Github user cloud-fan commented on a diff in the pull request:
https://github.com/apache/spark/pull/19783#discussion_r155233734
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/statsEstimation/EstimationUtils.scala
---
@@ -114,4 +114,171 @@ object EstimationUtils {
}
}
+ /**
+ * Returns the number of the first bin into which a column values falls
for a specified
+ * numeric equi-height histogram.
+ *
+ * @param value a literal value of a column
+ * @param bins an array of bins for a given numeric equi-height histogram
+ * @return the number of the first bin into which a column values falls.
+ */
+ def findFirstBinForValue(value: Double, bins: Array[HistogramBin]): Int
= {
+ var i = 0
+ while ((i < bins.length) && (value > bins(i).hi)) {
+ i +=1
+ }
+ i
+ }
+
+ /**
+ * Returns the number of the last bin into which a column values falls
for a specified
+ * numeric equi-height histogram.
+ *
+ * @param value a literal value of a column
+ * @param bins an array of bins for a given numeric equi-height histogram
+ * @return the number of the last bin into which a column values falls.
+ */
+ def findLastBinForValue(value: Double, bins: Array[HistogramBin]): Int =
{
+ var i = bins.length - 1
+ while ((i >= 0) && (value < bins(i).lo)) {
+ i -=1
+ }
+ i
+ }
+
+ /**
+ * Returns a percentage of a bin holding values for column value in the
range of
+ * [lowerValue, higherValue]
+ *
+ * @param binId a given bin id in a specified histogram
+ * @param higherValue a given upper bound value of a specified column
value range
+ * @param lowerValue a given lower bound value of a specified column
value range
+ * @param histogram a numeric equi-height histogram
+ * @return the percentage of a single bin holding values in [lowerValue,
higherValue].
+ */
+ private def getOccupation(
+ binId: Int,
+ higherValue: Double,
+ lowerValue: Double,
+ histogram: Histogram): Double = {
+ val curBin = histogram.bins(binId)
+ if (curBin.hi == curBin.lo) {
+ // the entire bin is covered in the range
+ 1.0
+ } else if (higherValue == lowerValue) {
+ // set percentage to 1/NDV
+ 1.0 / curBin.ndv.toDouble
+ } else {
+ // Use proration since the range falls inside this bin.
+ math.min((higherValue - lowerValue) / (curBin.hi - curBin.lo), 1.0)
--- End diff --
shouldn't check the overlapping percentage? Do I miss some assumption for
this method?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
