maropu commented on a change in pull request #24158: [SPARK-26847][SQL] Pruning
nested serializers from object serializers: MapType support
URL: https://github.com/apache/spark/pull/24158#discussion_r268032776
##########
File path:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/objects.scala
##########
@@ -132,49 +130,74 @@ object ObjectSerializerPruning extends Rule[LogicalPlan]
{
fields.map(f => collectStructType(f.dataType, structs))
case ArrayType(elementType, _) =>
collectStructType(elementType, structs)
+ case MapType(_, valueType, _) =>
+ // Because we can't select a field from struct in key, so we skip key
type.
+ collectStructType(valueType, structs)
case _ =>
}
structs
}
+ /**
+ * This method returns pruned `CreateNamedStruct` expression given an
original `CreateNamedStruct`
+ * and a pruned `StructType`.
+ */
+ private def pruneNamedStruct(struct: CreateNamedStruct, prunedType:
StructType) = {
+ // Filters out the pruned fields.
+ val prunedFields = struct.nameExprs.zip(struct.valExprs).filter { case
(nameExpr, _) =>
+ val name = nameExpr.eval(EmptyRow).toString
+ prunedType.fieldNames.exists { fieldName =>
+ if (SQLConf.get.caseSensitiveAnalysis) {
+ fieldName.equals(name)
+ } else {
+ fieldName.equalsIgnoreCase(name)
+ }
+ }
+ }.flatMap(pair => Seq(pair._1, pair._2))
+
+ CreateNamedStruct(prunedFields)
+ }
+
+ /**
+ * When we change nested serializer data type, `If` expression will be
unresolved because
+ * literal null's data type doesn't match now. We need to align it with new
data type.
+ * Note: we should do `transformUp` explicitly to change data types.
+ */
+ private def alignNullTypeInIf(expr: Expression) = expr.transformUp {
+ case i @ If(_: IsNull, Literal(null, dt), ser) if
!dt.sameType(ser.dataType) =>
+ i.copy(trueValue = Literal(null, ser.dataType))
+ }
+
/**
* This method prunes given serializer expression by given pruned data type.
For example,
* given a serializer creating struct(a int, b int) and pruned data type
struct(a int),
- * this method returns pruned serializer creating struct(a int). For now it
supports to
- * prune nested fields in struct and array of struct.
- * TODO(SPARK-26847): support to prune nested fields in key and value of map
type.
+ * this method returns pruned serializer creating struct(a int).
*/
def pruneSerializer(
serializer: NamedExpression,
prunedDataType: DataType): NamedExpression = {
val prunedStructTypes = collectStructType(prunedDataType,
ArrayBuffer.empty[StructType])
var structTypeIndex = 0
- val prunedSerializer = serializer.transformDown {
+ val transformedSerializer = serializer.transformDown {
+ case m: ExternalMapToCatalyst =>
+ val valueStructs = m.valueConverter.collect {
Review comment:
`valueStructs` not used?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
