Skip to content

[SPARK-32258][SQL] NormalizeFloatingNumbers directly normalizes IF/CaseWhen/Coalesce child expressions #29061

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
viirya wants to merge 5 commits into from
Closed

[SPARK-32258][SQL] NormalizeFloatingNumbers directly normalizes IF/CaseWhen/Coalesce child expressions #29061

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
f1652aa
NormalizeFloatingNumbers can directly normalize on certain children
viirya Jul 9, 2020
ba0ea32
Change test name.
viirya Jul 9, 2020
5046337
Remove unused import and add jira number.
viirya Jul 10, 2020
78b8667
Address comment.
viirya Jul 12, 2020
d5dce7c
Remove unused code.
viirya Jul 12, 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
11 changes: 10 additions & 1 deletion ...yst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/NormalizeFloatingNumbers.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,7 @@

package org.apache.spark.sql.catalyst.optimizer

import org.apache.spark.sql.catalyst.expressions.{Alias, And, ArrayTransform, CreateArray, CreateMap, CreateNamedStruct, CreateStruct, EqualTo, ExpectsInputTypes, Expression, GetStructField, If, IsNull, KnownFloatingPointNormalized, LambdaFunction, Literal, NamedLambdaVariable, UnaryExpression}
import org.apache.spark.sql.catalyst.expressions.{Alias, And, ArrayTransform, CaseWhen, Coalesce, CreateArray, CreateMap, CreateNamedStruct, CreateStruct, EqualTo, ExpectsInputTypes, Expression, GetStructField, If, IsNull, KnownFloatingPointNormalized, LambdaFunction, Literal, NamedLambdaVariable, UnaryExpression}
import org.apache.spark.sql.catalyst.expressions.codegen.{CodegenContext, ExprCode}
import org.apache.spark.sql.catalyst.planning.ExtractEquiJoinKeys
import org.apache.spark.sql.catalyst.plans.logical.{LogicalPlan, Subquery, Window}
Expand Down Expand Up @@ -116,6 +116,15 @@ object NormalizeFloatingNumbers extends Rule[LogicalPlan] {
case CreateMap(children, useStringTypeWhenEmpty) =>
CreateMap(children.map(normalize), useStringTypeWhenEmpty)

case If(cond, trueValue, falseValue) =>
If(cond, normalize(trueValue), normalize(falseValue))

case CaseWhen(branches, elseVale) =>
CaseWhen(branches.map(br => (br._1, normalize(br._2))), elseVale.map(normalize))

case Coalesce(children) =>
Coalesce(children.map(normalize))

case _ if expr.dataType == FloatType || expr.dataType == DoubleType =>
Copy link
Contributor

@cloud-fan cloud-fan Jul 13, 2020

Choose a reason for hiding this comment

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

Shall we put these new cases after this case? The main goal of this optimization is to avoid constructing a new CreateStruct during normalization. If it's just a float/double type If/CashWhen/Coalesce, it's actually an overhead to duplicate the normalization work in each child.

Copy link
Member Author

@viirya viirya Jul 13, 2020

Choose a reason for hiding this comment

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

Okay.

KnownFloatingPointNormalized(NormalizeNaNAndZero(expr))

Expand Down
41 changes: 38 additions & 3 deletions ...st/scala/org/apache/spark/sql/catalyst/optimizer/NormalizeFloatingPointNumbersSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@ package org.apache.spark.sql.catalyst.optimizer

import org.apache.spark.sql.catalyst.dsl.expressions._
import org.apache.spark.sql.catalyst.dsl.plans._
import org.apache.spark.sql.catalyst.expressions.{And, IsNull, KnownFloatingPointNormalized}
import org.apache.spark.sql.catalyst.expressions.{CaseWhen, If, IsNull, KnownFloatingPointNormalized}
import org.apache.spark.sql.catalyst.plans.PlanTest
import org.apache.spark.sql.catalyst.plans.logical._
import org.apache.spark.sql.catalyst.rules.RuleExecutor
Expand Down Expand Up @@ -85,8 +85,43 @@ class NormalizeFloatingPointNumbersSuite extends PlanTest {
val optimized = Optimize.execute(query)
val doubleOptimized = Optimize.execute(optimized)
val joinCond = IsNull(a) === IsNull(b) &&
KnownFloatingPointNormalized(NormalizeNaNAndZero(coalesce(a, 0.0))) ===
KnownFloatingPointNormalized(NormalizeNaNAndZero(coalesce(b, 0.0)))
coalesce(KnownFloatingPointNormalized(NormalizeNaNAndZero(a)),
KnownFloatingPointNormalized(NormalizeNaNAndZero(0.0))) ===
coalesce(KnownFloatingPointNormalized(NormalizeNaNAndZero(b)),
KnownFloatingPointNormalized(NormalizeNaNAndZero(0.0)))
val correctAnswer = testRelation1.join(testRelation2, condition = Some(joinCond))

comparePlans(doubleOptimized, correctAnswer)
}

test("SPARK-32258: normalize the children of If") {
val cond = If(a > 0.1D, a, a + 0.2D) === b
val query = testRelation1.join(testRelation2, condition = Some(cond))
val optimized = Optimize.execute(query)
val doubleOptimized = Optimize.execute(optimized)

val joinCond = If(a > 0.1D,
KnownFloatingPointNormalized(NormalizeNaNAndZero(a)),
KnownFloatingPointNormalized(NormalizeNaNAndZero(a + 0.2D))) ===
KnownFloatingPointNormalized(NormalizeNaNAndZero(b))
val correctAnswer = testRelation1.join(testRelation2, condition = Some(joinCond))

comparePlans(doubleOptimized, correctAnswer)
}

test("SPARK-32258: normalize the children of CaseWhen") {
val cond = CaseWhen(
Seq((a > 0.1D, a), (a > 0.2D, a + 0.2D)),
Some(a + 0.3D)) === b
val query = testRelation1.join(testRelation2, condition = Some(cond))
val optimized = Optimize.execute(query)
val doubleOptimized = Optimize.execute(optimized)

val joinCond = CaseWhen(
Seq((a > 0.1D, KnownFloatingPointNormalized(NormalizeNaNAndZero(a))),
(a > 0.2D, KnownFloatingPointNormalized(NormalizeNaNAndZero(a + 0.2D)))),
Some(KnownFloatingPointNormalized(NormalizeNaNAndZero(a + 0.3D)))) ===
KnownFloatingPointNormalized(NormalizeNaNAndZero(b))
val correctAnswer = testRelation1.join(testRelation2, condition = Some(joinCond))

comparePlans(doubleOptimized, correctAnswer)
Expand Down