Iurii Gerzhedovich created IGNITE-25001:
-------------------------------------------
Summary: Sql. ProjectScanMergeRule doesn't preserve name of columns
Key: IGNITE-25001
URL: https://issues.apache.org/jira/browse/IGNITE-25001
Project: Ignite
Issue Type: Improvement
Components: sql
Reporter: Iurii Gerzhedovich
During the merge project into scan, we create a new name for columns like a
$f0,$f1,...
It can lead to issues for correlation as we can have different names in
different parts of the plan.
An investigation can be started from the CorrelatedSubqueryPlannerTest#test by
disabling the rules. New names we derived in
org.apache.ignite.internal.sql.engine.rel.ProjectableFilterableTableScan#deriveRowType
Additional example of query to demonstrate the issue
{code:java}
select (select count(*) from t1 inner_t1 where inner_t1.id = doubled_id) FROM
(SELECT id, id * 2 as doubled_id FROM t1) as t1{code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
