I think it's not correct. The input need to be distributed by group key

Similarly, SimpleAgg needs input to be distributed in singoton

What if select has group by but no agg function within select items?

create table t (a char(10), b int, c bool);
select t.a from t group by t.c;

This query is invalid....

postgres=# select t.a from t group by t.c;
ERROR: column "t.a" must appear in the GROUP BY clause or be used in an aggregate function
LINE 1: select t.a from t group by t.c;

Yes, so are we able to prompt error on this sql?

Yes, we can.

dev=> select v1 from t group by v2;
ERROR: Invalid input syntax: column must appear in the GROUP BY clause or be used in an aggregate function

We can add some error cases in the planner test, although they are already covered by unit tests of LogicalAgg::create.

Let's be more compact and follow expr::AggCall:

            let mut builder = f.debug_tuple(&format!("{:?}", self.agg_kind));
            self.inputs.iter().for_each(|child| {
                builder.field(child);
            });
            builder.finish()

nit: groups before aggs

typo: to_stream

The output distribution of HashAgg and SimpleAgg are also by-groups and Single respectively.

Similar to logical, groups before aggs.

We can add some error cases in the planner test, although they are already covered by unit tests of LogicalAgg::create.