cloud-fan commented on code in PR #48596:
URL: https://github.com/apache/spark/pull/48596#discussion_r1815031070
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/AnsiTypeCoercion.scala:
##########
@@ -223,47 +205,13 @@ object AnsiTypeCoercion extends TypeCoercionBase {
override def canCast(from: DataType, to: DataType): Boolean =
Cast.canAnsiCast(from, to)
object PromoteStrings extends TypeCoercionRule {
- private def castExpr(expr: Expression, targetType: DataType): Expression =
{
- expr.dataType match {
- case NullType => Literal.create(null, targetType)
- case l if l != targetType => Cast(expr, targetType)
- case _ => expr
- }
- }
override def transform: PartialFunction[Expression, Expression] = {
// Skip nodes who's children have not been resolved yet.
case e if !e.childrenResolved => e
-
- case b @ BinaryOperator(left, right)
- if findWiderTypeForString(left.dataType, right.dataType).isDefined =>
- val promoteType = findWiderTypeForString(left.dataType,
right.dataType).get
- b.withNewChildren(Seq(castExpr(left, promoteType), castExpr(right,
promoteType)))
-
- case Abs(e @ StringTypeExpression(), failOnError) => Abs(Cast(e,
DoubleType), failOnError)
- case m @ UnaryMinus(e @ StringTypeExpression(), _) =>
- m.withNewChildren(Seq(Cast(e, DoubleType)))
- case UnaryPositive(e @ StringTypeExpression()) => UnaryPositive(Cast(e,
DoubleType))
-
- case d @ DateAdd(left @ StringTypeExpression(), _) =>
- d.copy(startDate = Cast(d.startDate, DateType))
- case d @ DateAdd(_, right @ StringTypeExpression()) =>
- d.copy(days = Cast(right, IntegerType))
- case d @ DateSub(left @ StringTypeExpression(), _) =>
- d.copy(startDate = Cast(d.startDate, DateType))
- case d @ DateSub(_, right @ StringTypeExpression()) =>
- d.copy(days = Cast(right, IntegerType))
-
- case s @ SubtractDates(left @ StringTypeExpression(), _, _) =>
- s.copy(left = Cast(s.left, DateType))
- case s @ SubtractDates(_, right @ StringTypeExpression(), _) =>
- s.copy(right = Cast(s.right, DateType))
- case t @ TimeAdd(left @ StringTypeExpression(), _, _) =>
- t.copy(start = Cast(t.start, TimestampType))
- case t @ SubtractTimestamps(left @ StringTypeExpression(), _, _, _) =>
- t.copy(left = Cast(t.left, t.right.dataType))
- case t @ SubtractTimestamps(_, right @ StringTypeExpression(), _, _) =>
- t.copy(right = Cast(right, t.left.dataType))
+ case withChildrenResolved
+ if
AnsiStringPromotionTypeCoercion.apply.isDefinedAt(withChildrenResolved) =>
Review Comment:
The`transform` is used in the base trait `TypeCoercionRule` and
`CombinedTypeCoercionRule`, and no one calls its `isDefinedAt` function.
```
withPropagatedTypes.transformExpressionsUpWithPruning(
AlwaysProcess.fn, ruleId)(typeCoercionFn)
...
override def transform: PartialFunction[Expression, Expression] = {
val transforms = rules.map(_.transform)
Function.unlift { e: Expression =>
val result = transforms.foldLeft(e) {
case (current, transform) => transform.applyOrElse(current,
identity[Expression])
}
if (result ne e) {
Some(result)
} else {
None
}
}
}
```
In the meanwhile, doing pattern match twice might cause performance
regressions.
--
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]
