fallintoplace opened a new pull request, #2377: URL: https://github.com/apache/iceberg-go/pull/2377
## What - Reduce work when planning split compaction tasks. ## Why - Splits of the same file often share delete metadata. ## Implementation - Index whole-file tasks by position instead of copying them into a map. - Compare optional values directly and skip deep comparisons for shared delete slices. - Keep deep comparisons for separately allocated slices. ## Benchmark Full `PlanCompaction` planner with prebuilt tasks and no rewrite work. Apple M1 Pro, Go 1.26.3, median of 3 runs, `GOMAXPROCS=1`: | Workload | Before | After | | --- | ---: | ---: | | 20,000 files, 4 splits, no deletes | 11.86 ms | 6.84 ms | | 5,000 files, 8 splits, 8 shared deletes | 21.11 ms | 3.52 ms | | 3,000 files, 8 splits, 64 shared deletes | 70.45 ms | 2.36 ms | | 5,000 files, 8 splits, 16 copied deletes | 35.83 ms | 34.34 ms | `GOMAXPROCS=1 go test ./table/compaction -run '^$' -bench '^BenchmarkPlanCompactionSplitNormalization$' -benchmem -benchtime=150ms -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]
