xuzifu666 commented on PR #5354: URL: https://github.com/apache/calcite/pull/5354#issuecomment-6383450343
> I’m wondering whether we still need to maintain the expand=true mode. We should try to address these issues in expand=false mode wherever possible. The issue exposed by the last SQL query in the comments on [CALCITE-4340](https://issues.apache.org/jira/browse/CALCITE-4340) is actually what [CALCITE-7278](https://issues.apache.org/jira/browse/CALCITE-7278) aims to resolve. Thanks for the suggestion. @iwanttobepowerful I investigated whether the fix can be moved to `expand=false` mode, but found it's not currently feasible: Under `expand=false`, sub-queries remain as `$SCALAR_QUERY` expressions. The decorrelator cannot process them — withDecorrelate(true) on the expand=false plan fails to produce a valid result. The current fix in convertCorrelatedScalarSubqueryInOn (expand=true path) already produces a decorrelation-friendly plan — running with withDecorrelate(true) yields an identical plan, confirming the rewrite is what the decorrelator expects. Making the `expand=false` path work would require enhancing the decorrelator / `SubQueryRemoveRule` to handle $SCALAR_QUERY in ON clauses, which is the scope of https://issues.apache.org/jira/browse/CALCITE-7278. I'd suggest keeping this fix as a stopgap for the expand=true path (still the default) until CALCITE-7278 finished. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
