[
https://issues.apache.org/jira/browse/CALCITE-3973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17100436#comment-17100436
]
Danny Chen commented on CALCITE-3973:
-------------------------------------
Thanks [~alexbaden] for reporting this ~
You are right, in SqlSelectOperator, the leftPrec and rightPrec should both be
hard code to 0 [1], after fixing that, please also add a test case in
SqlParserTest#testQueryHint.
[1]
https://github.com/apache/calcite/blob/d1aae57fad71bf10942d4c53b339363537927c29/core/src/main/java/org/apache/calcite/sql/SqlSelectOperator.java#L150
> Writing SQL hints to a string results in SQL that cannot be parsed
> -------------------------------------------------------------------
>
> Key: CALCITE-3973
> URL: https://issues.apache.org/jira/browse/CALCITE-3973
> Project: Calcite
> Issue Type: Bug
> Reporter: Alex Baden
> Priority: Minor
>
> When using the new SQL hints feature:
>
> {code:java}
> select /*+ cpu */ x from test limit 2;
> {code}
> If one calls to SqlString() on the node:
>
>
> {code:java}
> node.toSqlString(CalciteSqlDialect.DEFAULT).toString()
> {code}
>
> We get:
>
> {code:java}
> 'SELECT /*+ ("cpu") */ "x" FROM "test" FETCH NEXT 2 ROWS ONLY'
> {code}
> {{which is not valid SQL for a hint. It looks like the precedence options
> need to be modified when writing out the hint in SqlSelectOperator.unparse}}
>
> We do some rewriting of the sql statement and then re-parse it, so this is
> preventing us from using hints. I am happy to submit a patch if I am headed
> in the right direction with the description above.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
