zhengruifeng commented on code in PR #52303:
URL: https://github.com/apache/spark/pull/52303#discussion_r2345877031
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/python/FlatMapGroupsInBatchExec.scala:
##########
@@ -82,17 +83,31 @@ trait FlatMapGroupsInBatchExec extends SparkPlan with
UnaryExecNode with PythonS
val data = groupedData(iter, dedupAttributes)
- val runner = new ArrowPythonRunner(
- chainedFunc,
- pythonEvalType,
- Array(argOffsets),
- groupedSchema(dedupAttributes),
- sessionLocalTimeZone,
- largeVarTypes,
- pythonRunnerConf,
- pythonMetrics,
- jobArtifactUUID,
- conf.pythonUDFProfiler)
+ val runner = if (arrowBatchSlicingEnabled) {
Review Comment:
In general, it should be beneficial, so I enable the config
`spark.sql.execution.arrow.arrowBatchSlicing.enabled` by default.
But I still want to be conservative here, since in some cases the
performance is sensitive to the values of `maxBytesPerBatch` and
`maxRecordsPerBatch`. I am not very sure the default values are suitable for
all cases. With this config, users can quickly disable the change.
--
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]
