[
https://issues.apache.org/jira/browse/FLINK-24422?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17487099#comment-17487099
]
Marios Trivyzas commented on FLINK-24422:
-----------------------------------------
Did some debugging, so my findings are that this is currently allowed because
even though *LocalDateTime* is not a valid input class for
{*}LocalZonedTimestampType{*}, in
*ValuesOperationFactory#convertToExpectedType* we resort to the final block
{noformat}
if (supportsExplicitCast(sourceLogicalType.copy(true),
targetLogicalType.copy(true))) {
return Optional.of(postResolverFactory.cast(sourceExpression,
targetDataType));
} else {
return Optional.empty();
}{noformat}
which succeeds in casting the *LocalDateTime/TIMESTAMP()* literal input value
to {*}Instant/TIMESTAMP_LTZ(){*}.
So now I have second thoughts if we should actually fix this, maybe by just
rejecting this exact conversion?
[~slinkydeveloper] [~twalthr] What do you think?
> When creating a TIMESTAMP_LTZ only allow Instant
> ------------------------------------------------
>
> Key: FLINK-24422
> URL: https://issues.apache.org/jira/browse/FLINK-24422
> Project: Flink
> Issue Type: Sub-task
> Reporter: Marios Trivyzas
> Assignee: Francesco Guardiani
> Priority: Major
>
> Currently we also accept *LocalDateTime* objects:
> *.from(TIMESTAMP_LTZ(),
> LocalDateTime.parse("2021-09-24T12:34:56.123456")).resultsIn(LocalDate.of(2021,
> 9, 24))*
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
