[GitHub] [spark] peter-toth commented on pull request #40473: [SPARK-42851][SQL] Guard EquivalentExpressions.addExpr() with supportedExpression()
peter-toth commented on PR #40473: URL: https://github.com/apache/spark/pull/40473#issuecomment-1474880345 Hm, I think you are right @Kimahriman, `LambdaVariable` and `NamedLambdaVariable` are very different and `NamedLambdaVariable` seem to be used only in `LambdaFunction`s, so https://github.com/apache/spark/pull/39046 doesn't make sense and actually it can prevent pulling out higher order functions and so cause performance regression... I think that PR should be reverted. But I feel that is orthogonal to the issue that we use `EquivalentExpressions` for different purposes in `PhysicalAggregation` (the only place where we use `.addExpr()`) and in executors (`.addExprTree()` for subexpression elimination). -- 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: reviews-unsubscr...@spark.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] peter-toth commented on pull request #40473: [SPARK-42851][SQL] Guard EquivalentExpressions.addExpr() with supportedExpression()
peter-toth commented on PR #40473: URL: https://github.com/apache/spark/pull/40473#issuecomment-1474774246 Thanks @rednaxelafx for the fix and pinging me. I think you are right that `EquivalentExpressions.addExpr()` should be guarded by `supportedExpression()` if we guard `getExprState()`. But, I'm not sure it is right that we don't deduplicate the `max(transform(array(id), x -> x))` in your example query. Probably the real issue here is that in`PhysicalAggregation` the class `EquivalentExpressions` is used for simply deduplicating whole expressions while on executors we use it for common subexpression elimination. In the former case we don't need the `LambdaVariable ` guard but in the latter one we need it. So maybe we should add a argument to `EquivalentExpressions` to enable/disable the guards and in `PhysicalAggregation` we should disable it? -- 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: reviews-unsubscr...@spark.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org