Jackie-Jiang opened a new pull request, #20228: URL: https://github.com/apache/pinot/pull/20228
Fix the tuple sketch failure exposed by #20227 and related sketch lifecycle and serialization bugs. - Return single pending sketches before initializing tuple, CPC, or Theta unions. This avoids unset tuple parameters and preserves configured precision when deserialized results are serialized and then reused for merging. Simplify `getResult()` by removing the redundant empty-state branches. - Correct the integer tuple sketch size bound to include the summary's mode byte and round nominal entries as DataSketches does. This prevents under-allocation in raw index writers. - Deserialize KLL BYTES only for included, non-null rows, once per row for MV group-by. Null placeholders and unused backing-array entries are no longer passed to DataSketches. Adds regressions for all three areas. The dependency stays at 9.0.0 so these fixes can merge independently of #20227. Validation: - 225 scoped tests pass with DataSketches 9.0.0 and again with the 9.1.0 override, including accumulator, tuple aggregator, KLL unit/query, intermediate serialization, and legacy serialization compatibility tests. - Spotless, Checkstyle, license formatting, and license checks pass for `pinot-segment-local` and `pinot-core`. - Compilation with deprecation and warning reporting enabled; no warnings in changed files. -- 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]
