[postgres] Complaining about missing alias in CTE

[enote-kane]

Search before asking

• I searched the issues and found no similar issues.

What Happened

Linting errors raised where no errors are expected

Expected Behaviour

No errors.

Observed Behaviour

$ sqlfluff lint --nocolor --verbose --dialect postgres 2022-07-05.cte-column-alias.sql 
==== sqlfluff ====
sqlfluff:                1.1.0 python:                 3.8.10
implementation:        cpython verbosity:                   1
dialect:              postgres templater:               jinja
rules:                                all
==== readout ====

=== [ path: 2022-07-05.cte-column-alias.sql ] ===

== [2022-07-05.cte-column-alias.sql] FAIL                                                                                                                                         
L:   2 | P:  15 | L022 | Blank line expected but not found after CTE closing
                       | bracket.
L:   5 | P:   9 | L013 | Column expression without alias. Use explicit `AS`
                       | clause.
L:   6 | P:   9 | L013 | Column expression without alias. Use explicit `AS`
                       | clause.
==== summary ====
violations:        3 status:         FAIL
All Finished!

While also (still) raising issue #3474, it also complains about the missing alias as soon as one uses a non-column/non-constant expression.

How to reproduce

Use the following 2022-07-05.cte-column-alias.sql file:

WITH
cte (a, b, c) AS (
    SELECT
        column_a,
        min(column_b),
        sum(column_c)
    FROM my_table
    GROUP BY 1
)

SELECT
    a,
    b,
    c
FROM cte;

Dialect

postgres

Version

sqlfluff:                1.1.0 python:                 3.8.10
implementation:        cpython verbosity:                   1
dialect:              postgres templater:               jinja
rules:                                all

Configuration

defaults

Are you willing to work on and submit a PR to address the issue?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct

[barrywhart]

We merged a PR a couple days ago which may have fixed this.

[barrywhart]

This is not a bug. L013 is working as designed: It wants computed columns to have aliases. Why do you think this is a bug?

[tunetheweb]

It could be argued the computer columns have aliases here (a, b, c).

[tunetheweb]

Then again it could also be argued a more explicit alias within the select query itself would be more useful (e.g. it could be used in order by or group by queries).

[barrywhart]

That's true -- a different form of alias. I'm inclined to call this an enhancement rather than a bug.