rubenada commented on code in PR #6523:
URL: https://github.com/apache/hive/pull/6523#discussion_r3785750385
##########
ql/src/test/results/clientpositive/perf/tpcds30tb/json/query30.q.out:
##########
@@ -1947,7 +1947,7 @@
"name": null
}
],
- "rowCount": 36000000
Review Comment:
This should be consequence of
[CALCITE-7083](https://issues.apache.org/jira/browse/CALCITE-7083)
"RelMdDistinctRowCount aggregates implementation problems", specifically this
check that bails out and returns the agg rowcount as a most conservative
estimate:
```
public @Nullable Double getDistinctRowCount(Aggregate rel,
RelMetadataQuery mq,
ImmutableBitSet groupKey, @Nullable RexNode predicate) {
...
// the result of the aggregation function is difficult to infer, so if
the groupKey contains
// the aggregated column, return the full row count of Aggregate. This
is the most conservative
// estimate, and the actual ndv will not be greater than it.
if (groupKey.anyMatch(key -> key >= rel.getGroupCount())) {
return mq.getRowCount(rel);
}
...
```
Since we have a sequence of Aggregate - Project - Aggregate, the top-most
aggregate should compute its rowcount based on the distinctRowCount of its
input (Project), which should, in turn, call the distinctRowCount of its input
(bottom Aggregate), so the alteration of the `getDistinctRowCount` for
Aggregate propagates from there.
The increase seems huge, but if you compare it with the rowcount of the
surrounding operators (e.g. the input Project-Aggregate), it seems on the same
order of magnitude.
--
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]
