[Python-Dev] Support for PyGetSetDefs in pydoc

[Barry Warsaw]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Jul 13, 2006, at 9:15 AM, Nick Coghlan wrote:

> Could you include a "look up late-breaking types" function in  
> types.py that site.py calls after it finishes setting up the  
> standard library path?
>
> Still a little hackish, I know, but it seems less icky than your  
> suggestions above, as it means that fixing anything similar that  
> comes up in the future will only require modification of types.py.

I've come to the conclusion that it's much better to add a "helper"  
module that defines a few extension types solely for the purposes of  
exposing them in types.py, than it is to rely on accidents of  
implementation discovered in the existing interpreter.  To that end,  
I've implemented a _typesmodule.c that exposes a getset descriptor  
and a member descriptor on a type that can't be instantiated.  That  
module gets built into Python and provides a central place to add any  
future C extension types.  It also provides a handy place for Jython  
or IronPython to hang their own equivalents, if and when they ever  
get them.

Then in types.py, we import _types and pull the appropriate types  
from there.  types.py will not define GetSetType or  
MemberDescriptorType if that module can't be imported so that it's  
robust for other Python implementations.  I like not defining them  
better than setting them to None in that case.

Then inspect.py defines isgetset() and ismemberdescriptor() in terms  
of those types, but it always returns False if those types don't  
exist.  So inspect.py too is robust for other Python implementations.

I've updated SF patch #1520294 and assigned it back to Georg for  
another quick review.  Assuming he likes the updated patch and  
Anthony approves, please assign it back to me for final checkin.

In thinking about this more, re: back porting to Python 2.4, I don't  
think we can expose GetSetType, MemberDescriptorType, isgetset(), or  
ismemberdescriptor() in the appropriate modules.  That smells too  
much like the True/False thing.  However, I believe I can weedle a  
fix into pydoc for those types without exposing these new names, so  
I'd like to do that if there are no objections.

- -Barry

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (Darwin)

iQCVAwUBRLZw8XEjvBPtnXfVAQLkIAP+O0WXh+xO902KNah/kIYy/GSPsO/CfRUO
+YchtF2HdbyEXORwhsLXJVTKVROYgohCQrZp7LUX5vNUah0J0ycIPXhzpqrB0Gmc
8AaRChkFO8EGofzEGOvtunGUGdl9733u2mNUUAhGk15YFs8GDyiuzkL8SitcJQ3q
OHnCsdNRuSU=
=pbA8
-----END PGP SIGNATURE-----