Why does a Postgres recursive CTE need the list of columns as argument?

Question

This is kind of a subjective question.

If I have a recursive CTE something like:

WITH RECURSIVE r(x, y) AS ( ...
    SELECT x, y -- select1
    FROM t WHERE .. 
    UNION
    SELECT t.x, t.y -- select2
    FROM r JOIN t
    ON ...
)
SELECT x, y
FROM r

What exactly is the point of specifying the list of columns as an argument in the CTE definition? If I don't specify it, it anyways takes the list of columns in the SELECT.

If I do specify it, every time I want to make changes, I have to reflect that in both the SELECTs in the CTE and also in the argument list.

What benefit do I gain from writing WITH RECURSIVE r(x, y) AS instead of just WITH RECURSIVE r AS?

Answer 1

What exactly is the point of specifying the list of columns as an argument in the CTE definition?

This comes from the SQL Standard. It's another option to name the columns of the CTE.

If I don't specify it, it anyways takes the list of columns in the SELECT.

That is correct. If you don't specify it, the list of columns in the SELECT is used (and if it is a UNION/INTERSECT/EXCEPT query, the list of columns in the first of them. The 2nd and following lists are ignored).

If I do specify it, every time I want to make changes, I have to reflect that in both the SELECTs in the CTE and also in the argument list.

Incorrect. If you change the names in the "argument list", you don't have to make any change in the SELECT list. The argument list overrides the SELECT list names. You only need to change both if you add more or remove columns from the CTE.