[CALCITE-6087] EnumerableSortedAggregate returns incorrect result when input is empty

[1fanwang]

Jira Link

CALCITE-6087

Changes Proposed

An empty global aggregate implemented by EnumerableSortedAggregate returns no row. For example, MAX and COUNT over an empty input should return one row containing NULL and 0, but the result is empty.

Sorted grouping emits one row per observed key, and a global aggregate has no grouping key. The sorted aggregate rule now declines an empty group set so EnumerableAggregate handles it through its zero-key singleton path. Grouped sorted aggregates are unchanged.

Testing

Scenario	Result
Forced sorted aggregate before the fix	Reproduced an empty result instead of m=null; c=0
Focused regression test	1 passed
Core test suite	16,639 passed, 157 skipped

Raw logs

$ JAVA_HOME=$(/usr/libexec/java_home -v 21) ./gradlew :core:test --tests org.apache.calcite.test.enumerable.EnumerableSortedAggregateTest.sortedAggOnEmptyInput --no-daemon --console=plain
Expected: "m=null; c=0"
     but: was ""
1 test completed, 1 failed
BUILD FAILED

$ JAVA_HOME=$(/usr/libexec/java_home -v 21) ./gradlew :core:test --tests org.apache.calcite.test.enumerable.EnumerableSortedAggregateTest.sortedAggRuleOnEmptyInput --no-daemon --console=plain
1 completed, 0 failed, 0 skipped
BUILD SUCCESSFUL

$ JAVA_HOME=$(/usr/libexec/java_home -v 21) ./gradlew :core:test --no-daemon --console=plain
16639 completed, 0 failed, 157 skipped
BUILD SUCCESSFUL

Formatting, Checkstyle, and forbidden-apis checks pass.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
80.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[xuzifu666]

Thank you for the contribution! Please take a look to jira discussion.

[rubenada]

why adding this rule if it was just removed a few lines above?
UPDATE: ok, I understand, if we don't do it the query is un-implementable. But now I wonder if we can be really sure to be testing the fix here, or maybe it's the standard ENUMERABLE_AGGREGATE_RULE that kicks in....
I wonder if a more suitable test would be testing the without any hook (to confirm it passes with ENUMERABLE_AGGREGATE_RULE; and then testing it with Hook to remove ENUMERABLE_AGGREGATE_RULE, add ENUMERABLE_SORTED_AGGREGATE_RULE and verify that it fails because it cannot be implemented?

[rubenada]

Thanks for the contribution @1fanwang !
@xuzifu666 I opened this ticket long time ago, I had a suspicion about a probably fix (implemented in this PR), I asked the contributors that worked on the EnumerableSortedAggregate for their opinion about the fix, but got no answer.
I think we can move forward if we are indeed correcting a bug.