gabotechs commented on PR #23720:
URL: https://github.com/apache/datafusion/pull/23720#issuecomment-5035577028
I'm looking at ClickBench Q33, that should show some improvement, and even
though hashes are getting reused:
```sql
$ SELECT "URL", COUNT(*) AS c FROM hits GROUP BY "URL" ORDER BY c DESC LIMIT
10;
SortPreservingMergeExec: [c@1 DESC], fetch=10, metrics=[]
ProjectionExec: expr=[URL@0 as URL, count(Int64(1))@1 as c], metrics=[]
SortExec: TopK(fetch=10), expr=[count(Int64(1))@1 DESC],
preserve_partitioning=[true], filter=[count(Int64(1))@1 IS NULL OR
count(Int64(1))@1 > 55255], metrics=[]
AggregateExec: mode=FinalPartitioned, gby=[URL@0 as URL],
aggr=[count(Int64(1))], metrics=[hash_rows_computed=0, hash_rows_reused=22.98 M]
RepartitionExec: partitioning=Hash([URL@0], 16),
input_partitions=16, metrics=[hash_rows_computed=0, hash_rows_reused=22.98 M]
AggregateExec: mode=Partial, gby=[URL@0 as URL],
aggr=[count(Int64(1))], metrics=[hash_rows_computed=100.00 M,
hash_rows_reused=0]
DataSourceExec: ...
```
It seems to have 0 impact on performance:
```
│ QQuery 33 │ 1458.92 / 1482.58 ±12.28 / 1492.64 ms │ 1440.78 / 1462.94
±13.67 / 1478.61 ms │ no 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]
