wiedld commented on code in PR #14637: URL: https://github.com/apache/datafusion/pull/14637#discussion_r1957023902
########## datafusion/physical-optimizer/src/enforce_sorting/mod.rs: ########## @@ -126,29 +126,65 @@ fn update_sort_ctx_children( /// [`CoalescePartitionsExec`] descendant(s) for every child of a plan. The data /// attribute stores whether the plan is a `CoalescePartitionsExec` or is /// connected to a `CoalescePartitionsExec` via its children. +/// +/// The tracker halts at each [`SortExec`] (where the SPM will act to replace the coalesce). +/// +/// This requires a bottom-up traversal was previously performed, updating the +/// children previously. pub type PlanWithCorrespondingCoalescePartitions = PlanContext<bool>; Review Comment: I did a refactor. See [this commit](https://github.com/apache/datafusion/pull/14637/commits/0661ed7e8934e7f2a711416b85cbafde2a7b99e2). * The `parallelize_sorts` and its helper `remove_bottleneck_in_subplan` removed more nodes than it should, including needed nodes. Those nodes were then added back in a few conditionals. * Instead, if I tightened up the context `PlanWithCorrespondingCoalescePartitions` and how it was built in `update_coalesce_ctx_children` -- then (a) it avoids excess node removal, and also (b) we no longer need to add back nodes later. -- 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: github-unsubscr...@datafusion.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: github-unsubscr...@datafusion.apache.org For additional commands, e-mail: github-h...@datafusion.apache.org
