AdamGS opened a new issue, #25276: URL: https://github.com/apache/datafusion/issues/25276
Currently `RewriteSetComparison` turns set comparisons into `CASE` with multiple `EXISTS`, which ends up as multiple mark joins. For example, from https://github.com/apache/datafusion/pull/25273/ (and using that branch to get a correct plan) we get: ``` explain select id from emp WHERE salary = ANY (SELECT salary FROM emp WHERE id <= 2); +---------------+-------------------------------------------------------------------------------------------------------------------+ | plan_type | plan | +---------------+-------------------------------------------------------------------------------------------------------------------+ | logical_plan | Projection: emp.id | | | LeftMark Join: Filter: emp.salary = __correlated_sq_3.salary IS TRUE | | | Projection: emp.id, emp.salary | | | LeftMark Join: Filter: emp.salary = __correlated_sq_2.salary IS NULL | | | Projection: emp.id, emp.salary | | | Filter: __correlated_sq_1.mark | | | LeftMark Join: Filter: emp.salary = __correlated_sq_1.salary IS TRUE | | | TableScan: emp projection=[id, salary] | | | SubqueryAlias: __correlated_sq_1 | | | Projection: emp.salary | | | Filter: emp.id <= Int32(2) | | | TableScan: emp projection=[id, salary] | | | SubqueryAlias: __correlated_sq_2 | | | Projection: emp.salary | | | Filter: emp.id <= Int32(2) | | | TableScan: emp projection=[id, salary] | | | SubqueryAlias: __correlated_sq_3 | | | Projection: emp.salary | | | Filter: emp.id <= Int32(2) | | | TableScan: emp projection=[id, salary] | | physical_plan | NestedLoopJoinExec: join_type=LeftMark, filter=(salary@0 = salary@1) IS NOT DISTINCT FROM true, projection=[id@0] | | | CoalescePartitionsExec | | | NestedLoopJoinExec: join_type=LeftMark, filter=salary@0 = salary@1 IS NULL, projection=[id@0, salary@1] | | | CoalescePartitionsExec | | | FilterExec: mark@2, projection=[id@0, salary@1] | | | RepartitionExec: partitioning=RoundRobinBatch(14), input_partitions=1 | | | NestedLoopJoinExec: join_type=RightMark, filter=(salary@0 = salary@1) IS NOT DISTINCT FROM true | | | FilterExec: id@0 <= 2, projection=[salary@1] | | | DataSourceExec: partitions=1, partition_sizes=[1] | | | DataSourceExec: partitions=1, partition_sizes=[1] | | | RepartitionExec: partitioning=RoundRobinBatch(14), input_partitions=1 | | | FilterExec: id@0 <= 2, projection=[salary@1] | | | DataSourceExec: partitions=1, partition_sizes=[1] | | | RepartitionExec: partitioning=RoundRobinBatch(14), input_partitions=1 | | | FilterExec: id@0 <= 2, projection=[salary@1] | | | DataSourceExec: partitions=1, partition_sizes=[1] | | | | +---------------+-------------------------------------------------------------------------------------------------------------------+ ``` But for the equivalent `IN` query we get the obviously superior: ``` +---------------+---------------------------------------------------------------------------------------------------+ | plan_type | plan | +---------------+---------------------------------------------------------------------------------------------------+ | logical_plan | Projection: emp.id | | | LeftSemi Join: emp.salary = __correlated_sq_1.salary | | | TableScan: emp projection=[id, salary] | | | SubqueryAlias: __correlated_sq_1 | | | Projection: emp.salary | | | Filter: emp.id <= Int32(2) | | | TableScan: emp projection=[id, salary] | | physical_plan | HashJoinExec: mode=CollectLeft, join_type=RightSemi, on=[(salary@0, salary@1)], projection=[id@0] | | | FilterExec: id@0 <= 2, projection=[salary@1] | | | DataSourceExec: partitions=1, partition_sizes=[1] | | | DataSourceExec: partitions=1, partition_sizes=[1] | | | | +---------------+---------------------------------------------------------------------------------------------------+ ``` -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
