On 06/19/2014 07:18 PM, Tom Lane wrote:
> Jov <amutu@amutu.com> writes:
>> the doc say:
>>> ALTER USER is now an alias for ALTER ROLE<http://www.postgresql.org/docs/devel/static/sql-alterrole.html>
>
>> but alter user lack the following format:
>> ...
>
> If we're going to have a policy that these commands be exactly equivalent,
> it seems like this patch is just sticking a finger into the dike. It does
> nothing to prevent anyone from making the same mistake again in future.
>
> What about collapsing both sets of productions into one, along the lines
> of
>
> role_or_user: ROLE | USER;
>
> AlterRoleSetStmt:
> ALTER role_or_user RoleId opt_in_database SetResetClause
>
> (and similarly to the latter for every existing ALTER ROLE variant).
I thought about suggesting that, and it seems that I should have. I
also thought about suggesting adding GROUP as an alias, too. That's
probably not as good of an idea.
> Because MAPPING is an unreserved keyword, I think that this approach
> might force us to also change ALTER USER MAPPING to ALTER role_or_user
> MAPPING, which is not contemplated by the SQL standard. But hey,
> it would satisfy the principle of least surprise no? Anyway we don't
> have to document that that would work.
That's a small price to pay, so I'm all for accepting it. I agree that
it doesn't need to be documented.
--
Vik
