[CALCITE-7450] ValuesReduceRule incorrectly drops tuples when filter condition is irreducible

[darpan-e6]

Summary

• When a filter condition on VALUES cannot be reduced to RexLiteral by RexExecutorImpl,it returns the expression unchanged.
• ValuesReduceRule.apply() then calls isAlwaysTrue() on the unreduced RexCall, which returns false, causing the rule to incorrectly drop the tuple.
• The fix checks whether the reduced value is a RexLiteral before evaluating isAlwaysTrue(); if it is not reducible, the rule bails out and leaves the plan unchanged.

Test plan

• Added ValuesReduceRuleTest with a RAND() function that cannot be reduced by Janino.
• The test verifies the plan is unchanged after applying PROJECT_FILTER_VALUES_MERGE — tuples are preserved.

[caicancai]

@darpan-e6 As discussed in the email, we do not need to retain the email addresses of Claude's co-authors.

[darpan-e6]

@darpan-e6 As discussed in the email, we do not need to retain the email addresses of Claude's co-authors.

I'd like to keep it if that's okay, Claude did help me write the test, and it feels right to acknowledge that. Happy to discuss if there's a policy reason to remove it though!

Also, I haven't received any mail from you on this.

[caicancai]

@darpan-e6 As discussed in the email, we do not need to retain the email addresses of Claude's co-authors.

I'd like to keep it if that's okay, Claude did help me write the test, and it feels right to acknowledge that. Happy to discuss if there's a policy reason to remove it though!

Also, I haven't received any mail from you on this.

You can check the Calcite mailing list; there's discussion about this there.

[darpan-e6]

@darpan-e6 As discussed in the email, we do not need to retain the email addresses of Claude's co-authors.

I'd like to keep it if that's okay, Claude did help me write the test, and it feels right to acknowledge that. Happy to discuss if there's a policy reason to remove it though!
Also, I haven't received any mail from you on this.

You can check the Calcite mailing list; there's discussion about this there.

Okay, makes sense. I will remove it.

[sonarqubecloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
85.7% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[zabetak]

FYI: We have a whole framework in place for testing rules based on RelOptFixture and DiffRepository. The test in this PR would be much shorter and easier to read if it was re-using existing code. Ideally, the test should be either moved to RelOptRulesTest class or follow the paradigm of AggregateFilterToFilteredAggregateRuleTest.

@darpan-e6 If you have some time to rework the test raise a new PR and ping me for review.

[caicancai]

Oh, my question, I forgot, sorry.

[darpan-e6]

I was not aware of it. Thanks @zabetak, I will do that. Allow me some time (like 4-5 hours).