Github user jinfengni commented on a diff in the pull request:
https://github.com/apache/drill/pull/416#discussion_r55457004
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/planner/physical/ConvertCountToDirectScan.java
---
@@ -117,6 +117,24 @@ public void onMatch(RelOptRuleCall call) {
} else if (aggCall.getArgList().size() == 1) {
// count(columnName) ==> Agg ( Scan )) ==> columnValueCount
int index = aggCall.getArgList().get(0);
+
+ if (proj != null) {
+ // project in the middle of Agg and Scan : Only when input of
AggCall is a RexInputRef in Project, we find the index of Scan's field.
+ // For instance,
+ // Agg - count($0)
+ // \
+ // Proj - Exp={$1}
+ // \
+ // Scan (col1, col2).
+ // return count of "col2" in Scan's metadata, if found.
+
+ if (proj.getProjects().get(index) instanceof RexInputRef) {
+ index = ((RexInputRef)
proj.getProjects().get(index)).getIndex();
--- End diff --
I added two more case: count(100) and count(1+2) to check the count
pushdown happen.
Currently, count( 1 + non-nulllable-col) will not pushdown count. That's
because Drill return nullable-ANY for "+" function. With Sean's type exposure
patch, it should get the right type for "+" in planning time, and it will get
count pushddown as well.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
