提出问题
在某个C++应用程序中,我们用一组插件来实现一些具有统一接口的功能,我们使用Python来代替动态链接库形式的插件,这样可以方便地根据需求的变化改写脚本代码,而不是必须重新编译链接二进制的动态链接库。Python强大的功能足以胜任,但是有一些操作系统特定的功能需要用C++来实现,再由Python调用。所以,最基础地,我们需要做到:
- 1. 把Python嵌入到C++应用程序中,在C++程序中调用Python函数和获得变量的值;
- 2. 用C++为Python编写扩展模块(动态链接库),在Python程序中调用C++开发的扩展功能函数。
常用的Python/C API介绍
下面是例子中用到的几个Python/C API的简要介绍及示例代码。注意,这并不是这些函数的详细介绍,而仅仅是我们所用到的功能简介,更详细内容请参考文档[1]、[2]、[3]、[4]。
打开Microsoft Visual Studio .NET 2003,新建一个控制台程序,#include <Python.h>,并在main函数里加入示例代码。
//先定义一些变量
char *cstr;
PyObject *pstr, *pmod, *pdict;
PyObject *pfunc, *pargs;
1. void Py_Initialize( )
初始化Python解释器,在C++程序中使用其它Python/C API之前,必须调用此函数,如果调用失败,将产生一个致命的错误。例:
Py_Initialize();
2. int PyRun_SimpleString( const char *command)
执行一段Python代码,就好象是在__main__ 函数里面执行一样。例:
PyRun_SimpleString("from time import time,ctime\n"
"print ''Today is'',ctime(time())\n");
3. PyObject* PyImport_ImportModule( char *name)
导入一个Python模块,参数name可以是*.py文件的文件名。相当于Python内建函数__import__()。例:
pmod = PyImport_ImportModule("mymod"); //mymod.py
4. PyObject* PyModule_GetDict( PyObject *module)
相当于Python模块对象的__dict__ 属性,得到模块名称空间下的字典对象。例:
pdict = PyModule_GetDict(pmod);
5. PyObject* PyRun_String( const char *str, int start, PyObject *globals, PyObject *locals)
执行一段Python代码。
pstr = PyRun_String("message", Py_eval_input, pdict, pdict);
6. int PyArg_Parse( PyObject *args, char *format, ...)
解构Python数据为C的类型,这样C程序中才可以使用Python里的数据。例:
/* convert to C and print it*/
PyArg_Parse(pstr, "s", &cstr);
printf("%s\n", cstr);
7. PyObject* PyObject_GetAttrString( PyObject *o, char *attr_name)
返回模块对象o中的attr_name 属性或函数,相当于Python中表达式语句:o.attr_name。例:
/* to call mymod.transform(mymod.message) */
pfunc = PyObject_GetAttrString(pmod, "transform");
8. PyObject* Py_BuildValue( char *format, ...)
构建一个参数列表,把C类型转换为Python对象,使Python可以使用C类型数据,例:
cstr="this is hjs''s test, to uppercase";
pargs = Py_BuildValue("(s)", cstr);
9. PyEval_CallObject(PyObject* pfunc, PyObject* pargs)
此函数有两个参数,都指向Python对象指针,pfunc是要调用的Python 函数,通常可用PyObject_GetAttrString()获得;pargs是函数的参数列表,通常可用Py_BuildValue()构建。例:
pstr = PyEval_CallObject(pfunc, pargs);
PyArg_Parse(pstr, "s", &cstr);
printf("%s\n", cstr);
10. void Py_Finalize( )
关闭Python解释器,释放解释器所占用的资源。例:
Py_Finalize();
Python2.4环境没有提供调试版本的Python24d.lib,所以上述示例在release模式下编译。编译完成后,把可行文件和附2给出的mymod.py文件放在一起,再点击即可运行。为了简化编程,附3 给出了simplepy.h。这样,调用mymod.transform变成如下形式:
//#include”simplepy.h”
CSimplepy py;
py.ImportModule("mymod");
std::string str=py.CallObject("transform",
"this is hjs''s test, to uppercase");
printf("%s\n", str.c_str());
接下来,我们来用C++为Python编写扩展模块(动态链接库),并在Python程序中调用C++开发的扩展功能函数。生成一个取名为pyUtil的Win32 DLL工程,除了pyUtil.cpp文件以外,从工程中移除所有其它文件,并填入如下的代码:
// pyUtil.cpp
#ifdef PYUTIL_EXPORTS
#define PYUTIL_API __declspec(dllexport)
#else
#define PYUTIL_API __declspec(dllimport)
#endif
#include<windows.h>
#include<string>
#include<Python.h>
BOOL APIENTRY DllMain( HANDLE hModule,
DWORD ul_reason_for_call,
LPVOID lpReserved
?)
{
switch (ul_reason_for_call)
{
case DLL_PROCESS_ATTACH:
case DLL_THREAD_ATTACH:
case DLL_THREAD_DETACH:
case DLL_PROCESS_DETACH:
break;
}
return TRUE;
}
std::string Recognise_Img(const std::string url)
{
//返回结果
return "从dll中返回的数据... : " +url;
}
static PyObject* Recognise(PyObject *self, PyObject *args)
{
const char *url;
std::string sts;
if (!PyArg_ParseTuple(args, "s", &url))
return NULL;
sts = Recognise_Img(url);
return Py_BuildValue("s", sts.c_str() );
}
static PyMethodDef AllMyMethods[] = {
{"Recognise", Recognise, METH_VARARGS},//暴露给Python的函数
{NULL, NULL} /* Sentinel */
};
extern "C" PYUTIL_API void initpyUtil()
{
PyObject *m, *d;
m = Py_InitModule("pyUtil", AllMyMethods); //初始化本模块,并暴露函数
d = PyModule_GetDict(m);
}
在Python代码中调用这个动态链接库:
import pyUtil
result = pyUtil.Recognise("input url of specific data")
print "the result is: "+ result
用C++为Python写扩展时,如果您愿意使用Boost.Python库的话,开发过程会变得更开心J,要编写一个与上述pyUtil同样功能的动态链接库,只需把文件内容替换为下面的代码。当然,编译需要boost_python.lib支持,运行需要boost_python.dll支持。
#include<string>
#include <boost/python.hpp>
using namespace boost::python;
#pragma comment(lib, "boost_python.lib")
std::string strtmp;
char const* Recognise(const char* url)
{
strtmp ="从dll中返回的数据... : ";
strtmp+=url;
return strtmp.c_str();
}
BOOST_PYTHON_MODULE(pyUtil)
{
def("Recognise", Recognise);
}
所有示例都在Microsoft Windows XP Professional + Microsoft Visual Studio .NET 2003 + Python2.4环境下测试通过,本文所用的Boost库为1.33版本。
参考资料
- [1] Python Documentation Release 2.4.1. 2005.3.30,如果您以默认方式安装了Python2.4,那么该文档的位置在C:\Program Files\Python24\Doc\Python24.chm;
- [2] Michael Dawson. Python Programming for the Absolute Beginner. Premier Press. 2003;
- [3] Mark Lutz. Programming Python, 2nd Edition. O''Reilly. 2001.3 ;
- [4] Mark Hammond, Andy Robinson. Python Programming on Win32. O''Reilly. 2000.1 ;
- Python主页:http://www.python.org;
- Boost库主面:www.boost.org;
附1 text.txt
this is test text in text.txt.
附2 mymod.py
import string
message = ''original string''
message =message+message
msg_error=""
try:
text_file = open("text.txt", "r")
whole_thing = text_file.read()
print whole_thing
text_file.close()
except IOError, (errno, strerror):
print "I/O error(%s): %s" % (errno, strerror)
def transform(input):
#input = string.replace(input, ''life'', ''Python'')
return string.upper(input)
def change_msg(nul):
global message #如果没有此行,message是函数里头的局部变量
message=''string changed''
return message
def r_file(nul):
return whole_thing
def get_msg(nul):
return message
附3 simplepy.h
#ifndef _SIMPLEPY_H_
#define _SIMPLEPY_H_
// simplepy.h v1.0
// Purpose: facilities for Embedded Python.
// by hujinshan @2005年9月2日9:13:02
#include
using std::string;
#include
//--------------------------------------------------------------------
// Purpose: ease the job to embed Python into C++ applications
// by hujinshan @2005年9月2日9:13:18
//--------------------------------------------------------------------
class CSimplepy // : private noncopyable
{
public:
///constructor
CSimplepy()
{
Py_Initialize();
pstr=NULL, pmod=NULL, pdict=NULL;
pfunc=NULL, pargs=NULL;
}
///destructor
virtual ~CSimplepy()
{
Py_Finalize();
}
///import the user module
bool ImportModule(const char* mod_name)
{
try{
pmod = PyImport_ImportModule(const_cast(mod_name));
if(pmod==NULL)
return false;
pdict = PyModule_GetDict(pmod);
}
catch(...)
{
return false;
}
if(pmod!=NULL && pdict!=NULL)
return true;
else
return false;
}
///Executes the Python source code from command in the __main__ module.
///If __main__ does not already exist, it is created.
///Returns 0 on success or -1 if an exception was raised.
///If there was an error, there is no way to get the exception information.
int Run_SimpleString(const char* str)
{
return PyRun_SimpleString(const_cast(str) );
}
///PyRun_String("message", Py_eval_input, pdict, pdict);
///Execute Python source code from str in the context specified by the dictionaries globals.
///The parameter start specifies the start token that should be used to parse the source code.
///Returns the result of executing the code as a Python object, or NULL if an exception was raised.
string Run_String(const char* str)
{
char *cstr;
pstr = PyRun_String(str, Py_eval_input, pdict, pdict);
if(pstr==NULL)
throw ("when Run_String, there is an exception was raised by Python environment.");
PyArg_Parse(pstr, "s", &cstr);
return string(cstr);
}
///support olny one parameter for python function, I think it''s just enough.
string CallObject(const char* func_name, const char* parameter)
{
pfunc=NULL;
pfunc = PyObject_GetAttrString(pmod, const_cast(func_name));
if(pfunc==NULL)
throw (string("do not found in Python module for: ")
+func_name).c_str();
char* cstr;
pargs = Py_BuildValue("(s)", const_cast(parameter));
pstr = PyEval_CallObject(pfunc, pargs);
if(pstr==NULL)
throw ("when PyEval_CallObject, there is an exception was raised by Python environment");
PyArg_Parse(pstr, "s", &cstr);
return string(cstr);
}
//PyObject *args;
//args = Py_BuildValue("(si)", label, count); /* make arg-list */
//pres = PyEval_CallObject(Handler, args);
protected:
PyObject *pstr, *pmod, *pdict;
PyObject *pfunc, *pargs;
};
#endif // _SIMPLEPY_H_
// end of file