PythonCc++扩展资源-CSDN文库

Python

需积分: 10 199 浏览量 2017-11-17 11:20:41 上传评论收藏 327KB PDF 举报

资源推荐

资源详情

资源评论

2017/10/11 1.ExtendingPythonwithCorC++—Python3.6.3documentation

https://docs.python.org/3/extending/extending.html 1/23

1.ExtendingPythonwithCorC++

Itis quiteeasytoaddnewbuiltinmodulestoPython,ifyouknowhowtoprograminC.Such

extension modules can do two things that can’t be done directly in Python: they can

implementnewbuiltinobjecttypes,andtheycancallClibraryfunctionsandsystemcalls.

Tosupportextensions,thePythonAPI(ApplicationProgrammersInterface)definesasetof

functions,macrosandvariablesthatprovideaccesstomostaspectsofthePythonruntime

system.ThePythonAPIisincorporatedinaCsourcefilebyincludingtheheader"Python.h".

The compilation of an extension module depends on its intended use as well as on your

systemsetup;detailsaregiveninlaterchapters.

Note: TheCextensioninterfaceisspecifictoCPython,andextensionmodulesdonot

workonotherPythonimplementations.Inmanycases,itispossibletoavoidwritingC

extensionsandpreserveportabilitytootherimplementations.Forexample,ifyourusecase

iscallingClibraryfunctionsorsystemcalls,youshouldconsiderusingthectypesmodule

orthecffilibraryratherthanwritingcustomCcode.ThesemodulesletyouwritePython

codetointerfacewithCcodeandaremoreportablebetweenimplementationsofPython

thanwritingandcompilingaCextensionmodule.

1.1.ASimpleExample

Let’screateanextensionmodulecalled spam(thefavoritefoodofMontyPythonfans…)and

let’s say we want to create a Python interface to the C library function system(). [1] This

function takesanullterminatedcharacterstringasargumentandreturnsaninteger.Wewant

thisfunctiontobecallablefromPythonasfollows:

>>> import spam

>>> status = spam.system("ls -l")

Begin by creating a file spammodule.c. (Historically, if a module is called spam, the C file

containing its implementation is called spammodule.c; if the module name is very long, like

spammify,themodulenamecanbejustspammify.c.)

Thefirstlineofourfilecanbe:

#include

<Python.h>

whichpullsinthePythonAPI(youcanaddacommentdescribingthepurposeofthemodule

andacopyrightnoticeifyoulike).

Note: SincePythonmaydefinesomepreprocessordefinitionswhichaffectthestandard

headersonsomesystems,youmustincludePython.hbeforeanystandardheadersare

included.

2017/10/11 1.ExtendingPythonwithCorC++—Python3.6.3documentation

https://docs.python.org/3/extending/extending.html 2/23

AlluservisiblesymbolsdefinedbyPython.hhaveaprefixofPyorPY,exceptthosedefinedin

standardheader files.For convenience,andsince theyare usedextensivelyby the Python

interpreter, "Python.h" includes a few standard header files: <stdio.h>, <string.h>,

<errno.h>,and<stdlib.h>.Ifthelatterheaderfiledoesnotexistonyoursystem,itdeclares

thefunctionsmalloc(),free()andrealloc()directly.

ThenextthingweaddtoourmodulefileistheCfunctionthatwillbecalledwhenthePython

expressionspam.system(string)isevaluated(we’llseeshortlyhowitendsupbeingcalled):

static

PyObject *

spam_system(PyObject *self, PyObject *args)

{

const

char *command;

int sts;

(!PyArg_ParseTuple(args, "s", &command))

return

NULL;

sts = system(command);

return

PyLong_FromLong(sts);

}

ThereisastraightforwardtranslationfromtheargumentlistinPython(forexample,thesingle

expression "ls -l")totheargumentspassedtotheCfunction.TheC function alwayshas

twoarguments,conventionallynamedselfandargs.

The self argument points to the module object for modulelevel functions; for a method it

wouldpointtotheobjectinstance.

TheargsargumentwillbeapointertoaPythontupleobjectcontainingthearguments.Each

itemofthetuplecorrespondstoanargumentinthecall’sargumentlist.The arguments are

Pythonobjects—inordertodo anythingwiththeminourCfunctionwehavetoconvertthem

toCvalues.Thefunction PyArg_ParseTuple()inthePythonAPIcheckstheargumenttypes

andconvertsthemtoCvalues.Itusesatemplatestringtodeterminetherequiredtypesofthe

argumentsaswellasthetypesoftheCvariablesinto which to store the converted values.

Moreaboutthislater.

PyArg_ParseTuple() returns true (nonzero) if all arguments have the right type and its

componentshavebeenstoredinthevariableswhoseaddressesarepassed.Itreturnsfalse

(zero)ifan invalid argumentlistwaspassed.Inthelattercase it alsoraisesanappropriate

exceptionsothecallingfunctioncanreturnNULLimmediately(aswesawintheexample).

1.2.Intermezzo:ErrorsandExceptions

An important conventionthroughout thePython interpreteris thefollowing: when afunction

fails,itshouldsetanexceptionconditionandreturnanerrorvalue(usuallyaNULLpointer).

Exceptionsarestoredinastaticglobalvariableinsidetheinterpreter;ifthisvariableisNULL

no exception has occurred. A second global variable stores the “associated value” of the

exception (the second argument to raise). A third variable contains the stack traceback in

casetheerrororiginatedinPythoncode.ThesethreevariablesaretheCequivalentsofthe

2017/10/11 1.ExtendingPythonwithCorC++—Python3.6.3documentation

https://docs.python.org/3/extending/extending.html 3/23

result in Python of sys.exc_info() (see the section on module sys in the Python Library

Reference).Itisimportanttoknowaboutthemtounderstandhowerrorsarepassedaround.

ThePythonAPIdefinesanumberoffunctionstosetvarioustypesofexceptions.

Themostcommononeis PyErr_SetString().ItsargumentsareanexceptionobjectandaC

string.Theexceptionobjectisusuallyapredefinedobjectlike PyExc_ZeroDivisionError.The

CstringindicatesthecauseoftheerrorandisconvertedtoaPythonstringobjectandstored

asthe“associatedvalue”oftheexception.

AnotherusefulfunctionisPyErr_SetFromErrno(),whichonlytakesanexceptionargumentand

constructstheassociatedvaluebyinspectionoftheglobalvariable errno.Themostgeneral

function is PyErr_SetObject(), which takes two object arguments, the exception and its

associatedvalue.Youdon’tneedtoPy_INCREF()theobjectspassedtoanyofthesefunctions.

You can test nondestructively whether an exception has been set with PyErr_Occurred().

Thisreturnsthecurrentexceptionobject,orNULLif noexceptionhasoccurred.Younormally

don’tneedtocallPyErr_Occurred()toseewhetheranerroroccurredinafunctioncall,since

youshouldbeabletotellfromthereturnvalue.

When afunctionfthatcallsanotherfunctiongdetectsthatthelatterfails,fshould itselfreturn

anerrorvalue(usuallyNULLor -1).Itshouldnotcalloneofthe PyErr_*()functions—one

hasalreadybeencalledbyg.f’scalleristhensupposedtoalsoreturnanerrorindicationtoits

caller,againwithoutcallingPyErr_*(),andsoon—themostdetailedcauseoftheerrorwas

already reported by the function that first detected it. Once the error reaches the Python

interpreter’smain loop, this aborts thecurrently executing Python code and tries to findan

exceptionhandlerspecifiedbythePythonprogrammer.

(There are situations where a module can actually give a more detailed error message by

calling another PyErr_*() function,and insuch casesit isfine todo so.As ageneral rule,

however,thisisnotnecessary,andcancauseinformationaboutthecauseoftheerrortobe

lost:mostoperationscanfailforavarietyofreasons.)

To ignore an exception set by a function call that failed, the exception condition must be

clearedexplicitlybycallingPyErr_Clear().TheonlytimeCcodeshouldcallPyErr_Clear()is

ifitdoesn’twantto passtheerrorontotheinterpreterbutwantstohandleitcompletelyby

itself(possiblybytryingsomethingelse,orpretendingnothingwentwrong).

Everyfailingmalloc()callmustbeturnedintoanexception—thedirect callerofmalloc()(or

realloc()) must call PyErr_NoMemory() and return a failure indicator itself. All the object

creating functions (for example, PyLong_FromLong()) already do this, so this note is only

relevanttothosewhocallmalloc()directly.

Alsonotethat,withtheimportantexceptionof PyArg_ParseTuple()andfriends,functionsthat

returnanintegerstatususuallyreturnapositivevalueorzeroforsuccessand -1forfailure,

likeUnixsystemcalls.

Finally, be careful to clean up garbage (by making Py_XDECREF() or Py_DECREF() calls for

objectsyouhavealreadycreated)whenyoureturnanerrorindicator!

2017/10/11 1.ExtendingPythonwithCorC++—Python3.6.3documentation

https://docs.python.org/3/extending/extending.html 4/23

The choice of which exception to raise is entirely yours. There are predeclared C objects

correspondingtoallbuiltinPythonexceptions,suchas PyExc_ZeroDivisionError,whichyou

can use directly. Of course, you should choose exceptions wisely — don’t use

PyExc_TypeError to mean that a file couldn’t be opened (that should probably be

PyExc_IOError).Ifsomething’swrongwiththeargumentlist,the PyArg_ParseTuple()function

usuallyraisesPyExc_TypeError.Ifyouhaveanargumentwhosevaluemustbeinaparticular

rangeormustsatisfyotherconditions,PyExc_ValueErrorisappropriate.

You can also define a new exception that is unique to your module. For this, you usually

declareastaticobjectvariableatthebeginningofyourfile:

static

PyObject *SpamError;

andinitializeitinyourmodule’sinitializationfunction(PyInit_spam())withanexceptionobject

(leavingouttheerrorcheckingfornow):

PyMODINIT_FUNC

PyInit_spam(void)

{

PyObject *m;

m = PyModule_Create(&spammodule);

(m == NULL)

return

NULL;

SpamError = PyErr_NewException("spam.error", NULL, NULL);

Py_INCREF(SpamError);

PyModule_AddObject(m, "error", SpamError);

return

}

NotethatthePythonnamefortheexceptionobjectisspam.error.The PyErr_NewException()

function may create a class with the base class being Exception (unless another class is

passedininsteadofNULL),describedinBuiltinExceptions.

NotealsothattheSpamErrorvariableretainsareferencetothenewlycreatedexceptionclass;

thisisintentional!Sincetheexceptioncouldberemovedfromthemodulebyexternalcode,

an owned reference to the class is needed to ensure that it will not be discarded, causing

SpamErrortobecomeadanglingpointer.Shoulditbecomeadanglingpointer,Ccodewhich

raisestheexceptioncouldcauseacoredumporotherunintendedsideeffects.

WediscusstheuseofPyMODINIT_FUNCasafunctionreturntypelaterinthissample.

The spam.error exception can be raised in your extension module using a call to

PyErr_SetString()asshownbelow:

static

PyObject *

spam_system(PyObject *self, PyObject *args)

{

const

char *command;

int sts;

2017/10/11 1.ExtendingPythonwithCorC++—Python3.6.3documentation

https://docs.python.org/3/extending/extending.html 5/23

(!PyArg_ParseTuple(args, "s", &command))

return

NULL;

sts = system(command);

(sts < 0) {

PyErr_SetString(SpamError, "System command failed");

return

NULL;

}

return

PyLong_FromLong(sts);

}

1.3.BacktotheExample

Goingbacktoourexamplefunction,youshouldnowbeabletounderstandthisstatement:

(!PyArg_ParseTuple(args, "s", &command))

return

NULL;

It returns NULL (the error indicator for functions returning object pointers) if an error is

detectedintheargumentlist,relyingontheexceptionsetby PyArg_ParseTuple().Otherwise

the string value of the argument has been copied to the local variable command. This is a

pointer assignment and you are not supposed to modify the string to which it points (so in

StandardC,thevariablecommandshouldproperlybedeclaredasconst char *command).

The next statementis a callto the Unixfunction system(), passingit thestring wejust got

fromPyArg_ParseTuple():

sts = system(command);

Our spam.system() function must return the value of sts as a Python object. This is done

usingthefunctionPyLong_FromLong().

return

PyLong_FromLong(sts);

In this case, it will return an integer object. (Yes, even integers are objects on the heap in

Python!)

If you have a C function that returns no useful argument (a function returning void), the

corresponding Python function must return None. You need this idiom to do so (which is

implementedbythePy_RETURN_NONEmacro):

Py_INCREF(Py_None);

return

Py_None;

Py_NoneistheCnameforthespecialPythonobjectNone.ItisagenuinePythonobjectrather

thanaNULLpointer,whichmeans“error”inmostcontexts,aswehaveseen.

1.4.TheModule’sMethodTableandInitialization

Function

剩余22页未读，继续阅读

评论收藏

内容反馈

qq_40383507

粉丝: 0
资源: 1

Python C c++扩展

最新资源

Python C c++扩展

使用C++扩展python

使用C/C++扩展Python

python扩展c++示例

c/c++ python

从python转c++

python扩展需要安装的VC++环境（python2.x 和 python3.x 两个版本）

解决c++调用python中文乱码问题

vs2010Python用c++扩展模块详细教程

C++ 扩展python

使用C(C++)扩展 Python（中文文档）

Python扩展C/C++库的方法(C转换为Python)

Python调用C++,通过Pybind11制作Python接口

C/C++转Python库

Python3.6+中文文档.pdf

Python & C/C++联合编程实战

python-3.11.2-docs.epub Python学习资料 Python文档

python3.7.7官方汉化版文档-pdf.zip

python程序的格式以及扩展名

python 3.10.5 Windows 64位安装包

python 3.10.5 MacOS 64位安装包

DAWG-Python:由 dawgdic C++ 库或 DAWG Python 扩展创建的用于 DAWG 的纯 python 阅读器

python-3.90 文档 手册 中文版.chm

PyTorch中的C++扩展实现

python官网 python3.7压缩包

基于Python+pytorch的图像处理+附完整代码图像处理，能够轻松实现图像的读取、显示、裁剪等还有机器学习等操作

python大作业 含爬虫、数据可视化、地图、报告、及源码（2016-2021全国各地区粮食产量）.rar

《点燃我温暖你》中李峋的同款爱心代码

Python金融量化的高级库：TA-Lib-0.4.24（包含python3.7、3.8、3.9、3.10的32位和64位版本）

第十五届蓝桥杯大赛软件赛省赛-PythonB组题目

最新资源

python-3.90 文档手册中文版.chm

python大作业含爬虫、数据可视化、地图、报告、及源码（2016-2021全国各地区粮食产量）.rar