Cython &C++:通过引用传递

2022-01-05 00:00:00 python reference pass-by-reference cython c++

我是 Cython 和 C++ 的菜鸟，所以我有一个关于参数传递的问题.我想避免在以下场景中传递参数的副本:

I am a noob with Cython and C++, so I have a question on argument passing. I want to avoid passing a copy of an argument in the following scenario:

# somefile.pyx
#distutils: language = c++
from libcpp.vector cimport vector

def add_one(vector[int] vect):
    cdef int i
    n = vect.size()
    for i in range(n):
        vect[i] += 1

cdef vector[int] v
for i in range(100000):
    v.push_back(i)
add_one(v) # <-- ??

我希望方法 add_one 只是就地"修改 v.我相信在 C++ 中，您可以通过在参数前面加上 & 来实现这一点，这意味着对指针的任何更改都会传递给被指点者.这样，您就不必担心传递指针或实际对象，即

I want the method add_one to just modify v "in-place." I believe in C++, you can achieve this by pre-pending the argument with &, which means that any changes to the pointer is passed to the pointee. That way, you don't have to worry about passing a pointer or the actual object, i.e.

add_one(v); # in c++

我可以在 Cython 中做同样的事情，还是必须明确地将 arg 类型更改为引用，即 def add_one(vector[int]* vect)?

Can I do the same in Cython, or do I have to explicitly change the arg type to a reference instead, i.e. def add_one(vector[int]* vect)?

推荐答案

找到了我自己问题的答案.显然，您可以通过引用传递，但该函数必须是 cdef'ed，而不是 def'ed.即

Found the answer to my own question. Apparently, you can pass by reference, but the function MUST be cdef'ed, not def'ed. i.e.

# somefile.pyx
#distutils: language = c++
from libcpp.vector cimport vector

cdef void add_one(vector[int]& vect):
    cdef int i
    n = vect.size()
    for i in range(<int>n):
        vect[i] += 1

cdef vector[int] v
for i in range(100000):
    v.push_back(i)
add_one(v)

相关文章