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-8784] [SQL] Add Python API for hex and unhex #7181

Closed
wants to merge 6 commits into from
Closed
Show file tree
Hide file tree
Changes from 1 commit
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
28 changes: 28 additions & 0 deletions python/pyspark/sql/functions.py
Original file line number Diff line number Diff line change
Expand Up @@ -381,6 +381,34 @@ def randn(seed=None):
return Column(jc)


@ignore_unicode_prefix
@since(1.5)
def hex(col):
"""Computes hex value of the given column, which could be StringType,
BinaryType, IntegerType or LongType.

>>> sqlContext.createDataFrame([('ABC', 3)], ['a', 'b']).select(hex('a'), hex('b')).collect()
[Row(hex(a)=u'414243', hex(b)=u'3')]
"""
sc = SparkContext._active_spark_context
jc = sc._jvm.functions.hex(_to_java_column(col))
return Column(jc)


@ignore_unicode_prefix
@since(1.5)
def unhex(col):
"""Inverse of hex. Interprets each pair of characters as a hexadecimal number
and converts to the byte representation of number.

>>> sqlContext.createDataFrame([('414243',)], ['a']).select(unhex('a')).collect()
[Row(unhex(a)=u'ABC')]
"""
sc = SparkContext._active_spark_context
jc = sc._jvm.functions.unhex(_to_java_column(col))
return Column(jc)


@ignore_unicode_prefix
@since(1.5)
def sha1(col):
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -157,7 +157,7 @@ object FunctionRegistry {
expression[Substring]("substr"),
expression[Substring]("substring"),
expression[Upper]("ucase"),
expression[UnHex]("unhex"),
expression[Unhex]("unhex"),
expression[Upper]("upper")
)

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -259,30 +259,22 @@ case class Hex(child: Expression) extends UnaryExpression with Serializable {
case LongType => hex(num.asInstanceOf[Long])
case IntegerType => hex(num.asInstanceOf[Integer].toLong)
case BinaryType => hex(num.asInstanceOf[Array[Byte]])
case StringType => hex(num.asInstanceOf[UTF8String])
case StringType => hex(num.asInstanceOf[UTF8String].getBytes)
}
}
}

/**
* Converts every character in s to two hex digits.
*/
private def hex(str: UTF8String): UTF8String = {
hex(str.getBytes)
}
private[this] val hexDigits = Array[Char](
'0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E', 'F'
).map(_.toByte)

private def hex(bytes: Array[Byte]): UTF8String = {
doHex(bytes, bytes.length)
}

private def doHex(bytes: Array[Byte], length: Int): UTF8String = {
private[this] def hex(bytes: Array[Byte]): UTF8String = {
val length = bytes.length
val value = new Array[Byte](length * 2)
var i = 0
while (i < length) {
value(i * 2) = Character.toUpperCase(Character.forDigit(
(bytes(i) & 0xF0) >>> 4, 16)).toByte
value(i * 2 + 1) = Character.toUpperCase(Character.forDigit(
bytes(i) & 0x0F, 16)).toByte
value(i * 2) = hexDigits((bytes(i) & 0xF0) >> 4)
value(i * 2 + 1) = hexDigits((bytes(i) & 0x0F))
i += 1
}
UTF8String.fromBytes(value)
Expand All @@ -303,6 +295,66 @@ case class Hex(child: Expression) extends UnaryExpression with Serializable {
}
}

/**
* Performs the inverse operation of HEX.
* Resulting characters are returned as a byte array.
*/
case class Unhex(child: Expression)
extends UnaryExpression with AutoCastInputTypes with Serializable {
Copy link
Contributor

Choose a reason for hiding this comment

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

can you use ExpectsInputTypes here? since I will remove AutoCastInputTypes.


override def nullable: Boolean = true
override def dataType: DataType = BinaryType
override def inputTypes: Seq[DataType] = Seq(BinaryType)

override def eval(input: InternalRow): Any = {
val num = child.eval(input)
if (num == null) {
null
} else {
unhex(num.asInstanceOf[UTF8String].getBytes)
}
}

// lookup table to translate '0' -> 0 ... 'F'/'f' -> 15
private[this] val unhexDigits = {
Copy link
Contributor

Choose a reason for hiding this comment

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

can we move the two tables into some static field in an object?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

We can, but putting them here is more clear. Only one object per Expression, I think it's fine.

Copy link
Contributor

Choose a reason for hiding this comment

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

why would it be more clear? are you thinking about the distance between its definition and where it is used?

this is one case where java beats scala with static fields. Ideally all the string functions should just be member functions in UTF8String.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Less code is better than more code. From performance's point of view, there is never an end to stop optimize it. I think we could go with something that's good enough (won't be the bottle neck).

Copy link
Contributor

Choose a reason for hiding this comment

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

how is this less code? It is just a bad idea to create unnecessary state. Just move both tables into two fields in Hex object.

Copy link
Contributor

Choose a reason for hiding this comment

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

BTW I absolutely disagree that "less code is better than more code" as an ethos. While it can be true in many cases, there are plenty of counter examples:

  • putting everything into a single file without any namespacing reduces the import statements, but that creates bad logical structures
  • not writing any test cases reduces the amount of code also, but that's obviously bad
  • using arcane scala features can substantially reduce code size in certain cases, at the expense of readability

In this case, I don't see how this creates less code (it takes 2 lines of code to define a scala object -- you can even put it in an existing java class like in UTF8String as a static field).

val array = Array.fill[Byte](128)(-1)
(0 to 9).foreach(i => array('0' + i) = i.toByte)
(0 to 5).foreach(i => array('A' + i) = (i + 10).toByte)
(0 to 5).foreach(i => array('a' + i) = (i + 10).toByte)
array
}

private[this] def unhex(bytes: Array[Byte]): Array[Byte] = {
val out = new Array[Byte]((bytes.length + 1) >> 1)
var i = 0
if ((bytes.length & 0x01) != 0) {
// padding with '0'
if (bytes(0) < 0) {
return null
}
val v = unhexDigits(bytes(0))
if (v == -1) {
return null
}
out(0) = v
i += 1
}
// two characters form the hex value.
while (i < bytes.length) {
if (bytes(i) < 0 || bytes(i + 1) < 0) {
Copy link
Contributor

Choose a reason for hiding this comment

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

There would be exception on my previous logic when facing non-ascii character. Thanks for fixing this.

return null
}
val first = unhexDigits(bytes(i))
val second = unhexDigits(bytes(i + 1))
if (first == -1 || second == -1) {
return null
}
out(i / 2) = (((first << 4) | second) & 0xFF).toByte
i += 2
}
out
}
}

////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////////////////////////////////////////////////////////////
Expand Down Expand Up @@ -351,58 +403,6 @@ case class Pow(left: Expression, right: Expression)
}
}

/**
* Performs the inverse operation of HEX.
* Resulting characters are returned as a byte array.
*/
case class UnHex(child: Expression) extends UnaryExpression with Serializable {

override def dataType: DataType = BinaryType

override def checkInputDataTypes(): TypeCheckResult = {
if (child.dataType.isInstanceOf[StringType] || child.dataType == NullType) {
TypeCheckResult.TypeCheckSuccess
} else {
TypeCheckResult.TypeCheckFailure(s"unHex accepts String type, not ${child.dataType}")
}
}

override def eval(input: InternalRow): Any = {
val num = child.eval(input)
if (num == null) {
null
} else {
unhex(num.asInstanceOf[UTF8String].getBytes)
}
}

private val unhexDigits = {
val array = Array.fill[Byte](128)(-1)
(0 to 9).foreach(i => array('0' + i) = i.toByte)
(0 to 5).foreach(i => array('A' + i) = (i + 10).toByte)
(0 to 5).foreach(i => array('a' + i) = (i + 10).toByte)
array
}

private def unhex(inputBytes: Array[Byte]): Array[Byte] = {
var bytes = inputBytes
if ((bytes.length & 0x01) != 0) {
bytes = '0'.toByte +: bytes
}
val out = new Array[Byte](bytes.length >> 1)
// two characters form the hex value.
var i = 0
while (i < bytes.length) {
val first = unhexDigits(bytes(i))
val second = unhexDigits(bytes(i + 1))
if (first == -1 || second == -1) { return null}
out(i / 2) = (((first << 4) | second) & 0xFF).toByte
i += 2
}
out
}
}

case class Hypot(left: Expression, right: Expression)
extends BinaryMathExpression(math.hypot, "HYPOT")

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -239,9 +239,10 @@ class MathFunctionsSuite extends SparkFunSuite with ExpressionEvalHelper {
}

test("unhex") {
checkEvaluation(UnHex(Literal("737472696E67")), "string".getBytes)
checkEvaluation(UnHex(Literal("")), new Array[Byte](0))
checkEvaluation(UnHex(Literal("0")), Array[Byte](0))
checkEvaluation(Unhex(Literal("737472696E67")), "string".getBytes)
Copy link
Contributor

Choose a reason for hiding this comment

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

add a test for null literal of string type

checkEvaluation(Unhex(Literal("")), new Array[Byte](0))
checkEvaluation(Unhex(Literal("F")), Array[Byte](15))
checkEvaluation(Unhex(Literal("ff")), Array[Byte](-1))
}

test("hypot") {
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -1060,7 +1060,7 @@ object functions {
* @group math_funcs
* @since 1.5.0
*/
def unhex(column: Column): Column = UnHex(column.expr)
def unhex(column: Column): Column = Unhex(column.expr)

/**
* Inverse of hex. Interprets each pair of characters as a hexadecimal number
Expand Down