[
https://issues.apache.org/jira/browse/CALCITE-4754?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17418369#comment-17418369
]
Javier Rivas Rodriguez commented on CALCITE-4754:
-------------------------------------------------
Hi [~julianhyde], I have opened a [draft
PR|https://github.com/apache/calcite/pull/2532] with several commits as
suggested, to work on these changes.
I have left _supportsIsTrueInsideCaseWhen_ as it was temporarily. I think I'm
not able to completely understand the changes you are requesting and I'd need
some help clarifying them. At the moment in a filter operation, the IS TRUE
operator is added
[here|https://github.com/apache/calcite/blob/master/core/src/main/java/org/apache/calcite/sql2rel/SqlToRelConverter.java#L5733]
tracking it to CALCITE-1042 and it gets carried over to the CASE (WHEN
transformation. There are also
[tests|https://github.com/apache/calcite/blob/master/core/src/test/java/org/apache/calcite/rel/rel2sql/RelToSqlConverterTest.java#L224-L254]
in the codebase verifying it is added. That is what led me to think that I
could add the option of removing it based on the dialect.
As I have mentioned earlier, I'm keen to get some more directions on how to
approach the problem if this is not the right angle and make those changes.
> Invalid Redshift SQL when executing a PIVOT query
> -------------------------------------------------
>
> Key: CALCITE-4754
> URL: https://issues.apache.org/jira/browse/CALCITE-4754
> Project: Calcite
> Issue Type: Bug
> Components: core
> Affects Versions: 1.27.0
> Reporter: Javier Rivas Rodriguez
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.28.0
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Given a Calcite connection to Redshift
> When running a query like the one below
> {code:java}
> SELECT *
> FROM
> (SELECT field1, field2, field3
> FROM my_table)
> PIVOT SUM(field1) FOR field2 IN (val1, val2){code}
> The SQL produced is something like
> {code:java}
> SELECT
> field3,
> SUM(field1) FILTER (WHERE field2 = val1),
> SUM(field1) FILTER (WHERE field2 = val2)
> FROM my_table
> GROUP BY field3{code}
> This SQL transformation is not accepted by Redshift
> The expected result would be something like
> {code:java}
> SELECT
> field3,
> SUM(CASE WHEN field2 = val1 THEN field1 ELSE NULL END),
> SUM(CASE WHEN field2 = val2 THEN field1 ELSE NULL END)
> FROM my_table
> GROUP BY field3
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
