Skip to content

fix: implement lazy evaluation in Coalesce function #2270

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

Open
wants to merge 5 commits into
base: main
Choose a base branch
from
+62 −6
Open

fix: implement lazy evaluation in Coalesce function #2270

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
cd59b4e
lazy_eval_coalesce
coderfender Aug 31, 2025
bf82938
lazy_coalesce_fallback_case_statement
coderfender Aug 31, 2025
739ef0c
lazy_coalesce_fallback_case_statement
coderfender Aug 31, 2025
f6369dd
lazy_coalesce_fallback_case_statement
coderfender Aug 31, 2025
e0cdb86
lazy_coalesce_fallback_case_statement
coderfender Sep 2, 2025
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
7 changes: 2 additions & 5 deletions spark/src/main/scala/org/apache/comet/serde/QueryPlanSerde.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -191,7 +191,8 @@ object QueryPlanSerde extends Logging with CometExprShim {
classOf[Log2] -> CometLog2,
classOf[Pow] -> CometScalarFunction[Pow]("pow"),
classOf[If] -> CometIf,
classOf[CaseWhen] -> CometCaseWhen)
classOf[CaseWhen] -> CometCaseWhen,
classOf[Coalesce] -> CometCoalesce)

/**
* Mapping of Spark aggregate expression class to Comet expression handler.
Expand Down Expand Up @@ -1078,10 +1079,6 @@ object QueryPlanSerde extends Logging with CometExprShim {
None
}

case a @ Coalesce(_) =>
val exprChildren = a.children.map(exprToProtoInternal(_, inputs, binding))
scalarFunctionExprToProto("coalesce", exprChildren: _*)

// With Spark 3.4, CharVarcharCodegenUtils.readSidePadding gets called to pad spaces for
// char types.
// See https://github.com/apache/spark/pull/38151
Expand Down
47 changes: 46 additions & 1 deletion spark/src/main/scala/org/apache/comet/serde/conditional.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@ package org.apache.comet.serde

import scala.collection.JavaConverters._

import org.apache.spark.sql.catalyst.expressions.{Attribute, CaseWhen, Expression, If}
import org.apache.spark.sql.catalyst.expressions.{Attribute, CaseWhen, Coalesce, Expression, If, IsNotNull}

import org.apache.comet.CometSparkSessionExtensions.withInfo
import org.apache.comet.serde.QueryPlanSerde.exprToProtoInternal
Expand Down Expand Up @@ -91,3 +91,48 @@ object CometCaseWhen extends CometExpressionSerde[CaseWhen] {
}
}
}

object CometCoalesce extends CometExpressionSerde[Coalesce] {
override def convert(
expr: Coalesce,
inputs: Seq[Attribute],
binding: Boolean): Option[ExprOuterClass.Expr] = {
val branches = expr.children.dropRight(1).map { child =>
(IsNotNull(child), child)
}
val elseValue = Some(expr.children.last)
var allBranches: Seq[Expression] = Seq()
Copy link
Member

Choose a reason for hiding this comment

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

allBranches seems to be redundant. It ends up being populated with the same contents as branches, but in a different order. It is then only used in a withInfo call when falling back to Spark, so the order there shouldn't matter. I think we could just use branches there instead.

val whenSeq = branches.map(elements => {
allBranches = allBranches :+ elements._1
exprToProtoInternal(elements._1, inputs, binding)
})
val thenSeq = branches.map(elements => {
allBranches = allBranches :+ elements._2
exprToProtoInternal(elements._2, inputs, binding)
})
assert(whenSeq.length == thenSeq.length)
if (whenSeq.forall(_.isDefined) && thenSeq.forall(_.isDefined)) {
val builder = ExprOuterClass.CaseWhen.newBuilder()
builder.addAllWhen(whenSeq.map(_.get).asJava)
builder.addAllThen(thenSeq.map(_.get).asJava)
if (elseValue.isDefined) {
Copy link
Member

Choose a reason for hiding this comment

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

elseValue is explicitly set to Some earlier, so the condition elseValue.isDefined is always true. I don't think that elseValue needs to be an Option?

val elseValueExpr =
exprToProtoInternal(elseValue.get, inputs, binding)
if (elseValueExpr.isDefined) {
builder.setElseExpr(elseValueExpr.get)
} else {
withInfo(expr, elseValue.get)
return None
}
}
Some(
ExprOuterClass.Expr
.newBuilder()
.setCaseWhen(builder)
.build())
} else {
withInfo(expr, allBranches: _*)
None
}
}
}
14 changes: 14 additions & 0 deletions spark/src/test/scala/org/apache/comet/CometExpressionSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -394,6 +394,20 @@ class CometExpressionSuite extends CometTestBase with AdaptiveSparkPlanHelper {
}
}

test("test coalesce lazy eval") {
withSQLConf(
SQLConf.ANSI_ENABLED.key -> "true",
CometConf.COMET_CAST_ALLOW_INCOMPATIBLE.key -> "true") {
val data = Seq((9999999999999L, 0))
withParquetTable(data, "t1") {
val res = spark.sql("""
|SELECT coalesce(_1, CAST(_1 AS TINYINT)) from t1;
| """.stripMargin)
checkSparkAnswerAndOperator(res)
}
}
}

test("dictionary arithmetic") {
// TODO: test ANSI mode
withSQLConf(SQLConf.ANSI_ENABLED.key -> "false", "parquet.enable.dictionary" -> "true") {
Expand Down
Loading