Hello,
Now Spark UI does not show HashAggregateExec modes, could we add the
aggregate modes in SparkPlan? I think it's helpful when we analyze a very
complicated SparkPlan.
Am I right?
For example:
SELECT key2, sum(value2) as sum_value2
FROM ( SELECT id % 1 as key2, id as value2 FROM range(1, 1000) ) as skewData2
GROUP BY key2
== Physical Plan ==
AdaptiveSparkPlan isFinalPlan=true
+- *(2) HashAggregate(keys=[key2#286L], functions=[sum(value2#287L)],
modes=(Final), output=[key2#286L, sum_value2#288L])
+- CustomShuffleReader coalesced
+- ShuffleQueryStage 0
+- Exchange hashpartitioning(key2#286L, 5), ENSURE_REQUIREMENTS,
[id=#112]
+- *(1) HashAggregate(keys=[key2#286L],
functions=[partial_sum(value2#287L)], modes=(Partial), output=[key2#286L,
sum#294L])
+- *(1) Project [(id#289L % 1) AS key2#286L, id#289L AS
value2#287L]
+- *(1) Range (1, 1000, step=1, splits=2)
