gianm opened a new pull request, #18144: URL: https://github.com/apache/druid/pull/18144
This patch pushes shuffle logic from RunWorkOrder into the stage definition. The motivation is to enable optimizations that take advantage of data already being sorted or partitioned in the desired way. Main changes: - FrameProcessorFactory is renamed StageProcessor, and its "makeProcessors" method is replaced with a more powerful "execute" method. The new "execute" is expected to fully contain partitioning and shuffling logic. - RunWorkOrder, the worker class that manages execution, no longer has any shuffle-related processing code. It delegates this to the StageProcessor. - StandardStageOperations, StandardStageRunner, & StandardStageProcessor are introduced for StageProcessor that want to perform shuffling in the "standard" way that was formerly mandatory. Other changes: - Moved FrameContext and related classes from the "kernel" package to the "exec" package. - Remove "readableChannelUsableWhenWriting" from OutputChannel, to reduce complexity. It was only used by the standard hashPartition implementation, and wasn't strictly needed there. -- 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]
