Re: alter user/role CURRENT_USER

Kyotaro HORIGUCHI wrote:

> - Storage for new information
> 
> The new struct NameId stores an identifier which telling what it
> logically is using the new enum NameIdTypes.

I think NameId is a bad name for this.  My point is that NameId, as it
stands, might be a name for anything, not just a role; and the object it
identifies is not an Id either.  Maybe RoleSpec?  Do we need a public_ok
argument to get_nameid_oid() (get a better name for this function too)
so that callers don't have to check for InvalidOid argument?  I think
the arrangement you propose is not very convenient; it'd be best to
avoid duplicating the check for InvalidOid in all callers of the new
function, particularly where there was no check before.

-- 
Álvaro Herrera                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services