[ 
https://issues.apache.org/jira/browse/CALCITE-3630?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17003897#comment-17003897
 ] 

Chunwei Lei edited comment on CALCITE-3630 at 12/27/19 5:49 AM:
----------------------------------------------------------------

Of course, it can reduce the search space. And it also keeps the same behavior 
with the if branch[1]. I don't figure out why we cannot do the same thing if we 
already do it in the if branch. 

[1] 
[https://github.com/apache/calcite/blob/master/core/src/main/java/org/apache/calcite/rel/rules/ReduceExpressionsRule.java#L194-L216].


was (Author: chunwei lei):
Of course, it can reduce the search space. And it also keeps the same behavior 
with the if branch[1]. I don't figure out why we cannot do the same thing if we 
already do it in the if branch. 

 

[1] 
[https://github.com/apache/calcite/blob/master/core/src/main/java/org/apache/calcite/rel/rules/ReduceExpressionsRule.java#L194-L216].

> Improve ReduceExpressionsRule
> -----------------------------
>
>                 Key: CALCITE-3630
>                 URL: https://issues.apache.org/jira/browse/CALCITE-3630
>             Project: Calcite
>          Issue Type: Bug
>          Components: core
>    Affects Versions: 1.21.0
>            Reporter: Chunwei Lei
>            Assignee: Chunwei Lei
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 1.22.0
>
>          Time Spent: 1h
>  Remaining Estimate: 0h
>
> In {{FilterReduceExpressionsRule}}, if the expression is a IS [NOT] NULL on a 
> non-nullable column, then we can either remove the filter or replace it with 
> an Empty(namely {{reduceNotNullableFilter}}). In such a case, we can set 
> importance to 0 to reduce search space.
>  
> {code:java}
> // code placeholder
>   private void reduceNotNullableFilter(
>       RelOptRuleCall call,
>       Filter filter,
>       RexNode rexNode,
>       boolean reverse) {
>     // If the expression is a IS [NOT] NULL on a non-nullable
>     // column, then we can either remove the filter or replace
>     // it with an Empty.
>     boolean alwaysTrue;
>     switch (rexNode.getKind()) {
>     case IS_NULL:
>     case IS_UNKNOWN:
>       alwaysTrue = false;
>       break;
>     case IS_NOT_NULL:
>       alwaysTrue = true;
>       break;
>     default:
>       return;
>     }
>     if (reverse) {
>       alwaysTrue = !alwaysTrue;
>     }
>     RexNode operand = ((RexCall) rexNode).getOperands().get(0);
>     if (operand instanceof RexInputRef) {
>       RexInputRef inputRef = (RexInputRef) operand;
>       if (!inputRef.getType().isNullable()) {
>         if (alwaysTrue) {
>           call.transformTo(filter.getInput());
>         } else {
>           call.transformTo(createEmptyRelOrEquivalent(call, filter));
>         }
>       }
>     }
>   }
> }
> {code}
>  



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to