Michael Smith has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/21955 )

Change subject: IMPALA-13465: Trace TupleId further to reduce Agg cardinality
......................................................................


Patch Set 13:

(3 comments)

http://gerrit.cloudera.org:8080/#/c/21955/12/fe/src/main/java/org/apache/impala/planner/AggregationNode.java
File fe/src/main/java/org/apache/impala/planner/AggregationNode.java:

http://gerrit.cloudera.org:8080/#/c/21955/12/fe/src/main/java/org/apache/impala/planner/AggregationNode.java@286
PS12, Line 286:             "{}:AGGREGATE aggPhase={} aggInputCardinality={} 
aggIdx={} numGroups={} "
> I'd like to make it uniform with the rest of 3 TRACE log below. unlike getD
This makes the code more fragile, and that concerns me. It's uglier, but I'd 
rather keep the isTraceEnabled checks. Maybe make a little helper just for 
trace logging in this class?


http://gerrit.cloudera.org:8080/#/c/21955/13/fe/src/main/java/org/apache/impala/planner/AggregationNode.java
File fe/src/main/java/org/apache/impala/planner/AggregationNode.java:

http://gerrit.cloudera.org:8080/#/c/21955/13/fe/src/main/java/org/apache/impala/planner/AggregationNode.java@480
PS13, Line 480:     return lowerNumGroupsByInputCardinality(numGroups, 
aggInputCardinality);
This replaced taking Math.min after every multiply? I guess it caps the result 
at aggInputCardinality either way.


http://gerrit.cloudera.org:8080/#/c/21955/12/testdata/workloads/functional-planner/queries/PlannerTest/tpcds_cpu_cost/tpcds-q03.test
File 
testdata/workloads/functional-planner/queries/PlannerTest/tpcds_cpu_cost/tpcds-q03.test:

http://gerrit.cloudera.org:8080/#/c/21955/12/testdata/workloads/functional-planner/queries/PlannerTest/tpcds_cpu_cost/tpcds-q03.test@43
PS12, Line 43: |  tuple-ids=3 row-size=52B cardinality=9.00M cost=53096454
Is that covered by the statement

> This patch also makes tuple-based reduction more conservative by capping at 
> input cardinality/limit, or using output cardinality if the producer node is 
> a UnionNode or has hard estimates.

Can you update the commit message to specify what part of IMPALA-13405 is being 
reverted? Or if we're reverting the whole thing, put up a revert patch?



--
To view, visit http://gerrit.cloudera.org:8080/21955
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I11f59ccc469c24c1800abaad3774c56190306944
Gerrit-Change-Number: 21955
Gerrit-PatchSet: 13
Gerrit-Owner: Riza Suminto <[email protected]>
Gerrit-Reviewer: Aman Sinha <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Michael Smith <[email protected]>
Gerrit-Reviewer: Riza Suminto <[email protected]>
Gerrit-Reviewer: Yida Wu <[email protected]>
Gerrit-Comment-Date: Tue, 29 Oct 2024 20:48:27 +0000
Gerrit-HasComments: Yes

Reply via email to