Skip to content

[SPARK-40903][SQL] Avoid reordering decimal Add for canonicalization if data type is changed #38379

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
gengliangwang wants to merge 5 commits into from
Closed

[SPARK-40903][SQL] Avoid reordering decimal Add for canonicalization if data type is changed #38379

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
750b6a5
fix decimal add's canonicalized
gengliangwang Oct 24, 2022
81bb9c5
add comment
gengliangwang Oct 25, 2022
4b0d05b
Update sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala
gengliangwang Oct 25, 2022
3b79685
address comment
gengliangwang Oct 25, 2022
cf49066
address comments
gengliangwang Oct 25, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 7 additions & 1 deletion sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/Expression.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -242,7 +242,13 @@ abstract class Expression extends TreeNode[Expression] {
* This means that the lazy `cannonicalized` is called and computed only on the root of the
* adjacent expressions.
*/
lazy val canonicalized: Expression = {
lazy val canonicalized: Expression = withCanonicalizedChildren

/**
* The default process of canonicalization. It is a one pass, bottum-up expression tree
* computation based oncanonicalizing children before canonicalizing the current node.
*/
final protected def withCanonicalizedChildren: Expression = {
val canonicalizedChildren = children.map(_.canonicalized)
withNewChildren(canonicalizedChildren)
}
Expand Down
5 changes: 4 additions & 1 deletion sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/arithmetic.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -477,7 +477,10 @@ case class Add(
override protected def withNewChildrenInternal(newLeft: Expression, newRight: Expression): Add =
copy(left = newLeft, right = newRight)

override lazy val canonicalized: Expression = {
override lazy val canonicalized: Expression = dataType match {
case _: DecimalType =>
Copy link
Contributor

@cloud-fan cloud-fan Oct 25, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can we add some comments to explain the reason?

Copy link
Contributor

@ulysses-you ulysses-you Oct 25, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

thank you @gengliangwang for the catching.
can we make it more fine-grained ? Not all decimal add will fail, so we can check if we can reorder them safely. e.g., precision and scale in all left and right are same.

Copy link
Member Author

@gengliangwang gengliangwang Oct 25, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@cloud-fan comment added

Copy link
Member Author

@gengliangwang gengliangwang Oct 25, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ulysses-you Are you sure about that? My concern is that if both left and right contains integer contains decimal Adds, the result may still be different after sorting all the sub Adds

Copy link
Contributor

@peter-toth peter-toth Oct 25, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

How about adding an extra Cast into the canonicalized form if needed like:

  override lazy val canonicalized: Expression = {
    // TODO: do not reorder consecutive `Add`s with different `evalMode`
    val reordered = orderCommutative({ case Add(l, r, _) => Seq(l, r) }).reduce(Add(_, _, evalMode))
    if (dataType != reordered.dataType) {
      Cast(reordered, dataType)
    } else {
      reordered
    }
  }

Copy link
Contributor

@peter-toth peter-toth Oct 25, 2022 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hmm, maybe adding an extra Cast is not a good idea as the 2 expressions with different dataTypes shouldn't be considered equal, but if reordered's data type matches the original then why can't we reorder?

Copy link
Member Author

@gengliangwang gengliangwang Oct 25, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@peter-toth The ideal solution would be adding extra casts in all the canonicalization of the children ofComplexTypeMergingExpression if the data type is changed. However, there are also overriding in some of the ComplexTypeMergingExpression.
So I would take your suggestion to reorder the Add if the result data type is not changed. Thank you.

withCanonicalizedChildren
case _ =>
// TODO: do not reorder consecutive `Add`s with different `evalMode`
orderCommutative({ case Add(l, r, _) => Seq(l, r) }).reduce(Add(_, _, evalMode))
}
Expand Down
8 changes: 8 additions & 0 deletions sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4518,6 +4518,14 @@ class SQLQuerySuite extends QueryTest with SharedSparkSession with AdaptiveSpark
}
}
}

test("SPARK-40903: Don't reorder Add for canonicalize if it is decimal type") {
val tableName = "decimalTable"
withTable(tableName) {
sql(s"create table $tableName(a decimal(12, 5), b decimal(12, 6)) using orc")
checkAnswer(sql(s"select sum(coalesce(a+b+ 1.75, a)) from $tableName"), Row(null))
}
}
}

case class Foo(bar: Option[String])