emecii commented on PR #11052: URL: https://github.com/apache/arrow-rs/pull/11052#issuecomment-5807160814
The focused benchmark follow-up is implemented and pushed at [emecii:bench/variant-list-validity-11052](https://github.com/emecii/arrow-rs/commit/204ad24c6a6ca6fdbf6958aad90c80604ec8c5a4) (benchmark-only). Identical fixtures ran at `f9e02ba7` / `d881317d` in separate builds: 104 Criterion cases, plus five allocation samples per case in a separate executable. All head cases pass strict value/validity checks; 32 baseline cases are explicitly marked as the known missing-index bug. One concrete tradeoff: the 64-row all-OOB Int64 case has equivalent output but adds 5 allocations / 224 requested bytes. A reverse-order repeat measured 1.000 µs [95% CI 0.996–1.004] → 1.238 µs [1.230–1.246], about +24% (Apple M4, Rust 1.98.0, optimized; 50 samples, 1 s warm-up, 3 s measurement). At 8192 rows, mixed/nested List paths add 2/4 allocations and 1080/2160 bytes; their Variant outputs have corrected semantics. This is not a claim of overall performance neutrality. The benchmark branch passes 367 unit tests, Clippy, formatting, and all 104 benchmark preflights. Commands and measurement boundaries are in the added benchmark README. -- 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]
