Pure Python with Cython decorators: How to get attribute access at module level

Question

I would like to write some Pure Python with Cython decorator, but when I rename my NONE.PY to NONE.PYX I've got an error. To workaround this issue I need to wrap each attribute with a pure python definition call without decorator. I wonder why...

here the module none.pyx (if you rename it to none.py, you will have no issue at all)

import cython

@cython.cfunc
@cython.returns(cython.double)
@cython.locals(n=cython.int,i=cython.int,r=cython.int)
def ccrange(n):
  r=0
  for i in range(n):
    r+=i
  return r

def crange(n):  return ccrange(n)

and the python test file test_none.py:

import pyximport; pyximport.install()
import none
n=10000
print ">>pure python call>>",none.crange(n)
print ">>cython call>>",none.ccrange(n)

Result with none.pyx:

pure python call>> 49995000.0 cython call>> Traceback (most recent call last): File "C:\\Users\\damien\\python4d\\bacoland\\test_none.py", line 6, in print ">>cython call>>",none.ccrange(n)
AttributeError: 'module' object has no attribute 'ccrange'

Rename none.pyx to none.py, give:

pure python call>> 49995000 cython call>> 49995000

Thanks for Help! Have a NiceDay :-)

EDIT: Avoid the decorator @cython.cfunc is breaking the speed avantage of cython... Consider this following code with and without @cython.cfunc:

@cython.cfunc
@cython.returns(cython.double)
@cython.locals(n=cython.int)
def fibo_c(n):
  if n == 0 or n == 1:
      return n
  return fibo_c(n-2) + fibo_c(n-1)

Answer 1

@cython.cfunc decorator is an equivalent to cdef ing a function (see here for details), so this function is accessible only within C code. So, to make it accessible from Python get rid of @cython.cfunc .

Answer 2

此时应更换你的@cython.cfunc （相当于cdef ，即从仅C调用）由@cython.ccall （相当于cpdef ，即都可以访问使用慢速Python的调用约定和快速的C调用约定）。