Unexpected parent for node after query parsing with aggregation

[frensjan]

It seems that there is a bug in constructing the tuple expression for queries with aggregations. E.g., this code throws an exception:

String query =
        """
        SELECT (COUNT(?s) AS ?count)
        WHERE { ?s ?p ?o  }
        """;

TupleExpr expr = new SPARQLParser().parseQuery(query, null).getTupleExpr();
new ParentReferenceChecker(null).optimize(expr, new SimpleDataset(), new EmptyBindingSet());

The exception message:

java.lang.AssertionError: After query parsing there was an unexpected parent for node Count
   Var (name=s)
: ExtensionElem (count)
   Count
      Var (name=s)
 (expected GroupElem (count)
   Count
      Var (name=s)
)

There aren't any issues with the query execution. But it does cause issues with optimisers that e.g. replace nodes. The exact 'location' the parent reference is mixed up seems to depend on the query 'shape' a bit. E.g., with a more complex expression in the projection like SELECT (COUNT(?s)/30 AS ?count) it's the ?s ?p ?o StatementPattern that gets the wrong parent reference.

Version

4.3.8

Are you interested in contributing a solution yourself?

Perhaps?

[frensjan]

I've added parent reference checking in SPARQLParserTest in this branch:

These tests then fail:

• testGroupByProjectionHandling_Aggregate_Alias
• testGroupByProjectionHandling_Aggregate_Constant
• testGroupByProjectionHandling_Aggregate_SimpleExpr2
• testGroupByProjectionHandling_function
• testGroupByProjectionHandling_renameVariableWithAggregation
• testGroupByProjectionHandling_variableEffectivelyAggregationResult
• testProjectionHandling_CustomAggregateFunction
• testProjectionHandling_HavingCustomAggregateFunction
• testProjectionHandling_MultipleAggregateFunction

testParseWildcardSubselectInUpdate also fails, but not for the same reason.

I think I've been able to track down the issue to this line in TupleExprBuilder.java. At least that's where for a particular case I've seen the wrong link being made.