dtenedor commented on code in PR #48004:
URL: https://github.com/apache/spark/pull/48004#discussion_r1761268223
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/randomExpressions.scala:
##########
@@ -181,3 +188,237 @@ case class Randn(child: Expression, hideSeed: Boolean =
false) extends RDG {
object Randn {
def apply(seed: Long): Randn = Randn(Literal(seed, LongType))
}
+
+@ExpressionDescription(
+ usage = """
+ _FUNC_(min, max[, seed]) - Returns a random value with independent and
identically
+ distributed (i.i.d.) values with the specified range of numbers. The
random seed is optional.
+ The provided numbers specifying the minimum and maximum values of the
range must be constant.
+ If both of these numbers are integers, then the result will also be an
integer. Otherwise if
+ one or both of these are floating-point numbers, then the result will
also be a floating-point
+ number.
+ """,
+ examples = """
+ Examples:
+ > SELECT _FUNC_(10, 20, 0) > 0 AS result;
+ true
+ """,
+ since = "4.0.0",
+ group = "math_funcs")
+case class Uniform(min: Expression, max: Expression, seedExpression:
Expression)
+ extends RuntimeReplaceable with TernaryLike[Expression] with RDG {
+ def this(min: Expression, max: Expression) = this(min, max, UnresolvedSeed)
+
+ final override lazy val deterministic: Boolean = false
+ override val nodePatterns: Seq[TreePattern] =
+ Seq(RUNTIME_REPLACEABLE, EXPRESSION_WITH_RANDOM_SEED)
+
+ override val dataType: DataType = {
+ val first = min.dataType
+ val second = max.dataType
+ (min.dataType, max.dataType) match {
+ case _ if !seedExpression.resolved || seedExpression.dataType ==
NullType =>
+ NullType
+ case (_, NullType) | (NullType, _) => NullType
+ case (_, LongType) | (LongType, _)
+ if Seq(first, second).forall(integer) => LongType
+ case (_, IntegerType) | (IntegerType, _)
+ if Seq(first, second).forall(integer) => IntegerType
+ case (_, ShortType) | (ShortType, _)
+ if Seq(first, second).forall(integer) => ShortType
+ case (_, DoubleType) | (DoubleType, _) => DoubleType
+ case (_, FloatType) | (FloatType, _) => FloatType
+ case _ =>
+ throw SparkException.internalError(
+ s"Unexpected argument data types: ${min.dataType}, ${max.dataType}")
+ }
+ }
+
+ private def integer(t: DataType): Boolean = t match {
+ case _: ShortType | _: IntegerType | _: LongType => true
+ case _ => false
+ }
+
+ override def checkInputDataTypes(): TypeCheckResult = {
+ var result: TypeCheckResult = TypeCheckResult.TypeCheckSuccess
+ def requiredType = "integer or floating-point"
+ Seq((min, "min", 0),
+ (max, "max", 1),
+ (seedExpression, "seed", 2)).foreach {
+ case (expr: Expression, name: String, index: Int) =>
+ if (result == TypeCheckResult.TypeCheckSuccess) {
+ if (!expr.foldable) {
+ result = DataTypeMismatch(
+ errorSubClass = "NON_FOLDABLE_INPUT",
+ messageParameters = Map(
+ "inputName" -> name,
+ "inputType" -> requiredType,
+ "inputExpr" -> toSQLExpr(expr)))
+ } else expr.dataType match {
+ case _: ShortType | _: IntegerType | _: LongType | _: FloatType |
_: DoubleType |
+ _: NullType =>
+ case _ =>
+ result = DataTypeMismatch(
+ errorSubClass = "UNEXPECTED_INPUT_TYPE",
+ messageParameters = Map(
+ "paramIndex" -> ordinalNumber(index),
+ "requiredType" -> requiredType,
+ "inputSql" -> toSQLExpr(expr),
+ "inputType" -> toSQLType(expr.dataType)))
+ }
+ }
+ }
+ result
+ }
+
+ override def first: Expression = min
+ override def second: Expression = max
+ override def third: Expression = seedExpression
+
+ override def withNewSeed(newSeed: Long): Expression =
+ Uniform(min, max, Literal(newSeed, LongType))
+
+ override def withNewChildrenInternal(
+ newFirst: Expression, newSecond: Expression, newThird: Expression):
Expression =
+ Uniform(newFirst, newSecond, newThird)
+
+ override def replacement: Expression = {
+ if (Seq(min, max, seedExpression).exists(_.dataType == NullType)) {
+ Literal(null)
+ } else {
+ def cast(e: Expression, to: DataType): Expression = if (e.dataType ==
to) e else Cast(e, to)
+ cast(Add(
+ cast(min, DoubleType),
+ Multiply(
+ Subtract(
+ cast(max, DoubleType),
+ cast(min, DoubleType)),
+ Rand(seed))),
+ dataType)
+ }
+ }
+}
+
+@ExpressionDescription(
+ usage = """
+ _FUNC_(length[, seed]) - Returns a string of the specified length whose
characters are chosen
+ uniformly at random from the following pool of characters: 0-9, a-z,
A-Z. The random seed is
+ optional. The string length must be a constant two-byte or four-byte
integer (SMALLINT or INT,
+ respectively).
+ """,
+ examples =
+ """
+ Examples:
+ > SELECT _FUNC_(3, 0) AS result;
+ 8i7
+ """,
+ since = "4.0.0",
+ group = "string_funcs")
+case class RandStr(length: Expression, override val seedExpression: Expression)
+ extends ExpressionWithRandomSeed with BinaryLike[Expression] with
Nondeterministic {
+ def this(length: Expression) = this(length, UnresolvedSeed)
+
+ override def nullable: Boolean = false
+ override def dataType: DataType = StringType
+ override def stateful: Boolean = true
+ override def left: Expression = length
+ override def right: Expression = seedExpression
+
+ /**
+ * Record ID within each partition. By being transient, the Random Number
Generator is
+ * reset every time we serialize and deserialize and initialize it.
+ */
+ @transient protected var rng: XORShiftRandom = _
+
+ @transient protected lazy val seed: Long = seedExpression match {
+ case e if e.dataType == IntegerType => e.eval().asInstanceOf[Int]
+ case e if e.dataType == LongType => e.eval().asInstanceOf[Long]
+ }
+ override protected def initializeInternal(partitionIndex: Int): Unit = {
+ rng = new XORShiftRandom(seed + partitionIndex)
+ }
+
+ override def withNewSeed(newSeed: Long): Expression = RandStr(length,
Literal(newSeed, LongType))
+ override def withNewChildrenInternal(newFirst: Expression, newSecond:
Expression): Expression =
+ RandStr(newFirst, newSecond)
+
+ override def checkInputDataTypes(): TypeCheckResult = {
+ var result: TypeCheckResult = TypeCheckResult.TypeCheckSuccess
+ def requiredType = "INT or SMALLINT"
+ Seq((length, "length", 0),
+ (seedExpression, "seedExpression", 1)).foreach {
+ case (expr: Expression, name: String, index: Int) =>
+ if (result == TypeCheckResult.TypeCheckSuccess) {
+ if (!expr.foldable) {
+ result = DataTypeMismatch(
+ errorSubClass = "NON_FOLDABLE_INPUT",
+ messageParameters = Map(
+ "inputName" -> name,
+ "inputType" -> requiredType,
+ "inputExpr" -> toSQLExpr(expr)))
+ } else expr.dataType match {
+ case _: ShortType | _: IntegerType =>
+ case _: LongType if index == 1 =>
+ case _ =>
+ result = DataTypeMismatch(
+ errorSubClass = "UNEXPECTED_INPUT_TYPE",
+ messageParameters = Map(
+ "paramIndex" -> ordinalNumber(index),
+ "requiredType" -> requiredType,
+ "inputSql" -> toSQLExpr(expr),
+ "inputType" -> toSQLType(expr.dataType)))
+ }
+ }
+ }
+ result
+ }
+
+ override def evalInternal(input: InternalRow): Any = {
+ val numChars: Int = length.eval(input) match {
+ case i: Int => i
+ case n: Long => n.toInt
+ case s: Short => s.toInt
+ }
+ val bytes = new Array[Byte](numChars)
+ (0 until numChars).foreach { i =>
+ val num = (rng.nextInt() % 30).abs
+ num match {
+ case _ if num < 10 =>
+ bytes.update(i, ('0' + num).toByte)
+ case _ if num < 20 =>
+ bytes.update(i, ('a' + num - 10).toByte)
Review Comment:
You are correct, this was a bug in the algorithm. The random number range
should actually be between 0 and 61, inclusive. Then the range [0, 9]
corresponds to digits 0-9, [10-35] corresponds to characters a-z, and [36-61]
corresponds to A-Z. I updated the algorithm to fix it and added a comment here.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
