Github user ron8hu commented on a diff in the pull request:
https://github.com/apache/spark/pull/17415#discussion_r109471156
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/statsEstimation/FilterEstimation.scala
---
@@ -550,6 +565,220 @@ case class FilterEstimation(plan: Filter,
catalystConf: CatalystConf) extends Lo
Some(percent.toDouble)
}
+ /**
+ * Returns a percentage of rows meeting a binary comparison expression
containing two columns.
+ * In SQL queries, we also see predicate expressions involving two
columns
+ * such as "column-1 (op) column-2" where column-1 and column-2 belong
to same table.
+ * Note that, if column-1 and column-2 belong to different tables, then
it is a join
+ * operator's work, NOT a filter operator's work.
+ *
+ * @param op a binary comparison operator, including =, <=>, <, <=, >, >=
+ * @param attrLeft the left Attribute (or a column)
+ * @param attrRight the right Attribute (or a column)
+ * @param update a boolean flag to specify if we need to update
ColumnStat of the given columns
+ * for subsequent conditions
+ * @return an optional double value to show the percentage of rows
meeting a given condition
+ */
+ def evaluateBinaryForTwoColumns(
+ op: BinaryComparison,
+ attrLeft: Attribute,
+ attrRight: Attribute,
+ update: Boolean): Option[Double] = {
+
+ if (!colStatsMap.contains(attrLeft)) {
+ logDebug("[CBO] No statistics for " + attrLeft)
+ return None
+ }
+ if (!colStatsMap.contains(attrRight)) {
+ logDebug("[CBO] No statistics for " + attrRight)
+ return None
+ }
+
+ attrLeft.dataType match {
+ case StringType | BinaryType =>
+ // TODO: It is difficult to support other binary comparisons for
String/Binary
+ // type without min/max and advanced statistics like histogram.
+ logDebug("[CBO] No range comparison statistics for String/Binary
type " + attrLeft)
+ return None
+ case _ =>
+ }
+
+ val colStatLeft = colStatsMap(attrLeft)
+ val statsRangeLeft = Range(colStatLeft.min, colStatLeft.max,
attrLeft.dataType)
+ .asInstanceOf[NumericRange]
+ val maxLeft = BigDecimal(statsRangeLeft.max)
+ val minLeft = BigDecimal(statsRangeLeft.min)
+
+ val colStatRight = colStatsMap(attrRight)
+ val statsRangeRight = Range(colStatRight.min, colStatRight.max,
attrRight.dataType)
+ .asInstanceOf[NumericRange]
+ val maxRight = BigDecimal(statsRangeRight.max)
+ val minRight = BigDecimal(statsRangeRight.min)
+
+ // determine the overlapping degree between predicate range and
column's range
+ val allNotNull = (colStatLeft.nullCount == 0) &&
(colStatRight.nullCount == 0)
+ val (noOverlap: Boolean, completeOverlap: Boolean) = op match {
+ // Left < Right or Left <= Right
+ // - no overlap:
+ // minRight maxRight minLeft maxLeft
+ // --------+------------------+------------+-------------+------->
+ // - complete overlap: (If null values exists, we set it to partial
overlap.)
+ // minLeft maxLeft minRight maxRight
+ // --------+------------------+------------+-------------+------->
+ case _: LessThan =>
+ (minLeft >= maxRight, (maxLeft < minRight) && allNotNull)
+ case _: LessThanOrEqual =>
+ (minLeft > maxRight, (maxLeft <= minRight) && allNotNull)
+
+ // Left > Right or Left >= Right
+ // - no overlap:
+ // minLeft maxLeft minRight maxRight
+ // --------+------------------+------------+-------------+------->
+ // - complete overlap: (If null values exists, we set it to partial
overlap.)
+ // minRight maxRight minLeft maxLeft
+ // --------+------------------+------------+-------------+------->
+ case _: GreaterThan =>
+ (maxLeft <= minRight, (minLeft > maxRight) && allNotNull)
+ case _: GreaterThanOrEqual =>
+ (maxLeft < minRight, (minLeft >= maxRight) && allNotNull)
+
+ // Left = Right or Left <=> Right
+ // - no overlap:
+ // minLeft maxLeft minRight maxRight
+ // --------+------------------+------------+-------------+------->
+ // minRight maxRight minLeft maxLeft
+ // --------+------------------+------------+-------------+------->
+ // - complete overlap:
+ // minLeft maxLeft
+ // minRight maxRight
+ // --------+------------------+------->
+ case _: EqualTo =>
+ ((maxLeft < minRight) || (maxRight < minLeft),
+ (minLeft == minRight) && (maxLeft == maxRight) && allNotNull
--- End diff --
Agreed. We prefer over estimation to under estimation in order to avoid
out-of-memory error.
---
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
