Steve Carlin has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/24528 )

Change subject: IMPALA-15135: Calcite planner: added support for Kudu tables
......................................................................


Patch Set 12:

(2 comments)

http://gerrit.cloudera.org:8080/#/c/24528/12/testdata/workloads/functional-query/queries/QueryTest/kudu_runtime_filter_with_timestamp_conversion.test
File 
testdata/workloads/functional-query/queries/QueryTest/kudu_runtime_filter_with_timestamp_conversion.test:

http://gerrit.cloudera.org:8080/#/c/24528/12/testdata/workloads/functional-query/queries/QueryTest/kudu_runtime_filter_with_timestamp_conversion.test@30
PS12, Line 30: .*.ts
> We need to think about the Calcite changes needed for introducing the alias
Heh, I was thinking about this right before I read your comment.

I agree this is definitely needed. It's possible that it already exists in 
Calcite and I've missed it, but the plugins to our code allows us to add tables 
to a general schema, and I haven't yet seen a way to propagate that information 
when the RelNodes are created.  So yeah, as you mentioned, we might need 
Calcite community involvement here.

As for your second comment, I answered that in the other comment you left with 
this version, so it's probably better to respond to that issue in the other 
comment.


http://gerrit.cloudera.org:8080/#/c/24528/12/testdata/workloads/functional-query/queries/QueryTest/partitioned-top-n-high-ndv.test
File 
testdata/workloads/functional-query/queries/QueryTest/partitioned-top-n-high-ndv.test:

http://gerrit.cloudera.org:8080/#/c/24528/12/testdata/workloads/functional-query/queries/QueryTest/partitioned-top-n-high-ndv.test@39
PS12, Line 39: row_regex:'|  partition by: .*part1, .*part2, .*part3, .*col4, 
.*col5'
> Could we use CALCITE_PLANNER_RESULTS section for the Calcite plan ? The ali
I have had this problem in other explain plans.  While the number of explain 
plans isn't trivial and slightly pervasive, it also was manageable to change 
them all.

Relative note:  The junit tests which test the explain plans do not check every 
line, so I didn't have to change anything there.

For the most part, I've been doing it this way.  I actually kinda felt the 
reverse of your last statement.  Introducing a CALCITE_PLANNER_RESULTS section 
duplicates a lot of information.  I felt that, for the most part, the explain 
plans are more concerned with the general Plan Node layout, runtime filters are 
generated, etc...

Also, by having a second section, it would be easy for someone to change the 
original planner text here without changing the Calcite planner text.  This 
wouldn't be found until later in testing, or even worse, with our current 
jenkins monitoring, after the commit is done.

One last note: while it is prone to potentially passing a check on a field on 
the wrong table, I think an error like that would most likely be picked up 
causing incorrect results and failing other tests.



--
To view, visit http://gerrit.cloudera.org:8080/24528
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ie22a66f0edad195cade74a8447efc83a8ce526f7
Gerrit-Change-Number: 24528
Gerrit-PatchSet: 12
Gerrit-Owner: Steve Carlin <[email protected]>
Gerrit-Reviewer: Aman Sinha <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Jason Fehr <[email protected]>
Gerrit-Reviewer: Joe McDonnell <[email protected]>
Gerrit-Reviewer: Michael Smith <[email protected]>
Gerrit-Reviewer: Quanlong Huang <[email protected]>
Gerrit-Reviewer: Steve Carlin <[email protected]>
Gerrit-Reviewer: Xuebin Su <[email protected]>
Gerrit-Comment-Date: Mon, 13 Jul 2026 09:46:49 +0000
Gerrit-HasComments: Yes

Reply via email to