Add AggregationNode.hasSingleGlobalAggregation

[lukasz-stec]

Description

Extract single global aggregation (GROUP BY ()) from multiple places to a method in AggregationNode

Non-technical explanation

Code refactoring

Release notes

( X) This is not user-visible or docs only and no release notes are required.
( ) Release notes are required, please propose a release note for me.
( ) Release notes are required, with the following suggested text:

[raunaqmorarka]

lgtm, @martint @sopel39 can you also PTAL ?

[sopel39]

lgtm % comment

[sopel39]

I would remove Returns {@code false} case.

I wondder why

even if all are empty e.g. GROUP BY GROUPING SETS ((), ()).

is relevant

[lukasz-stec]

some places look only at getGroupingKeys().isEmpty() and consider this a global aggregation. This includes any number of empty grouping sets.
This method only allows for a single empty grouping set and I wanted to emphasize that in the comment so it's easier to spot the difference.

[martint]

Which places do that? That would be a bug.

There's only one possible interpretation of "single global aggregation". It must collapse all rows into a single group and produce a single row. The presence of multiple groups would immediately disqualify it as a "single" aggregation.

[lukasz-stec]

Which places do that?

The most prominent example:

io.trino.sql.planner.LocalExecutionPlanner.Visitor#visitAggregation
...
         if (node.getGroupingKeys().isEmpty()) {
                return planGlobalAggregation(node, source, context);
            }

[martint]

This should be documented using @return in the javadocs:

@return true if the aggregation collapses all rows into a single global group (e.g., as a result of a GROUP BY () query). Otherwise, false.

[lukasz-stec]

I updated the comment

[lukasz-stec]

flaky TestParquetReader #13633

[lukasz-stec]

rebased on the master to restart CI