idea: multiple arguments to_regclass function

Pavel Stehule <pavel.stehule@gmail.com> writes:

I missing some variants of to_regclass

to_regclass(schemaname, objectname)
to_regclass(catalogname, schemaname, objectname)

Can do that already:

to_regclass(quote_ident(schemaname) || '.' || quote_ident(objectname))

I'm not eager to build nine more to_reg* functions to do the equivalent
of that, and even less eager to build eighteen.

regards, tom lane

ne 16. 4. 2023 v 16:23 odesílatel Tom Lane <tgl@sss.pgh.pa.us> napsal:

Pavel Stehule <pavel.stehule@gmail.com> writes:

I missing some variants of to_regclass

to_regclass(schemaname, objectname)
to_regclass(catalogname, schemaname, objectname)

Can do that already:

to_regclass(quote_ident(schemaname) || '.' || quote_ident(objectname))

I'm not eager to build nine more to_reg* functions to do the equivalent
of that, and even less eager to build eighteen.

Yes, I can. But there is overhead with escaping and string concatenation.
And it is a little bit sad, so immediately the parser has to do an inverse
process.

Maybe we can introduce only three functions

anyelement get_object(catalogname name, schemaname name, objectname name,
returntype anyelement)
anyelement get_object(schemaname name, objectname name, returntype
anyelement)
anyelement get_object(objectname name, returntype anyelement)

so usage can be like

DECLATE _tab regclass;
BEGIN
_tab := get_object('public', 'mytab', _tab);
..

?

Regards

Pavel