Riza Suminto has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/20040 )

Change subject: IMPALA-12200: Cap stats NDV from SetOperationStmt.createMetadata
......................................................................


Patch Set 2:

(1 comment)

Tried reran TpcdsPlannerTest after rebase, and I found some query plan change 
shape with this patch.

http://gerrit.cloudera.org:8080/#/c/20040/2/testdata/workloads/functional-planner/queries/PlannerTest/tpcds/tpcds-q11.test
File 
testdata/workloads/functional-planner/queries/PlannerTest/tpcds/tpcds-q11.test:

http://gerrit.cloudera.org:8080/#/c/20040/2/testdata/workloads/functional-planner/queries/PlannerTest/tpcds/tpcds-q11.test@103
PS2, Line 103: cardinality=87.18K
Join Cardinality increase here, I think because JoinNode switch its cardinality 
estimate due to lower build side NDV.
In JoinNode.getJoinCardinality, there is this comment:

   * Generic estimation:
   * cardinality = |child(0)| * |child(1)| / max(NDV(L.c), NDV(R.d))
   * - case A: NDV(L.c) <= NDV(R.d)
   *   every row from child(0) joins with |child(1)| / NDV(R.d) rows
   * - case B: NDV(L.c) > NDV(R.d)
   *   every row from child(1) joins with |child(0)| / NDV(L.c) rows
   * - we adjust the NDVs from both sides to account for predicates that may
   *   might have reduce the cardinality and NDVs

It still in case A, but with lower NDV(R.d).



--
To view, visit http://gerrit.cloudera.org:8080/20040
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ic0bb2eff5005fdfb11adf31499214c63dd552c05
Gerrit-Change-Number: 20040
Gerrit-PatchSet: 2
Gerrit-Owner: Riza Suminto <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Kurt Deschler <[email protected]>
Gerrit-Reviewer: Quanlong Huang <[email protected]>
Gerrit-Reviewer: Riza Suminto <[email protected]>
Gerrit-Reviewer: Wenzhe Zhou <[email protected]>
Gerrit-Comment-Date: Mon, 12 Jun 2023 22:50:03 +0000
Gerrit-HasComments: Yes

Reply via email to