Comment by Exuma

At least in PostgreSQL, both by alias and ordinal are possible:

  localhost(from SCB-MUSE-BOXX).postgres.scb.5432 [Sun Nov 16 12:02:15 PST 2025]
  > create table test (a_key integer primary key, a_group integer, a_val numeric);
  CREATE TABLE
  Time: 3.102 ms

  localhost(from SCB-MUSE-BOXX).postgres.scb.5432 [Sun Nov 16 12:02:25 PST 2025]
  > insert into test (a_key, a_group, a_val) values (1, 1, 5.5), (2, 1, 2.6), (3, 2, 1.1), (4, 2, 6.5);
  INSERT 0 4
  Time: 2.302 ms

  localhost(from SCB-MUSE-BOXX).postgres.scb.5432 [Sun Nov 16 12:02:58 PST 2025]
  > select a_group AS my_group, sum(a_val) from test group by my_group;
   my_group | sum
  ----------+-----
          2 | 7.6
          1 | 8.1
  (2 rows)
  
  Time: 4.124 ms
  localhost(from SCB-MUSE-BOXX).postgres.scb.5432 [Sun Nov 16 12:03:15 PST 2025]
  > select a_group AS my_group, sum(a_val) from test group by 1;
   my_group | sum
  ----------+-----
          2 | 7.6
          1 | 8.1
  (2 rows)
  
  Time: 0.360 ms

Some do. It would also be nice to reference by ordinal number similar to order by. Very handy for quick and dirty queries. I can see the issue though that people start to lean on it too much.

The problem with this and similar requests is that it would change the identifier scoping in incompatible ways and therefore potentially break a lot of existing SQL code.

I think it should be not only in GROUP BY, but in every context, e.g., inside expressions in SELECT, WHERE, etc.

PostgreSQL and DuckDB support this, which makes MSSQL feel like a dinosaur in context.