epsio-banay commented on code in PR #12943:
URL: https://github.com/apache/datafusion/pull/12943#discussion_r1801471152
##########
datafusion/optimizer/src/push_down_filter.rs:
##########
@@ -685,7 +685,8 @@ impl OptimizerRule for PushDownFilter {
.map(LogicalPlan::Filter)?;
insert_below(LogicalPlan::Repartition(repartition), new_filter)
}
- LogicalPlan::Distinct(distinct) => {
+ LogicalPlan::Distinct(distinct @ Distinct::All(_)) => {
+ // note that we check for distinct all as distinct on is not
commutable
Review Comment:
You are right.
I think the push down should be allowed only if the predicate contains only
expressions that exist in the on_expr vec or constants. Do you know any tool or
function that can help with that?
Another thing I missed is that the push down should be allowed if the filter
is on the same logical select with the distinct i.e. in this query push down
should be allowed: `select distinct on (a) a, b from foo where b <> 2 order by
a, b desc;`.
I think this optimization bug occurs when you pushdown the filter into a
subquery with distinct, for example:
`select * from (select distinct on (a) a, b from foo order by a, b desc) sub
where b <> 2;`. I need to add tests for that aswell
--
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]
