[
https://issues.apache.org/jira/browse/CALCITE-4510?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17290783#comment-17290783
]
Ruben Q L commented on CALCITE-4510:
------------------------------------
Thanks for the explanation, [~fan_li_ya].
I agree removing this hard-coded " NOT NULL" that is all around the place and
centralize it in a single constant is a good idea (plus it solves your problem
with user-defined type systems).
Searching the code, it seems the PR could go a bit further and do the
replacement in other classes too:
- RelDatatypeImpl (already covered in the initial PR)
- RexLiteral (already covered in the initial PR)
- SqlTypeUtil#equalSansNullability
- RelOptUtil#TypeDumper#accept (x2)
And even in some tests classes:
- SqlTests#getTypeString
- SqlOperatorBaseTest (x4): checkCastToApproxOkay, checkCastToStringOkay,
checkCastToScalarOkay, assertSubFunReturns
- CalciteAssert#typeString
- RexProgramTestBase#assertNode
- RexProgramTest#assertTypeAndToString
WDYT?
> Weird digests for literals with some user defined types
> -------------------------------------------------------
>
> Key: CALCITE-4510
> URL: https://issues.apache.org/jira/browse/CALCITE-4510
> Project: Calcite
> Issue Type: Bug
> Components: core
> Reporter: Liya Fan
> Assignee: Liya Fan
> Priority: Minor
> Labels: pull-request-available
> Time Spent: 10m
> Remaining Estimate: 0h
>
> We find weird literals for some user defined non-nullable types. Some
> investigation shows that the problem lies in the {{RexLiteral#toJavaString}}
> method.
> In particular, it checks the type string suffix with an 8-character string:
> {noformat}
> if (!fullTypeString.endsWith("NOT NULL")) {
> {noformat}
> However, it trims the last 9 characters from the end of the string:
> {noformat}
> sb.append(fullTypeString, 0, fullTypeString.length() - 9);
> {noformat}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
