Re: Fwd: Proposal: variant of regclass

Andres Freund <andres@2ndquadrant.com> writes:
> There's actually another good reason to not copy regclass's behaviour:

> postgres=# CREATE TABLE "123"();
> CREATE TABLE
> postgres=# SELECT '123'::regclass;
>  regclass
>  ----------
>   123
>   (1 row)

> I don't think that's fixable for ::regclass, but we shouldn't copy it.

I think that's not proving what you thought; the case is correctly handled
if you quote:

regression=# create table "123"(z int);
CREATE TABLE
regression=# select '123'::regclass;regclass 
----------123
(1 row)

regression=# select '"123"'::regclass;regclass 
----------"123"
(1 row)

But I agree that we don't want these functions accepting numeric OIDs,
even though ::regclass must.
        regards, tom lane