Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1 | \documentstyle[twoside,11pt,myformat]{report} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 2 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 3 | % XXX PM Modulator |
| 4 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 5 | \title{Extending and Embedding the Python Interpreter} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 6 | |
Guido van Rossum | 16cd7f9 | 1994-10-06 10:29:26 +0000 | [diff] [blame] | 7 | \input{boilerplate} |
Guido van Rossum | 83eb962 | 1993-11-23 16:28:45 +0000 | [diff] [blame] | 8 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 9 | % Tell \index to actually write the .idx file |
| 10 | \makeindex |
| 11 | |
| 12 | \begin{document} |
| 13 | |
| 14 | \pagenumbering{roman} |
| 15 | |
| 16 | \maketitle |
| 17 | |
Guido van Rossum | 16cd7f9 | 1994-10-06 10:29:26 +0000 | [diff] [blame] | 18 | \input{copyright} |
| 19 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 20 | \begin{abstract} |
| 21 | |
| 22 | \noindent |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 23 | This document describes how to write modules in C or \Cpp{} to extend the |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 24 | Python interpreter. It also describes how to use Python as an |
| 25 | `embedded' language, and how extension modules can be loaded |
| 26 | dynamically (at run time) into the interpreter, if the operating |
| 27 | system supports this feature. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 28 | |
| 29 | \end{abstract} |
| 30 | |
| 31 | \pagebreak |
| 32 | |
| 33 | { |
| 34 | \parskip = 0mm |
| 35 | \tableofcontents |
| 36 | } |
| 37 | |
| 38 | \pagebreak |
| 39 | |
| 40 | \pagenumbering{arabic} |
| 41 | |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 42 | |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 43 | \chapter{Extending Python with C or \Cpp{} code} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 44 | |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 45 | |
| 46 | \section{Introduction} |
| 47 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 48 | It is quite easy to add non-standard built-in modules to Python, if |
| 49 | you know how to program in C. A built-in module known to the Python |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 50 | programmer as \code{spam} is generally implemented by a file called |
| 51 | \file{spammodule.c} (if the module name is very long, like |
| 52 | \samp{spammify}, you can drop the \samp{module}, leaving a file name |
| 53 | like \file{spammify.c}). The standard built-in modules also adhere to |
| 54 | this convention, and in fact some of them are excellent examples of |
| 55 | how to create an extension. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 56 | |
| 57 | Extension modules can do two things that can't be done directly in |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 58 | Python: they can implement new data types (which are different from |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 59 | classes, by the way), and they can make system calls or call C library |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 60 | functions. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 61 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 62 | To support extensions, the Python API (Application Programmers |
| 63 | Interface) defines many functions, macros and variables that provide |
| 64 | access to almost every aspect of the Python run-time system. |
| 65 | Most of the Python API is imported by including the single header file |
| 66 | \code{"Python.h"}. All user-visible symbols defined by including this |
| 67 | file have a prefix of \samp{Py} or \samp{PY}, except those defined in |
| 68 | standard header files --- for convenience, and since they are needed by |
| 69 | the Python interpreter, \file{"Python.h"} includes a few standard |
| 70 | header files: \file{<stdio.h>}, \file{<string.h>}, \file{<errno.h>}, |
| 71 | and \file{<stdlib.h>}. If the latter header file does not exist on |
| 72 | your system, it declares the functions \code{malloc()}, \code{free()} |
| 73 | and \code{realloc()} itself. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 74 | |
| 75 | The compilation of an extension module depends on your system setup |
| 76 | and the intended use of the module; details are given in a later |
| 77 | section. |
| 78 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 79 | Note: unless otherwise mentioned, all file references in this |
| 80 | document are relative to the Python toplevel directory |
| 81 | (the directory that contains the \file{configure} script). |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 82 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 83 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 84 | \section{A Simple Example} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 85 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 86 | Let's create an extension module called \samp{spam}. Create a file |
| 87 | \samp{spammodule.c}. The first line of this file can be: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 88 | |
| 89 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 90 | #include "Python.h" |
| 91 | \end{verbatim} |
| 92 | |
| 93 | which pulls in the Python API (you can add a comment describing the |
| 94 | purpose of the module and a copyright notice if you like). |
| 95 | |
| 96 | Let's create a Python interface to the C library function |
| 97 | \code{system()}.\footnote{An interface for this function already |
| 98 | exists in the \code{posix} module --- it was chosen as a simple and |
| 99 | straightfoward example.} This function takes a zero-terminated |
| 100 | character string as argument and returns an integer. We will want |
| 101 | this function to be callable from Python as follows: |
| 102 | |
| 103 | \begin{verbatim} |
| 104 | >>> import spam |
| 105 | >>> status = spam.system("ls -l") |
| 106 | \end{verbatim} |
| 107 | |
| 108 | The next thing we add to our module file is the C function that will |
| 109 | be called when the Python expression \samp{spam.system(\var{string})} |
| 110 | is evaluated (well see shortly how it ends up being called): |
| 111 | |
| 112 | \begin{verbatim} |
| 113 | static PyObject * |
| 114 | spam_system(self, args) |
| 115 | PyObject *self; |
| 116 | PyObject *args; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 117 | { |
| 118 | char *command; |
| 119 | int sts; |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 120 | if (!PyArg_ParseTuple(args, "s", &command)) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 121 | return NULL; |
| 122 | sts = system(command); |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 123 | return Py_BuildValue("i", sts); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 124 | } |
| 125 | \end{verbatim} |
| 126 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 127 | There is a straightforward translation from the argument list in |
| 128 | Python (here the single expression \code{"ls -l"}) to the arguments |
| 129 | that are passed to the C function. The C function always has two |
| 130 | arguments, conventionally named \var{self} and \var{args}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 131 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 132 | The \var{self} argument is only used when the C function implements a |
| 133 | builtin method --- this will be discussed later. In the example, |
| 134 | \var{self} will always be a \code{NULL} pointer, since we are defining |
| 135 | a function, not a method. (This is done so that the interpreter |
| 136 | doesn't have to understand two different types of C functions.) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 137 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 138 | The \var{args} argument will be a pointer to a Python tuple object |
| 139 | containing the arguments --- the length of the tuple will be the |
| 140 | number of arguments. It is necessary to do full argument type |
| 141 | checking in each call, since otherwise the Python user would be able |
| 142 | to cause the Python interpreter to crash (rather than raising an |
| 143 | exception) by passing invalid arguments to a function in an extension |
| 144 | module. Because argument checking and converting arguments to C are |
| 145 | such common tasks, there's a general function in the Python |
| 146 | interpreter that combines them: \code{PyArg_ParseTuple()}. It uses a |
| 147 | template string to determine the types of the Python argument and the |
| 148 | types of the C variables into which it should store the converted |
| 149 | values (more about this later). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 150 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 151 | \code{PyArg_ParseTuple()} returns nonzero if all arguments have the |
| 152 | right type and its components have been stored in the variables whose |
| 153 | addresses are passed. It returns zero if an invalid argument was |
| 154 | passed. In the latter case it also raises an appropriate exception by |
| 155 | so the calling function can return \code{NULL} immediately. Here's |
| 156 | why: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 157 | |
| 158 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 159 | \section{Intermezzo: Errors and Exceptions} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 160 | |
| 161 | An important convention throughout the Python interpreter is the |
| 162 | following: when a function fails, it should set an exception condition |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 163 | and return an error value (usually a \code{NULL} pointer). Exceptions |
| 164 | are stored in a static global variable inside the interpreter; if |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 165 | this variable is \code{NULL} no exception has occurred. A second |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 166 | global variable stores the `associated value' of the exception |
| 167 | --- the second argument to \code{raise}. A third variable contains |
| 168 | the stack traceback in case the error originated in Python code. |
| 169 | These three variables are the C equivalents of the Python variables |
| 170 | \code{sys.exc_type}, \code{sys.exc_value} and \code{sys.exc_traceback} |
| 171 | --- see the section on module \code{sys} in the Library Reference |
| 172 | Manual. It is important to know about them to understand how errors |
| 173 | are passed around. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 174 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 175 | The Python API defines a host of functions to set various types of |
| 176 | exceptions. The most common one is \code{PyErr_SetString()} --- its |
| 177 | arguments are an exception object (e.g. \code{PyExc_RuntimeError} --- |
| 178 | actually it can be any object that is a legal exception indicator), |
| 179 | and a C string indicating the cause of the error (this is converted to |
| 180 | a string object and stored as the `associated value' of the |
| 181 | exception). Another useful function is \code{PyErr_SetFromErrno()}, |
| 182 | which only takes an exception argument and constructs the associated |
| 183 | value by inspection of the (\UNIX{}) global variable \code{errno}. The |
| 184 | most general function is \code{PyErr_SetObject()}, which takes two |
| 185 | object arguments, the exception and its associated value. You don't |
| 186 | need to \code{Py_INCREF()} the objects passed to any of these |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 187 | functions. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 188 | |
| 189 | You can test non-destructively whether an exception has been set with |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 190 | \code{PyErr_Occurred()} --- this returns the current exception object, |
| 191 | or \code{NULL} if no exception has occurred. Most code never needs to |
| 192 | call \code{PyErr_Occurred()} to see whether an error occurred or not, |
| 193 | but relies on error return values from the functions it calls instead. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 194 | |
| 195 | When a function that calls another function detects that the called |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 196 | function fails, it should return an error value (e.g. \code{NULL} or |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 197 | \code{-1}). It shouldn't call one of the \code{PyErr_*} functions --- |
| 198 | one has already been called. The caller is then supposed to also |
| 199 | return an error indication to {\em its} caller, again {\em without} |
| 200 | calling \code{PyErr_*()}, and so on --- the most detailed cause of the |
| 201 | error was already reported by the function that first detected it. |
| 202 | Once the error has reached Python's interpreter main loop, this aborts |
| 203 | the currently executing Python code and tries to find an exception |
| 204 | handler specified by the Python programmer. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 205 | |
| 206 | (There are situations where a module can actually give a more detailed |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 207 | error message by calling another \code{PyErr_*} function, and in such |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 208 | cases it is fine to do so. As a general rule, however, this is not |
| 209 | necessary, and can cause information about the cause of the error to |
| 210 | be lost: most operations can fail for a variety of reasons.) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 211 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 212 | To ignore an exception set by a function call that failed, the exception |
| 213 | condition must be cleared explicitly by calling \code{PyErr_Clear()}. |
| 214 | The only time C code should call \code{PyErr_Clear()} is if it doesn't |
| 215 | want to pass the error on to the interpreter but wants to handle it |
| 216 | completely by itself (e.g. by trying something else or pretending |
| 217 | nothing happened). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 218 | |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 219 | Note that a failing \code{malloc()} call must also be turned into an |
| 220 | exception --- the direct caller of \code{malloc()} (or |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 221 | \code{realloc()}) must call \code{PyErr_NoMemory()} and return a |
| 222 | failure indicator itself. All the object-creating functions |
| 223 | (\code{PyInt_FromLong()} etc.) already do this, so only if you call |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 224 | \code{malloc()} directly this note is of importance. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 225 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 226 | Also note that, with the important exception of |
| 227 | \code{PyArg_ParseTuple()}, functions that return an integer status |
| 228 | usually return \code{0} or a positive value for success and \code{-1} |
| 229 | for failure (like \UNIX{} system calls). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 230 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 231 | Finally, be careful about cleaning up garbage (making \code{Py_XDECREF()} |
| 232 | or \code{Py_DECREF()} calls for objects you have already created) when |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 233 | you return an error! |
| 234 | |
| 235 | The choice of which exception to raise is entirely yours. There are |
| 236 | predeclared C objects corresponding to all built-in Python exceptions, |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 237 | e.g. \code{PyExc_ZeroDevisionError} which you can use directly. Of |
| 238 | course, you should chose exceptions wisely --- don't use |
| 239 | \code{PyExc_TypeError} to mean that a file couldn't be opened (that |
| 240 | should probably be \code{PyExc_IOError}). If something's wrong with |
| 241 | the argument list, the \code{PyArg_ParseTuple()} function usually |
| 242 | raises \code{PyExc_TypeError}. If you have an argument whose value |
| 243 | which must be in a particular range or must satisfy other conditions, |
| 244 | \code{PyExc_ValueError} is appropriate. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 245 | |
| 246 | You can also define a new exception that is unique to your module. |
| 247 | For this, you usually declare a static object variable at the |
| 248 | beginning of your file, e.g. |
| 249 | |
| 250 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 251 | static PyObject *SpamError; |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 252 | \end{verbatim} |
| 253 | |
| 254 | and initialize it in your module's initialization function |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 255 | (\code{initspam()}) with a string object, e.g. (leaving out the error |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 256 | checking for simplicity): |
| 257 | |
| 258 | \begin{verbatim} |
| 259 | void |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 260 | initspam() |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 261 | { |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 262 | PyObject *m, *d; |
| 263 | m = Py_InitModule("spam", spam_methods); |
| 264 | d = PyModule_GetDict(m); |
| 265 | SpamError = PyString_FromString("spam.error"); |
| 266 | PyDict_SetItemString(d, "error", SpamError); |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 267 | } |
| 268 | \end{verbatim} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 269 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 270 | Note that the Python name for the exception object is \code{spam.error} |
| 271 | --- it is conventional for module and exception names to be spelled in |
| 272 | lower case. It is also conventional that the \emph{value} of the |
| 273 | exception object is the same as its name, e.g.\ the string |
| 274 | \code{"spam.error"}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 275 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 276 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 277 | \section{Back to the Example} |
| 278 | |
| 279 | Going back to our example function, you should now be able to |
| 280 | understand this statement: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 281 | |
| 282 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 283 | if (!PyArg_ParseTuple(args, "s", &command)) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 284 | return NULL; |
| 285 | \end{verbatim} |
| 286 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 287 | It returns \code{NULL} (the error indicator for functions returning |
| 288 | object pointers) if an error is detected in the argument list, relying |
| 289 | on the exception set by \code{PyArg_ParseTuple()}. Otherwise the |
| 290 | string value of the argument has been copied to the local variable |
| 291 | \code{command}. This is a pointer assignment and you are not supposed |
| 292 | to modify the string to which it points (so in ANSI C, the variable |
| 293 | \code{command} should properly be declared as \code{const char |
| 294 | *command}). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 295 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 296 | The next statement is a call to the \UNIX{} function \code{system()}, |
| 297 | passing it the string we just got from \code{PyArg_ParseTuple()}: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 298 | |
| 299 | \begin{verbatim} |
| 300 | sts = system(command); |
| 301 | \end{verbatim} |
| 302 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 303 | Our \code{spam.system()} function must return a value: the integer |
| 304 | \code{sts} which contains the return value of the \UNIX{} |
| 305 | \code{system()} function. This is done using the function |
| 306 | \code{Py_BuildValue()}, which is something like the inverse of |
| 307 | \code{PyArg_ParseTuple()}: it takes a format string and an arbitrary |
| 308 | number of C values, and returns a new Python object. More info on |
| 309 | \code{Py_BuildValue()} is given later. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 310 | |
| 311 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 312 | return Py_BuildValue("i", sts); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 313 | \end{verbatim} |
| 314 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 315 | In this case, it will return an integer object. (Yes, even integers |
| 316 | are objects on the heap in Python!) |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 317 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 318 | If you have a C function that returns no useful argument (a function |
| 319 | returning \code{void}), the corresponding Python function must return |
| 320 | \code{None}. You need this idiom to do so: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 321 | |
| 322 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 323 | Py_INCREF(Py_None); |
| 324 | return Py_None; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 325 | \end{verbatim} |
| 326 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 327 | \code{Py_None} is the C name for the special Python object |
| 328 | \code{None}. It is a genuine Python object (not a \code{NULL} |
| 329 | pointer, which means `error' in most contexts, as we have seen). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 330 | |
| 331 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 332 | \section{The Module's Method Table and Initialization Function} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 333 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 334 | I promised to show how \code{spam_system()} is called from Python |
| 335 | programs. First, we need to list its name and address in a ``method |
| 336 | table'': |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 337 | |
| 338 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 339 | static PyMethodDef spam_methods[] = { |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 340 | ... |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 341 | {"system", spam_system, 1}, |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 342 | ... |
| 343 | {NULL, NULL} /* Sentinel */ |
| 344 | }; |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 345 | \end{verbatim} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 346 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 347 | Note the third entry (\samp{1}). This is a flag telling the |
| 348 | interpreter the calling convention to be used for the C function. It |
| 349 | should normally always be \samp{1}; a value of \samp{0} means that an |
| 350 | obsolete variant of \code{PyArg_ParseTuple()} is used. |
| 351 | |
| 352 | The method table must be passed to the interpreter in the module's |
| 353 | initialization function (which should be the only non-\code{static} |
| 354 | item defined in the module file): |
| 355 | |
| 356 | \begin{verbatim} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 357 | void |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 358 | initspam() |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 359 | { |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 360 | (void) Py_InitModule("spam", spam_methods); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 361 | } |
| 362 | \end{verbatim} |
| 363 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 364 | When the Python program imports module \code{spam} for the first time, |
| 365 | \code{initspam()} is called. It calls \code{Py_InitModule()}, which |
| 366 | creates a ``module object'' (which is inserted in the dictionary |
| 367 | \code{sys.modules} under the key \code{"spam"}), and inserts built-in |
| 368 | function objects into the newly created module based upon the table |
| 369 | (an array of \code{PyMethodDef} structures) that was passed as its |
| 370 | second argument. \code{Py_InitModule()} returns a pointer to the |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 371 | module object that it creates (which is unused here). It aborts with |
| 372 | a fatal error if the module could not be initialized satisfactorily, |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 373 | so the caller doesn't need to check for errors. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 374 | |
| 375 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 376 | \section{Compilation and Linkage} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 377 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 378 | There are two more things to do before you can use your new extension |
| 379 | module: compiling and linking it with the Python system. If you use |
| 380 | dynamic loading, the details depend on the style of dynamic loading |
| 381 | your system uses; see the chapter on Dynamic Loading for more info |
| 382 | about this. |
| 383 | |
| 384 | If you can't use dynamic loading, or if you want to make your module a |
| 385 | permanent part of the Python interpreter, you will have to change the |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 386 | configuration setup and rebuild the interpreter. Luckily, this is |
| 387 | very simple: just place your file (\file{spammodule.c} for example) in |
| 388 | the \file{Modules} directory, add a line to the file |
| 389 | \file{Modules/Setup} describing your file: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 390 | |
| 391 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 392 | spam spammodule.o |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 393 | \end{verbatim} |
| 394 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 395 | and rebuild the interpreter by running \code{make} in the toplevel |
| 396 | directory. You can also run \code{make} in the \file{Modules} |
| 397 | subdirectory, but then you must first rebuilt the \file{Makefile} |
| 398 | there by running \code{make Makefile}. (This is necessary each time |
| 399 | you change the \file{Setup} file.) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 400 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 401 | If your module requires additional libraries to link with, these can |
| 402 | be listed on the line in the \file{Setup} file as well, for instance: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 403 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 404 | \begin{verbatim} |
| 405 | spam spammodule.o -lX11 |
| 406 | \end{verbatim} |
| 407 | |
| 408 | |
| 409 | \section{Calling Python Functions From C} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 410 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 411 | So far we have concentrated on making C functions callable from |
| 412 | Python. The reverse is also useful: calling Python functions from C. |
| 413 | This is especially the case for libraries that support so-called |
| 414 | `callback' functions. If a C interface makes use of callbacks, the |
| 415 | equivalent Python often needs to provide a callback mechanism to the |
| 416 | Python programmer; the implementation will require calling the Python |
| 417 | callback functions from a C callback. Other uses are also imaginable. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 418 | |
| 419 | Fortunately, the Python interpreter is easily called recursively, and |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 420 | there is a standard interface to call a Python function. (I won't |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 421 | dwell on how to call the Python parser with a particular string as |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 422 | input --- if you're interested, have a look at the implementation of |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 423 | the \samp{-c} command line option in \file{Python/pythonmain.c}.) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 424 | |
| 425 | Calling a Python function is easy. First, the Python program must |
| 426 | somehow pass you the Python function object. You should provide a |
| 427 | function (or some other interface) to do this. When this function is |
| 428 | called, save a pointer to the Python function object (be careful to |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 429 | \code{Py_INCREF()} it!) in a global variable --- or whereever you see fit. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 430 | For example, the following function might be part of a module |
| 431 | definition: |
| 432 | |
| 433 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 434 | static PyObject *my_callback = NULL; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 435 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 436 | static PyObject * |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 437 | my_set_callback(dummy, arg) |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 438 | PyObject *dummy, *arg; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 439 | { |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 440 | Py_XDECREF(my_callback); /* Dispose of previous callback */ |
| 441 | Py_XINCREF(arg); /* Add a reference to new callback */ |
| 442 | my_callback = arg; /* Remember new callback */ |
| 443 | /* Boilerplate to return "None" */ |
| 444 | Py_INCREF(Py_None); |
| 445 | return Py_None; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 446 | } |
| 447 | \end{verbatim} |
| 448 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 449 | The macros \code{Py_XINCREF()} and \code{Py_XDECREF()} increment/decrement |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 450 | the reference count of an object and are safe in the presence of |
| 451 | \code{NULL} pointers. More info on them in the section on Reference |
| 452 | Counts below. |
| 453 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 454 | Later, when it is time to call the function, you call the C function |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 455 | \code{PyEval_CallObject()}. This function has two arguments, both |
| 456 | pointers to arbitrary Python objects: the Python function, and the |
| 457 | argument list. The argument list must always be a tuple object, whose |
| 458 | length is the number of arguments. To call the Python function with |
| 459 | no arguments, pass an empty tuple; to call it with one argument, pass |
| 460 | a singleton tuple. \code{Py_BuildValue()} returns a tuple when its |
| 461 | format string consists of zero or more format codes between |
| 462 | parentheses. For example: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 463 | |
| 464 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 465 | int arg; |
| 466 | PyObject *arglist; |
| 467 | PyObject *result; |
| 468 | ... |
| 469 | arg = 123; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 470 | ... |
| 471 | /* Time to call the callback */ |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 472 | arglist = Py_BuildValue("(i)", arg); |
| 473 | result = PyEval_CallObject(my_callback, arglist); |
| 474 | Py_DECREF(arglist); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 475 | \end{verbatim} |
| 476 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 477 | \code{PyEval_CallObject()} returns a Python object pointer: this is |
| 478 | the return value of the Python function. \code{PyEval_CallObject()} is |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 479 | `reference-count-neutral' with respect to its arguments. In the |
| 480 | example a new tuple was created to serve as the argument list, which |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 481 | is \code{Py_DECREF()}-ed immediately after the call. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 482 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 483 | The return value of \code{PyEval_CallObject()} is ``new'': either it |
| 484 | is a brand new object, or it is an existing object whose reference |
| 485 | count has been incremented. So, unless you want to save it in a |
| 486 | global variable, you should somehow \code{Py_DECREF()} the result, |
| 487 | even (especially!) if you are not interested in its value. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 488 | |
| 489 | Before you do this, however, it is important to check that the return |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 490 | value isn't \code{NULL}. If it is, the Python function terminated by raising |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 491 | an exception. If the C code that called \code{PyEval_CallObject()} is |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 492 | called from Python, it should now return an error indication to its |
| 493 | Python caller, so the interpreter can print a stack trace, or the |
| 494 | calling Python code can handle the exception. If this is not possible |
| 495 | or desirable, the exception should be cleared by calling |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 496 | \code{PyErr_Clear()}. For example: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 497 | |
| 498 | \begin{verbatim} |
| 499 | if (result == NULL) |
| 500 | return NULL; /* Pass error back */ |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 501 | ...use result... |
| 502 | Py_DECREF(result); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 503 | \end{verbatim} |
| 504 | |
| 505 | Depending on the desired interface to the Python callback function, |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 506 | you may also have to provide an argument list to \code{PyEval_CallObject()}. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 507 | In some cases the argument list is also provided by the Python |
| 508 | program, through the same interface that specified the callback |
| 509 | function. It can then be saved and used in the same manner as the |
| 510 | function object. In other cases, you may have to construct a new |
| 511 | tuple to pass as the argument list. The simplest way to do this is to |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 512 | call \code{Py_BuildValue()}. For example, if you want to pass an integral |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 513 | event code, you might use the following code: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 514 | |
| 515 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 516 | PyObject *arglist; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 517 | ... |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 518 | arglist = Py_BuildValue("(l)", eventcode); |
| 519 | result = PyEval_CallObject(my_callback, arglist); |
| 520 | Py_DECREF(arglist); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 521 | if (result == NULL) |
| 522 | return NULL; /* Pass error back */ |
| 523 | /* Here maybe use the result */ |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 524 | Py_DECREF(result); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 525 | \end{verbatim} |
| 526 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 527 | Note the placement of \code{Py_DECREF(argument)} immediately after the call, |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 528 | before the error check! Also note that strictly spoken this code is |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 529 | not complete: \code{Py_BuildValue()} may run out of memory, and this should |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 530 | be checked. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 531 | |
| 532 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 533 | \section{Format Strings for {\tt PyArg_ParseTuple()}} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 534 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 535 | The \code{PyArg_ParseTuple()} function is declared as follows: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 536 | |
| 537 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 538 | int PyArg_ParseTuple(PyObject *arg, char *format, ...); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 539 | \end{verbatim} |
| 540 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 541 | The \var{arg} argument must be a tuple object containing an argument |
| 542 | list passed from Python to a C function. The \var{format} argument |
| 543 | must be a format string, whose syntax is explained below. The |
| 544 | remaining arguments must be addresses of variables whose type is |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 545 | determined by the format string. For the conversion to succeed, the |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 546 | \var{arg} object must match the format and the format must be |
| 547 | exhausted. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 548 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 549 | Note that while \code{PyArg_ParseTuple()} checks that the Python |
| 550 | arguments have the required types, it cannot check the validity of the |
| 551 | addresses of C variables passed to the call: if you make mistakes |
| 552 | there, your code will probably crash or at least overwrite random bits |
| 553 | in memory. So be careful! |
| 554 | |
| 555 | A format string consists of zero or more ``format units''. A format |
| 556 | unit describes one Python object; it is usually a single character or |
| 557 | a parenthesized sequence of format units. With a few exceptions, a |
| 558 | format unit that is not a parenthesized sequence normally corresponds |
| 559 | to a single address argument to \code{PyArg_ParseTuple()}. In the |
| 560 | following description, the quoted form is the format unit; the entry |
| 561 | in (round) parentheses is the Python object type that matches the |
| 562 | format unit; and the entry in [square] brackets is the type of the C |
| 563 | variable(s) whose address should be passed. (Use the \samp{\&} |
| 564 | operator to pass a variable's address.) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 565 | |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 566 | \begin{description} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 567 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 568 | \item[\samp{s} (string) [char *]] |
| 569 | Convert a Python string to a C pointer to a character string. You |
| 570 | must not provide storage for the string itself; a pointer to an |
| 571 | existing string is stored into the character pointer variable whose |
| 572 | address you pass. The C string is null-terminated. The Python string |
| 573 | must not contain embedded null bytes; if it does, a \code{TypeError} |
| 574 | exception is raised. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 575 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 576 | \item[\samp{s\#} (string) {[char *, int]}] |
| 577 | This variant on \code{'s'} stores into two C variables, the first one |
| 578 | a pointer to a character string, the second one its length. In this |
| 579 | case the Python string may contain embedded null bytes. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 580 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 581 | \item[\samp{z} (string or \code{None}) {[char *]}] |
| 582 | Like \samp{s}, but the Python object may also be \code{None}, in which |
| 583 | case the C pointer is set to \code{NULL}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 584 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 585 | \item[\samp{z\#} (string or \code{None}) {[char *, int]}] |
| 586 | This is to \code{'s\#'} as \code{'z'} is to \code{'s'}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 587 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 588 | \item[\samp{b} (integer) {[char]}] |
| 589 | Convert a Python integer to a tiny int, stored in a C \code{char}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 590 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 591 | \item[\samp{h} (integer) {[short int]}] |
| 592 | Convert a Python integer to a C \code{short int}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 593 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 594 | \item[\samp{i} (integer) {[int]}] |
| 595 | Convert a Python integer to a plain C \code{int}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 596 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 597 | \item[\samp{l} (integer) {[long int]}] |
| 598 | Convert a Python integer to a C \code{long int}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 599 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 600 | \item[\samp{c} (string of length 1) {[char]}] |
| 601 | Convert a Python character, represented as a string of length 1, to a |
| 602 | C \code{char}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 603 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 604 | \item[\samp{f} (float) {[float]}] |
| 605 | Convert a Python floating point number to a C \code{float}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 606 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 607 | \item[\samp{d} (float) {[double]}] |
| 608 | Convert a Python floating point number to a C \code{double}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 609 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 610 | \item[\samp{O} (object) {[PyObject *]}] |
| 611 | Store a Python object (without any conversion) in a C object pointer. |
| 612 | The C program thus receives the actual object that was passed. The |
| 613 | object's reference count is not increased. The pointer stored is not |
| 614 | \code{NULL}. |
| 615 | |
| 616 | \item[\samp{O!} (object) {[\var{typeobject}, PyObject *]}] |
| 617 | Store a Python object in a C object pointer. This is similar to |
| 618 | \samp{O}, but takes two C arguments: the first is the address of a |
| 619 | Python type object, the second is the address of the C variable (of |
| 620 | type \code{PyObject *}) into which the object pointer is stored. |
| 621 | If the Python object does not have the required type, a |
| 622 | \code{TypeError} exception is raised. |
| 623 | |
| 624 | \item[\samp{O\&} (object) {[\var{converter}, \var{anything}]}] |
| 625 | Convert a Python object to a C variable through a \var{converter} |
| 626 | function. This takes two arguments: the first is a function, the |
| 627 | second is the address of a C variable (of arbitrary type), converted |
| 628 | to \code{void *}. The \var{converter} function in turn is called as |
| 629 | follows: |
| 630 | |
| 631 | \code{\var{status} = \var{converter}(\var{object}, \var{address});} |
| 632 | |
| 633 | where \var{object} is the Python object to be converted and |
| 634 | \var{address} is the \code{void *} argument that was passed to |
| 635 | \code{PyArg_ConvertTuple()}. The returned \var{status} should be |
| 636 | \code{1} for a successful conversion and \code{0} if the conversion |
| 637 | has failed. When the conversion fails, the \var{converter} function |
| 638 | should raise an exception. |
| 639 | |
| 640 | \item[\samp{S} (string) {[PyStringObject *]}] |
| 641 | Like \samp{O} but raises a \code{TypeError} exception that the object |
| 642 | is a string object. The C variable may also be declared as |
| 643 | \code{PyObject *}. |
| 644 | |
| 645 | \item[\samp{(\var{items})} (tuple) {[\var{matching-items}]}] |
| 646 | The object must be a Python tuple whose length is the number of format |
| 647 | units in \var{items}. The C arguments must correspond to the |
| 648 | individual format units in \var{items}. Format units for tuples may |
| 649 | be nested. |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 650 | |
| 651 | \end{description} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 652 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 653 | It is possible to pass Python long integers where integers are |
| 654 | requested; however no proper range checking is done -- the most |
| 655 | significant bits are silently truncated when the receiving field is |
| 656 | too small to receive the value (actually, the semantics are inherited |
| 657 | from downcasts in C --- your milage may vary). |
| 658 | |
| 659 | A few other characters have a meaning in a format string. These may |
| 660 | not occur inside nested parentheses. They are: |
| 661 | |
| 662 | \begin{description} |
| 663 | |
| 664 | \item[\samp{|}] |
| 665 | Indicates that the remaining arguments in the Python argument list are |
| 666 | optional. The C variables corresponding to optional arguments should |
| 667 | be initialized to their default value --- when an optional argument is |
| 668 | not specified, the \code{PyArg_ParseTuple} does not touch the contents |
| 669 | of the corresponding C variable(s). |
| 670 | |
| 671 | \item[\samp{:}] |
| 672 | The list of format units ends here; the string after the colon is used |
| 673 | as the function name in error messages (the ``associated value'' of |
| 674 | the exceptions that \code{PyArg_ParseTuple} raises). |
| 675 | |
| 676 | \item[\samp{;}] |
| 677 | The list of format units ends here; the string after the colon is used |
| 678 | as the error message \emph{instead} of the default error message. |
| 679 | Clearly, \samp{:} and \samp{;} mutually exclude each other. |
| 680 | |
| 681 | \end{description} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 682 | |
| 683 | Some example calls: |
| 684 | |
| 685 | \begin{verbatim} |
| 686 | int ok; |
| 687 | int i, j; |
| 688 | long k, l; |
| 689 | char *s; |
| 690 | int size; |
| 691 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 692 | ok = PyArg_ParseTuple(args, ""); /* No arguments */ |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 693 | /* Python call: f() */ |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 694 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 695 | ok = PyArg_ParseTuple(args, "s", &s); /* A string */ |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 696 | /* Possible Python call: f('whoops!') */ |
| 697 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 698 | ok = PyArg_ParseTuple(args, "lls", &k, &l, &s); /* Two longs and a string */ |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 699 | /* Possible Python call: f(1, 2, 'three') */ |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 700 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 701 | ok = PyArg_ParseTuple(args, "(ii)s#", &i, &j, &s, &size); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 702 | /* A pair of ints and a string, whose size is also returned */ |
| 703 | /* Possible Python call: f(1, 2, 'three') */ |
| 704 | |
| 705 | { |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 706 | char *file; |
| 707 | char *mode = "r"; |
| 708 | int bufsize = 0; |
| 709 | ok = PyArg_ParseTuple(args, "s|si", &file, &mode, &bufsize); |
| 710 | /* A string, and optionally another string and an integer */ |
| 711 | /* Possible Python calls: |
| 712 | f('spam') |
| 713 | f('spam', 'w') |
| 714 | f('spam', 'wb', 100000) */ |
| 715 | } |
| 716 | |
| 717 | { |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 718 | int left, top, right, bottom, h, v; |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 719 | ok = PyArg_ParseTuple(args, "((ii)(ii))(ii)", |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 720 | &left, &top, &right, &bottom, &h, &v); |
| 721 | /* A rectangle and a point */ |
| 722 | /* Possible Python call: |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 723 | f(((0, 0), (400, 300)), (10, 10)) */ |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 724 | } |
| 725 | \end{verbatim} |
| 726 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 727 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 728 | \section{The {\tt Py_BuildValue()} Function} |
| 729 | |
| 730 | This function is the counterpart to \code{PyArg_ParseTuple()}. It is |
| 731 | declared as follows: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 732 | |
| 733 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 734 | PyObject *Py_BuildValue(char *format, ...); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 735 | \end{verbatim} |
| 736 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 737 | It recognizes a set of format units similar to the ones recognized by |
| 738 | \code{PyArg_ParseTuple()}, but the arguments (which are input to the |
| 739 | function, not output) must not be pointers, just values. It returns a |
| 740 | new Python object, suitable for returning from a C function called |
| 741 | from Python. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 742 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 743 | One difference with \code{PyArg_ParseTuple()}: while the latter |
| 744 | requires its first argument to be a tuple (since Python argument lists |
| 745 | are always represented as tuples internally), \code{BuildValue()} does |
| 746 | not always build a tuple. It builds a tuple only if its format string |
| 747 | contains two or more format units. If the format string is empty, it |
| 748 | returns \code{None}; if it contains exactly one format unit, it |
| 749 | returns whatever object is described by that format unit. To force it |
| 750 | to return a tuple of size 0 or one, parenthesize the format string. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 751 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 752 | In the following description, the quoted form is the format unit; the |
| 753 | entry in (round) parentheses is the Python object type that the format |
| 754 | unit will return; and the entry in [square] brackets is the type of |
| 755 | the C value(s) to be passed. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 756 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 757 | The characters space, tab, colon and comma are ignored in format |
| 758 | strings (but not within format units such as \samp{s\#}). This can be |
| 759 | used to make long format strings a tad more readable. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 760 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 761 | \begin{description} |
| 762 | |
| 763 | \item[\samp{s} (string) {[char *]}] |
| 764 | Convert a null-terminated C string to a Python object. If the C |
| 765 | string pointer is \code{NULL}, \code{None} is returned. |
| 766 | |
| 767 | \item[\samp{s\#} (string) {[char *, int]}] |
| 768 | Convert a C string and its length to a Python object. If the C string |
| 769 | pointer is \code{NULL}, the length is ignored and \code{None} is |
| 770 | returned. |
| 771 | |
| 772 | \item[\samp{z} (string or \code{None}) {[char *]}] |
| 773 | Same as \samp{s}. |
| 774 | |
| 775 | \item[\samp{z\#} (string or \code{None}) {[char *, int]}] |
| 776 | Same as \samp{s\#}. |
| 777 | |
| 778 | \item[\samp{i} (integer) {[int]}] |
| 779 | Convert a plain C \code{int} to a Python integer object. |
| 780 | |
| 781 | \item[\samp{b} (integer) {[char]}] |
| 782 | Same as \samp{i}. |
| 783 | |
| 784 | \item[\samp{h} (integer) {[short int]}] |
| 785 | Same as \samp{i}. |
| 786 | |
| 787 | \item[\samp{l} (integer) {[long int]}] |
| 788 | Convert a C \code{long int} to a Python integer object. |
| 789 | |
| 790 | \item[\samp{c} (string of length 1) {[char]}] |
| 791 | Convert a C \code{int} representing a character to a Python string of |
| 792 | length 1. |
| 793 | |
| 794 | \item[\samp{d} (float) {[double]}] |
| 795 | Convert a C \code{double} to a Python floating point number. |
| 796 | |
| 797 | \item[\samp{f} (float) {[float]}] |
| 798 | Same as \samp{d}. |
| 799 | |
| 800 | \item[\samp{O} (object) {[PyObject *]}] |
| 801 | Pass a Python object untouched (except for its reference count, which |
| 802 | is incremented by one). If the object passed in is a \code{NULL} |
| 803 | pointer, it is assumed that this was caused because the call producing |
| 804 | the argument found an error and set an exception. Therefore, |
| 805 | \code{Py_BuildValue()} will return \code{NULL} but won't raise an |
| 806 | exception. If no exception has been raised yet, |
| 807 | \code{PyExc_SystemError} is set. |
| 808 | |
| 809 | \item[\samp{S} (object) {[PyObject *]}] |
| 810 | Same as \samp{O}. |
| 811 | |
| 812 | \item[\samp{O\&} (object) {[\var{converter}, \var{anything}]}] |
| 813 | Convert \var{anything} to a Python object through a \var{converter} |
| 814 | function. The function is called with \var{anything} (which should be |
| 815 | compatible with \code{void *}) as its argument and should return a |
| 816 | ``new'' Python object, or \code{NULL} if an error occurred. |
| 817 | |
| 818 | \item[\samp{(\var{items})} (tuple) {[\var{matching-items}]}] |
| 819 | Convert a sequence of C values to a Python tuple with the same number |
| 820 | of items. |
| 821 | |
| 822 | \item[\samp{[\var{items}]} (list) {[\var{matching-items}]}] |
| 823 | Convert a sequence of C values to a Python list with the same number |
| 824 | of items. |
| 825 | |
| 826 | \item[\samp{\{\var{items}\}} (dictionary) {[\var{matching-items}]}] |
| 827 | Convert a sequence of C values to a Python dictionary. Each pair of |
| 828 | consecutive C values adds one item to the dictionary, serving as key |
| 829 | and value, respectively. |
| 830 | |
| 831 | \end{description} |
| 832 | |
| 833 | If there is an error in the format string, the |
| 834 | \code{PyExc_SystemError} exception is raised and \code{NULL} returned. |
| 835 | |
| 836 | Examples (to the left the call, to the right the resulting Python value): |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 837 | |
| 838 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 839 | Py_BuildValue("") None |
| 840 | Py_BuildValue("i", 123) 123 |
Guido van Rossum | f23e0fe | 1995-03-18 11:04:29 +0000 | [diff] [blame^] | 841 | Py_BuildValue("iii", 123, 456, 789) (123, 456, 789) |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 842 | Py_BuildValue("s", "hello") 'hello' |
| 843 | Py_BuildValue("ss", "hello", "world") ('hello', 'world') |
| 844 | Py_BuildValue("s#", "hello", 4) 'hell' |
| 845 | Py_BuildValue("()") () |
| 846 | Py_BuildValue("(i)", 123) (123,) |
| 847 | Py_BuildValue("(ii)", 123, 456) (123, 456) |
| 848 | Py_BuildValue("(i,i)", 123, 456) (123, 456) |
| 849 | Py_BuildValue("[i,i]", 123, 456) [123, 456] |
Guido van Rossum | f23e0fe | 1995-03-18 11:04:29 +0000 | [diff] [blame^] | 850 | Py_BuildValue("{s:i,s:i}", |
| 851 | "abc", 123, "def", 456) {'abc': 123, 'def': 456} |
| 852 | Py_BuildValue("((ii)(ii)) (ii)", |
| 853 | 1, 2, 3, 4, 5, 6) (((1, 2), (3, 4)), (5, 6)) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 854 | \end{verbatim} |
| 855 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 856 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 857 | \section{Reference Counts} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 858 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 859 | \subsection{Introduction} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 860 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 861 | In languages like C or \Cpp{}, the programmer is responsible for |
| 862 | dynamic allocation and deallocation of memory on the heap. In C, this |
| 863 | is done using the functions \code{malloc()} and \code{free()}. In |
| 864 | \Cpp{}, the operators \code{new} and \code{delete} are used with |
| 865 | essentially the same meaning; they are actually implemented using |
| 866 | \code{malloc()} and \code{free()}, so we'll restrict the following |
| 867 | discussion to the latter. |
| 868 | |
| 869 | Every block of memory allocated with \code{malloc()} should eventually |
| 870 | be returned to the pool of available memory by exactly one call to |
| 871 | \code{free()}. It is important to call \code{free()} at the right |
| 872 | time. If a block's address is forgotten but \code{free()} is not |
| 873 | called for it, the memory it occupies cannot be reused until the |
| 874 | program terminates. This is called a \dfn{memory leak}. On the other |
| 875 | hand, if a program calls \code{free()} for a block and then continues |
| 876 | to use the block, it creates a conflict with re-use of the block |
| 877 | through another \code{malloc()} call. This is called \dfn{using freed |
| 878 | memory} has the same bad consequences as referencing uninitialized |
| 879 | data --- core dumps, wrong results, mysterious crashes. |
| 880 | |
| 881 | Common causes of memory leaks are unusual paths through the code. For |
| 882 | instance, a function may allocate a block of memory, do some |
| 883 | calculation, and then free the block again. Now a change in the |
| 884 | requirements for the function may add a test to the calculation that |
| 885 | detects an error condition and can return prematurely from the |
| 886 | function. It's easy to forget to free the allocated memory block when |
| 887 | taking this premature exit, especially when it is added later to the |
| 888 | code. Such leaks, once introduced, often go undetected for a long |
| 889 | time: the error exit is taken only in a small fraction of all calls, |
| 890 | and most modern machines have plenty of virtual memory, so the leak |
| 891 | only becomes apparent in a long-running process that uses the leaking |
| 892 | function frequently. Therefore, it's important to prevent leaks from |
| 893 | happening by having a coding convention or strategy that minimizes |
| 894 | this kind of errors. |
| 895 | |
| 896 | Since Python makes heavy use of \code{malloc()} and \code{free()}, it |
| 897 | needs a strategy to avoid memory leaks as well as the use of freed |
| 898 | memory. The chosen method is called \dfn{reference counting}. The |
| 899 | principle is simple: every object contains a counter, which is |
| 900 | incremented when a reference to the object is stored somewhere, and |
| 901 | which is decremented when a reference to it is deleted. When the |
| 902 | counter reaches zero, the last reference to the object has been |
| 903 | deleted and the object is freed. |
| 904 | |
| 905 | An alternative strategy is called \dfn{automatic garbage collection}. |
| 906 | (Sometimes, reference counting is also referred to as a garbage |
| 907 | collection strategy, hence my use of ``automatic'' to distinguish the |
| 908 | two.) The big advantage of automatic garbage collection is that the |
| 909 | user doesn't need to call \code{free()} explicitly. (Another claimed |
| 910 | advantage is an improvement in speed or memory usage --- this is no |
| 911 | hard fact however.) The disadvantage is that for C, there is no |
| 912 | truly portable automatic garbage collector, while reference counting |
| 913 | can be implemented portably (as long as the functions \code{malloc()} |
| 914 | and \code{free()} are available --- which the C Standard guarantees). |
| 915 | Maybe some day a sufficiently portable automatic garbage collector |
| 916 | will be available for C. Until then, we'll have to live with |
| 917 | reference counts. |
| 918 | |
| 919 | \subsection{Reference Counting in Python} |
| 920 | |
| 921 | There are two macros, \code{Py_INCREF(x)} and \code{Py_DECREF(x)}, |
| 922 | which handle the incrementing and decrementing of the reference count. |
| 923 | \code{Py_DECREF()} also frees the object when the count reaches zero. |
| 924 | For flexibility, it doesn't call \code{free()} directly --- rather, it |
| 925 | makes a call through a function pointer in the object's \dfn{type |
| 926 | object}. For this purpose (and others), every object also contains a |
| 927 | pointer to its type object. |
| 928 | |
| 929 | The big question now remains: when to use \code{Py_INCREF(x)} and |
| 930 | \code{Py_DECREF(x)}? Let's first introduce some terms. Nobody |
| 931 | ``owns'' an object; however, you can \dfn{own a reference} to an |
| 932 | object. An object's reference count is now defined as the number of |
| 933 | owned references to it. The owner of a reference is responsible for |
| 934 | calling \code{Py_DECREF()} when the reference is no longer needed. |
| 935 | Ownership of a reference can be transferred. There are three ways to |
| 936 | dispose of an owned reference: pass it on, store it, or call |
| 937 | \code{Py_DECREF()}. Forgetting to dispose of an owned reference creates |
| 938 | a memory leak. |
| 939 | |
| 940 | It is also possible to \dfn{borrow}\footnote{The metaphor of |
| 941 | ``borrowing'' a reference is not completely correct: the owner still |
| 942 | has a copy of the reference.} a reference to an object. The borrower |
| 943 | of a reference should not call \code{Py_DECREF()}. The borrower must |
| 944 | not hold on to the object longer than the owner from which it was |
| 945 | borrowed. Using a borrowed reference after the owner has disposed of |
| 946 | it risks using freed memory and should be avoided |
| 947 | completely.\footnote{Checking that the reference count is at least 1 |
| 948 | \strong{does not work} --- the reference count itself could be in |
| 949 | freed memory and may thus be reused for another object!} |
| 950 | |
| 951 | The advantage of borrowing over owning a reference is that you don't |
| 952 | need to take care of disposing of the reference on all possible paths |
| 953 | through the code --- in other words, with a borrowed reference you |
| 954 | don't run the risk of leaking when a premature exit is taken. The |
| 955 | disadvantage of borrowing over leaking is that there are some subtle |
| 956 | situations where in seemingly correct code a borrowed reference can be |
| 957 | used after the owner from which it was borrowed has in fact disposed |
| 958 | of it. |
| 959 | |
| 960 | A borrowed reference can be changed into an owned reference by calling |
| 961 | \code{Py_INCREF()}. This does not affect the status of the owner from |
| 962 | which the reference was borrowed --- it creates a new owned reference, |
| 963 | and gives full owner responsibilities (i.e., the new owner must |
| 964 | dispose of the reference properly, as well as the previous owner). |
| 965 | |
| 966 | \subsection{Ownership Rules} |
| 967 | |
| 968 | Whenever an object reference is passed into or out of a function, it |
| 969 | is part of the function's interface specification whether ownership is |
| 970 | transferred with the reference or not. |
| 971 | |
| 972 | Most functions that return a reference to an object pass on ownership |
| 973 | with the reference. In particular, all functions whose function it is |
| 974 | to create a new object, e.g.\ \code{PyInt_FromLong()} and |
| 975 | \code{Py_BuildValue()}, pass ownership to the receiver. Even if in |
| 976 | fact, in some cases, you don't receive a reference to a brand new |
| 977 | object, you still receive ownership of the reference. For instance, |
| 978 | \code{PyInt_FromLong()} maintains a cache of popular values and can |
| 979 | return a reference to a cached item. |
| 980 | |
| 981 | Many functions that extract objects from other objects also transfer |
| 982 | ownership with the reference, for instance |
| 983 | \code{PyObject_GetAttrString()}. The picture is less clear, here, |
| 984 | however, since a few common routines are exceptions: |
| 985 | \code{PyTuple_GetItem()}, \code{PyList_GetItem()} and |
| 986 | \code{PyDict_GetItem()} (and \code{PyDict_GetItemString()}) all return |
| 987 | references that you borrow from the tuple, list or dictionary. |
| 988 | |
| 989 | The function \code{PyImport_AddModule()} also returns a borrowed |
| 990 | reference, even though it may actually create the object it returns: |
| 991 | this is possible because an owned reference to the object is stored in |
| 992 | \code{sys.modules}. |
| 993 | |
| 994 | When you pass an object reference into another function, in general, |
| 995 | the function borrows the reference from you --- if it needs to store |
| 996 | it, it will use \code{Py_INCREF()} to become an independent owner. |
| 997 | There are exactly two important exceptions to this rule: |
| 998 | \code{PyTuple_SetItem()} and \code{PyList_SetItem()}. These functions |
| 999 | take over ownership of the item passed to them --- even if they fail! |
| 1000 | (Note that \code{PyDict_SetItem()} and friends don't take over |
| 1001 | ownership --- they are ``normal''.) |
| 1002 | |
| 1003 | When a C function is called from Python, it borrows references to its |
| 1004 | arguments from the caller. The caller owns a reference to the object, |
| 1005 | so the borrowed reference's lifetime is guaranteed until the function |
| 1006 | returns. Only when such a borrowed reference must be stored or passed |
| 1007 | on, it must be turned into an owned reference by calling |
| 1008 | \code{Py_INCREF()}. |
| 1009 | |
| 1010 | The object reference returned from a C function that is called from |
| 1011 | Python must be an owned reference --- ownership is tranferred from the |
| 1012 | function to its caller. |
| 1013 | |
| 1014 | \subsection{Thin Ice} |
| 1015 | |
| 1016 | There are a few situations where seemingly harmless use of a borrowed |
| 1017 | reference can lead to problems. These all have to do with implicit |
| 1018 | invocations of the interpreter, which can cause the owner of a |
| 1019 | reference to dispose of it. |
| 1020 | |
| 1021 | The first and most important case to know about is using |
| 1022 | \code{Py_DECREF()} on an unrelated object while borrowing a reference |
| 1023 | to a list item. For instance: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1024 | |
| 1025 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1026 | bug(PyObject *list) { |
| 1027 | PyObject *item = PyList_GetItem(list, 0); |
| 1028 | PyList_SetItem(list, 1, PyInt_FromLong(0L)); |
| 1029 | PyObject_Print(item, stdout, 0); /* BUG! */ |
| 1030 | } |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1031 | \end{verbatim} |
| 1032 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1033 | This function first borrows a reference to \code{list[0]}, then |
| 1034 | replaces \code{list[1]} with the value \code{0}, and finally prints |
| 1035 | the borrowed reference. Looks harmless, right? But it's not! |
| 1036 | |
| 1037 | Let's follow the control flow into \code{PyList_SetItem()}. The list |
| 1038 | owns references to all its items, so when item 1 is replaced, it has |
| 1039 | to dispose of the original item 1. Now let's suppose the original |
| 1040 | item 1 was an instance of a user-defined class, and let's further |
| 1041 | suppose that the class defined a \code{__del__()} method. If this |
| 1042 | class instance has a reference count of 1, disposing of it will call |
| 1043 | its \code{__del__()} method. |
| 1044 | |
| 1045 | Since it is written in Python, the \code{__del__()} method can execute |
| 1046 | arbitrary Python code. Could it perhaps do something to invalidate |
| 1047 | the reference to \code{item} in \code{bug()}? You bet! Assuming that |
| 1048 | the list passed into \code{bug()} is accessible to the |
| 1049 | \code{__del__()} method, it could execute a statement to the effect of |
| 1050 | \code{del list[0]}, and assuming this was the last reference to that |
| 1051 | object, it would free the memory associated with it, thereby |
| 1052 | invalidating \code{item}. |
| 1053 | |
| 1054 | The solution, once you know the source of the problem, is easy: |
| 1055 | temporarily increment the reference count. The correct version of the |
| 1056 | function reads: |
| 1057 | |
| 1058 | \begin{verbatim} |
| 1059 | no_bug(PyObject *list) { |
| 1060 | PyObject *item = PyList_GetItem(list, 0); |
| 1061 | Py_INCREF(item); |
| 1062 | PyList_SetItem(list, 1, PyInt_FromLong(0L)); |
| 1063 | PyObject_Print(item, stdout, 0); |
| 1064 | Py_DECREF(item); |
| 1065 | } |
| 1066 | \end{verbatim} |
| 1067 | |
| 1068 | This is a true story. An older version of Python contained variants |
| 1069 | of this bug and someone spent a considerable amount of time in a C |
| 1070 | debugger to figure out why his \code{__del__()} methods would fail... |
| 1071 | |
| 1072 | The second case of problems with a borrowed reference is a variant |
| 1073 | involving threads. Normally, multiple threads in the Python |
| 1074 | interpreter can't get in each other's way, because there is a global |
| 1075 | lock protecting Python's entire object space. However, it is possible |
| 1076 | to temporarily release this lock using the macro |
| 1077 | \code{Py_BEGIN_ALLOW_THREADS}, and to re-acquire it using |
| 1078 | \code{Py_END_ALLOW_THREADS}. This is common around blocking I/O |
| 1079 | calls, to let other threads use the CPU while waiting for the I/O to |
| 1080 | complete. Obviously, the following function has the same problem as |
| 1081 | the previous one: |
| 1082 | |
| 1083 | \begin{verbatim} |
| 1084 | bug(PyObject *list) { |
| 1085 | PyObject *item = PyList_GetItem(list, 0); |
| 1086 | Py_BEGIN_ALLOW_THREADS |
| 1087 | ...some blocking I/O call... |
| 1088 | Py_END_ALLOW_THREADS |
| 1089 | PyObject_Print(item, stdout, 0); /* BUG! */ |
| 1090 | } |
| 1091 | \end{verbatim} |
| 1092 | |
| 1093 | \subsection{NULL Pointers} |
| 1094 | |
| 1095 | In general, functions that take object references as arguments don't |
| 1096 | expect you to pass them \code{NULL} pointers, and will dump core (or |
| 1097 | cause later core dumps) if you do so. Functions that return object |
| 1098 | references generally return \code{NULL} only to indicate that an |
| 1099 | exception occurred. The reason for not testing for \code{NULL} |
| 1100 | arguments is that functions often pass the objects they receive on to |
| 1101 | other function --- if each function were to test for \code{NULL}, |
| 1102 | there would be a lot of redundant tests and the code would run slower. |
| 1103 | |
| 1104 | It is better to test for \code{NULL} only at the ``source'', i.e.\ |
| 1105 | when a pointer that may be \code{NULL} is received, e.g.\ from |
| 1106 | \code{malloc()} or from a function that may raise an exception. |
| 1107 | |
| 1108 | The macros \code{Py_INCREF()} and \code{Py_DECREF()} |
| 1109 | don't check for \code{NULL} pointers --- however, their variants |
| 1110 | \code{Py_XINCREF()} and \code{Py_XDECREF()} do. |
| 1111 | |
| 1112 | The macros for checking for a particular object type |
| 1113 | (\code{Py\var{type}_Check()}) don't check for \code{NULL} pointers --- |
| 1114 | again, there is much code that calls several of these in a row to test |
| 1115 | an object against various different expected types, and this would |
| 1116 | generate redundant tests. There are no variants with \code{NULL} |
| 1117 | checking. |
| 1118 | |
| 1119 | The C function calling mechanism guarantees that the argument list |
| 1120 | passed to C functions (\code{args} in the examples) is never |
| 1121 | \code{NULL} --- in fact it guarantees that it is always a tuple.% |
| 1122 | \footnote{These guarantees don't hold when you use the ``old'' style |
| 1123 | calling convention --- this is still found in much existing code.} |
| 1124 | |
| 1125 | It is a severe error to ever let a \code{NULL} pointer ``escape'' to |
| 1126 | the Python user. |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1127 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1128 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1129 | \section{Writing Extensions in \Cpp{}} |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1130 | |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 1131 | It is possible to write extension modules in \Cpp{}. Some restrictions |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1132 | apply: since the main program (the Python interpreter) is compiled and |
| 1133 | linked by the C compiler, global or static objects with constructors |
| 1134 | cannot be used. All functions that will be called directly or |
| 1135 | indirectly (i.e. via function pointers) by the Python interpreter will |
| 1136 | have to be declared using \code{extern "C"}; this applies to all |
| 1137 | `methods' as well as to the module's initialization function. |
| 1138 | It is unnecessary to enclose the Python header files in |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1139 | \code{extern "C" \{...\}} --- they use this form already if the symbol |
| 1140 | \samp{__cplusplus} is defined (all recent C++ compilers define this |
| 1141 | symbol). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1142 | |
| 1143 | \chapter{Embedding Python in another application} |
| 1144 | |
| 1145 | Embedding Python is similar to extending it, but not quite. The |
| 1146 | difference is that when you extend Python, the main program of the |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 1147 | application is still the Python interpreter, while if you embed |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1148 | Python, the main program may have nothing to do with Python --- |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1149 | instead, some parts of the application occasionally call the Python |
| 1150 | interpreter to run some Python code. |
| 1151 | |
| 1152 | So if you are embedding Python, you are providing your own main |
| 1153 | program. One of the things this main program has to do is initialize |
| 1154 | the Python interpreter. At the very least, you have to call the |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1155 | function \code{Py_Initialize()}. There are optional calls to pass command |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1156 | line arguments to Python. Then later you can call the interpreter |
| 1157 | from any part of the application. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1158 | |
| 1159 | There are several different ways to call the interpreter: you can pass |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1160 | a string containing Python statements to \code{PyRun_SimpleString()}, |
| 1161 | or you can pass a stdio file pointer and a file name (for |
| 1162 | identification in error messages only) to \code{PyRun_SimpleFile()}. You |
| 1163 | can also call the lower-level operations described in the previous |
| 1164 | chapters to construct and use Python objects. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1165 | |
| 1166 | A simple demo of embedding Python can be found in the directory |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1167 | \file{Demo/embed}. |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1168 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1169 | |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 1170 | \section{Embedding Python in \Cpp{}} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1171 | |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 1172 | It is also possible to embed Python in a \Cpp{} program; precisely how this |
| 1173 | is done will depend on the details of the \Cpp{} system used; in general you |
| 1174 | will need to write the main program in \Cpp{}, and use the \Cpp{} compiler |
| 1175 | to compile and link your program. There is no need to recompile Python |
| 1176 | itself using \Cpp{}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1177 | |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1178 | |
| 1179 | \chapter{Dynamic Loading} |
| 1180 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1181 | On most modern systems it is possible to configure Python to support |
| 1182 | dynamic loading of extension modules implemented in C. When shared |
| 1183 | libraries are used dynamic loading is configured automatically; |
| 1184 | otherwise you have to select it as a build option (see below). Once |
| 1185 | configured, dynamic loading is trivial to use: when a Python program |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1186 | executes \code{import spam}, the search for modules tries to find a |
| 1187 | file \file{spammodule.o} (\file{spammodule.so} when using shared |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1188 | libraries) in the module search path, and if one is found, it is |
| 1189 | loaded into the executing binary and executed. Once loaded, the |
| 1190 | module acts just like a built-in extension module. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1191 | |
| 1192 | The advantages of dynamic loading are twofold: the `core' Python |
| 1193 | binary gets smaller, and users can extend Python with their own |
| 1194 | modules implemented in C without having to build and maintain their |
| 1195 | own copy of the Python interpreter. There are also disadvantages: |
| 1196 | dynamic loading isn't available on all systems (this just means that |
| 1197 | on some systems you have to use static loading), and dynamically |
| 1198 | loading a module that was compiled for a different version of Python |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1199 | (e.g. with a different representation of objects) may dump core. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1200 | |
| 1201 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1202 | \section{Configuring and Building the Interpreter for Dynamic Loading} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1203 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1204 | There are three styles of dynamic loading: one using shared libraries, |
| 1205 | one using SGI IRIX 4 dynamic loading, and one using GNU dynamic |
| 1206 | loading. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1207 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1208 | \subsection{Shared Libraries} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1209 | |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 1210 | The following systems support dynamic loading using shared libraries: |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1211 | SunOS 4; Solaris 2; SGI IRIX 5 (but not SGI IRIX 4!); and probably all |
| 1212 | systems derived from SVR4, or at least those SVR4 derivatives that |
| 1213 | support shared libraries (are there any that don't?). |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1214 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1215 | You don't need to do anything to configure dynamic loading on these |
| 1216 | systems --- the \file{configure} detects the presence of the |
| 1217 | \file{<dlfcn.h>} header file and automatically configures dynamic |
| 1218 | loading. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1219 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1220 | \subsection{SGI IRIX 4 Dynamic Loading} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1221 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1222 | Only SGI IRIX 4 supports dynamic loading of modules using SGI dynamic |
| 1223 | loading. (SGI IRIX 5 might also support it but it is inferior to |
| 1224 | using shared libraries so there is no reason to; a small test didn't |
| 1225 | work right away so I gave up trying to support it.) |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1226 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1227 | Before you build Python, you first need to fetch and build the \code{dl} |
| 1228 | package written by Jack Jansen. This is available by anonymous ftp |
| 1229 | from host \file{ftp.cwi.nl}, directory \file{pub/dynload}, file |
| 1230 | \file{dl-1.6.tar.Z}. (The version number may change.) Follow the |
| 1231 | instructions in the package's \file{README} file to build it. |
| 1232 | |
| 1233 | Once you have built \code{dl}, you can configure Python to use it. To |
| 1234 | this end, you run the \file{configure} script with the option |
| 1235 | \code{--with-dl=\var{directory}} where \var{directory} is the absolute |
| 1236 | pathname of the \code{dl} directory. |
| 1237 | |
| 1238 | Now build and install Python as you normally would (see the |
| 1239 | \file{README} file in the toplevel Python directory.) |
| 1240 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1241 | \subsection{GNU Dynamic Loading} |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1242 | |
| 1243 | GNU dynamic loading supports (according to its \file{README} file) the |
| 1244 | following hardware and software combinations: VAX (Ultrix), Sun 3 |
| 1245 | (SunOS 3.4 and 4.0), Sparc (SunOS 4.0), Sequent Symmetry (Dynix), and |
| 1246 | Atari ST. There is no reason to use it on a Sparc; I haven't seen a |
| 1247 | Sun 3 for years so I don't know if these have shared libraries or not. |
| 1248 | |
| 1249 | You need to fetch and build two packages. One is GNU DLD 3.2.3, |
| 1250 | available by anonymous ftp from host \file{ftp.cwi.nl}, directory |
| 1251 | \file{pub/dynload}, file \file{dld-3.2.3.tar.Z}. (As far as I know, |
| 1252 | no further development on GNU DLD is being done.) The other is an |
| 1253 | emulation of Jack Jansen's \code{dl} package that I wrote on top of |
| 1254 | GNU DLD 3.2.3. This is available from the same host and directory, |
| 1255 | file dl-dld-1.1.tar.Z. (The version number may change --- but I doubt |
| 1256 | it will.) Follow the instructions in each package's \file{README} |
| 1257 | file to configure build them. |
| 1258 | |
| 1259 | Now configure Python. Run the \file{configure} script with the option |
| 1260 | \code{--with-dl-dld=\var{dl-directory},\var{dld-directory}} where |
| 1261 | \var{dl-directory} is the absolute pathname of the directory where you |
| 1262 | have built the \file{dl-dld} package, and \var{dld-directory} is that |
| 1263 | of the GNU DLD package. The Python interpreter you build hereafter |
| 1264 | will support GNU dynamic loading. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1265 | |
| 1266 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1267 | \section{Building a Dynamically Loadable Module} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1268 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1269 | Since there are three styles of dynamic loading, there are also three |
| 1270 | groups of instructions for building a dynamically loadable module. |
| 1271 | Instructions common for all three styles are given first. Assuming |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1272 | your module is called \code{spam}, the source filename must be |
| 1273 | \file{spammodule.c}, so the object name is \file{spammodule.o}. The |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1274 | module must be written as a normal Python extension module (as |
| 1275 | described earlier). |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1276 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1277 | Note that in all cases you will have to create your own Makefile that |
| 1278 | compiles your module file(s). This Makefile will have to pass two |
| 1279 | \samp{-I} arguments to the C compiler which will make it find the |
| 1280 | Python header files. If the Make variable \var{PYTHONTOP} points to |
| 1281 | the toplevel Python directory, your \var{CFLAGS} Make variable should |
| 1282 | contain the options \samp{-I\$(PYTHONTOP) -I\$(PYTHONTOP)/Include}. |
| 1283 | (Most header files are in the \file{Include} subdirectory, but the |
| 1284 | \file{config.h} header lives in the toplevel directory.) You must |
| 1285 | also add \samp{-DHAVE_CONFIG_H} to the definition of \var{CFLAGS} to |
| 1286 | direct the Python headers to include \file{config.h}. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1287 | |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1288 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1289 | \subsection{Shared Libraries} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1290 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1291 | You must link the \samp{.o} file to produce a shared library. This is |
| 1292 | done using a special invocation of the \UNIX{} loader/linker, {\em |
| 1293 | ld}(1). Unfortunately the invocation differs slightly per system. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1294 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1295 | On SunOS 4, use |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1296 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1297 | ld spammodule.o -o spammodule.so |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1298 | \end{verbatim} |
| 1299 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1300 | On Solaris 2, use |
| 1301 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1302 | ld -G spammodule.o -o spammodule.so |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1303 | \end{verbatim} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1304 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1305 | On SGI IRIX 5, use |
| 1306 | \begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1307 | ld -shared spammodule.o -o spammodule.so |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1308 | \end{verbatim} |
| 1309 | |
| 1310 | On other systems, consult the manual page for {\em ld}(1) to find what |
| 1311 | flags, if any, must be used. |
| 1312 | |
| 1313 | If your extension module uses system libraries that haven't already |
| 1314 | been linked with Python (e.g. a windowing system), these must be |
| 1315 | passed to the {\em ld} command as \samp{-l} options after the |
| 1316 | \samp{.o} file. |
| 1317 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1318 | The resulting file \file{spammodule.so} must be copied into a directory |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1319 | along the Python module search path. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1320 | |
| 1321 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1322 | \subsection{SGI IRIX 4 Dynamic Loading} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1323 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1324 | {bf IMPORTANT:} You must compile your extension module with the |
| 1325 | additional C flag \samp{-G0} (or \samp{-G 0}). This instruct the |
| 1326 | assembler to generate position-independent code. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1327 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1328 | You don't need to link the resulting \file{spammodule.o} file; just |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1329 | copy it into a directory along the Python module search path. |
| 1330 | |
| 1331 | The first time your extension is loaded, it takes some extra time and |
| 1332 | a few messages may be printed. This creates a file |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1333 | \file{spammodule.ld} which is an image that can be loaded quickly into |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1334 | the Python interpreter process. When a new Python interpreter is |
| 1335 | installed, the \code{dl} package detects this and rebuilds |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1336 | \file{spammodule.ld}. The file \file{spammodule.ld} is placed in the |
| 1337 | directory where \file{spammodule.o} was found, unless this directory is |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1338 | unwritable; in that case it is placed in a temporary |
| 1339 | directory.\footnote{Check the manual page of the \code{dl} package for |
| 1340 | details.} |
| 1341 | |
| 1342 | If your extension modules uses additional system libraries, you must |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1343 | create a file \file{spammodule.libs} in the same directory as the |
| 1344 | \file{spammodule.o}. This file should contain one or more lines with |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1345 | whitespace-separated options that will be passed to the linker --- |
| 1346 | normally only \samp{-l} options or absolute pathnames of libraries |
| 1347 | (\samp{.a} files) should be used. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1348 | |
| 1349 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1350 | \subsection{GNU Dynamic Loading} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1351 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1352 | Just copy \file{spammodule.o} into a directory along the Python module |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1353 | search path. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1354 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1355 | If your extension modules uses additional system libraries, you must |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1356 | create a file \file{spammodule.libs} in the same directory as the |
| 1357 | \file{spammodule.o}. This file should contain one or more lines with |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1358 | whitespace-separated absolute pathnames of libraries (\samp{.a} |
| 1359 | files). No \samp{-l} options can be used. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1360 | |
| 1361 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1362 | \input{ext.ind} |
| 1363 | |
| 1364 | \end{document} |