Thread: oddity with ALTER ROLE/USER
I noticed that ALTER ROLE/USER succeeds even when called without any options: postgres=# alter user foo; ALTER ROLE postgres=# alter role foo; ALTER ROLE postgres=# alter group foo; ERROR: syntax error at or near ";" LINE 1: alter group foo; That seems odd, does nothing useful, and is inconsistent with, for example, ALTER GROUP as shown above. Proposed patch attached. Comments/thoughts? Joe -- Crunchy Data - http://crunchydata.com PostgreSQL Support for Secure Enterprises Consulting, Training, & Open Source Development
Attachment
Joe Conway <mail@joeconway.com> writes:
> I noticed that ALTER ROLE/USER succeeds even when called without any
> options:
> postgres=# alter user foo;
> ALTER ROLE
> postgres=# alter role foo;
> ALTER ROLE
> postgres=# alter group foo;
> ERROR: syntax error at or near ";"
> LINE 1: alter group foo;
> That seems odd, does nothing useful, and is inconsistent with, for
> example, ALTER GROUP as shown above.
> Proposed patch attached.
If you want to make it act like alter group, why not make it act
like alter group? That is, the way to fix this is to change the
grammar so that AlterOptRoleList doesn't permit an expansion with
zero list elements.
Having said that, I can't get excited about changing this at all.
Nobody will thank us for it, and someone might complain.
regards, tom lane
On 2/22/19 4:19 PM, Tom Lane wrote: > Joe Conway <mail@joeconway.com> writes: >> I noticed that ALTER ROLE/USER succeeds even when called without any >> options: > >> postgres=# alter user foo; >> ALTER ROLE >> postgres=# alter role foo; >> ALTER ROLE >> postgres=# alter group foo; >> ERROR: syntax error at or near ";" >> LINE 1: alter group foo; > >> That seems odd, does nothing useful, and is inconsistent with, for >> example, ALTER GROUP as shown above. > >> Proposed patch attached. > > If you want to make it act like alter group, why not make it act > like alter group? That is, the way to fix this is to change the > grammar so that AlterOptRoleList doesn't permit an expansion with > zero list elements. I considered that but liked the more specific error message. > Having said that, I can't get excited about changing this at all. > Nobody will thank us for it, and someone might complain. The other route is change the documentation to reflect reality I guess. Joe -- Crunchy Data - http://crunchydata.com PostgreSQL Support for Secure Enterprises Consulting, Training, & Open Source Development
Attachment
Greetings,
* Tom Lane (tgl@sss.pgh.pa.us) wrote:
> Joe Conway <mail@joeconway.com> writes:
> > I noticed that ALTER ROLE/USER succeeds even when called without any
> > options:
>
> > postgres=# alter user foo;
> > ALTER ROLE
> > postgres=# alter role foo;
> > ALTER ROLE
> > postgres=# alter group foo;
> > ERROR: syntax error at or near ";"
> > LINE 1: alter group foo;
>
> > That seems odd, does nothing useful, and is inconsistent with, for
> > example, ALTER GROUP as shown above.
>
> > Proposed patch attached.
>
> If you want to make it act like alter group, why not make it act
> like alter group? That is, the way to fix this is to change the
> grammar so that AlterOptRoleList doesn't permit an expansion with
> zero list elements.
>
> Having said that, I can't get excited about changing this at all.
> Nobody will thank us for it, and someone might complain.
Is there no chance that someone's issueing such an ALTER ROLE foo;
statement and thinking it's actually doing something? I suppose it's
pretty unlikely, but we do complain in some cases where we've been asked
to do something and we end up not actually doing anything for various
reasons ("no privileges were granted..."), though that's only a warning.
Maybe that'd be useful to have here though? At least then we're making
it clear to the user that nothing is happening and we don't break
existing things.
Thanks!
Stephen