Github user viirya commented on a diff in the pull request:
https://github.com/apache/spark/pull/8587#discussion_r43395879
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/functions.scala
---
@@ -524,6 +525,133 @@ case class Sum(child: Expression) extends
DeclarativeAggregate {
override val evaluateExpression = Cast(currentSum, resultType)
}
+/**
+ * Compute Pearson correlation between two expressions.
+ * When applied on empty data (i.e., count is zero), it returns NaN.
+ *
+ * Definition of Pearson correlation can be found at
+ *
http://en.wikipedia.org/wiki/Pearson_product-moment_correlation_coefficient
+ *
+ * @param left one of the expressions to compute correlation with.
+ * @param right another expression to compute correlation with.
+ */
+case class Corr(
+ left: Expression,
+ right: Expression,
+ mutableAggBufferOffset: Int = 0,
+ inputAggBufferOffset: Int = 0)
+ extends ImperativeAggregate {
+
+ def children: Seq[Expression] = Seq(left, right)
+
+ def nullable: Boolean = false
+
+ def dataType: DataType = DoubleType
+
+ def inputTypes: Seq[AbstractDataType] = Seq(DoubleType)
+
+ def aggBufferSchema: StructType =
StructType.fromAttributes(aggBufferAttributes)
+
+ def inputAggBufferAttributes: Seq[AttributeReference] =
aggBufferAttributes.map(_.newInstance())
+
+ val aggBufferAttributes: Seq[AttributeReference] = Seq(
+ AttributeReference("xAvg", DoubleType)(),
+ AttributeReference("yAvg", DoubleType)(),
+ AttributeReference("Ck", DoubleType)(),
+ AttributeReference("MkX", DoubleType)(),
+ AttributeReference("MkY", DoubleType)(),
+ AttributeReference("count", LongType)())
+
+ override def withNewMutableAggBufferOffset(newMutableAggBufferOffset:
Int): ImperativeAggregate =
+ copy(mutableAggBufferOffset = newMutableAggBufferOffset)
+
+ override def withNewInputAggBufferOffset(newInputAggBufferOffset: Int):
ImperativeAggregate =
+ copy(inputAggBufferOffset = newInputAggBufferOffset)
+
+ override def initialize(buffer: MutableRow): Unit = {
+ (0 until 5).map(idx => buffer.setDouble(mutableAggBufferOffset + idx,
0.0))
+ buffer.setLong(mutableAggBufferOffset + 5, 0L)
+ }
+
+ override def update(buffer: MutableRow, input: InternalRow): Unit = {
+ val x = left.eval(input).asInstanceOf[Double]
+ val y = right.eval(input).asInstanceOf[Double]
+
+ var xAvg = buffer.getDouble(mutableAggBufferOffset)
+ var yAvg = buffer.getDouble(mutableAggBufferOffset + 1)
+ var Ck = buffer.getDouble(mutableAggBufferOffset + 2)
+ var MkX = buffer.getDouble(mutableAggBufferOffset + 3)
+ var MkY = buffer.getDouble(mutableAggBufferOffset + 4)
+ var count = buffer.getLong(mutableAggBufferOffset + 5)
+
+ val deltaX = x - xAvg
+ val deltaY = y - yAvg
+ count += 1
+ xAvg += deltaX / count
+ yAvg += deltaY / count
+ Ck += deltaX * (y - yAvg)
+ MkX += deltaX * (x - xAvg)
+ MkY += deltaY * (y - yAvg)
+
+ buffer.setDouble(mutableAggBufferOffset, xAvg)
+ buffer.setDouble(mutableAggBufferOffset + 1, yAvg)
+ buffer.setDouble(mutableAggBufferOffset + 2, Ck)
+ buffer.setDouble(mutableAggBufferOffset + 3, MkX)
+ buffer.setDouble(mutableAggBufferOffset + 4, MkY)
+ buffer.setLong(mutableAggBufferOffset + 5, count)
+ }
+
+ // Merge counters from other partitions. Formula can be found at:
+ // http://en.wikipedia.org/wiki/Algorithms_for_calculating_variance
+ override def merge(buffer1: MutableRow, buffer2: InternalRow): Unit = {
+ val count2 = buffer2.getLong(inputAggBufferOffset + 5)
+
+ if (count2 > 0) {
--- End diff --
We only need to consider count in buffer2. I will add document for it.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
