为什么在Python中将mmap实现为共享库？

Question

I'm new to python. But I'm curious that why the mmap in python is implemented as shared library, while not a .py file. This makes my IDE can't index the source codes of mmap.

Here is the output in my python3.2 environment(ubuntu 12.04):

markz@markz-hp6200:~$ python3.2
Python 3.2.3 (default, Oct 19 2012, 20:10:41) 
[GCC 4.6.3] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import mmap
>>> print(mmap.__file__)
/usr/lib/python3.2/lib-dynload/mmap.cpython-32mu.so
>>> 

Answer 1

mmap uses OS-specific facilities such as the mmap function on Linux (and MapViewOfFile function on Windows). These are not directly available in Python (eg through the os module), so at least part of the mmap module has to be written in C to call those functions.

This particular use has nothing to do with efficiency -- the module had to be written in C (or use ctypes ) in order to use this OS functionality.

Answer 2

There are lots of things in the standard library of Cpython written in C. Other examples are itertools , parts of csv ( _csv ), parts of collections ( _collections )... I would think that one of the main reasons is for efficiency. The developers decided they could get the code to go faster if they used a compiled language.

Answer 3

mmap is implemented in C, not python. An extension module implemented in C has to be shared library.

Why? In this case, probably because you can't implement the functionality without access to C. The modules which are implemented in python are implemented using the functionality provided in other modules. But the mmap functionality is implemented by calling the C standard library functions. There is no way it could be a python file.