Calling Cython C functions from Python

Question

I have a Cython file called foo.pyx containing the following functions:

def add_one(int n):
    cdef int m = n + 1
    return m

cdef int c_add_one(int n):
    return n + 1

I build this pyx file using cython -a foo.pyx and can then do:

>>> import foo
>>> foo.add_one(5)
6
>>> foo.c_add_one(5)
AttributeError: 'module' object has no attribute 'c_add_one'

So it looks like I can't call c_add_one from python. What are the advantages of declaring a function using cdef?

cdef defines a function that can be accessed only from C. If you want c_add_one to be accessible from Python, you should use cpdef instead. The advantage is of cdef is to remove the overhead of the python layer when calling it. — William Zhang
– William Zhang, Commented Nov 4, 2014 at 23:49

Kevin · Accepted Answer · 2014-11-05 21:04:13Z

7

Speed: Cython and C can call cdef and cpdef functions much faster than regular functions. Regular def functions are full-blown Python objects. They require reference counting, the GIL, etc.
Encapsulation: If a function is declared cdef, it will not be exposed to Python users. This is beneficial if the function was never intended for public consumption (e.g. because it doesn't check its parameters, or relates to an implementation detail which might change in the future). Users can still circumvent this via C/Cython, of course, but doing so is more of a hassle than most will bother with. In this way, it's similar to double-underscore name mangling.

answered Nov 5, 2014 at 21:04

Kevin

30.4k11 gold badges67 silver badges89 bronze badges

Sign up to request clarification or add additional context in comments.

1 Answer 1