Skip to content
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.

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

[SPARK-35207][SQL] Normalize hash function behavior with negative zero (floating point types) #32496

Closed
Closed
Show file tree
Hide file tree
Changes from 2 commits
Commits
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
Original file line number Diff line number Diff line change
Expand Up @@ -369,11 +369,25 @@ abstract class HashExpression[E] extends Expression {
protected def genHashBoolean(input: String, result: String): String =
genHashInt(s"$input ? 1 : 0", result)

protected def genHashFloat(input: String, result: String): String =
genHashInt(s"Float.floatToIntBits($input)", result)
protected def genHashFloat(input: String, result: String): String = {
s"""
|if(Float.floatToIntBits($input) == Float.floatToIntBits(-0.0f)) {
Copy link
Member

Choose a reason for hiding this comment

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

Why do we need to use floatToIntBits here? $input == -0.0f instead?

Copy link
Contributor

@cloud-fan cloud-fan May 13, 2021

Choose a reason for hiding this comment

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

+1, $input == 0.0f should be good enough.

| ${genHashInt(s"Float.floatToIntBits(0.0f)", result)}
Copy link
Member

Choose a reason for hiding this comment

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

Although this has the semantic, shall we use simply "0" instead of s"Float.floatToIntBits(0.0f)"? We may add some comment for the semantic instead.

jshell> Float.floatToIntBits(0.0f)
$1 ==> 0

Copy link
Contributor Author

Choose a reason for hiding this comment

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

You are right, I have tested it to be sure.
Murmur3HashFunction.hashInt(java.lang.Float.floatToIntBits(0.0f), 42) == Murmur3HashFunction.hashInt(0, 42)
It is simpler.

|} else {
| ${genHashInt(s"Float.floatToIntBits($input)", result)}
|}
""".stripMargin
}

protected def genHashDouble(input: String, result: String): String =
genHashLong(s"Double.doubleToLongBits($input)", result)
protected def genHashDouble(input: String, result: String): String = {
s"""
|if(Double.doubleToLongBits($input) == Double.doubleToLongBits(-0.0d)) {
Copy link
Member

Choose a reason for hiding this comment

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

ditto

| ${genHashLong(s"Double.doubleToLongBits(0.0d)", result)}
Copy link
Member

Choose a reason for hiding this comment

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

ditto. We had better use the simplest constant here instead of s"Double.doubleToLongBits(0.0d)".

Copy link
Member

Choose a reason for hiding this comment

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

In this case, 0L?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

The same as the previous point, thanks

|} else {
| ${genHashLong(s"Double.doubleToLongBits($input)", result)}
|}
""".stripMargin
}

protected def genHashDecimal(
ctx: CodegenContext,
Expand Down Expand Up @@ -523,7 +537,9 @@ abstract class InterpretedHashFunction {
case s: Short => hashInt(s, seed)
case i: Int => hashInt(i, seed)
case l: Long => hashLong(l, seed)
case f: Float if (f == -0.0f) => hashInt(java.lang.Float.floatToIntBits(0.0f), seed)
case f: Float => hashInt(java.lang.Float.floatToIntBits(f), seed)
case d: Double if (d == -0.0d) => hashLong(java.lang.Double.doubleToLongBits(0.0d), seed)
case d: Double => hashLong(java.lang.Double.doubleToLongBits(d), seed)
case d: Decimal =>
val precision = dataType.asInstanceOf[DecimalType].precision
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -708,6 +708,16 @@ class HashExpressionsSuite extends SparkFunSuite with ExpressionEvalHelper {
checkEvaluation(HiveHash(Seq(yearMonth)), 1234)
}

test("SPARK-35207: Compute hash consistent between -0.0 and 0.0") {
def checkResult(exprs1: Expression, exprs2: Expression): Unit = {
assert(Murmur3Hash(Seq(exprs1), 42).eval() == Murmur3Hash(Seq(exprs2), 42).eval())
assert(XxHash64(Seq(exprs1), 42).eval() == XxHash64(Seq(exprs2), 42).eval())
assert(HiveHash(Seq(exprs1)).eval() == HiveHash(Seq(exprs2)).eval())
}
checkResult(Literal.create(0D, DoubleType), Literal.create(-0D, DoubleType))
Copy link
Member

Choose a reason for hiding this comment

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

Please use checkEvaluation instead.

Copy link
Member

Choose a reason for hiding this comment

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

Could you add float tests here, too?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Oh! thanks, I pretend to put float instead long

checkResult(Literal.create(0L, LongType), Literal.create(-0L, LongType))
}

private def testHash(inputSchema: StructType): Unit = {
val inputGenerator = RandomDataGenerator.forType(inputSchema, nullable = false).get
val toRow = RowEncoder(inputSchema).createSerializer()
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -654,4 +654,30 @@ class WholeStageCodegenSuite extends QueryTest with SharedSparkSession
}
}
}

test("SPARK-35207: Compute hash consistent between -0.0 and 0.0 doubles with Codegen") {
Copy link
Member

Choose a reason for hiding this comment

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

I think we don't need to add tests here (It's okay just to add tests in HashExprSuite.

Copy link
Contributor

Choose a reason for hiding this comment

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

+1, if you use checkEvaluation, both codegen and interpreted are checked.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

very useful function!

val data = Seq((0.0d, -1.0d, 1.0d))
withTempPath { dir =>
val path = dir.getCanonicalPath
data.toDF("col1", "col2", "col3").write.parquet(path)
sql(s"create table testHash(col1 double, col2 double, col3 double) " +
s"using parquet location '$path'")
sql("select hash(col1 / col2) == hash(col1 / col3) from testHash").collect()
.foreach(row => assert(row.getBoolean(0) == true))
sql("drop table testHash")
}
}

test("SPARK-35207: Compute hash consistent between -0.0 and 0.0 floats with Codegen") {
val data = Seq((0.0f, -1.0f, 1.0f))
withTempPath { dir =>
val path = dir.getCanonicalPath
data.toDF("col1", "col2", "col3").write.parquet(path)
sql(s"create table testHash(col1 float, col2 float, col3 float) " +
s"using parquet location '$path'")
sql("select hash(col1 / col2) == hash(col1 / col3) from testHash").collect()
.foreach(row => assert(row.getBoolean(0) == true))
sql("drop table testHash")
}
}
}