Alex Behm has posted comments on this change. Change subject: IMPALA-4076: Fix runtime filter sort compare method ......................................................................
Patch Set 2: (2 comments) http://gerrit.cloudera.org:8080/#/c/4652/2/fe/src/main/java/org/apache/impala/planner/RuntimeFilterGenerator.java File fe/src/main/java/org/apache/impala/planner/RuntimeFilterGenerator.java: Line 341: } > We are checking the cardinality of the child left child (the potential dest The equation is trying to get the selectivity from cardinalities. Only doing a division is not sufficient because it doesn't correctly capture the 0 cardinality case. So the math is wrong, and we need an extra check to compensate. I maintain that the selectivity should be 0 if one side of an inner join has a 0 cardinality. Think about it this way: Returning -1 will discard the most selective runtime filters. http://gerrit.cloudera.org:8080/#/c/4652/2/testdata/workloads/functional-planner/queries/PlannerTest/runtime-filter-propagation.test File testdata/workloads/functional-planner/queries/PlannerTest/runtime-filter-propagation.test: Line 1123: # so the number of runtime filters has to be greater than 32 and they have to be in a > I don't think it would silently fail, because the plan would change. But ju You are right, let's remove setting the query option in PlannerTest, sorry. -- To view, visit http://gerrit.cloudera.org:8080/4652 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: Iad433f2ece423ea29e79e81b68fa53cb0af18378 Gerrit-PatchSet: 2 Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-Owner: Taras Bobrovytsky <tbobrovyt...@cloudera.com> Gerrit-Reviewer: Alex Behm <alex.b...@cloudera.com> Gerrit-Reviewer: Taras Bobrovytsky <tbobrovyt...@cloudera.com> Gerrit-HasComments: Yes
