Alex Behm has posted comments on this change. Change subject: IMPALA-2502: don't redundantly repartition grouping aggregations ......................................................................
Patch Set 4: (1 comment) http://gerrit.cloudera.org:8080/#/c/2414/4/testdata/workloads/functional-planner/queries/PlannerTest/aggregation.test File testdata/workloads/functional-planner/queries/PlannerTest/aggregation.test: Line 947: select straight_join count(distinct c_custkey) > I'm not sure that I understand which slots and what would qualify as intere By "interesting" expression I mean something that is not a SlotRef. Example: select coalesce(t1.id, 123), count(*) from t1 join t2 on coalesce(t1.id, 123) = ifnull(t2.id, 20) group by 1 or select ifnull(t2.id, 20), count(*) from t1 join t2 on coalesce(t1.id, 123) = ifnull(t2.id, 20) group by 1 Ideally, we would be able to make both cases work, but I have a feeling only one of those variants will. I'm not saying we need to fix this in this patch, but it might be good to try it out, maybe even file a follow-on JIRA. I think the join hoisting has a similar limitation. You can imagine more tricky expressions with multiple slotrefs for which we cannot rely on our equivalence computation. -- To view, visit http://gerrit.cloudera.org:8080/2414 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: Iffdcfd3629b8a69bd23915e1adba3b8323cbbaef Gerrit-PatchSet: 4 Gerrit-Project: Impala Gerrit-Branch: cdh5-trunk Gerrit-Owner: Tim Armstrong <[email protected]> Gerrit-Reviewer: Alex Behm <[email protected]> Gerrit-Reviewer: Marcel Kornacker <[email protected]> Gerrit-Reviewer: Tim Armstrong <[email protected]> Gerrit-HasComments: Yes
