Thread: Weird...but correct?

It's bizarre how you can comment on columns in composite types!

CREATE TYPE test (a int4, b int4);
COMMENT ON COLUMN test.a IS 'A column';

Seems harmless, but should we allow it?

Chris

> CREATE TYPE test (a int4, b int4);
> COMMENT ON COLUMN test.a IS 'A column';
> 
> Seems harmless, but should we allow it?

Actually, currently it's bad because such comments will not be dumped by 
pg_dump.  Shall I fix pg_dump?

Chris

Christopher Kings-Lynne <chriskl@familyhealth.com.au> writes:
> It's bizarre how you can comment on columns in composite types!
> CREATE TYPE test (a int4, b int4);
> COMMENT ON COLUMN test.a IS 'A column';

And not only that, but:

regression=# \d+ testComposite type "public.test"Column |  Type   | Description
--------+---------+-------------a      | integer | A columnb      | integer |

> Seems harmless, but should we allow it?

Why not?

> Shall I fix pg_dump?

Yes please, if it doesn't seem outlandishly hard.  (From memory it might
just be adding a subroutine call to dumpType...)
        regards, tom lane