sunchao commented on PR #11276: URL: https://github.com/apache/arrow-rs/pull/11276#issuecomment-6580460381
Rebased on top of #11157 and dropped the duplicate `Skip`/`Skip` change in db07a741. The remaining change caches the byte range while the decoder is waiting. All three mask setters now invalidate that range, and the regression test changes the selection mid-request and checks the requested bytes and decoded result against a fresh decoder. I reran the benchmarks against current main and updated the PR description. With 1,000 column chunks, a poll while waiting for bytes went from 4.90 µs to 40.7 ns. The full in-memory request/push/decode case went from 434.6 µs to 416.7 µs, with identical requested ranges and decoded indexes. The two requested error-path tests are also in this revision. Ready for another look. -- 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]
