andygrove commented on issue #1737:
URL:
https://github.com/apache/datafusion-comet/issues/1737#issuecomment-2891444222
Some notes from debugging this:
`CometExecRule` performs the following transformation:
```
INPUT: Exchange hashpartitioning(date_id#5283, product_id#5284,
units_sold#5285, store_id#5286, 5), ENSURE_REQUIREMENTS, [plan_id=860]
+- HashAggregate(keys=[date_id#5283, product_id#5284, units_sold#5285,
store_id#5286], functions=[], output=[date_id#5283, product_id#5284,
units_sold#5285, store_id#5286])
+- CometScan parquet
spark_catalog.default.fact_sk[date_id#5283,product_id#5284,units_sold#5285,store_id#5286]
Batched: true, DataFilters: [], Format: CometParquet, Location:
InMemoryFileIndex(25
paths)[file:/Users/andy/git/apache/apache-spark/sql/core/spark-warehouse/org...,
PartitionFilters: [isnotnull(store_id#5286), dynamicpruningexpression(true)],
PushedFilters: [], ReadSchema: struct<date_id:int,product_id:int,units_sold:int>
OUTPUT CometExchange hashpartitioning(date_id#5283, product_id#5284,
units_sold#5285, store_id#5286, 5), ENSURE_REQUIREMENTS, CometNativeShuffle,
[plan_id=870]
+- CometHashAggregate [date_id#5283, product_id#5284, units_sold#5285,
store_id#5286], [date_id#5283, product_id#5284, units_sold#5285, store_id#5286]
+- CometScan parquet
spark_catalog.default.fact_sk[date_id#5283,product_id#5284,units_sold#5285,store_id#5286]
Batched: true, DataFilters: [], Format: CometParquet, Location:
InMemoryFileIndex(25
paths)[file:/Users/andy/git/apache/apache-spark/sql/core/spark-warehouse/org...,
PartitionFilters: [isnotnull(store_id#5286), dynamicpruningexpression(true)],
PushedFilters: [], ReadSchema: struct<date_id:int,product_id:int,units_sold:int>
```
Because the original plan does not support columner, AQE is applying
columnar transition rules when invoking `postStageCreationRules(outputsColumnar
= plan.supportsColumnar)` and this causes `CometColumnarToRow` to get added,
which then gets wrapped in `WholeStageCodegenExec`, at which point AQE fails
because it was expecting an exchange.
--
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]
