Add support for select distinct on(a, b, c) ...

The issue in Presto is that on one side, one can’t use

select distinct on (a, b)
  c
from d

but one also cannot use:

select
  c
from d
group by a, b

Combining these two limitations together, makes deduplicating rows a relatively cumbersome process, needing resorting either to subqueries with window functions and retrieving the row number, or to array aggregations. Again, a lot of context to be carried over, a complexity which adds up exponentially as more elements get in, and much more error-prone than either of the cleaner solutions above.

Postgres implementation of select distinct on is very straightforward and even allows for custom sorting, e.g:

select distinct on (a, b)
  c
from d
order by
  e desc,
  f asc

[Israel-Kli]

@JivanRoquet The solution from here work for me on Athena:

SELECT Name, MAX(Address), MAX(other field)...
FROM MyTable
GROUP BY Name

Will give you one row per Name.

https://stackoverflow.com/a/6792357/9225626

[NicolasGuary]

Found a solution from https://redshift-support.matillion.com/s/article/2822021

ROW_NUMBER() OVER ( PARTITION BY <<unique columns>> ORDER BY <<sort columns>>) as counts

And then select where counts=1 only.

Hope this can help