Am 22.09.2012 15:30, schrieb Daniele Varrazzo:
> On Sat, Sep 22, 2012 at 2:25 PM, Tobias Oberstein
> <tobias.oberstein@gmail.com> wrote:
>
>> the _from_db class method on CompositeCaster takes a name argument and
>> parsed that into "schema" and "typename".
>>
>> It uses both to retrieve Oids etc, but then only forwards "typename", and
>> not "schema" to the CompositeCaster constructor.
>>
>> If a have 2 composite types defined "public.t_foo" and "bar.t_foo", and
>> register both, one will be overwritten ..
>
> Uhm... why overwritten? The two CompositeCaster will register two
> different typecasters on two different oids. The name is only used as
Ok. So new_type/new_array_type have no issue with having the same "name"
used twice on different OIDs?
> name namedtuple name. What would the schema be used for?
My use case: have a CompositeDictCaster that spits out:
{'record': 'public.t_station', 'x': 10, 'y': 8}
that is injects a 'record' field containing the composite type name.
Hence I need the schema, not only the typename.
Background:
I am experimenting with that in combination with a custom dict adapter
that is able to adapt such value again into
(10, 8, NULL)::public.t_station
so this is fully "round trippable". That way the type info is embedded
and I don't need to do complex object mangling.
The motivation for all this: I have a network server that has a
JSON/WebSocket based API for RPC (and PubSub), and I want to forward
those to PostgreSQL stored procedure calls.
Call PG stored procs from browser JS ..
Cheers,
Tobias
>
> -- Daniele
>