flexi adaption/casting scheme

Hi,

I'd like to implement a certain type adaption/casting scheme
and struggle with the right approach. Maybe someone can give
me a hint? Would be great and welcome!

Context: I am _only_ concerned with calling stored procedures,
so there is the complete type metadata from PG catalog regarding
the PG side available to use for mapping.


PG => Py (typecasting):
-----------------------

1.1
hstore => plain Python dict

1.2
JSON => json.loads() .. whatever Python object that gives

1.3
composite type => plain Python dict with key/value pairs only
for all attributes in the composite type that have non-NULL values

1.4
everything else => as per-default with Psycopg


Py => PG (adaption):
--------------------

2.1
plain Python dict ..:

2.1.1
PG target type = hstore => dict-to-hstore with conversion of keys/values
to str repr. if needed

2.1.2
PG target type = JSON => json.dumps() whatever str that produces

2.1.3
PG target type = composite type => for every key in the Python dict that
is an attribute in the composite type, fill in the value from the dict;
for every attribute in the composite type where there is no key in the
Python dict, fill in NULL

2.2
everything else => as per-default with Psycopg

==

Above should work with nested PG types (array of composite type with
an attribute again composite type etc etc).

It should work with IN, OUT, INOUT parameters and array, setof, etc
returning procedures.

==

How do I tackle this? Or even more fundamental: is it sane / doable at
all (using public Psycopg hooks only)?

Thanks alot,
Tobias