Alex Behm has posted comments on this change.

Change subject: IMPALA-3167: Fix assignment of WHERE-clause predicate through 
grouping agg + outer join.

Patch Set 2:

Commit Message:

Line 7: IMPALA-3167: Fix assignment of WHERE-clause predicate through grouping 
agg + outer join.
> long lines
I though we allowed up to 90 chars. If that's the case then our gerrit is 

Fixed anyway.

Line 28: The fix is to conservatively not mark bound predicates as assigned if 
there exists
> if there are

Line 30: duplicate assignments of predicates. Those are simply deduped now.
> i don't understand that last part.
Expanded commit msg with an example that shows how duplicate assignment can 
File fe/src/main/java/org/apache/impala/analysis/

Line 1548:               (evalAfterJoin(srcConjunct)
> move to preceding line and fix indentation, like l1551

Line 1550:                 != globalState_.outerJoinedTupleIds.get(srcTid)))
> fix indentation

Line 1553:                    != 
> same here
File fe/src/main/java/org/apache/impala/planner/

Line 1194:     Expr.removeDuplicates(conjuncts);
> how do you end up with duplicate predicates in a *scan* node due to the cha
Explained duplicates in commit msg with an example.

Your point about HDFS scans only is a good one. Looks like we don't pick up 
bound predicates in other scan nodes! That's a bug. I think we should address 
that very soon, at least for Kudu. But I'd prefer to separate those changes 
from this bugfix.

To view, visit
To unsubscribe, visit

Gerrit-MessageType: comment
Gerrit-Change-Id: I774d13a13ad1e8fe82512df98dc29983bdd232eb
Gerrit-PatchSet: 2
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Alex Behm <>
Gerrit-Reviewer: Alex Behm <>
Gerrit-Reviewer: Anonymous Coward #27
Gerrit-Reviewer: Marcel Kornacker <>
Gerrit-HasComments: Yes

Reply via email to