如何把C/C++程序编译成实用的python模块！

作者: 14e61d025165 | 来源:发表于2019-05-25 14:47 被阅读0次

如何把C/C++程序编译成实用的python模块！
Python和C++语言的整合
在Clion中实现C++调用Python模块的功能
探索Objective-C中的class和对象在C++中的原理实
AOSP-编译C/C++可执行程序/库
java编译过程
python加速--如何调用C/C++代码
protobuffer
windows下python3调用C++接口
python进阶-12-opencv人脸识别

<tt-image data-tteditor-tag="tteditorTag" contenteditable="false" class="syl1558766826266" data-render-status="finished" data-syl-blot="image" style="box-sizing: border-box; cursor: text; color: rgb(34, 34, 34); font-family: "PingFang SC", "Hiragino Sans GB", "Microsoft YaHei", "WenQuanYi Micro Hei", "Helvetica Neue", Arial, sans-serif; font-size: 16px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; background-color: rgb(255, 255, 255); text-decoration-style: initial; text-decoration-color: initial; display: block;">

image

在Python遇到性能瓶颈时怎么办？答案是找对应功能的C/C++程序，把它编译成CPython模块，供Python调用来提高性能。Python学习交流群：1004391443，这里有资源共享，技术解答，还有小编从最基础的Python资料到项目实战的学习资料都有整理，希望能帮助你更了解python，学习python

如何把C/C++程序编译成Python模块

比如Python中做科学计算，数据处理的Numpy模块就是使用C语言编写的，Numpy处理速度比Pandas快数倍。Numpy的处理速度一点都不比go语言差。

本文就是介绍如何把C/C++程序编译成Python模块。本文偏技术，需要耐着性质看。

Python 作为一个胶水语言，可以很方便的通过C/C++来进行扩展，提高性能。前面我写了一篇文章介绍如何通过Python的ctypes加载普通的.so库。

其实，这还不算真正的用C/C++写Python的扩展模块。

本文将介绍如何使用C语言和C++写Python模块。

一、Python的C语言接

Python语言最初是用C语言实现的一种脚本语言，后来被称为CPython，是因为后来它语言实现的Python，比如Python实现的Python——PyPy，Java语言实现的Python——Jython，.Net实现的Python——IronPython。

CPython具有优良的开放性和可扩展性，并提供了方便灵活的应用程序接口(API)，从而使得C/C++程序员能够对Python解释器的功能进行扩展。

Python的C语言接口很适合封装C语言实现的各种函数，如果要封装C++的类，使用boost_python或者SWIG更方便和合适，还有一个类似boost_python的支持C++11的pybind11。

1 模块封装

假设我们有一个C函数：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">/* 文件名：mylib.c */
int addone(int a) {
return a+1;
}
</pre>

如果想在Python解释器中调用该函数，则应该首先将其实现为Python中的一个模块，这需要编写相应的封装接口，如下所示：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">/* wrap_mylib.c */

include

include "mylib.h"

PyObject* wrap_addone(PyObject* self, PyObject* args)
{
int n, result;
if (! PyArg_ParseTuple(args, "i:fact", &n))
return NULL;
result = addone(n); /*这里调用C函数 /
return Py_BuildValue("i", result);
}
static PyMethodDef mylibMethods[] =
{
{"addone", wrap_addone, METH_VARARGS, "Add one to N"},
{NULL, NULL}
};
void initmylib()
{
PyObject m;
m = Py_InitModule("mylib", mylibMethods);
}
</pre>

上面就是一个典型的Python扩展模块，它至少应该包含三个部分：导出函数、方法列表和初始化函数。

2 导出函数

要在Python解释器中调用C语言中的某个函数，首先要为它编写对应的导出函数，上述例子中的导出函数为wrap_addone。在Python的C语言扩展中，所有的导出函数都具有相同的函数原型：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">PyObject* wrap_method(PyObject* self, PyObject* args);
</pre>

这个函数是Python解释器和C函数进行交互的接口，一般以wrap_开头后面跟上C语言的函数名，这样命名把导出函数和C语言函数对应起来使得代码更加清晰。它带有两个参数：self和args。

参数self 只在C函数被实现为内联方法(built-in method)时才被用到，通常该参数的值为空(NULL)。

参数args 中包含了Python解释器要传递给C函数的所有参数，通常使用Python的C语言扩展接口提供的函数PyArg_ParseTuple()来获得这些参数值。

所有的导出函数都返回一个PyObject指针，如果对应的C函数没有真正的返回值(即返回值类型为void)，则应返回一个全局的None对象(Py_None)，并将其引用计数增1，如下所示：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">PyObject* wrap_method(PyObject *self, PyObject *args)
{
Py_INCREF(Py_None);
return Py_None;
}
</pre>

3 方法列表

方法列表中列出了所有可以被Python解释器使用的方法，上述例子对应的方法列表为：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">static PyMethodDef mylibMethods[] =
{
{"addone", wrap_addone, METH_VARARGS, "Add one to N"},
{NULL, NULL}
};
</pre>

方法列表中的每项由四个部分组成：

方法名

导出函数

参数传递方式

方法描述

方法名是从Python解释器中调用该方法时所使用的名字。

参数传递方式则规定了Python向C函数传递参数的具体形式，可选的两种方式是METH_VARARGS和METH_KEYWORDS，其中METH_VARARGS是参数传递的标准形式，它通过Python的元组在Python解释器和C函数之间传递参数，若采用METH_KEYWORD方式，则Python解释器和C函数之间将通过Python的字典类型在两者之间进行参数传递。

4 初始化函数

所有的Python扩展模块都必须要有一个初始化函数，以便Python解释器能够对模块进行正确的初始化。Python解释器规定所有的初始化函数的函数名都必须以init开头，并加上模块的名字。对于模块mylib来说，则相应的初始化函数为:

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">void initmylib()
{
PyObject* m;
m = Py_InitModule("mylib", mylibMethods);
}
</pre>

当Python解释器需要导入该模块时，将根据该模块的名称查找相应的初始化函数，一旦找到则调用该函数进行相应的初始化工作，初始化函数则通过调用Python的C语言扩展接口所提供的函数Py_InitModule()，来向Python解释器注册该模块中所有可以用到的方法。

5 编译链接

要在Python解释器中使用C语言编写的扩展模块，必须将其编译成动态链接库的形式。下面以Linux为例，介绍如何将C编写的Python扩展模块编译成动态链接库：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">$ gcc -fpic -shared -o mylib.so
-I/usr/include/python2.7
mylib.c wrap_mylib.c
</pre>

6 在Python中调用

上面编译生成的Python扩展模块的动态链接库，可以在Python中直接import。如下所示：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">veelion@gtx:~$ python
Python 2.7.12 (default, Nov 19 2016, 06:48:10)
[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.

import example
example.addone(7)
8

</pre>

这里生成的.so动态库和上一篇中不用Python的C语言生成的动态库是不一样的，从生成过程和使用方法就可以看出来，这里的动态库使用起来感觉就是一个Python模块，直接import就可以了。

二、用boost_python库封装C++类

安装boost python库：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">sudo aptitude install libboost-python-dev
</pre>

示例

下面代码简单实现了一个普通函数maxab()和一个Student类：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">#include

include

int maxab(int a, int b) { return a>b?a:b; }
class Student {
private:
int age;
std::string name;
public:
Student() {}
Student(std::string const& _name, int _age) { name=_name; age=_age; }
static void myrole() { std::cout << "I'm a student!" << std::endl; }
void whoami() { std::cout << "I am " << name << std::endl; }
bool operator==(Student const& s) const { return age == s.age; }
bool operator!=(Student const& s) const { return age != s.age; }
};
</pre>

使用boost.python库封装也很简单，如下代码所示：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">#include

include <boost/python.hpp>

include <boost/python/suite/indexing/vector_indexing_suite.hpp>

include

include "student.h"

using namespace boost::python;
BOOST_PYTHON_MODULE(student) {
// This will enable user-defined docstrings and python signatures,
// while disabling the C++ signatures
scope().attr("version") = "1.0.0";
scope().attr("doc") = "a demo module to use boost_python.";
docstring_options local_docstring_options(true, false, false);
def(
"maxab", &maxab, "return max of two numbers.
"
);
class_("Student", "a class of student")
.def(init<>())
.def(init<std::string, int>())
// methods for Chinese word segmentation
.def(
"whoami", &Student::whoami, "method's doc string..."
)
.def(
"myrole", &Student::myrole, "method's doc string..."
)
.staticmethod("myrole");
// 封装STL
class_<std::vector >("StudentVec")
.def(vector_indexing_suite<std::vector >())
;
}
</pre>

上述代码还是include了Python.h文件，如果不include的话，会报错误：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">wrap_python.hpp:50:23: fatal error: pyconfig.h: No such file or directory
</pre>

编译

编译以上代码有两种方式，一种是在命令行下面直接使用g++编译：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">g++ -I/usr/include/python2.7 -fPIC wrap_student.cpp -lboost_python -shared -o student.so
</pre>

首先指定Python.h的路径，如果是Python 3的话就要修改为相应的路径，编译wrap_student.cpp要指定-fPIC参数，链接(-lboost_python)生成动态库(-shared)。生成的student.so动态库就可以被python直接import使用了

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">In [1]: import student
In [2]: student.maxab(2, 5)
Out[2]: 5
In [3]: s = student.Student('Tom', 12)
In [4]: s.whoami()
I am Tom
In [5]: s.myrole()
I'm a student!
</pre>

另外一直方法是用python的setuptools编写setup.py脚本：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">#!/usr/bin/env python
from setuptools import setup, Extension
setup(name="student",
ext_modules=[
Extension("student", ["wrap_student.cpp"],
libraries = ["boost_python"])
])
</pre>

然后执行命令编译：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">python setup.py build
or
sudo python setup.py install
</pre>

三、SWIG封装C++类

Python调用C/C++代码的利器除了boost_python外，还有SWIG（Simplified Wrapper and Interface Generator），它是用来为脚本语言调用C和C++程序的软件开发工具，它实际上是一个编译器，获取C/C++的声明和定义，用一个壳封装起来，以便其它脚本语言访问这些声明。所以，SWIG 最大的好处就是将脚本语言的开发效率和 C/C++ 的运行效率有机的结合起来。

一个双数组Trie Tree的实现：cedar在中文分词、新词发现等算法中可以y用于词典的创建。本文以cedar的SWIG封装实现来说明SWIG的使用。

0. 安装swig

工欲善其事必先利其器，首先要安装swig，Ubuntu安装swig很简单：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">sudo aptitude install swig
</pre>

1. 声明和定义C/C++代码

在cedar的swig目录下面有cedar的C++声明和实现代码trie.h，但是这个实现里面没有遍历所有key的函数方法，所以我添加了一个实现，首先定义一个数据结构来定义key：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">// key-value pair return type for next_key()
class kv_t {
public:
std::string key;
int value;
};
</pre>

添加一个函数每次返回一个key，当key字符串为空时表示遍历结束，继续调用的话就又从头开始遍历：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">// to iterate all keys
kv_t next_key() const {
static size_t from = 0, p = 0;
union { int i; int x; } b;
char key[256] = {0};
kv_t kv;
if(from == 0) {
b.i = _t->begin(from, p);
}else{
b.i = _t->next(from, p);
}
if (b.i == trie_t::CEDAR_NO_PATH) {
kv.key = "";
kv.value = 0;
from = 0;
p = 0;
return kv;
}
_t->suffix(key, p, from);
kv.key = key;
kv.value = b.x;
return kv;
}
</pre>

2. 编写接口文件.i

查看cedar.i可以看到SWIG的接口文件的编写规则：

首先在 %module 后面声明模块名称，这就是Python在import时使用的模块名称；

在%{ … %}之间包含相关头文件

在%include 后面可以声明对STL的支持

最后声明要封装的函数和变量，也可以之间包含头文件：%include “trie.h”

3. 封装代码

可以在Makefile里面看到python-bindings：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">python-bindings:
swig -Wall -python -builtin -outdir python -c++ cedar.i
mv -f cedar_wrap.cxx python
</pre>

直接make或者单独运行上面的swig命令，就可以生成cedar.py和cedar_wrap.cxx文件。

4. 编译生成动态库

编译生成的cedar_wrap.cxx使用python distutils的setup，可以参考python/setup.py的编写。setup.py的build如下：

<pre spellcheck="false" style="box-sizing: border-box; margin: 5px 0px; padding: 5px 10px; border: 0px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-variant-numeric: inherit; font-variant-east-asian: inherit; font-weight: 400; font-stretch: inherit; font-size: 16px; line-height: inherit; font-family: inherit; vertical-align: baseline; cursor: text; counter-reset: list-1 0 list-2 0 list-3 0 list-4 0 list-5 0 list-6 0 list-7 0 list-8 0 list-9 0; background-color: rgb(240, 240, 240); border-radius: 3px; white-space: pre-wrap; color: rgb(34, 34, 34); letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-style: initial; text-decoration-color: initial;">python setup.py build
</pre>

就会在当前目录下面创建目录build，下面生成lib.linux-x86_64-2.7/cedar.py 和 _cedar.so

四、 pybind11封装C++

从pybind11的名字可以看出，它是用来封装C++ 11代码为Python模块的库。它的目标和用法都是想Boost_python库看齐，但是它又比庞大的Boost库精简。我知道这个库的时间不长，也没有具体实践过。以前都是写C++，然后用boost封装。但是，感觉pybind11更简洁，所以下一个项目可以试试它。到时候再分享使用心得给大家。