[
https://issues.apache.org/jira/browse/CALCITE-7758?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18109977#comment-18109977
]
Julian Hyde commented on CALCITE-7758:
--------------------------------------
Solving this will require matching expressions that are not atoms, and any
strategy for subexpression matching will increase the cost of simplification.
Even a cost that is {{O(n)}} on the size of the expression becomes {{O(n^2)}}
for the planner as a whole, and small cost increases become an unseen tax on
planning.
I don't recall whether {{RexSimplify}} builds tables of expression fragments as
it goes, but the benefit of this change should be weighed against increased
cost.
> RexSimplify does not absorb redundant IS NOT NULL on non-input-ref
> sub-expressions in AND simplification
> --------------------------------------------------------------------------------------------------------
>
> Key: CALCITE-7758
> URL: https://issues.apache.org/jira/browse/CALCITE-7758
> Project: Calcite
> Issue Type: Improvement
> Components: core
> Reporter: Ruben Q L
> Assignee: Ruben Q L
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.43.0
>
>
> The following simplification:
> {noformat}
> IS_NOT_NULL(x) AND (x + 1) < 10
> =>
> (x + 1) < 10
> {noformat}
> is correctly applied on RexSimplify (Unknown=FALSE) for RexInputRef and
> RexFieldAccess (via simplifyAnd2ForUnknownAsFalse).
> However, if the expression X is e.g. a CAST, the simplification does not
> occur:
> {noformat}
> IS_NOT_NULL(CAST(a)) AND (CAST(a) + 1) < 10
> =>
> unchanged; should be: (CAST(a) + 1) < 10
> {noformat}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
