How would one go about specifying a static method or variable in a python class, in CPython C++?
In the PyTypeObject structure, tp_getset, tp_methods, tp_members all seem to assume an instance of the class.
Thanks very much, Rob.
5
-
see the answers of stackoverflow.com/questions/68645/…bilash.saha– bilash.saha2011-11-04 21:49:15 +00:00Commented Nov 4, 2011 at 21:49
-
docs.python.org/extending/extending.htmlbilash.saha– bilash.saha2011-11-04 22:06:55 +00:00Commented Nov 4, 2011 at 22:06
-
1The first link does not refer to C++ while the second link does not refer to static variables in python classes. Is there something in those links Im missing that relates to my question? Thanksrbairos– rbairos2011-11-04 22:10:59 +00:00Commented Nov 4, 2011 at 22:10
-
make your question a bit clear.may be with a example about what you want...bilash.saha– bilash.saha2011-11-04 22:12:07 +00:00Commented Nov 4, 2011 at 22:12
-
In CPython, you can extend python with new types etc, by setting up a PyTypeObject structure. There does not seem to be any way to define static methods and variables in this structure, only regular methods and variables (tp_getset, tp_methods, tp_members, etc).rbairos– rbairos2011-11-04 22:14:44 +00:00Commented Nov 4, 2011 at 22:14
Add a comment
|
1 Answer
Static and class methods can be defined in tp_methods by adding METH_STATIC or METH_CLASS to the ml_flags field of the PyMethodDef structure. This is equivalent to @staticmethod and @classmethod decorators.
The first parameter of the method, which normally is the instance pointer, will be NULL for static methods and PyTypeObject* for class methods.
http://docs.python.org/c-api/structures.html#PyMethodDef
Class attributes can be added by setting the tp_dict to a dictionary with these attributes before calling PyType_Ready() (in your module initialization function). Alternatively tp_dict can be left as NULL in which case PyType_Ready() will create the dict for you. The attributes can then be added to that dict.
http://docs.python.org/c-api/typeobj.html#tp_dict
Computed class attributes require a descriptor class, exactly as in Python. An instance of the descriptor should then be added to the tp_dict as with other class attributes.
