[R3gardless]

With remove_pg_catalog_from_functions=True, the schema is kept only for extract, position and xmlexists. The other functions named after a column name keyword lose it as well:

SELECT pg_catalog.coalesce(1, 2)           ->  SELECT coalesce(1, 2)            -- different statement
SELECT pg_catalog.normalize('abc', 'NFC')  ->  SELECT normalize('abc', 'NFC')   -- syntax error
SELECT pg_catalog.int(1)                   ->  SELECT int(1)                    -- syntax error

This replaces the hardcoded list with a check against COL_NAME_KEYWORDS, except overlay and substring, which work unqualified. I added the new cases to the existing test.

[lelit]

Thanks, apparently this is already ok in both v7 (where I applied a backport of your #229) and v8 (after #224). Reopen should that not be the case.

[R3gardless]

Hi @lelit, thanks for checking! If I'm not mistaken, with remove_pg_catalog_from_functions=True these still come out wrong on the current v7 and v8 heads:

SELECT pg_catalog.coalesce(1, 2)          -> SELECT coalesce(1, 2)           (parsed back as a COALESCE expression)
SELECT pg_catalog.normalize('abc', 'NFC') -> SELECT normalize('abc', 'NFC')  (syntax error)
SELECT pg_catalog.int(1)                  -> SELECT int(1)                   (syntax error)

I think it's because _is_pg_catalog_func() still keeps the schema only for extract, position and xmlexists. If I got something wrong, please just ignore this.

[lelit]

Ok, it seems I conflated two different problems, sorry. As the others, I'm going to transplant this on v7 first.