Github user wzhfy commented on a diff in the pull request:
https://github.com/apache/spark/pull/19783#discussion_r153972295
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/statsEstimation/FilterEstimation.scala
---
@@ -332,8 +332,45 @@ case class FilterEstimation(plan: Filter) extends
Logging {
colStatsMap.update(attr, newStats)
}
- Some(1.0 / BigDecimal(ndv))
- } else {
+ // We compute filter selectivity using Histogram information
+ attr.dataType match {
+ case StringType | BinaryType =>
+ Some(1.0 / BigDecimal(ndv))
+
+ case _ =>
+ // returns 1/ndv if there is no histogram
+ if (colStat.histogram.isEmpty) return Some(1.0 / BigDecimal(ndv))
+
+ // We traverse histogram bins to locate the literal value
--- End diff --
This comment is not accurate, here we want to get the bins occupied by the
literal value, because if the value is skewed, it can occupy multiple bins.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
