uros-b commented on code in PR #57629:
URL: https://github.com/apache/spark/pull/57629#discussion_r3700550155
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/InsertMapSortExpression.scala:
##########
@@ -17,33 +17,46 @@
package org.apache.spark.sql.catalyst.optimizer
+import scala.collection.immutable.VectorMap
import scala.collection.mutable
-import org.apache.spark.sql.catalyst.expressions.{Alias, ArrayTransform,
CreateNamedStruct, Expression, GetStructField, If, IsNull, LambdaFunction,
Literal, MapFromArrays, MapKeys, MapSort, MapValues, NamedExpression,
NamedLambdaVariable}
+import org.apache.spark.sql.catalyst.expressions.{Alias, ArrayTransform,
Attribute, CreateNamedStruct, Expression, GetStructField, If, IsNull,
LambdaFunction, Literal, MapFromArrays, MapKeys, MapSort, MapValues,
NamedExpression, NamedLambdaVariable}
+import
org.apache.spark.sql.catalyst.expressions.aggregate.{AggregateExpression,
AggregateFunction}
import org.apache.spark.sql.catalyst.plans.logical.{Aggregate, LogicalPlan,
Project, RepartitionByExpression}
import org.apache.spark.sql.catalyst.rules.Rule
import org.apache.spark.sql.catalyst.trees.TreePattern.{AGGREGATE,
REPARTITION_OPERATION}
+import org.apache.spark.sql.internal.SQLConf
import org.apache.spark.sql.types.{ArrayType, MapType, StructType}
import org.apache.spark.util.ArrayImplicits.SparkArrayOps
/**
- * Adds [[MapSort]] to [[Aggregate]] expressions containing map columns,
- * as the key/value pairs need to be in the correct order before grouping:
+ * Adds [[MapSort]] to grouping expressions and distinct aggregate arguments
that contain maps,
+ * ensuring key/value pairs have a consistent order before aggregation:
*
* SELECT map_column, COUNT(*) FROM TABLE GROUP BY map_column =>
* SELECT _groupingmapsort as map_column, COUNT(*) FROM (
* SELECT map_sort(map_column) as _groupingmapsort FROM TABLE
* ) GROUP BY _groupingmapsort
+ *
+ * SELECT COUNT(DISTINCT map_column) FROM TABLE =>
+ * SELECT COUNT(DISTINCT _distinctmapsort) FROM (
+ * SELECT map_sort(map_column) as _distinctmapsort FROM TABLE
+ * )
*/
Review Comment:
Could we also note here why the normalization lives in this rule rather than
in RewriteDistinctAggregates, which is where distinct arguments become grouping
keys? A single distinct group without a FILTER never reaches that rule -
RewriteDistinctAggregates L249 bails out and
AggUtils.planAggregateWithOneDistinct handles it in the physical planner, so
normalizing there would miss the most common case. That's the load-bearing
reason for the placement, and it isn't apparent from the code.
--
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]
