Github user frreiss commented on a diff in the pull request:
https://github.com/apache/spark/pull/13155#discussion_r65584546
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
---
@@ -1695,16 +1695,176 @@ object RewriteCorrelatedScalarSubquery extends
Rule[LogicalPlan] {
}
/**
+ * Statically evaluate an expression containing zero or more
placeholders, given a set
+ * of bindings for placeholder values.
+ */
+ private def evalExpr(expr : Expression, bindings : Map[Long,
Option[Any]]) : Option[Any] = {
+ val rewrittenExpr = expr transform {
+ case r @ AttributeReference(_, dataType, _, _) =>
+ bindings(r.exprId.id) match {
+ case Some(v) => Literal.create(v, dataType)
+ case None => Literal.default(NullType)
+ }
+ }
+ Option(rewrittenExpr.eval())
+ }
+
+ /**
+ * Statically evaluate an expression containing one or more aggregates
on an empty input.
+ */
+ private def evalAggOnZeroTups(expr : Expression) : Option[Any] = {
+ // AggregateExpressions are Unevaluable, so we need to replace all
aggregates
+ // in the expression with the value they would return for zero input
tuples.
+ val rewrittenExpr = expr transform {
+ case a @ AggregateExpression(aggFunc, _, _, resultId) =>
+ aggFunc.defaultResult.getOrElse(Literal.default(NullType))
+ }
+ Option(rewrittenExpr.eval())
+ }
+
+ /**
+ * Statically evaluate a scalar subquery on an empty input.
+ *
+ * <b>WARNING:</b> This method only covers subqueries that pass the
checks under
+ * [[org.apache.spark.sql.catalyst.analysis.CheckAnalysis]]. If the
checks in
+ * CheckAnalysis become less restrictive, this method will need to
change.
+ */
+ private def evalSubqueryOnZeroTups(plan: LogicalPlan) : Option[Any] = {
+ // Inputs to this method will start with a chain of zero or more
SubqueryAlias
+ // and Project operators, followed by an optional Filter, followed by
an
+ // Aggregate. Traverse the operators recursively.
+ def evalPlan(lp : LogicalPlan) : Map[Long, Option[Any]] = {
+ lp match {
+ case SubqueryAlias(_, child) => evalPlan(child)
+ case Filter(condition, child) =>
+ val bindings = evalPlan(child)
+ if (bindings.size == 0) bindings
+ else {
+ val exprResult = evalExpr(condition, bindings).getOrElse(false)
+ .asInstanceOf[Boolean]
+ if (exprResult) bindings else Map()
+ }
+
+ case Project(projectList, child) =>
+ val bindings = evalPlan(child)
+ if (bindings.size == 0) {
+ bindings
+ } else {
+ projectList.map(ne => (ne.exprId.id, evalExpr(ne,
bindings))).toMap
+ }
+
+ case Aggregate(_, aggExprs, _) =>
+ // Some of the expressions under the Aggregate node are the join
columns
+ // for joining with the outer query block. Fill those
expressions in with
+ // nulls and statically evaluate the remainder.
+ aggExprs.map(ne => ne match {
+ case AttributeReference(_, _, _, _) => (ne.exprId.id, None)
--- End diff --
The join attributes that the Analyzer adds to the Aggregate node are
AttributeReference nodes, and the `evalAggOnZeroTups` method as currently
written can't process them. A more comprehensive facility for statically
evaluating expressions would be nice to have; but I'm hesitant to add such a
mechanism as part of a bug fix. Perhaps a follow-on JIRA is in order?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
