[
https://issues.apache.org/jira/browse/IMPALA-13480?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17904272#comment-17904272
]
Riza Suminto commented on IMPALA-13480:
---------------------------------------
While working on the enabling VALIDATE_CARDINALITY test option
([https://gerrit.cloudera.org/c/22184/),] I found the following 3 bugs:
First, in IMPALA-13405, cardinality estimate of MERGE phase aggregation
is not capped against the output cardinality of the EXCHANGE node.
Second, tuple-based optimization IMPALA-13405 can cause cardinality
underestimation if HAVING predicate exist. This is due to the default
selectivity of 10% applied for each HAVING predicate. We should consider
skipping tuple-based optimization if predicate exist, or raising the default
selectivity for
HAVING predicate and undo this skipping logic (IMPALA-13542).
Third, is missing stats recompute after conjunct transfer in multi-phase
aggregation. I plan to fix it separately in IMPALA-13526.
> PlannerTest.testAggregation should VALIDATE_CARDINALITY
> -------------------------------------------------------
>
> Key: IMPALA-13480
> URL: https://issues.apache.org/jira/browse/IMPALA-13480
> Project: IMPALA
> Issue Type: Improvement
> Components: Test
> Affects Versions: Impala 4.4.0
> Reporter: Riza Suminto
> Assignee: Riza Suminto
> Priority: Minor
>
> PlannerTest.testAggregation does not VALIDATE_CARDINALITY today. Validating
> cardinality will allow us to track our estimation quality and capture
> behavior change like
> https://github.com/apache/impala/blob/c83e5d97693fd3035b33622512d1584a5e56ce8b/fe/src/main/java/org/apache/impala/planner/AggregationNode.java#L74-L76
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
