[CALCITE-7595] Support FILTER clause with window functions

[xuzifu666]

jira: https://issues.apache.org/jira/browse/CALCITE-7595

[mihaibudiu]

Where is the documentation about the new grammar and its semantics?

[mihaibudiu]

Please add here the following:

• next 4 tests are related to this issue
• results were validated on Postgres

[xuzifu666]

OK, done.

[iwanttobepowerful]

The execution results differ from PostgreSQL.
@xuzifu666
https://onecompiler.com/postgresql/44smkpfxb

[xuzifu666]

This data source should be different. I used the standard Scott data source for the test set, and it passed the syntax verification. I'll investigate the Calcite data source later. @iwanttobepowerful

[mihaibudiu]

I don't understand where this is used. Can you please explain?

[xuzifu666]

The FILTER operator implements conditional aggregation in SQL, such as:
SUM(salary) FILTER (WHERE department = 'Sales') COUNT(*) FILTER (WHERE age > 30)
Processing Pipeline:

1. SQL Parsing → FILTER (WHERE condition) parsed as SqlKind.FILTER node
2. Validation → Checks aggregate-FILTER compatibility (we reject COUNT(DISTINCT) + FILTER in window functions)
3. SQL-to-Relational Conversion → FILTER condition extracted and stored
4. Code Generation → Registers FilterImplementor to translate FILTER into executable code
   I had added a comment about it.

[mihaibudiu]

How come this does not affect other aggregates that have filters and only applies to window aggregates?

[xuzifu666]

In my view restriction is in SqlOverOperator.validateCall(), which is only called for window functions (the OVER operator).

if (hasFilter && aggCall.getKind() == SqlKind.COUNT
      && aggCall.getFunctionQuantifier() != null) {
    throw validator.newValidationError(aggCall, RESOURCE.overNonAggregate());
  }

1. Normal aggregates bypass SqlOverOperator entirely—they're validated by their own operator (e.g., SqlCountAggFunction)
2. Window aggregates MUST go through SqlOverOperator because they have the OVER clause
3. Our restriction is injected into the SqlOverOperator validation, so it naturally only affects window functions

[mihaibudiu]

I have approved, but I still have a question

[mihaibudiu]

Interesting, can you add a comment in the RexImplTable about this?

[xuzifu666]

Interesting, can you add a comment in the RexImplTable about this?

OK, I had added comments about it.

[xuzifu666]

Do you happy with these changes? If OK, I would squash the commits~ @mihaibudiu

[mihaibudiu]

I have approved

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
92.7% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud