Ethan Furman wrote: > On 02/26/2013 09:21 AM, Robin Becker wrote: >> In python 2 I was able to improve speed of reportlab using a C extension >> to optimize some heavily used methods. >> >> so I was able to do this >> >> >> class A: >> ..... >> def method(self,...): >> .... >> >> >> try: >> from extension import c_method >> import new >> A.method = new.instancemethod(c_method,None,A) >> except: >> pass >> >> and if the try succeeds our method is bound as a class method ie is >> unbound and works fine when I call it. >> >> In python 3 this doesn't seem to work at all. In fact the new module is >> gone. The types.MethodType stuff doesn't seem to work. >> >> Is there a way in Python 3.3 to make this happen? This particular method >> is short, but is called many times so adding python wrapping layers is >> not a good way forward. > > Dumb question, but have you tried just assigning it? In Py3 methods are > just normal functions... > > 8<---------------------- > class A(): > pass > > A.method = c_method > 8<----------------------
The problem is that functions implemented in C don't support the descriptor protocol (they don't have a __get__() method). So >>> from math import sqrt >>> class A(int): ... pass ... >>> A.c = sqrt >>> A.py = lambda self: sqrt(self) >>> a = A(42) >>> a.py() 6.48074069840786 >>> a.c() Traceback (most recent call last): File "<stdin>", line 1, in <module> TypeError: sqrt() takes exactly one argument (0 given) -- http://mail.python.org/mailman/listinfo/python-list