712 B/op | 48,537,888 B/op | | Allocations | 159,023/op | 119,050/op | **18.6x faster** on this selective workload, with **7.5% fewer bytes** and **25% fewer allocations**. This measures equality-delete matching, not full scan execution. ```sh go test ./table -run '^$' -bench '^BenchmarkEqualityDeleteIndexMetrics/metrics$' -benchmem -benchtime=1x -count=3 -cpu=1 ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
