Kikyou1997 commented on code in PR #17790:
URL: https://github.com/apache/doris/pull/17790#discussion_r1136987551
##########
fe/fe-core/src/main/java/org/apache/doris/nereids/stats/StatsCalculator.java:
##########
@@ -436,12 +437,44 @@ private Statistics computeAggregate(Aggregate<? extends
Plan> aggregate) {
// TODO: since we have no column stats here. just use a fix ratio to
compute the row count.
List<Expression> groupByExpressions =
aggregate.getGroupByExpressions();
Statistics childStats = groupExpression.childStatistics(0);
- Map<Expression, ColumnStatistic> childSlotToColumnStats =
childStats.columnStatistics();
- double resultSetCount = groupByExpressions.stream().flatMap(expr ->
expr.getInputSlots().stream())
-
.filter(childSlotToColumnStats::containsKey).map(childSlotToColumnStats::get).map(s
-> s.ndv)
- .reduce(1d, (a, b) -> a * b);
- if (resultSetCount <= 0) {
- resultSetCount = 1L;
+ double resultSetCount = 1;
+ if (!groupByExpressions.isEmpty()) {
+ Map<Expression, ColumnStatistic> childSlotToColumnStats =
childStats.columnStatistics();
+ double inputRowCount = childStats.getRowCount();
+ if (inputRowCount == 0) {
+ //on empty relation, Agg output 1 tuple
+ resultSetCount = 1;
+ } else {
+ //TODO: add column correlation for agg estimation
+ // group by A1, A2
+ // currently, the estimated row count is between max(ndv_A1,
ndv_A2) and ndv_A1 * ndv_A2
+ resultSetCount = groupByExpressions.stream().flatMap(expr ->
expr.getInputSlots().stream())
+ .map(Slot::getExprId)
+ .filter(childSlotToColumnStats::containsKey)
+ .map(childSlotToColumnStats::get)
+ .map(s -> {
+ double adjustedNdv = s.ndv;
+ if (s.isUnKnown) {
Review Comment:
最好不要这么做，如果本身ndv就是0,最后也会以default计算。我建议在前面用一个单独的循环去做这件事情。
--
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]
