[
https://issues.apache.org/jira/browse/DERBY-4391?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12764130#action_12764130
]
Dag H. Wanvik commented on DERBY-4391:
--------------------------------------
I looked at this patch and verified that the patch solved the issue, and that
the new test case fails without the patch.
Thanks!
I wonder if it is meaningful for the visitor in
PredicateList.markReferencedColumns() to go down into subqueries. This
ColumnReference with the empty source here, stems from inside the subquery, and
looking at the logic in FromBaseTable#changeAccessPath, what we are doing here
is to collect needed columns from the outer table. I think that means that we
do not need to look inside a subquery?
Changing to this visitor declaration makes the query work too:
CollectNodesVisitor collectCRs =
new CollectNodesVisitor(ColumnReference.class,
SubqueryNode.class);
As for the question of the empty source,
looking at the query tree, I see that there is indeed a column reference which
has an empty source, which I also
see, e.g in this query:
select * from (values 2 union values 2) x
At the corresponding place (column reference) in a query tree for
select * from (select * from t union select * from t) x
I see that the column reference source actually points to a result column of
the PRN wrapping the base table in the *left* query tree of the UNION. (That
also seems a bit arbitrary, why not the right? :) If it is sound for the VALUE
case to have an empty column reference, I can't answer that, because I don't
fully understand how the column references are used in later phases.
> NullPointerException when comparing indexed column with result from a set
> operation
> -----------------------------------------------------------------------------------
>
> Key: DERBY-4391
> URL: https://issues.apache.org/jira/browse/DERBY-4391
> Project: Derby
> Issue Type: Bug
> Components: SQL
> Affects Versions: 10.2.2.0, 10.3.3.0, 10.4.2.0, 10.5.3.0, 10.6.0.0
> Reporter: Knut Anders Hatlen
> Assignee: Knut Anders Hatlen
> Attachments: d4391-1a.diff, d4391-1a.stat
>
>
> I'm reporting this issue on behalf of Bernt M. Johnsen.
> If an indexed column is compared with a UNION query (or some other set
> operation), a NullPointerException is raised, as can be seen by this sequence
> of statements in ij:
> ij> create table t(a int not null primary key, b int);
> 0 rows inserted/updated/deleted
> ij> select * from t where a < (values 4 union values 4);
> ERROR XJ001: Java exception: ': java.lang.NullPointerException'.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
