Initializing new types

[Martin v. Loewis]

cpitaar at yahoo.com.ar (Carlos P.) writes:

> The most important question is: am I using PyType_Ready() as I
> should?

Yes, that looks alright.

> But I´m not sure if the way to explicitly initialize the type is
> the one I´m following. 

What makes you think you might not?

> Besides this, the code in xxsubtype.c confuses
> me a little more because of the double initialization with different
> base types:
> 
> """
> DL_EXPORT(void)
> initxxsubtype(void)
> {
> 	PyObject *m, *d;
> 
> 	spamdict_type.tp_base = &PyDict_Type;
> 	if (PyType_Ready(&spamdict_type) < 0)
> 		return;
> 
> 	spamlist_type.tp_base = &PyList_Type;
> 	if (PyType_Ready(&spamlist_type) < 0)
> 		return;

That, in itself, is not double-initialization. Notice that these are
two different types: spam*dict* and spam*list*, each of which is
initialized but once in this fragment. There is another initialization
of these types later on in initxxsubtype, which is superfluous.

> And a last thing: PEP 253 says something about PyType_InitDict(). But
> I couldn´t find this function in the sources. Perhaps the PEP is a
> little outdated, isn´t it?

That is very possible. Please submit a bug report or patch to
sf.net/projects/python.

> Well, as I´ve said before, my sample code is listed below. And after
> it, an alternative one that shows what I think is not the right way
> (or at least the best way) to initialize a type.

Even if you fill out all fields of a type, you should still call
PyType_Ready; future Python versions may add additional fields with
default contents. So you should put contents into the fields only
if this is specific to your type; no need to put in generic routines.

HTH,
Martin