Aleksandr Efimov has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/24426 )

Change subject: IMPALA-14601: Support HBO for JoinNode cardinality
......................................................................


Patch Set 23:

(1 comment)

I ran this change together with 24693 on a cluster, since they meet in the same 
goldens.

24693 appends ", partition constants ignored" to an HBO annotation whose match 
came through IGNORE_PARTITION_CONSTANTS. That caveat does not stay on the scan: 
a join or an aggregate over a partition-predicated scan canonicalizes the same 
predicates into its own key, so it matches through that strategy itself and 
reads the same way. In the first query of hbo-multiple-scans.test that is three 
of the lines this change newly annotates - 06:AGGREGATE, 03:AGGREGATE and 
02:HASH JOIN.

Cherry-picking either onto the other is clean, so this is a golden regeneration 
for whichever of us lands second, not a merge conflict. Flagging it so it 
doesn't arrive as a surprise failure.

http://gerrit.cloudera.org:8080/#/c/24426/23/tests/query_test/test_hbo.py
File tests/query_test/test_hbo.py:

http://gerrit.cloudera.org:8080/#/c/24426/23/tests/query_test/test_hbo.py@62
PS23, Line 62:       # To avoid the test being fragile, we only compare the 
cardinality lines.
> The data size calculation not just depends on cardinalities, but also on av
The run above is the case I had in mind here. It only failed because the caveat 
happens to land on the cardinality line, which the filter keeps. Had the same 
interaction moved one line up - a join changing side, a distribution mode 
flipping - the goldens would have displayed the new text and the test would 
still have passed.

I'm not asking HBO to drive the distribution mode; avgRowSize is a fair reason 
to leave that for later. The narrower thing is that the mode lines this patch 
adds to the goldens are shown but not compared. Your call whether that is worth 
widening the filter now.



--
To view, visit http://gerrit.cloudera.org:8080/24426
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I70b655ae7027d0d9eb8e9fae9ba2e1b7ad9876b4
Gerrit-Change-Number: 24426
Gerrit-PatchSet: 23
Gerrit-Owner: Quanlong Huang <[email protected]>
Gerrit-Reviewer: Aleksandr Efimov <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Quanlong Huang <[email protected]>
Gerrit-Reviewer: Steve Carlin <[email protected]>
Gerrit-Comment-Date: Thu, 27 Aug 2026 10:33:33 +0000
Gerrit-HasComments: Yes

Reply via email to