Skip to content

[SPARK-32868][SQL] Add more order irrelevant aggregates to EliminateSorts #29740

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
tanelk wants to merge 3 commits into from
Closed

[SPARK-32868][SQL] Add more order irrelevant aggregates to EliminateSorts #29740

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
5838716
Mark HyperLogLogPlusPlus and BitAggregate as order irrelevant
tanelk Sep 13, 2020
c2f191e
Mark HyperLogLogPlusPlus and BitAggregate as order irrelevant
tanelk Sep 13, 2020
20323c8
Remove HyperLogLogPlusPlus
tanelk Sep 14, 2020
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
6 changes: 6 additions & 0 deletions sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/dsl/package.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -197,6 +197,12 @@ package object dsl {
Max(e).toAggregateExpression(isDistinct = false, filter = filter)
def maxDistinct(e: Expression, filter: Option[Expression] = None): Expression =
Max(e).toAggregateExpression(isDistinct = true, filter = filter)
def bitAnd(e: Expression, filter: Option[Expression] = None): Expression =
BitAndAgg(e).toAggregateExpression(isDistinct = false, filter = filter)
def bitOr(e: Expression, filter: Option[Expression] = None): Expression =
BitOrAgg(e).toAggregateExpression(isDistinct = false, filter = filter)
def bitXor(e: Expression, filter: Option[Expression] = None): Expression =
BitXorAgg(e).toAggregateExpression(isDistinct = false, filter = filter)
def upper(e: Expression): Expression = Upper(e)
def lower(e: Expression): Expression = Lower(e)
def coalesce(args: Expression*): Expression = Coalesce(args)
Expand Down
2 changes: 1 addition & 1 deletion sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1044,7 +1044,7 @@ object EliminateSorts extends Rule[LogicalPlan] {

private def isOrderIrrelevantAggs(aggs: Seq[NamedExpression]): Boolean = {
def isOrderIrrelevantAggFunction(func: AggregateFunction): Boolean = func match {
case _: Min | _: Max | _: Count => true
case _: Min | _: Max | _: Count | _: BitAggregate | _: HyperLogLogPlusPlus => true

@cloud-fan cloud-fan Sep 14, 2020

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

HyperLogLogPlusPlus is an estimation, will the input order affect the estimation?

@tanelk tanelk Sep 14, 2020

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Internally it uses the hash values. From the hash it calculates the bucket id and the max number of leading zeros for all the inputs for each buckets. All of these are deterministic operations, that do not depend on the order of insertion.

@maropu maropu Sep 14, 2020

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

(IMHO) even if so, we don't have any restriction about the deteminisity of the HyperLogLogPlusPlus impmentation. So, adding it in this rule look a bit dangerous.

@tanelk tanelk Sep 14, 2020

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

I see your point. I'll remove that then, but keep the BitAggregate

// Arithmetic operations for floating-point values are order-sensitive
// (they are not associative).
case _: Sum | _: Average | _: CentralMomentAgg =>
Expand Down
27 changes: 20 additions & 7 deletions ...catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/EliminateSortsSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -197,13 +197,26 @@ class EliminateSortsSuite extends PlanTest {
comparePlans(optimizedThrice, correctAnswerThrice)
}

test("remove orderBy in groupBy clause with count aggs") {
val projectPlan = testRelation.select('a, 'b)
val unnecessaryOrderByPlan = projectPlan.orderBy('a.asc, 'b.desc)
val groupByPlan = unnecessaryOrderByPlan.groupBy('a)(count(1))
val optimized = Optimize.execute(groupByPlan.analyze)
val correctAnswer = projectPlan.groupBy('a)(count(1)).analyze
comparePlans(optimized, correctAnswer)
test("remove orderBy in groupBy clause with order irrelevant aggs") {
Seq(
(e : Expression) => min(e),
(e : Expression) => minDistinct(e),
(e : Expression) => max(e),
(e : Expression) => maxDistinct(e),
(e : Expression) => count(e),
(e : Expression) => countDistinct(e),
(e : Expression) => approxCountDistinct(e),
(e : Expression) => bitAnd(e),
(e : Expression) => bitOr(e),
(e : Expression) => bitXor(e)
).foreach(agg => {
val projectPlan = testRelation.select('a, 'b)
val unnecessaryOrderByPlan = projectPlan.orderBy('a.asc, 'b.desc)
val groupByPlan = unnecessaryOrderByPlan.groupBy('a)(agg('b))
val optimized = Optimize.execute(groupByPlan.analyze)
val correctAnswer = projectPlan.groupBy('a)(agg('b)).analyze
comparePlans(optimized, correctAnswer)
})
}

test("remove orderBy in groupBy clause with sum aggs") {
Expand Down