[
https://issues.apache.org/jira/browse/CALCITE-5343?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17623443#comment-17623443
]
Julian Hyde commented on CALCITE-5343:
--------------------------------------
Your proposed workaround sounds to me like several layers of duct tape. As I
said, the semantics of '/' should be orthogonal from the type system. You could
view it as a new operator, or you could think of it as a new error-handling
mode.
> Type of division operator
> -------------------------
>
> Key: CALCITE-5343
> URL: https://issues.apache.org/jira/browse/CALCITE-5343
> Project: Calcite
> Issue Type: Bug
> Components: core
> Reporter: Mihai Budiu
> Priority: Minor
>
> The Calcite type checker assigns a type of INTEGER (not nullable) to the
> DIVIDE RexCall in the following statement: `SELECT 1/0`. Some databases, such
> as MySQL, evaluate this expression to NULL.
> This result is obtained even when using SqlConformanceEnum.MYSQL_5 for the
> parser and the validator.
> My question is whether there is some other way to influence how type
> inference is performed for division.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
