On 2021-Mar-26, Joel Jacobson wrote: > On Fri, Mar 26, 2021, at 11:30, Alvaro Herrera wrote: > > On 2021-Mar-26, Joel Jacobson wrote: > > > > > On Thu, Mar 25, 2021, at 17:51, Tom Lane wrote: > > > > > > I wonder what performance will be like with lots o' objects. > > > > > > I guess pg_get_acl() would need to be implemented using a switch(classid) > > > with 36 cases (one for each class)? > > > > No, we have a generalized object query mechanism, see objectaddress.c > > That's where I was looking actually and noticed the switch with 36 cases, in > the function getObjectDescription().
Ah! well, you don't have to repeat that. AFAICS the way to do it is like AlterObjectOwner_internal obtains data -- first do get_catalog_object_by_oid (gives you the HeapTuple that represents the object), then heap_getattr( ..., get_object_attnum_acl(), ..), and there you have the ACL which you can "explode" (or maybe just return as-is). AFAICS if you do this, it's just one cache lookups per object, or one indexscan for the cases with no by-OID syscache. It should be much cheaper than the UNION ALL query. And you use pg_shdepend to guide this, so you only do it for the objects that you already know are interesting. -- Álvaro Herrera Valdivia, Chile "La conclusión que podemos sacar de esos estudios es que no podemos sacar ninguna conclusión de ellos" (Tanenbaum)
