jorgecarleitao commented on a change in pull request #8172:
URL: https://github.com/apache/arrow/pull/8172#discussion_r487403087
##########
File path: rust/datafusion/src/physical_plan/planner.rs
##########
@@ -218,32 +217,18 @@ impl DefaultPhysicalPlanner {
.collect::<Result<Vec<_>>>()?;
let aggregates = aggr_expr
.iter()
- .map(|e| {
- tuple_err((
- self.create_aggregate_expr(e, &input_schema,
ctx_state),
- e.name(&input_schema),
- ))
- })
+ .map(|e| self.create_aggregate_expr(e, &input_schema,
ctx_state))
.collect::<Result<Vec<_>>>()?;
- let initial_aggr = HashAggregateExec::try_new(
+ let initial_aggr = Arc::new(HashAggregateExec::try_new(
AggregateMode::Partial,
groups.clone(),
aggregates.clone(),
input,
- )?;
+ )?);
- if initial_aggr.output_partitioning().partition_count() == 1 {
Review comment:
Sorry. What I meant is that compared to the previous implementation, we
double the amount of work, even when we did not need because there is a single
partition.
IMO we should handle this in another way, e.g. via a physical optimizer that
re-writes the partial aggregation by a final aggregation when the input is a
single partition.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
