cloud-fan commented on a change in pull request #20965: [SPARK-21870][SQL]
Split aggregation code into small functions
URL: https://github.com/apache/spark/pull/20965#discussion_r319951300
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/HashAggregateExec.scala
##########
@@ -255,41 +261,153 @@ case class HashAggregateExec(
""".stripMargin
}
+ private def isValidParamLength(paramLength: Int): Boolean = {
+ // This config is only for testing
+ sqlContext.getConf("spark.sql.HashAggregateExec.isValidParamLength", null)
match {
+ case null | "" => CodeGenerator.isValidParamLength(paramLength)
+ case validLength => paramLength <= validLength.toInt
+ }
+ }
+
+ // Splits aggregate code into small functions because the most of JVM
implementations
+ // can not compile too long functions.
+ //
+ // Note: The difference from `CodeGenerator.splitExpressions` is that we
define an individual
+ // function for each aggregation function (e.g., SUM and AVG). For example,
in a query
+ // `SELECT SUM(a), AVG(a) FROM VALUES(1) t(a)`, we define two functions
+ // for `SUM(a)` and `AVG(a)`.
+ private def splitAggregateExpressions(
+ ctx: CodegenContext,
+ aggNames: Seq[String],
+ aggBufferUpdatingExprs: Seq[Seq[Expression]],
+ aggCodeBlocks: Seq[Block],
+ subExprs: Map[Expression, SubExprEliminationState]): Option[String] = {
+ val exprValsInSubExprs = subExprs.flatMap { case (_, s) => s.value ::
s.isNull :: Nil }
+ if (exprValsInSubExprs.exists(_.isInstanceOf[SimpleExprValue])) {
+ // `SimpleExprValue`s cannot be used as an input variable for split
functions, so
Review comment:
is it because `ExprValue` is not a real tree format? Otherwise we can get
all the referred variables of an expr value and put them on the parameter list.
----------------------------------------------------------------
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:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
