Aman Sinha has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/17615 )

Change subject: IMPALA-10755: Fix migration of analytic predicate to inline 
view.
......................................................................


Patch Set 2:

(7 comments)

> Patch Set 1:
>
> (1 comment)

Thanks Quanlong for the quick review. Pls see responses.

http://gerrit.cloudera.org:8080/#/c/17615/1//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/17615/1//COMMIT_MSG@20
PS1, Line 20: Testing:
> TODO: run all FE and relevant end-to-end tests.
Done


http://gerrit.cloudera.org:8080/#/c/17615/1/fe/src/main/java/org/apache/impala/planner/SingleNodePlanner.java
File fe/src/main/java/org/apache/impala/planner/SingleNodePlanner.java:

http://gerrit.cloudera.org:8080/#/c/17615/1/fe/src/main/java/org/apache/impala/planner/SingleNodePlanner.java@1397
PS1, Line 1397:       Expr viewPred = pred.substitute(inlineViewRef.getSmap(), 
analyzer, false);
              :       tupleIds.clear();
> nit: this can be replaced as
Done


http://gerrit.cloudera.org:8080/#/c/17615/1/fe/src/main/java/org/apache/impala/planner/SingleNodePlanner.java@1402
PS1, Line 1402: ere MAX
> nit: this can be null since we don't need the slotIds.
Done


http://gerrit.cloudera.org:8080/#/c/17615/1/fe/src/main/java/org/apache/impala/planner/SingleNodePlanner.java@1408
PS1, Line 1408:     }
> nit: Can we simply use isBound()? i.e.
I did not change this because after looking at the implementation of 
SlotRef.isBoundByTupleIds() and SlotRef.referencesTuple().  I see there's an 
extra Preconditions check in referencesTuple:
       Preconditions.checkState(type_.isValid());

I think it is useful to do that check, so I left this as-is. Let me know if you 
think otherwise.


http://gerrit.cloudera.org:8080/#/c/17615/1/testdata/workloads/tpch/queries/analytic-fns.test
File testdata/workloads/tpch/queries/analytic-fns.test:

http://gerrit.cloudera.org:8080/#/c/17615/1/testdata/workloads/tpch/queries/analytic-fns.test@23
PS1, Line 23: ---- RESULTS
> nit: we don't need the ORDER  BY since the RESULTS section is not specified
Makes sense. Done.


http://gerrit.cloudera.org:8080/#/c/17615/1/testdata/workloads/tpch/queries/analytic-fns.test@52
PS1, Line 52: ULTS
> nit: could you cast this to BIGINT? Just concerning if this will be flaky s
Good point. Added the cast.


http://gerrit.cloudera.org:8080/#/c/17615/1/testdata/workloads/tpch/queries/analytic-fns.test@53
PS1, Line 53: 1.00,1,11,7,12
> nit: we don't need the ORDER  BY since the RESULTS section is not specified
Done



--
To view, visit http://gerrit.cloudera.org:8080/17615
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ib5cad3d408ee3695cafb35f66a4f19b4e8d0529e
Gerrit-Change-Number: 17615
Gerrit-PatchSet: 2
Gerrit-Owner: Aman Sinha <[email protected]>
Gerrit-Reviewer: Aman Sinha <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Quanlong Huang <[email protected]>
Gerrit-Comment-Date: Wed, 23 Jun 2021 23:18:19 +0000
Gerrit-HasComments: Yes

Reply via email to