[
https://issues.apache.org/jira/browse/CALCITE-5923?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17755547#comment-17755547
]
Runkang He commented on CALCITE-5923:
-------------------------------------
After fix this issue, the result check for `CAST` operator is turned on, and
then, I find a test case `SqlOperatorTest.testCastStringToDateTime` fails due
to the failed test below with latest calcite-avatica branch:
{code:java}
f.checkScalar("cast('1945-2-2 12:2:5' as TIMESTAMP)",
"1945-02-02 12:02:05", "TIMESTAMP(0) NOT NULL");
{code}
The stacktrace is:
{code:java}
Caused by: java.lang.IllegalArgumentException: Invalid DATE value, '1945-2-2
12:2:5'
at
org.apache.calcite.avatica.util.DateTimeUtils.invalidType(DateTimeUtils.java:809)
at
org.apache.calcite.avatica.util.DateTimeUtils.validateDate(DateTimeUtils.java:752)
at
org.apache.calcite.avatica.util.DateTimeUtils.timestampStringToUnixDate0(DateTimeUtils.java:828)
at
org.apache.calcite.avatica.util.DateTimeUtils.timestampStringToUnixDate(DateTimeUtils.java:815)
... 113 more
{code}
It seems related to the fix in CALCITE-5678 in avatica-1.24, which use ISO8601
to validate the timestamp string.
I think we can safely modify this test case to use `checkFails` with message
`Invalid DATE value`.
[~zstan] What do you think? Looking forward to your reply.
> Some test cases in `SqlOperatorTest` violates the test fixture's design
> principle
> ---------------------------------------------------------------------------------
>
> Key: CALCITE-5923
> URL: https://issues.apache.org/jira/browse/CALCITE-5923
> Project: Calcite
> Issue Type: Bug
> Components: core
> Affects Versions: 1.35.0
> Reporter: Runkang He
> Assignee: Runkang He
> Priority: Minor
> Labels: pull-request-available
>
> There are some test cases in `SqlOperatorTest` directly use the
> `SqlOperatorFixtureImpl.DEFAULT` to get the `SqlOperatorFixture`, including
> `SqlOperatorTest.testCast` and many other test cases related with `CAST`
> operator. This causes that the result check is missing when execute
> `CalciteSqlOperatorTest`, which should has result check.
> This violates the design principle introduced by CALCITE-4885, which we
> should alway use `SqlOperatorTest.fixture()` to get the `SqlOperatorFixture`.
> This principle allows us to override`fixture()` method in subclasses to run
> tests in a different environment.
> So I think we should fix these related test cases to keep consistent with the
> principle.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
