[
https://issues.apache.org/jira/browse/CALCITE-6523?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17873019#comment-17873019
]
Julian Hyde commented on CALCITE-6523:
--------------------------------------
Currently this case is merely an observation, "Oracle SUBSTR semantics
different from Calcite", so is not actionable. Can you re-phrase summary and
description so that it reads like a bug? I think there is a test case where
someone submits a query to Calcite (expecting Calcite semantics), and Calcite's
JDBC adapter translates the query to Oracle SQL, and Oracle does not return the
results that the user expected.
> Oracle SUBSTR semantics different from Calcite
> ----------------------------------------------
>
> Key: CALCITE-6523
> URL: https://issues.apache.org/jira/browse/CALCITE-6523
> Project: Calcite
> Issue Type: Bug
> Components: core
> Reporter: suibianwanwan
> Priority: Minor
>
> Oracle's `SUBSTR` has very different semantics. For example, if the length of
> the character is less than the second argument, null is returned.
> {code:java}
> select substr('a', 2, 1);{code}
> I tested Spark sql, presto, mysql all return empty string, same in calcite. I
> noticed that in CALCITE-4427, the substr is handled in sqlToRel, should we
> also do some conversion in reltosqlConverter?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
