[
https://issues.apache.org/jira/browse/CALCITE-7210?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18023952#comment-18023952
]
Mihai Budiu commented on CALCITE-7210:
--------------------------------------
The root cause seems to be an incorrect type inference for LEAST/GREATEST.
Calcite generally expands LEAST into a CASE, but the type inferred for the same
expression using LEAST or its CASE expansion is different.
> BINARY literal values may not match their type
> ----------------------------------------------
>
> Key: CALCITE-7210
> URL: https://issues.apache.org/jira/browse/CALCITE-7210
> Project: Calcite
> Issue Type: Bug
> Components: core
> Affects Versions: 1.40.0
> Reporter: Mihai Budiu
> Assignee: Mihai Budiu
> Priority: Minor
>
> The following SqlOperatorTest fails:
> {code:java}
> final SqlOperatorFixture f = fixture()
> .setFor(SqlLibraryOperators.LEAST, VmName.EXPAND)
> .withLibrary(SqlLibrary.BIG_QUERY);
> f.checkScalar("least(x'01', x'0202')", "0100", "BINARY(2) NOT NULL");
> {code}
> The error is:
> Expected: is "0100"
> but: was "01"
> However, the type of the result BINARY(2) is correct. But "01" is not a legal
> value for this type.
> This occurs because BINARY literal values do not always match their declared
> type. I think there are several paths in simplification code which can lead
> to this outcome.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
