kasakrisz commented on code in PR #3761:
URL: https://github.com/apache/hive/pull/3761#discussion_r1035633744
##########
ql/src/java/org/apache/hadoop/hive/ql/optimizer/calcite/rules/HiveSubQueryRemoveRule.java:
##########
@@ -104,44 +106,43 @@ private HiveSubQueryRemoveRule(RelOptRuleOperand operand,
String description, Hi
// if subquery is in FILTER
if (relNode instanceof HiveFilter) {
final HiveFilter filter = call.rel(0);
- final RexSubQuery e = RexUtil.SubQueryFinder.find(filter.getCondition());
- assert e != null;
-
- final RelOptUtil.Logic logic =
- LogicVisitor.find(RelOptUtil.Logic.TRUE,
ImmutableList.of(filter.getCondition()), e);
+ // Since there is a RexSubQuery, there should be a HiveCorrelationInfo
+ // in the RelNode
+ Preconditions.checkState(filter.getCorrelationInfos().size() > 0);
+ HiveCorrelationInfo correlationInfo =
filter.getCorrelationInfos().get(0);
+ final RelOptUtil.Logic logic = LogicVisitor.find(RelOptUtil.Logic.TRUE,
+ ImmutableList.of(filter.getCondition()),
correlationInfo.rexSubQuery);
builder.push(filter.getInput());
final int fieldCount = builder.peek().getRowType().getFieldCount();
- SubqueryConf subqueryConfig =
filter.getCluster().getPlanner().getContext().unwrap(SubqueryConf.class);
- boolean isCorrScalarQuery =
subqueryConfig.getCorrScalarRexSQWithAgg().contains(e.rel);
+ boolean isCorrScalarQuery = correlationInfo.isCorrScalarQuery();
final RexNode target =
- apply(call.getMetadataQuery(), e, HiveFilter.getVariablesSet(e),
logic, builder, 1,
- fieldCount, isCorrScalarQuery);
- final RexShuttle shuttle = new ReplaceSubQueryShuttle(e, target);
+ apply(call.getMetadataQuery(), correlationInfo.rexSubQuery,
+ correlationInfo.correlationIds, logic, builder, 1, fieldCount,
isCorrScalarQuery);
+ final RexShuttle shuttle = new
ReplaceSubQueryShuttle(correlationInfo.rexSubQuery, target);
builder.filter(shuttle.apply(filter.getCondition()));
builder.project(fields(builder, filter.getRowType().getFieldCount()));
RelNode newRel = builder.build();
call.transformTo(newRel);
} else if (relNode instanceof HiveProject) {
- // if subquery is in PROJECT
final HiveProject project = call.rel(0);
- final RexSubQuery e = RexUtil.SubQueryFinder.find(project.getProjects());
- assert e != null;
+ // Since there is a RexSubQuery, there should be a HiveCorrelationInfo
+ // in the RelNode
+ Preconditions.checkState(project.getCorrelationInfos().size() > 0);
+ HiveCorrelationInfo correlationInfo =
project.getCorrelationInfos().get(0);
Review Comment:
I managed to have more than one `CorrelationInfo`:
```
explain cbo
select
year in (select year from t_test2 where t_test2.id = t_test1.id),
month in (select month from t_test2 where t_test2.id = t_test1.id)
from t_test1;
```
but here we always get the 1st element of the list and
`project.getCorrelationInfos()` is nowhere else is used or I'm missing
something.
Could you please check why do we have to collect all `CorrelationInfo`
objects
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
