Peter Eisentraut <peter_e@gmx.net> writes:
>> Could we make the assumption that table names in catalogs
>> will be of type 'name'?
> I wouldn't want to guarantee it for the information schema.
Your objections are not without merit, and in the interest of bringing
this thing to closure I'll concede for now. I want to get on with this
so that I can wrap up the pg_statistic view that started the whole
thread.
What I suggest we do is apply the portions of Joe's latest patch that
support has_table_privilege with OID inputs and with NAME inputs,
omitting the combinations that take TEXT inputs and do casefolding.
We can add that part later if it proves that people do indeed want it.
I have specific reasons for wanting to keep the functions accepting
NAME rather than TEXT: that will save a run-time type conversion in the
common case where one is reading the input from a system catalog, and
it will at least provide automatic truncation of overlength names when
one is accepting a literal. (I trust Peter won't object to that ;-).)
We will probably have to revisit this territory when we implement
schemas: there will need to be a way to input qualified table names
like foo.bar, and a way to input NON qualified names like "foo.bar".
But we can cross that bridge when we come to it.
Comments, objections?
regards, tom lane