[
https://issues.apache.org/jira/browse/DERBY-2370?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12491606
]
Dyre Tjeldvoll commented on DERBY-2370:
---------------------------------------
I have looked at the writeup and and also looked at the
patch. While I am certainly no expert in this area I think the
explanation and solution in the writeup looks very sound. The
patch seem to implement what is suggested, it is clear, well
documented and tested. +1 from me.
> EXISTS may return the wrong value for sub-queries involving set operations
> --------------------------------------------------------------------------
>
> Key: DERBY-2370
> URL: https://issues.apache.org/jira/browse/DERBY-2370
> Project: Derby
> Issue Type: Bug
> Components: SQL
> Affects Versions: 10.2.2.0
> Reporter: Dyre Tjeldvoll
> Assigned To: A B
> Attachments: d2370_engine_v1.patch, d2370_tests_v1.patch,
> d2370_v1.stat, d2370_writeup_v1.html, repro.sql
>
>
> It seems like EXISTS on a SELECT returning zero rows returns false (as
> expected), but EXISTS on INTERSECT of two disjunct sets returns true,
> e.g EXISTS (values 1 intersect values 2).
> Yip Ng wrote on derby-dev:
> I believe its probably got to do with the EXISTS subquery transforming
> the original RCL to
> a TRUE boolean value for the INTERSECT. So during row comparison at
> execution time
> for INTERSECT processing since true == true(thus intersects), so it
> will always return 'BAD'. Likewise,
> select * from ( values 'OK' ) as T where exists (values 1 except values 2);
> This supposedly should return 'OK' but because of the boolean
> transformation mentioned
> above for EXISTS subquery, it will return no rows for EXCEPT
> processing.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
