Marti Raudsepp wrote:
> On Fri, Oct 24, 2014 at 11:29 AM, Kyotaro HORIGUCHI
> <horiguchi.kyotaro@lab.ntt.co.jp> wrote:
> > - 0001-ALTER-ROLE-CURRENT_USER_v2.patch - the patch.
>
> +RoleId: CURRENT_USER { $$ = "current_user";}
> + | USER { $$ = "current_user";}
> + | CURRENT_ROLE { $$ = "current_user";}
> + | SESSION_USER { $$ = "session_user";}
>
> This is kind of ugly, and it means you can't distinguish between a
> CURRENT_USER keyword and a quoted user name "current_user". It's a
> legitimate user name, so the behavior of the following now changes:
>
> CREATE ROLE "current_user";
> ALTER ROLE "current_user" SET work_mem='10MB';
>
> There ought to be a better way to represent this than using magic string values.
Agreed. Since the current_user disease has already infected the USER
MAPPING stuff, I think we need to solve that problem -- how about having
this production return a new node which has either a string name or
flags for the various acceptable keywords?
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services