fallintoplace opened a new pull request, #2155: URL: https://github.com/apache/iceberg-go/pull/2155
**What** - Speed up binary partition grouping in partitioned writers. **Why** - Binary partition values use interface-backed map keys on every row. - Repeated values can avoid repeated key conversion. **Implementation** - Add a lazy `map[string]any` for binary partition keys. - Copy each key only when its partition is first created. - Add multi-level binary coverage and a grouping benchmark. **Benchmark** `go test ./table -run '^$' -bench '^BenchmarkBinaryPartitionGrouping$' -benchmem -benchtime=300ms -count=5 -cpu=1` Base `b7232461`, PR `7c57e2df`; Go 1.26.3 on Apple M1 Pro. Each iteration groups 65,536 rows. Times are medians of five runs. | Case | Time (ms/op, base → PR) | Allocs/op (base → PR) | Speedup | | --- | ---: | ---: | ---: | | 16 B / 16 partitions | 7.73 → 4.38 | 196,854 → 65,800 | 1.76× | | 16 B / 1,024 partitions | 8.30 → 5.05 | 203,797 → 73,751 | 1.64× | | 64 B / 1,024 partitions | 8.98 → 5.08 | 203,797 → 73,751 | 1.77× | | 16 B / 32,768 partitions | 25.14 → 22.07 | 426,258 → 327,956 | 1.14× | -- 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]
