andygrove opened a new pull request, #3244: URL: https://github.com/apache/datafusion-comet/pull/3244
## Summary - Serialize native query plans once and broadcast to all executors - Avoid repeated protobuf serialization for each partition ## Changes - Add `serializePlan()` method to serialize an `Operator` once - Add `getCometIterator()` overload accepting pre-serialized bytes - Update `getNativeLimitRDD` to broadcast the serialized plan - Update `CometTakeOrderedAndProjectExec` to broadcast the topK plan ## Impact For a query with 1000 partitions across 10 executors: - Plan serialization: 1000x → 1x - Plan transfer: 1000x → 10x (once per executor via broadcast) ## Test plan - [ ] Existing tests pass - [ ] Benchmark with high partition count queries 🤖 Generated with [Claude Code](https://claude.ai/claude-code) -- 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]
