[CALCITE-5400] EnumerableCalc should not generate Enumerable.current with a non-idempotent function inside

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.32.0
• Component/s: core
• Assignee: Unassigned
• Reporter: Dmitry Sysolyatin
• Created: 24/Nov/22 20:48
• Updated: 13/Feb/23 15:15

Description

The problem has been found inside CALCITE-5388. The following query can return wrong result:

with
    CTE1(rand1, val1) as ( select RAND_INTEGER(2), id from (values (1), (2)) as Vals1(id) ),
    CTE2(rand2, val2) as ( select RAND_INTEGER(2), id from (values (1), (2)) as Vals2(id) )
select
    CTE1.rand1,    CTE1.val1,
    CTE2.rand2,
    CTE2.val2 
from
    CTE1,
    CTE2 
where
    CTE1.rand1 = CTE2.rand2 

For instance it can return:

RAND1	VAL1	RAND2	VAL2
1	1	1	1
0	1	1	2

The problem is that EnumerableCalc generates Enumerable class that uses non-idempotent function `randInteger` inside 'current()' method: 

new org.apache.calcite.linq4j.AbstractEnumerable() {
          public org.apache.calcite.linq4j.Enumerator enumerator() {
            return new org.apache.calcite.linq4j.Enumerator() {
              <.........>
              public Object current() {
                return new Object[]{
                    $L4J$C$new_org_apache_calcite_runtime_RandomFunction_.randInteger(2),
                    org.apache.calcite.runtime.SqlFunctions.toInt(inputEnumerator.current())};
              }

              static final org.apache.calcite.runtime.RandomFunction $L4J$C$new_org_apache_calcite_runtime_RandomFunction_ = new org.apache.calcite.runtime.RandomFunction();
            };

          }
        };

if current() is called twice it can produce different results. What exactly happens inside EnumerableDefault.hashEquiJoin_ function (outers.current() is called inside moveNext() and current())

Issue Links

• is related to: CALCITE-5388 tempList expression inside EnumerableWindow.getPartitionIterator should be unoptimized