xumingming commented on PR #56987:
URL: https://github.com/apache/spark/pull/56987#issuecomment-4884553198
The golden files changes are due to some partial-final aggs collapsed. Take
q23a as an example, before the change, the agg related part are:
```
: : : +- * Filter (22)
: : : +- * HashAggregate (21)
: : : +- * HashAggregate (20)
: : : +- * Project (19)
: : : +- * SortMergeJoin Inner
(18)
(20) HashAggregate [codegen id : 8]
Input [3]: [d_date#11, i_item_sk#12, _groupingexpression#14]
Keys [3]: [_groupingexpression#14, i_item_sk#12, d_date#11]
Functions [1]: [partial_count(1)]
Aggregate Attributes [1]: [count#15]
Results [4]: [_groupingexpression#14, i_item_sk#12, d_date#11, count#16]
(21) HashAggregate [codegen id : 8]
Input [4]: [_groupingexpression#14, i_item_sk#12, d_date#11, count#16]
Keys [3]: [_groupingexpression#14, i_item_sk#12, d_date#11]
Functions [1]: [count(1)]
Aggregate Attributes [1]: [count(1)#17]
Results [2]: [i_item_sk#12 AS item_sk#18, count(1)#17 AS cnt#19]
```
`(20) HashAggregate` and `(21) HashAggregate` are partial-final agg pairs
that does not have exchange operator in between, so they are collapsed to a
single complete mode agg:
```
: : : +- * Filter (21)
: : : +- * HashAggregate (20)
: : : +- * Project (19)
: : : +- * SortMergeJoin Inner (18)
(20) HashAggregate [codegen id : 8]
Input [3]: [d_date#11, i_item_sk#12, _groupingexpression#14]
Keys [3]: [_groupingexpression#14, i_item_sk#12, d_date#11]
Functions [1]: [count(1)]
Aggregate Attributes [1]: [count(1)#15]
Results [2]: [i_item_sk#12 AS item_sk#16, count(1)#15 AS cnt#17]
```
Since there is less operators now, the operator numbers change, then a lot
of diffs in the golden files.
--
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]