Question

I have the following structure in my code:

#test.pyx
cdef class BClass:
    cdef str x
    cdef set_args(self, x):
        self.x = x

    cdef func(self, x):  
        print "BClass" , self.x, "--", x


fit = BClass()

def my_func():
    global fit
    fit.set_args("hello")
    fit.func("world")


#tester.py
import test
test.my_func()

I build pyx into a pyd without problem, but when I run tester.py I get the following error:

AttributeError: ‘test.BClass’ object has no attribute ‘set_args’

If I instantiate fit inside my_func all works as expected. Does this mean I can’t define a module level object and then reuse it in a function in the same module?

In my original code my_func is called repeatedly in a loop and instantiating a BClass has some overhead due to allocating some arrays, that’s why I want to instantiate fit outside my_func. Any thoughts how to do this?

Answer 1

Well, the solution is embarrassingly simple. Just declaring fit as

cdef BClass fit = BClass()

at a module level makes it work without any errors.

I still don’t understand why if fit is instantiated the Python way with

fit = BClass()

then I get this AttributeError and why if I instantiate it in the Python way inside my_func() it works without error???