如何在Python中调用C/C++代码

如何在Python中调用C/C++代码
How to mix C/C++ code in Python

本文介绍一种手动的、简单的在Python中使用C/C++代码的方式。这个方法主要使用了ctypes模块。其他的混合Python,C/C++编程的方法还有Swig Boost.Python。前一种方法需要写一个接口文件(interface),而后一种需要使用庞大、深奥的boost类库,后两者适合可能适合更复杂的情况,这里只介绍第一种方法。

混合C/C++代码需要这几步:
1. 包装接口 C/C++ wrap functions up
2. 打包成共享库 Compiling C/C++ code and pack it to shared library
3. Python中导入共享库 Python imports shared library

先介绍一下北京,这里我的C++类GenomeSequence使用了模板(Template)和Memorymap,这是一个访问基因序列的类,比如如果一个生物序列是GAGTTTTATCGCTTCCATGACGCAGAAGTTAACACT… 我们的类是gs,那么gs[0] = ‘G’, gs[1]=’A’ …. 摘录相关的函数如下:

class GenomeSequence : public genomeSequenceArray
{
public:
    /// Simple constructor - no implicit file open
    GenomeSequence();
    /// set the reference name that will be used in open()
    /// \param referenceFilename the name of the reference fasta file to open
    /// \return false for success, true otherwise
    ///
    /// \sa open()
    bool setReferenceName(std::string referenceFilename);
    /// return the number of bases represented in this reference
    /// \return count of bases
    genomeIndex_t   getNumberBases() const
    {
        return getElementCount();
    }
    inline char operator[](genomeIndex_t index) const
    {
        uint8_t val;
        if (index < getNumberBases())
        {
            if ((index&1)==0)
            {
                val = ((uint8_t *) data)[index>>1] & 0xf;
            }
            else
            {
                val = (((uint8_t *) data)[index>>1] & 0xf0) >> 4;
            }
        }
        else
        {
            val = baseNIndex;
        }
        val = isColorSpace() ? int2colorSpace[val] : int2base[val];
        return val;
    }
    /* ........... more codes omitted ................ */
}

但实际上这些细节并不重要,重要是如何包装,我们编写GenomeSequence_wrap.cpp文件,包括对上述4个函数的封装,源码如下:

#include "GenomeSequence.h"
#include <string>

extern "C"{
    GenomeSequence* GenomeSequence_new(){ return new GenomeSequence();}
    bool GenomeSequence_setReferenceName(GenomeSequence* gs, char* s) { 
        if (!gs) return false;
        std::string str = s;
        //printf("Loading %s ...\n", s);
        if (!gs->setReferenceName(str)){
            gs->open();
        } else {
            printf("Loading FAIL\n");
        }
        return (gs->setReferenceName(str));
    }
    void GenomeSequence_close(GenomeSequence* gs) {if (gs) gs->close();};
    int GenomeSequence_getNumBase(GenomeSequence* gs) {
        if (!gs) {
            printf("invalid gs\n");
            return -1;
        }
        return (gs->getNumberBases());
    }
    char GenomeSequence_getBase(GenomeSequence* gs, unsigned int i) { 
        if (gs) {
            return (*gs)[i];
        };
    };
}

第二步是编译,记住单个C/C++文件编译时使用-fPIC参数,最后打包的时候编译成共享库,摘录Makefile文件中片段如下:

lib:
	g++ -c -fPIC -I./lib GenomeSequence_wrap.c
	g++ -shared -Wl,-soname,libstatgen.so -o libstatgen.so  lib/*.o lib/samtools/k*.o lib/samtools/bgzf.o *.o

最后一步是在Python中写一个封装类,注意前两行引入ctypes库,之后就用这个库调用包装函数就行。
注意:我在GenomeSequence类的__getitem__中使用了如何扩展Python的容器类一文中介绍的一些技巧,这样可以更灵活的使用下标来访问数组中的元素。

from ctypes import cdll
lib = cdll.LoadLibrary("./libstatgen.so")

class GenomeSequence:
    def __init__ (self):
        self.obj = lib.GenomeSequence_new()
    def open(self, filename):
        lib.GenomeSequence_setReferenceName(self.obj, filename)
    def __len__ (self):
        return lib.GenomeSequence_getNumBase(self.obj)
    def __getitem__(self, key):
        if isinstance(key, int):
            return chr(lib.GenomeSequence_getBase(self.obj, key))
        elif isinstance(key, slice):
            return ''.join([self[x] for x in xrange(*key.indices(len(self)))])
        elif isinstance(key, tuple):
            return ''.join([self[i] for i in key])

    def at(self, i):
        return chr(lib.GenomeSequence_getBase(self.obj, i))
    def close(self):
        lib.GenomeSequence_close(self.obj)
    
if __name__ == '__main__':
    gs = GenomeSequence ()
    gs.open("/home/zhanxw/statgen/src/karma/test/phiX.fa");
    print len(gs)
    seq = [(gs.at(i)) for i in xrange(60)]
    print ''.join(seq)
    print gs[0:10],gs[20:30]
    print gs[0:10, 20:30]
    print gs[-10:]
    gs.close()
    print "DONE"

本文主要参考【1】。这里的方法基本重复了【1】中的步骤。写出本文中的代码在于进一步验证ctypes库可以灵活的处理C/C++和Python中的简单数据类型int, char*。

【1】Calling C/C++ from python?

Leave a Reply

Your email address will not be published. Required fields are marked *

*

This site uses Akismet to reduce spam. Learn how your comment data is processed.