Fred Drake | dca8792 | 1998-01-13 16:53:23 +0000 | [diff] [blame] | 1 | \documentclass[twoside,openright]{report} |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 2 | \usepackage{myformat} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 3 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 4 | % XXX PM Modulator |
| 5 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 6 | \title{Extending and Embedding the Python Interpreter} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 7 | |
Guido van Rossum | 16cd7f9 | 1994-10-06 10:29:26 +0000 | [diff] [blame] | 8 | \input{boilerplate} |
Guido van Rossum | 83eb962 | 1993-11-23 16:28:45 +0000 | [diff] [blame] | 9 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 10 | % Tell \index to actually write the .idx file |
| 11 | \makeindex |
| 12 | |
| 13 | \begin{document} |
| 14 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 15 | \pagestyle{empty} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 16 | \pagenumbering{roman} |
| 17 | |
| 18 | \maketitle |
| 19 | |
Guido van Rossum | 16cd7f9 | 1994-10-06 10:29:26 +0000 | [diff] [blame] | 20 | \input{copyright} |
| 21 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 22 | \begin{abstract} |
| 23 | |
| 24 | \noindent |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 25 | Python is an interpreted, object-oriented programming language. This |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 26 | document describes how to write modules in \C{} or \Cpp{} to extend the |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 27 | Python interpreter with new modules. Those modules can define new |
| 28 | functions but also new object types and their methods. The document |
| 29 | also describes how to embed the Python interpreter in another |
| 30 | application, for use as an extension language. Finally, it shows how |
| 31 | to compile and link extension modules so that they can be loaded |
| 32 | dynamically (at run time) into the interpreter, if the underlying |
| 33 | operating system supports this feature. |
| 34 | |
| 35 | This document assumes basic knowledge about Python. For an informal |
| 36 | introduction to the language, see the Python Tutorial. The Python |
| 37 | Reference Manual gives a more formal definition of the language. The |
| 38 | Python Library Reference documents the existing object types, |
| 39 | functions and modules (both built-in and written in Python) that give |
| 40 | the language its wide application range. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 41 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 42 | For a detailed description of the whole Python/\C{} API, see the separate |
| 43 | Python/\C{} API Reference Manual. \strong{Note:} While that manual is |
Guido van Rossum | fdacc58 | 1997-10-07 14:40:16 +0000 | [diff] [blame] | 44 | still in a state of flux, it is safe to say that it is much more up to |
| 45 | date than the manual you're reading currently (which has been in need |
| 46 | for an upgrade for some time now). |
| 47 | |
| 48 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 49 | \end{abstract} |
| 50 | |
Fred Drake | 4d4f9e7 | 1998-01-13 22:25:02 +0000 | [diff] [blame] | 51 | \tableofcontents |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 52 | |
| 53 | \pagenumbering{arabic} |
| 54 | |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 55 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 56 | \chapter{Extending Python with \C{} or \Cpp{} code} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 57 | |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 58 | |
| 59 | \section{Introduction} |
| 60 | |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 61 | It is quite easy to add new built-in modules to Python, if you know |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 62 | how to program in \C{}. Such \dfn{extension modules} can do two things |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 63 | that can't be done directly in Python: they can implement new built-in |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 64 | object types, and they can call \C{} library functions and system calls. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 65 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 66 | To support extensions, the Python API (Application Programmers |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 67 | Interface) defines a set of functions, macros and variables that |
| 68 | provide access to most aspects of the Python run-time system. The |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 69 | Python API is incorporated in a \C{} source file by including the header |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 70 | \code{"Python.h"}. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 71 | |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 72 | The compilation of an extension module depends on its intended use as |
| 73 | well as on your system setup; details are given in a later section. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 74 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 75 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 76 | \section{A Simple Example} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 77 | |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 78 | Let's create an extension module called \samp{spam} (the favorite food |
| 79 | of Monty Python fans...) and let's say we want to create a Python |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 80 | interface to the \C{} library function \code{system()}.\footnote{An |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 81 | interface for this function already exists in the standard module |
| 82 | \code{os} --- it was chosen as a simple and straightfoward example.} |
| 83 | This function takes a null-terminated character string as argument and |
| 84 | returns an integer. We want this function to be callable from Python |
| 85 | as follows: |
| 86 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 87 | \bcode\begin{verbatim} |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 88 | >>> import spam |
| 89 | >>> status = spam.system("ls -l") |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 90 | \end{verbatim}\ecode |
| 91 | % |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 92 | Begin by creating a file \samp{spammodule.c}. (In general, if a |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 93 | module is called \samp{spam}, the \C{} file containing its implementation |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 94 | is called \file{spammodule.c}; if the module name is very long, like |
| 95 | \samp{spammify}, the module name can be just \file{spammify.c}.) |
| 96 | |
| 97 | The first line of our file can be: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 98 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 99 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 100 | #include "Python.h" |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 101 | \end{verbatim}\ecode |
| 102 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 103 | which pulls in the Python API (you can add a comment describing the |
| 104 | purpose of the module and a copyright notice if you like). |
| 105 | |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 106 | All user-visible symbols defined by \code{"Python.h"} have a prefix of |
| 107 | \samp{Py} or \samp{PY}, except those defined in standard header files. |
| 108 | For convenience, and since they are used extensively by the Python |
| 109 | interpreter, \code{"Python.h"} includes a few standard header files: |
| 110 | \code{<stdio.h>}, \code{<string.h>}, \code{<errno.h>}, and |
| 111 | \code{<stdlib.h>}. If the latter header file does not exist on your |
| 112 | system, it declares the functions \code{malloc()}, \code{free()} and |
| 113 | \code{realloc()} directly. |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 114 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 115 | The next thing we add to our module file is the \C{} function that will |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 116 | be called when the Python expression \samp{spam.system(\var{string})} |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 117 | is evaluated (we'll see shortly how it ends up being called): |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 118 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 119 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 120 | static PyObject * |
| 121 | spam_system(self, args) |
| 122 | PyObject *self; |
| 123 | PyObject *args; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 124 | { |
| 125 | char *command; |
| 126 | int sts; |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 127 | if (!PyArg_ParseTuple(args, "s", &command)) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 128 | return NULL; |
| 129 | sts = system(command); |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 130 | return Py_BuildValue("i", sts); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 131 | } |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 132 | \end{verbatim}\ecode |
| 133 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 134 | There is a straightforward translation from the argument list in |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 135 | Python (e.g.\ the single expression \code{"ls -l"}) to the arguments |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 136 | passed to the \C{} function. The \C{} function always has two arguments, |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 137 | conventionally named \var{self} and \var{args}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 138 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 139 | The \var{self} argument is only used when the \C{} function implements a |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 140 | builtin method. This will be discussed later. In the example, |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 141 | \var{self} will always be a \NULL{} pointer, since we are defining |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 142 | a function, not a method. (This is done so that the interpreter |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 143 | doesn't have to understand two different types of \C{} functions.) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 144 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 145 | The \var{args} argument will be a pointer to a Python tuple object |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 146 | containing the arguments. Each item of the tuple corresponds to an |
| 147 | argument in the call's argument list. The arguments are Python |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 148 | objects -- in order to do anything with them in our \C{} function we have |
| 149 | to convert them to \C{} values. The function \code{PyArg_ParseTuple()} |
| 150 | in the Python API checks the argument types and converts them to \C{} |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 151 | values. It uses a template string to determine the required types of |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 152 | the arguments as well as the types of the \C{} variables into which to |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 153 | store the converted values. More about this later. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 154 | |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 155 | \code{PyArg_ParseTuple()} returns true (nonzero) if all arguments have |
| 156 | the right type and its components have been stored in the variables |
| 157 | whose addresses are passed. It returns false (zero) if an invalid |
| 158 | argument list was passed. In the latter case it also raises an |
| 159 | appropriate exception by so the calling function can return |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 160 | \NULL{} immediately (as we saw in the example). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 161 | |
| 162 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 163 | \section{Intermezzo: Errors and Exceptions} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 164 | |
| 165 | An important convention throughout the Python interpreter is the |
| 166 | following: when a function fails, it should set an exception condition |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 167 | and return an error value (usually a \NULL{} pointer). Exceptions |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 168 | are stored in a static global variable inside the interpreter; if this |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 169 | variable is \NULL{} no exception has occurred. A second global |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 170 | variable stores the ``associated value'' of the exception (the second |
| 171 | argument to \code{raise}). A third variable contains the stack |
| 172 | traceback in case the error originated in Python code. These three |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 173 | variables are the \C{} equivalents of the Python variables |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 174 | \code{sys.exc_type}, \code{sys.exc_value} and \code{sys.exc_traceback} |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 175 | (see the section on module \code{sys} in the Library Reference |
| 176 | Manual). It is important to know about them to understand how errors |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 177 | are passed around. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 178 | |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 179 | The Python API defines a number of functions to set various types of |
| 180 | exceptions. |
| 181 | |
| 182 | The most common one is \code{PyErr_SetString()}. Its arguments are an |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 183 | exception object and a \C{} string. The exception object is usually a |
| 184 | predefined object like \code{PyExc_ZeroDivisionError}. The \C{} string |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 185 | indicates the cause of the error and is converted to a Python string |
| 186 | object and stored as the ``associated value'' of the exception. |
| 187 | |
| 188 | Another useful function is \code{PyErr_SetFromErrno()}, which only |
| 189 | takes an exception argument and constructs the associated value by |
| 190 | inspection of the (\UNIX{}) global variable \code{errno}. The most |
| 191 | general function is \code{PyErr_SetObject()}, which takes two object |
| 192 | arguments, the exception and its associated value. You don't need to |
| 193 | \code{Py_INCREF()} the objects passed to any of these functions. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 194 | |
| 195 | You can test non-destructively whether an exception has been set with |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 196 | \code{PyErr_Occurred()}. This returns the current exception object, |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 197 | or \NULL{} if no exception has occurred. You normally don't need |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 198 | to call \code{PyErr_Occurred()} to see whether an error occurred in a |
| 199 | function call, since you should be able to tell from the return value. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 200 | |
Guido van Rossum | d16ddb6 | 1996-12-13 02:38:17 +0000 | [diff] [blame] | 201 | When a function \var{f} that calls another function \var{g} detects |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 202 | that the latter fails, \var{f} should itself return an error value |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 203 | (e.g. \NULL{} or \code{-1}). It should \emph{not} call one of the |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 204 | \code{PyErr_*()} functions --- one has already been called by \var{g}. |
| 205 | \var{f}'s caller is then supposed to also return an error indication |
| 206 | to \emph{its} caller, again \emph{without} calling \code{PyErr_*()}, |
| 207 | and so on --- the most detailed cause of the error was already |
| 208 | reported by the function that first detected it. Once the error |
| 209 | reaches the Python interpreter's main loop, this aborts the currently |
| 210 | executing Python code and tries to find an exception handler specified |
| 211 | by the Python programmer. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 212 | |
| 213 | (There are situations where a module can actually give a more detailed |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 214 | error message by calling another \code{PyErr_*()} function, and in |
| 215 | such cases it is fine to do so. As a general rule, however, this is |
| 216 | not necessary, and can cause information about the cause of the error |
| 217 | to be lost: most operations can fail for a variety of reasons.) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 218 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 219 | To ignore an exception set by a function call that failed, the exception |
| 220 | condition must be cleared explicitly by calling \code{PyErr_Clear()}. |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 221 | The only time \C{} code should call \code{PyErr_Clear()} is if it doesn't |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 222 | want to pass the error on to the interpreter but wants to handle it |
| 223 | completely by itself (e.g. by trying something else or pretending |
| 224 | nothing happened). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 225 | |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 226 | Note that a failing \code{malloc()} call must be turned into an |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 227 | exception --- the direct caller of \code{malloc()} (or |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 228 | \code{realloc()}) must call \code{PyErr_NoMemory()} and return a |
| 229 | failure indicator itself. All the object-creating functions |
| 230 | (\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] | 231 | \code{malloc()} directly this note is of importance. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 232 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 233 | Also note that, with the important exception of |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 234 | \code{PyArg_ParseTuple()} and friends, functions that return an |
| 235 | integer status usually return a positive value or zero for success and |
| 236 | \code{-1} for failure, like \UNIX{} system calls. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 237 | |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 238 | Finally, be careful to clean up garbage (by making \code{Py_XDECREF()} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 239 | or \code{Py_DECREF()} calls for objects you have already created) when |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 240 | you return an error indicator! |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 241 | |
| 242 | The choice of which exception to raise is entirely yours. There are |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 243 | predeclared \C{} objects corresponding to all built-in Python exceptions, |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 244 | e.g. \code{PyExc_ZeroDevisionError} which you can use directly. Of |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 245 | course, you should choose exceptions wisely --- don't use |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 246 | \code{PyExc_TypeError} to mean that a file couldn't be opened (that |
| 247 | should probably be \code{PyExc_IOError}). If something's wrong with |
| 248 | the argument list, the \code{PyArg_ParseTuple()} function usually |
| 249 | raises \code{PyExc_TypeError}. If you have an argument whose value |
| 250 | which must be in a particular range or must satisfy other conditions, |
| 251 | \code{PyExc_ValueError} is appropriate. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 252 | |
| 253 | You can also define a new exception that is unique to your module. |
| 254 | For this, you usually declare a static object variable at the |
| 255 | beginning of your file, e.g. |
| 256 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 257 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 258 | static PyObject *SpamError; |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 259 | \end{verbatim}\ecode |
| 260 | % |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 261 | and initialize it in your module's initialization function |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 262 | (\code{initspam()}) with a string object, e.g. (leaving out the error |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 263 | checking for now): |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 264 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 265 | \bcode\begin{verbatim} |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 266 | void |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 267 | initspam() |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 268 | { |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 269 | PyObject *m, *d; |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 270 | m = Py_InitModule("spam", SpamMethods); |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 271 | d = PyModule_GetDict(m); |
| 272 | SpamError = PyString_FromString("spam.error"); |
| 273 | PyDict_SetItemString(d, "error", SpamError); |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 274 | } |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 275 | \end{verbatim}\ecode |
| 276 | % |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 277 | Note that the Python name for the exception object is |
| 278 | \code{spam.error}. It is conventional for module and exception names |
| 279 | to be spelled in lower case. It is also conventional that the |
| 280 | \emph{value} of the exception object is the same as its name, e.g.\ |
| 281 | the string \code{"spam.error"}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 282 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 283 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 284 | \section{Back to the Example} |
| 285 | |
| 286 | Going back to our example function, you should now be able to |
| 287 | understand this statement: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 288 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 289 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 290 | if (!PyArg_ParseTuple(args, "s", &command)) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 291 | return NULL; |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 292 | \end{verbatim}\ecode |
| 293 | % |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 294 | It returns \NULL{} (the error indicator for functions returning |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 295 | object pointers) if an error is detected in the argument list, relying |
| 296 | on the exception set by \code{PyArg_ParseTuple()}. Otherwise the |
| 297 | string value of the argument has been copied to the local variable |
| 298 | \code{command}. This is a pointer assignment and you are not supposed |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 299 | to modify the string to which it points (so in Standard \C{}, the variable |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 300 | \code{command} should properly be declared as \samp{const char |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 301 | *command}). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 302 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 303 | The next statement is a call to the \UNIX{} function \code{system()}, |
| 304 | passing it the string we just got from \code{PyArg_ParseTuple()}: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 305 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 306 | \bcode\begin{verbatim} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 307 | sts = system(command); |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 308 | \end{verbatim}\ecode |
| 309 | % |
Guido van Rossum | d16ddb6 | 1996-12-13 02:38:17 +0000 | [diff] [blame] | 310 | Our \code{spam.system()} function must return the value of \code{sts} |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 311 | as a Python object. This is done using the function |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 312 | \code{Py_BuildValue()}, which is something like the inverse of |
| 313 | \code{PyArg_ParseTuple()}: it takes a format string and an arbitrary |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 314 | number of \C{} values, and returns a new Python object. More info on |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 315 | \code{Py_BuildValue()} is given later. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 316 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 317 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 318 | return Py_BuildValue("i", sts); |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 319 | \end{verbatim}\ecode |
| 320 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 321 | In this case, it will return an integer object. (Yes, even integers |
| 322 | are objects on the heap in Python!) |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 323 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 324 | If you have a \C{} function that returns no useful argument (a function |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 325 | returning \code{void}), the corresponding Python function must return |
| 326 | \code{None}. You need this idiom to do so: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 327 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 328 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 329 | Py_INCREF(Py_None); |
| 330 | return Py_None; |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 331 | \end{verbatim}\ecode |
| 332 | % |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 333 | \code{Py_None} is the \C{} name for the special Python object |
| 334 | \code{None}. It is a genuine Python object (not a \NULL{} |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 335 | pointer, which means ``error'' in most contexts, as we have seen). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 336 | |
| 337 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 338 | \section{The Module's Method Table and Initialization Function} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 339 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 340 | I promised to show how \code{spam_system()} is called from Python |
| 341 | programs. First, we need to list its name and address in a ``method |
| 342 | table'': |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 343 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 344 | \bcode\begin{verbatim} |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 345 | static PyMethodDef SpamMethods[] = { |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 346 | ... |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 347 | {"system", spam_system, METH_VARARGS}, |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 348 | ... |
| 349 | {NULL, NULL} /* Sentinel */ |
| 350 | }; |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 351 | \end{verbatim}\ecode |
| 352 | % |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 353 | Note the third entry (\samp{METH_VARARGS}). This is a flag telling |
| 354 | the interpreter the calling convention to be used for the \C{} |
| 355 | function. It should normally always be \samp{METH_VARARGS} or |
| 356 | \samp{METH_VARARGS | METH_KEYWORDS}; a value of \samp{0} means that an |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 357 | obsolete variant of \code{PyArg_ParseTuple()} is used. |
| 358 | |
Fred Drake | b6e5032 | 1998-02-04 20:26:31 +0000 | [diff] [blame^] | 359 | When using only \samp{METH_VARARGS}, the function should expect |
| 360 | the Python-level parameters to be passed in as a tuple acceptable for |
| 361 | parsing via \cfunction{PyArg_ParseTuple()}; more information on this |
| 362 | function is provided below. |
| 363 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 364 | The \code{METH_KEYWORDS} bit may be set in the third field if keyword |
| 365 | arguments should be passed to the function. In this case, the \C{} |
| 366 | function should accept a third \samp{PyObject *} parameter which will |
| 367 | be a dictionary of keywords. Use \code{PyArg_ParseTupleAndKeywords()} |
| 368 | to parse the arguemts to such a function. |
| 369 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 370 | The method table must be passed to the interpreter in the module's |
| 371 | initialization function (which should be the only non-\code{static} |
| 372 | item defined in the module file): |
| 373 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 374 | \bcode\begin{verbatim} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 375 | void |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 376 | initspam() |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 377 | { |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 378 | (void) Py_InitModule("spam", SpamMethods); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 379 | } |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 380 | \end{verbatim}\ecode |
| 381 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 382 | When the Python program imports module \code{spam} for the first time, |
| 383 | \code{initspam()} is called. It calls \code{Py_InitModule()}, which |
| 384 | creates a ``module object'' (which is inserted in the dictionary |
| 385 | \code{sys.modules} under the key \code{"spam"}), and inserts built-in |
| 386 | function objects into the newly created module based upon the table |
| 387 | (an array of \code{PyMethodDef} structures) that was passed as its |
| 388 | second argument. \code{Py_InitModule()} returns a pointer to the |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 389 | module object that it creates (which is unused here). It aborts with |
| 390 | a fatal error if the module could not be initialized satisfactorily, |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 391 | so the caller doesn't need to check for errors. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 392 | |
| 393 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 394 | \section{Compilation and Linkage} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 395 | |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 396 | There are two more things to do before you can use your new extension: |
| 397 | compiling and linking it with the Python system. If you use dynamic |
| 398 | loading, the details depend on the style of dynamic loading your |
| 399 | system uses; see the chapter on Dynamic Loading for more info about |
| 400 | this. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 401 | |
| 402 | If you can't use dynamic loading, or if you want to make your module a |
| 403 | 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] | 404 | configuration setup and rebuild the interpreter. Luckily, this is |
| 405 | very simple: just place your file (\file{spammodule.c} for example) in |
| 406 | the \file{Modules} directory, add a line to the file |
| 407 | \file{Modules/Setup} describing your file: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 408 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 409 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 410 | spam spammodule.o |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 411 | \end{verbatim}\ecode |
| 412 | % |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 413 | and rebuild the interpreter by running \code{make} in the toplevel |
| 414 | directory. You can also run \code{make} in the \file{Modules} |
| 415 | subdirectory, but then you must first rebuilt the \file{Makefile} |
| 416 | there by running \code{make Makefile}. (This is necessary each time |
| 417 | you change the \file{Setup} file.) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 418 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 419 | If your module requires additional libraries to link with, these can |
| 420 | 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] | 421 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 422 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 423 | spam spammodule.o -lX11 |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 424 | \end{verbatim}\ecode |
| 425 | % |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 426 | \section{Calling Python Functions From \C{}} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 427 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 428 | So far we have concentrated on making \C{} functions callable from |
| 429 | Python. The reverse is also useful: calling Python functions from \C{}. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 430 | This is especially the case for libraries that support so-called |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 431 | ``callback'' functions. If a \C{} interface makes use of callbacks, the |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 432 | equivalent Python often needs to provide a callback mechanism to the |
| 433 | Python programmer; the implementation will require calling the Python |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 434 | callback functions from a \C{} callback. Other uses are also imaginable. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 435 | |
| 436 | Fortunately, the Python interpreter is easily called recursively, and |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 437 | 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] | 438 | 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] | 439 | 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] | 440 | the \samp{-c} command line option in \file{Python/pythonmain.c}.) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 441 | |
| 442 | Calling a Python function is easy. First, the Python program must |
| 443 | somehow pass you the Python function object. You should provide a |
| 444 | function (or some other interface) to do this. When this function is |
| 445 | 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] | 446 | \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] | 447 | For example, the following function might be part of a module |
| 448 | definition: |
| 449 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 450 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 451 | static PyObject *my_callback = NULL; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 452 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 453 | static PyObject * |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 454 | my_set_callback(dummy, arg) |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 455 | PyObject *dummy, *arg; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 456 | { |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 457 | Py_XDECREF(my_callback); /* Dispose of previous callback */ |
| 458 | Py_XINCREF(arg); /* Add a reference to new callback */ |
| 459 | my_callback = arg; /* Remember new callback */ |
| 460 | /* Boilerplate to return "None" */ |
| 461 | Py_INCREF(Py_None); |
| 462 | return Py_None; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 463 | } |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 464 | \end{verbatim}\ecode |
| 465 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 466 | The macros \code{Py_XINCREF()} and \code{Py_XDECREF()} increment/decrement |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 467 | the reference count of an object and are safe in the presence of |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 468 | \NULL{} pointers. More info on them in the section on Reference |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 469 | Counts below. |
| 470 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 471 | 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] | 472 | \code{PyEval_CallObject()}. This function has two arguments, both |
| 473 | pointers to arbitrary Python objects: the Python function, and the |
| 474 | argument list. The argument list must always be a tuple object, whose |
| 475 | length is the number of arguments. To call the Python function with |
| 476 | no arguments, pass an empty tuple; to call it with one argument, pass |
| 477 | a singleton tuple. \code{Py_BuildValue()} returns a tuple when its |
| 478 | format string consists of zero or more format codes between |
| 479 | parentheses. For example: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 480 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 481 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 482 | int arg; |
| 483 | PyObject *arglist; |
| 484 | PyObject *result; |
| 485 | ... |
| 486 | arg = 123; |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 487 | ... |
| 488 | /* Time to call the callback */ |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 489 | arglist = Py_BuildValue("(i)", arg); |
| 490 | result = PyEval_CallObject(my_callback, arglist); |
| 491 | Py_DECREF(arglist); |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 492 | \end{verbatim}\ecode |
| 493 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 494 | \code{PyEval_CallObject()} returns a Python object pointer: this is |
| 495 | the return value of the Python function. \code{PyEval_CallObject()} is |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 496 | ``reference-count-neutral'' with respect to its arguments. In the |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 497 | 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] | 498 | is \code{Py_DECREF()}-ed immediately after the call. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 499 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 500 | The return value of \code{PyEval_CallObject()} is ``new'': either it |
| 501 | is a brand new object, or it is an existing object whose reference |
| 502 | count has been incremented. So, unless you want to save it in a |
| 503 | global variable, you should somehow \code{Py_DECREF()} the result, |
| 504 | even (especially!) if you are not interested in its value. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 505 | |
| 506 | Before you do this, however, it is important to check that the return |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 507 | value isn't \NULL{}. If it is, the Python function terminated by raising |
| 508 | 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] | 509 | called from Python, it should now return an error indication to its |
| 510 | Python caller, so the interpreter can print a stack trace, or the |
| 511 | calling Python code can handle the exception. If this is not possible |
| 512 | or desirable, the exception should be cleared by calling |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 513 | \code{PyErr_Clear()}. For example: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 514 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 515 | \bcode\begin{verbatim} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 516 | if (result == NULL) |
| 517 | return NULL; /* Pass error back */ |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 518 | ...use result... |
| 519 | Py_DECREF(result); |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 520 | \end{verbatim}\ecode |
| 521 | % |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 522 | Depending on the desired interface to the Python callback function, |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 523 | 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] | 524 | In some cases the argument list is also provided by the Python |
| 525 | program, through the same interface that specified the callback |
| 526 | function. It can then be saved and used in the same manner as the |
| 527 | function object. In other cases, you may have to construct a new |
| 528 | 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] | 529 | 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] | 530 | event code, you might use the following code: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 531 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 532 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 533 | PyObject *arglist; |
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 | arglist = Py_BuildValue("(l)", eventcode); |
| 536 | result = PyEval_CallObject(my_callback, arglist); |
| 537 | Py_DECREF(arglist); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 538 | if (result == NULL) |
| 539 | return NULL; /* Pass error back */ |
| 540 | /* Here maybe use the result */ |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 541 | Py_DECREF(result); |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 542 | \end{verbatim}\ecode |
| 543 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 544 | 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] | 545 | 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] | 546 | 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] | 547 | be checked. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 548 | |
| 549 | |
Fred Drake | 53396f6 | 1998-01-19 02:48:37 +0000 | [diff] [blame] | 550 | \section{Format Strings for \sectcode{PyArg_ParseTuple()}} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 551 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 552 | The \code{PyArg_ParseTuple()} function is declared as follows: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 553 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 554 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 555 | int PyArg_ParseTuple(PyObject *arg, char *format, ...); |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 556 | \end{verbatim}\ecode |
| 557 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 558 | The \var{arg} argument must be a tuple object containing an argument |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 559 | list passed from Python to a \C{} function. The \var{format} argument |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 560 | must be a format string, whose syntax is explained below. The |
| 561 | remaining arguments must be addresses of variables whose type is |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 562 | determined by the format string. For the conversion to succeed, the |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 563 | \var{arg} object must match the format and the format must be |
| 564 | exhausted. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 565 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 566 | Note that while \code{PyArg_ParseTuple()} checks that the Python |
| 567 | arguments have the required types, it cannot check the validity of the |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 568 | addresses of \C{} variables passed to the call: if you make mistakes |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 569 | there, your code will probably crash or at least overwrite random bits |
| 570 | in memory. So be careful! |
| 571 | |
| 572 | A format string consists of zero or more ``format units''. A format |
| 573 | unit describes one Python object; it is usually a single character or |
| 574 | a parenthesized sequence of format units. With a few exceptions, a |
| 575 | format unit that is not a parenthesized sequence normally corresponds |
| 576 | to a single address argument to \code{PyArg_ParseTuple()}. In the |
| 577 | following description, the quoted form is the format unit; the entry |
| 578 | in (round) parentheses is the Python object type that matches the |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 579 | format unit; and the entry in [square] brackets is the type of the \C{} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 580 | variable(s) whose address should be passed. (Use the \samp{\&} |
| 581 | operator to pass a variable's address.) |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 582 | |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 583 | \begin{description} |
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{s} (string) [char *]] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 586 | Convert a Python string to a \C{} pointer to a character string. You |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 587 | must not provide storage for the string itself; a pointer to an |
| 588 | existing string is stored into the character pointer variable whose |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 589 | address you pass. The \C{} string is null-terminated. The Python string |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 590 | must not contain embedded null bytes; if it does, a \code{TypeError} |
| 591 | exception is raised. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 592 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 593 | \item[\samp{s\#} (string) {[char *, int]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 594 | This variant on \code{'s'} stores into two \C{} variables, the first one |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 595 | a pointer to a character string, the second one its length. In this |
| 596 | case the Python string may contain embedded null bytes. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 597 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 598 | \item[\samp{z} (string or \code{None}) {[char *]}] |
| 599 | Like \samp{s}, but the Python object may also be \code{None}, in which |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 600 | case the \C{} pointer is set to \NULL{}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 601 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 602 | \item[\samp{z\#} (string or \code{None}) {[char *, int]}] |
| 603 | 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] | 604 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 605 | \item[\samp{b} (integer) {[char]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 606 | 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] | 607 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 608 | \item[\samp{h} (integer) {[short int]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 609 | Convert a Python integer to a \C{} \code{short int}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 610 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 611 | \item[\samp{i} (integer) {[int]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 612 | Convert a Python integer to a plain \C{} \code{int}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 613 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 614 | \item[\samp{l} (integer) {[long int]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 615 | Convert a Python integer to a \C{} \code{long int}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 616 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 617 | \item[\samp{c} (string of length 1) {[char]}] |
| 618 | Convert a Python character, represented as a string of length 1, to a |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 619 | \C{} \code{char}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 620 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 621 | \item[\samp{f} (float) {[float]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 622 | Convert a Python floating point number to a \C{} \code{float}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 623 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 624 | \item[\samp{d} (float) {[double]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 625 | Convert a Python floating point number to a \C{} \code{double}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 626 | |
Fred Drake | b6e5032 | 1998-02-04 20:26:31 +0000 | [diff] [blame^] | 627 | \item[\samp{D} (complex) {[Py_complex]}] |
| 628 | Convert a Python complex number to a \C{} \code{Py_complex} structure. |
| 629 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 630 | \item[\samp{O} (object) {[PyObject *]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 631 | Store a Python object (without any conversion) in a \C{} object pointer. |
| 632 | The \C{} program thus receives the actual object that was passed. The |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 633 | object's reference count is not increased. The pointer stored is not |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 634 | \NULL{}. |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 635 | |
| 636 | \item[\samp{O!} (object) {[\var{typeobject}, PyObject *]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 637 | Store a Python object in a \C{} object pointer. This is similar to |
| 638 | \samp{O}, but takes two \C{} arguments: the first is the address of a |
| 639 | Python type object, the second is the address of the \C{} variable (of |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 640 | type \code{PyObject *}) into which the object pointer is stored. |
| 641 | If the Python object does not have the required type, a |
| 642 | \code{TypeError} exception is raised. |
| 643 | |
| 644 | \item[\samp{O\&} (object) {[\var{converter}, \var{anything}]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 645 | Convert a Python object to a \C{} variable through a \var{converter} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 646 | function. This takes two arguments: the first is a function, the |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 647 | second is the address of a \C{} variable (of arbitrary type), converted |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 648 | to \code{void *}. The \var{converter} function in turn is called as |
| 649 | follows: |
| 650 | |
| 651 | \code{\var{status} = \var{converter}(\var{object}, \var{address});} |
| 652 | |
| 653 | where \var{object} is the Python object to be converted and |
| 654 | \var{address} is the \code{void *} argument that was passed to |
| 655 | \code{PyArg_ConvertTuple()}. The returned \var{status} should be |
| 656 | \code{1} for a successful conversion and \code{0} if the conversion |
| 657 | has failed. When the conversion fails, the \var{converter} function |
| 658 | should raise an exception. |
| 659 | |
| 660 | \item[\samp{S} (string) {[PyStringObject *]}] |
| 661 | Like \samp{O} but raises a \code{TypeError} exception that the object |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 662 | is a string object. The \C{} variable may also be declared as |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 663 | \code{PyObject *}. |
| 664 | |
| 665 | \item[\samp{(\var{items})} (tuple) {[\var{matching-items}]}] |
| 666 | The object must be a Python tuple whose length is the number of format |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 667 | units in \var{items}. The \C{} arguments must correspond to the |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 668 | individual format units in \var{items}. Format units for tuples may |
| 669 | be nested. |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 670 | |
| 671 | \end{description} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 672 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 673 | It is possible to pass Python long integers where integers are |
| 674 | requested; however no proper range checking is done -- the most |
| 675 | significant bits are silently truncated when the receiving field is |
| 676 | too small to receive the value (actually, the semantics are inherited |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 677 | from downcasts in \C{} --- your milage may vary). |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 678 | |
| 679 | A few other characters have a meaning in a format string. These may |
| 680 | not occur inside nested parentheses. They are: |
| 681 | |
| 682 | \begin{description} |
| 683 | |
| 684 | \item[\samp{|}] |
| 685 | Indicates that the remaining arguments in the Python argument list are |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 686 | optional. The \C{} variables corresponding to optional arguments should |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 687 | be initialized to their default value --- when an optional argument is |
| 688 | not specified, the \code{PyArg_ParseTuple} does not touch the contents |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 689 | of the corresponding \C{} variable(s). |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 690 | |
| 691 | \item[\samp{:}] |
| 692 | The list of format units ends here; the string after the colon is used |
| 693 | as the function name in error messages (the ``associated value'' of |
| 694 | the exceptions that \code{PyArg_ParseTuple} raises). |
| 695 | |
| 696 | \item[\samp{;}] |
| 697 | The list of format units ends here; the string after the colon is used |
| 698 | as the error message \emph{instead} of the default error message. |
| 699 | Clearly, \samp{:} and \samp{;} mutually exclude each other. |
| 700 | |
| 701 | \end{description} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 702 | |
| 703 | Some example calls: |
| 704 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 705 | \begin{verbatim} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 706 | int ok; |
| 707 | int i, j; |
| 708 | long k, l; |
| 709 | char *s; |
| 710 | int size; |
| 711 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 712 | ok = PyArg_ParseTuple(args, ""); /* No arguments */ |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 713 | /* Python call: f() */ |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 714 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 715 | ok = PyArg_ParseTuple(args, "s", &s); /* A string */ |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 716 | /* Possible Python call: f('whoops!') */ |
| 717 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 718 | 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] | 719 | /* Possible Python call: f(1, 2, 'three') */ |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 720 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 721 | ok = PyArg_ParseTuple(args, "(ii)s#", &i, &j, &s, &size); |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 722 | /* A pair of ints and a string, whose size is also returned */ |
Guido van Rossum | 7e924dd | 1997-02-10 16:51:52 +0000 | [diff] [blame] | 723 | /* Possible Python call: f((1, 2), 'three') */ |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 724 | |
| 725 | { |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 726 | char *file; |
| 727 | char *mode = "r"; |
| 728 | int bufsize = 0; |
| 729 | ok = PyArg_ParseTuple(args, "s|si", &file, &mode, &bufsize); |
| 730 | /* A string, and optionally another string and an integer */ |
| 731 | /* Possible Python calls: |
| 732 | f('spam') |
| 733 | f('spam', 'w') |
| 734 | f('spam', 'wb', 100000) */ |
| 735 | } |
| 736 | |
| 737 | { |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 738 | int left, top, right, bottom, h, v; |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 739 | ok = PyArg_ParseTuple(args, "((ii)(ii))(ii)", |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 740 | &left, &top, &right, &bottom, &h, &v); |
| 741 | /* A rectangle and a point */ |
| 742 | /* Possible Python call: |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 743 | f(((0, 0), (400, 300)), (10, 10)) */ |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 744 | } |
Fred Drake | b6e5032 | 1998-02-04 20:26:31 +0000 | [diff] [blame^] | 745 | |
| 746 | { |
| 747 | Py_complex c; |
| 748 | ok = PyArg_ParseTuple(args, "D:myfunction", &c); |
| 749 | /* a complex, also providing a function name for errors */ |
| 750 | /* Possible Python call: myfunction(1+2j) */ |
| 751 | } |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 752 | \end{verbatim} |
Fred Drake | b6e5032 | 1998-02-04 20:26:31 +0000 | [diff] [blame^] | 753 | |
| 754 | |
| 755 | \section{Keyword Parsing with \sectcode{PyArg_ParseTupleAndKeywords()}} |
| 756 | |
| 757 | The \cfunction{PyArg_ParseTupleAndKeywords()} function is declared as |
| 758 | follows: |
| 759 | |
| 760 | \bcode\begin{verbatim} |
| 761 | int PyArg_ParseTupleAndKeywords(PyObject *arg, PyObject *kwdict, |
| 762 | char *format, char **kwlist, ...); |
| 763 | \end{verbatim}\ecode |
| 764 | |
| 765 | The \var{arg} and \var{format} parameters are identical to those of the |
| 766 | \cfunction{PyArg_ParseTuple()} function. The \var{kwdict} parameter |
| 767 | is the dictionary of keywords received as the third parameter from the |
| 768 | Python runtime. The \var{kwlist} parameter is a \NULL{}-terminated |
| 769 | list of strings which identify the parameters; the names are matched |
| 770 | with the type information from \var{format} from left to right. |
| 771 | |
| 772 | \strong{Note:} Nested tuples cannot be parsed when using keyword |
| 773 | arguments! Keyword parameters passed in which are not present in the |
| 774 | \var{kwlist} will cause a \exception{TypeError} to be raised. |
| 775 | |
| 776 | Here is an example module which uses keywords, based on an example by |
| 777 | Geoff Philbrick (\email{philbrick@hks.com}): |
| 778 | |
| 779 | \begin{verbatim} |
| 780 | #include <stdio.h> |
| 781 | #include "Python.h" |
| 782 | |
| 783 | static PyObject * |
| 784 | keywdarg_parrot(self, args, keywds) |
| 785 | PyObject *self; |
| 786 | PyObject *args; |
| 787 | PyObject *keywds; |
| 788 | { |
| 789 | int voltage; |
| 790 | char *state = "a stiff"; |
| 791 | char *action = "voom"; |
| 792 | char *type = "Norwegian Blue"; |
| 793 | |
| 794 | static char *kwlist[] = {"voltage", "state", "action", "type", NULL}; |
| 795 | |
| 796 | if (!PyArg_ParseTupleAndKeywords(args, keywds, "i|sss", kwlist, |
| 797 | &voltage, &state, &action, &type)) |
| 798 | return NULL; |
| 799 | |
| 800 | printf("-- This parrot wouldn't %s if you put %i Volts through it.\n", |
| 801 | action, voltage); |
| 802 | printf("-- Lovely plumage, the %s -- It's %s!\n", type, state); |
| 803 | |
| 804 | Py_INCREF(Py_None); |
| 805 | |
| 806 | return Py_None; |
| 807 | } |
| 808 | |
| 809 | static PyMethodDef keywdarg_methods[] = { |
| 810 | {"parrot", (PyCFunction)keywdarg_parrot, METH_VARARGS|METH_KEYWORDS}, |
| 811 | {NULL, NULL} /* sentinel */ |
| 812 | }; |
| 813 | |
| 814 | void |
| 815 | initkeywdarg() |
| 816 | { |
| 817 | /* Create the module and add the functions */ |
| 818 | Py_InitModule("keywdarg", keywdarg_methods); |
| 819 | |
| 820 | } |
| 821 | \end{verbatim} |
| 822 | |
| 823 | |
Fred Drake | 53396f6 | 1998-01-19 02:48:37 +0000 | [diff] [blame] | 824 | \section{The \sectcode{Py_BuildValue()} Function} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 825 | |
| 826 | This function is the counterpart to \code{PyArg_ParseTuple()}. It is |
| 827 | declared as follows: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 828 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 829 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 830 | PyObject *Py_BuildValue(char *format, ...); |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 831 | \end{verbatim}\ecode |
| 832 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 833 | It recognizes a set of format units similar to the ones recognized by |
| 834 | \code{PyArg_ParseTuple()}, but the arguments (which are input to the |
| 835 | function, not output) must not be pointers, just values. It returns a |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 836 | new Python object, suitable for returning from a \C{} function called |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 837 | from Python. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 838 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 839 | One difference with \code{PyArg_ParseTuple()}: while the latter |
| 840 | requires its first argument to be a tuple (since Python argument lists |
| 841 | are always represented as tuples internally), \code{BuildValue()} does |
| 842 | not always build a tuple. It builds a tuple only if its format string |
| 843 | contains two or more format units. If the format string is empty, it |
| 844 | returns \code{None}; if it contains exactly one format unit, it |
| 845 | returns whatever object is described by that format unit. To force it |
| 846 | 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] | 847 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 848 | In the following description, the quoted form is the format unit; the |
| 849 | entry in (round) parentheses is the Python object type that the format |
| 850 | unit will return; and the entry in [square] brackets is the type of |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 851 | the \C{} value(s) to be passed. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 852 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 853 | The characters space, tab, colon and comma are ignored in format |
| 854 | strings (but not within format units such as \samp{s\#}). This can be |
| 855 | used to make long format strings a tad more readable. |
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 | \begin{description} |
| 858 | |
| 859 | \item[\samp{s} (string) {[char *]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 860 | Convert a null-terminated \C{} string to a Python object. If the \C{} |
| 861 | string pointer is \NULL{}, \code{None} is returned. |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 862 | |
| 863 | \item[\samp{s\#} (string) {[char *, int]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 864 | Convert a \C{} string and its length to a Python object. If the \C{} string |
| 865 | pointer is \NULL{}, the length is ignored and \code{None} is |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 866 | returned. |
| 867 | |
| 868 | \item[\samp{z} (string or \code{None}) {[char *]}] |
| 869 | Same as \samp{s}. |
| 870 | |
| 871 | \item[\samp{z\#} (string or \code{None}) {[char *, int]}] |
| 872 | Same as \samp{s\#}. |
| 873 | |
| 874 | \item[\samp{i} (integer) {[int]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 875 | Convert a plain \C{} \code{int} to a Python integer object. |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 876 | |
| 877 | \item[\samp{b} (integer) {[char]}] |
| 878 | Same as \samp{i}. |
| 879 | |
| 880 | \item[\samp{h} (integer) {[short int]}] |
| 881 | Same as \samp{i}. |
| 882 | |
| 883 | \item[\samp{l} (integer) {[long int]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 884 | Convert a \C{} \code{long int} to a Python integer object. |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 885 | |
| 886 | \item[\samp{c} (string of length 1) {[char]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 887 | Convert a \C{} \code{int} representing a character to a Python string of |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 888 | length 1. |
| 889 | |
| 890 | \item[\samp{d} (float) {[double]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 891 | Convert a \C{} \code{double} to a Python floating point number. |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 892 | |
| 893 | \item[\samp{f} (float) {[float]}] |
| 894 | Same as \samp{d}. |
| 895 | |
| 896 | \item[\samp{O} (object) {[PyObject *]}] |
| 897 | Pass a Python object untouched (except for its reference count, which |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 898 | is incremented by one). If the object passed in is a \NULL{} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 899 | pointer, it is assumed that this was caused because the call producing |
| 900 | the argument found an error and set an exception. Therefore, |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 901 | \code{Py_BuildValue()} will return \NULL{} but won't raise an |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 902 | exception. If no exception has been raised yet, |
| 903 | \code{PyExc_SystemError} is set. |
| 904 | |
| 905 | \item[\samp{S} (object) {[PyObject *]}] |
| 906 | Same as \samp{O}. |
| 907 | |
| 908 | \item[\samp{O\&} (object) {[\var{converter}, \var{anything}]}] |
| 909 | Convert \var{anything} to a Python object through a \var{converter} |
| 910 | function. The function is called with \var{anything} (which should be |
| 911 | compatible with \code{void *}) as its argument and should return a |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 912 | ``new'' Python object, or \NULL{} if an error occurred. |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 913 | |
| 914 | \item[\samp{(\var{items})} (tuple) {[\var{matching-items}]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 915 | Convert a sequence of \C{} values to a Python tuple with the same number |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 916 | of items. |
| 917 | |
| 918 | \item[\samp{[\var{items}]} (list) {[\var{matching-items}]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 919 | Convert a sequence of \C{} values to a Python list with the same number |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 920 | of items. |
| 921 | |
| 922 | \item[\samp{\{\var{items}\}} (dictionary) {[\var{matching-items}]}] |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 923 | Convert a sequence of \C{} values to a Python dictionary. Each pair of |
| 924 | consecutive \C{} values adds one item to the dictionary, serving as key |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 925 | and value, respectively. |
| 926 | |
| 927 | \end{description} |
| 928 | |
| 929 | If there is an error in the format string, the |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 930 | \code{PyExc_SystemError} exception is raised and \NULL{} returned. |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 931 | |
| 932 | 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] | 933 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 934 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 935 | Py_BuildValue("") None |
| 936 | Py_BuildValue("i", 123) 123 |
Guido van Rossum | f23e0fe | 1995-03-18 11:04:29 +0000 | [diff] [blame] | 937 | Py_BuildValue("iii", 123, 456, 789) (123, 456, 789) |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 938 | Py_BuildValue("s", "hello") 'hello' |
| 939 | Py_BuildValue("ss", "hello", "world") ('hello', 'world') |
| 940 | Py_BuildValue("s#", "hello", 4) 'hell' |
| 941 | Py_BuildValue("()") () |
| 942 | Py_BuildValue("(i)", 123) (123,) |
| 943 | Py_BuildValue("(ii)", 123, 456) (123, 456) |
| 944 | Py_BuildValue("(i,i)", 123, 456) (123, 456) |
| 945 | Py_BuildValue("[i,i]", 123, 456) [123, 456] |
Guido van Rossum | f23e0fe | 1995-03-18 11:04:29 +0000 | [diff] [blame] | 946 | Py_BuildValue("{s:i,s:i}", |
| 947 | "abc", 123, "def", 456) {'abc': 123, 'def': 456} |
| 948 | Py_BuildValue("((ii)(ii)) (ii)", |
| 949 | 1, 2, 3, 4, 5, 6) (((1, 2), (3, 4)), (5, 6)) |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 950 | \end{verbatim}\ecode |
| 951 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 952 | \section{Reference Counts} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 953 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 954 | \subsection{Introduction} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 955 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 956 | In languages like \C{} or \Cpp{}, the programmer is responsible for |
| 957 | dynamic allocation and deallocation of memory on the heap. In \C{}, this |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 958 | is done using the functions \code{malloc()} and \code{free()}. In |
| 959 | \Cpp{}, the operators \code{new} and \code{delete} are used with |
| 960 | essentially the same meaning; they are actually implemented using |
| 961 | \code{malloc()} and \code{free()}, so we'll restrict the following |
| 962 | discussion to the latter. |
| 963 | |
| 964 | Every block of memory allocated with \code{malloc()} should eventually |
| 965 | be returned to the pool of available memory by exactly one call to |
| 966 | \code{free()}. It is important to call \code{free()} at the right |
| 967 | time. If a block's address is forgotten but \code{free()} is not |
| 968 | called for it, the memory it occupies cannot be reused until the |
| 969 | program terminates. This is called a \dfn{memory leak}. On the other |
| 970 | hand, if a program calls \code{free()} for a block and then continues |
| 971 | to use the block, it creates a conflict with re-use of the block |
| 972 | through another \code{malloc()} call. This is called \dfn{using freed |
Guido van Rossum | debf2e8 | 1997-07-17 15:58:43 +0000 | [diff] [blame] | 973 | memory}. It has the same bad consequences as referencing uninitialized |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 974 | data --- core dumps, wrong results, mysterious crashes. |
| 975 | |
| 976 | Common causes of memory leaks are unusual paths through the code. For |
| 977 | instance, a function may allocate a block of memory, do some |
| 978 | calculation, and then free the block again. Now a change in the |
| 979 | requirements for the function may add a test to the calculation that |
| 980 | detects an error condition and can return prematurely from the |
| 981 | function. It's easy to forget to free the allocated memory block when |
| 982 | taking this premature exit, especially when it is added later to the |
| 983 | code. Such leaks, once introduced, often go undetected for a long |
| 984 | time: the error exit is taken only in a small fraction of all calls, |
| 985 | and most modern machines have plenty of virtual memory, so the leak |
| 986 | only becomes apparent in a long-running process that uses the leaking |
| 987 | function frequently. Therefore, it's important to prevent leaks from |
| 988 | happening by having a coding convention or strategy that minimizes |
| 989 | this kind of errors. |
| 990 | |
| 991 | Since Python makes heavy use of \code{malloc()} and \code{free()}, it |
| 992 | needs a strategy to avoid memory leaks as well as the use of freed |
| 993 | memory. The chosen method is called \dfn{reference counting}. The |
| 994 | principle is simple: every object contains a counter, which is |
| 995 | incremented when a reference to the object is stored somewhere, and |
| 996 | which is decremented when a reference to it is deleted. When the |
| 997 | counter reaches zero, the last reference to the object has been |
| 998 | deleted and the object is freed. |
| 999 | |
| 1000 | An alternative strategy is called \dfn{automatic garbage collection}. |
| 1001 | (Sometimes, reference counting is also referred to as a garbage |
| 1002 | collection strategy, hence my use of ``automatic'' to distinguish the |
| 1003 | two.) The big advantage of automatic garbage collection is that the |
| 1004 | user doesn't need to call \code{free()} explicitly. (Another claimed |
| 1005 | advantage is an improvement in speed or memory usage --- this is no |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1006 | hard fact however.) The disadvantage is that for \C{}, there is no |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1007 | truly portable automatic garbage collector, while reference counting |
| 1008 | can be implemented portably (as long as the functions \code{malloc()} |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1009 | and \code{free()} are available --- which the \C{} Standard guarantees). |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1010 | Maybe some day a sufficiently portable automatic garbage collector |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1011 | will be available for \C{}. Until then, we'll have to live with |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1012 | reference counts. |
| 1013 | |
| 1014 | \subsection{Reference Counting in Python} |
| 1015 | |
| 1016 | There are two macros, \code{Py_INCREF(x)} and \code{Py_DECREF(x)}, |
| 1017 | which handle the incrementing and decrementing of the reference count. |
| 1018 | \code{Py_DECREF()} also frees the object when the count reaches zero. |
| 1019 | For flexibility, it doesn't call \code{free()} directly --- rather, it |
| 1020 | makes a call through a function pointer in the object's \dfn{type |
| 1021 | object}. For this purpose (and others), every object also contains a |
| 1022 | pointer to its type object. |
| 1023 | |
| 1024 | The big question now remains: when to use \code{Py_INCREF(x)} and |
| 1025 | \code{Py_DECREF(x)}? Let's first introduce some terms. Nobody |
| 1026 | ``owns'' an object; however, you can \dfn{own a reference} to an |
| 1027 | object. An object's reference count is now defined as the number of |
| 1028 | owned references to it. The owner of a reference is responsible for |
| 1029 | calling \code{Py_DECREF()} when the reference is no longer needed. |
| 1030 | Ownership of a reference can be transferred. There are three ways to |
| 1031 | dispose of an owned reference: pass it on, store it, or call |
| 1032 | \code{Py_DECREF()}. Forgetting to dispose of an owned reference creates |
| 1033 | a memory leak. |
| 1034 | |
| 1035 | It is also possible to \dfn{borrow}\footnote{The metaphor of |
| 1036 | ``borrowing'' a reference is not completely correct: the owner still |
| 1037 | has a copy of the reference.} a reference to an object. The borrower |
| 1038 | of a reference should not call \code{Py_DECREF()}. The borrower must |
| 1039 | not hold on to the object longer than the owner from which it was |
| 1040 | borrowed. Using a borrowed reference after the owner has disposed of |
| 1041 | it risks using freed memory and should be avoided |
| 1042 | completely.\footnote{Checking that the reference count is at least 1 |
| 1043 | \strong{does not work} --- the reference count itself could be in |
| 1044 | freed memory and may thus be reused for another object!} |
| 1045 | |
| 1046 | The advantage of borrowing over owning a reference is that you don't |
| 1047 | need to take care of disposing of the reference on all possible paths |
| 1048 | through the code --- in other words, with a borrowed reference you |
| 1049 | don't run the risk of leaking when a premature exit is taken. The |
| 1050 | disadvantage of borrowing over leaking is that there are some subtle |
| 1051 | situations where in seemingly correct code a borrowed reference can be |
| 1052 | used after the owner from which it was borrowed has in fact disposed |
| 1053 | of it. |
| 1054 | |
| 1055 | A borrowed reference can be changed into an owned reference by calling |
| 1056 | \code{Py_INCREF()}. This does not affect the status of the owner from |
| 1057 | which the reference was borrowed --- it creates a new owned reference, |
| 1058 | and gives full owner responsibilities (i.e., the new owner must |
| 1059 | dispose of the reference properly, as well as the previous owner). |
| 1060 | |
| 1061 | \subsection{Ownership Rules} |
| 1062 | |
| 1063 | Whenever an object reference is passed into or out of a function, it |
| 1064 | is part of the function's interface specification whether ownership is |
| 1065 | transferred with the reference or not. |
| 1066 | |
| 1067 | Most functions that return a reference to an object pass on ownership |
| 1068 | with the reference. In particular, all functions whose function it is |
| 1069 | to create a new object, e.g.\ \code{PyInt_FromLong()} and |
| 1070 | \code{Py_BuildValue()}, pass ownership to the receiver. Even if in |
| 1071 | fact, in some cases, you don't receive a reference to a brand new |
| 1072 | object, you still receive ownership of the reference. For instance, |
| 1073 | \code{PyInt_FromLong()} maintains a cache of popular values and can |
| 1074 | return a reference to a cached item. |
| 1075 | |
| 1076 | Many functions that extract objects from other objects also transfer |
| 1077 | ownership with the reference, for instance |
| 1078 | \code{PyObject_GetAttrString()}. The picture is less clear, here, |
| 1079 | however, since a few common routines are exceptions: |
| 1080 | \code{PyTuple_GetItem()}, \code{PyList_GetItem()} and |
| 1081 | \code{PyDict_GetItem()} (and \code{PyDict_GetItemString()}) all return |
| 1082 | references that you borrow from the tuple, list or dictionary. |
| 1083 | |
| 1084 | The function \code{PyImport_AddModule()} also returns a borrowed |
| 1085 | reference, even though it may actually create the object it returns: |
| 1086 | this is possible because an owned reference to the object is stored in |
| 1087 | \code{sys.modules}. |
| 1088 | |
| 1089 | When you pass an object reference into another function, in general, |
| 1090 | the function borrows the reference from you --- if it needs to store |
| 1091 | it, it will use \code{Py_INCREF()} to become an independent owner. |
| 1092 | There are exactly two important exceptions to this rule: |
| 1093 | \code{PyTuple_SetItem()} and \code{PyList_SetItem()}. These functions |
| 1094 | take over ownership of the item passed to them --- even if they fail! |
| 1095 | (Note that \code{PyDict_SetItem()} and friends don't take over |
| 1096 | ownership --- they are ``normal''.) |
| 1097 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1098 | When a \C{} function is called from Python, it borrows references to its |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1099 | arguments from the caller. The caller owns a reference to the object, |
| 1100 | so the borrowed reference's lifetime is guaranteed until the function |
| 1101 | returns. Only when such a borrowed reference must be stored or passed |
| 1102 | on, it must be turned into an owned reference by calling |
| 1103 | \code{Py_INCREF()}. |
| 1104 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1105 | The object reference returned from a \C{} function that is called from |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1106 | Python must be an owned reference --- ownership is tranferred from the |
| 1107 | function to its caller. |
| 1108 | |
| 1109 | \subsection{Thin Ice} |
| 1110 | |
| 1111 | There are a few situations where seemingly harmless use of a borrowed |
| 1112 | reference can lead to problems. These all have to do with implicit |
| 1113 | invocations of the interpreter, which can cause the owner of a |
| 1114 | reference to dispose of it. |
| 1115 | |
| 1116 | The first and most important case to know about is using |
| 1117 | \code{Py_DECREF()} on an unrelated object while borrowing a reference |
| 1118 | to a list item. For instance: |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1119 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1120 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1121 | bug(PyObject *list) { |
| 1122 | PyObject *item = PyList_GetItem(list, 0); |
| 1123 | PyList_SetItem(list, 1, PyInt_FromLong(0L)); |
| 1124 | PyObject_Print(item, stdout, 0); /* BUG! */ |
| 1125 | } |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1126 | \end{verbatim}\ecode |
| 1127 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1128 | This function first borrows a reference to \code{list[0]}, then |
| 1129 | replaces \code{list[1]} with the value \code{0}, and finally prints |
| 1130 | the borrowed reference. Looks harmless, right? But it's not! |
| 1131 | |
| 1132 | Let's follow the control flow into \code{PyList_SetItem()}. The list |
| 1133 | owns references to all its items, so when item 1 is replaced, it has |
| 1134 | to dispose of the original item 1. Now let's suppose the original |
| 1135 | item 1 was an instance of a user-defined class, and let's further |
| 1136 | suppose that the class defined a \code{__del__()} method. If this |
| 1137 | class instance has a reference count of 1, disposing of it will call |
| 1138 | its \code{__del__()} method. |
| 1139 | |
| 1140 | Since it is written in Python, the \code{__del__()} method can execute |
| 1141 | arbitrary Python code. Could it perhaps do something to invalidate |
| 1142 | the reference to \code{item} in \code{bug()}? You bet! Assuming that |
| 1143 | the list passed into \code{bug()} is accessible to the |
| 1144 | \code{__del__()} method, it could execute a statement to the effect of |
| 1145 | \code{del list[0]}, and assuming this was the last reference to that |
| 1146 | object, it would free the memory associated with it, thereby |
| 1147 | invalidating \code{item}. |
| 1148 | |
| 1149 | The solution, once you know the source of the problem, is easy: |
| 1150 | temporarily increment the reference count. The correct version of the |
| 1151 | function reads: |
| 1152 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1153 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1154 | no_bug(PyObject *list) { |
| 1155 | PyObject *item = PyList_GetItem(list, 0); |
| 1156 | Py_INCREF(item); |
| 1157 | PyList_SetItem(list, 1, PyInt_FromLong(0L)); |
| 1158 | PyObject_Print(item, stdout, 0); |
| 1159 | Py_DECREF(item); |
| 1160 | } |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1161 | \end{verbatim}\ecode |
| 1162 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1163 | This is a true story. An older version of Python contained variants |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1164 | of this bug and someone spent a considerable amount of time in a \C{} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1165 | debugger to figure out why his \code{__del__()} methods would fail... |
| 1166 | |
| 1167 | The second case of problems with a borrowed reference is a variant |
| 1168 | involving threads. Normally, multiple threads in the Python |
| 1169 | interpreter can't get in each other's way, because there is a global |
| 1170 | lock protecting Python's entire object space. However, it is possible |
| 1171 | to temporarily release this lock using the macro |
| 1172 | \code{Py_BEGIN_ALLOW_THREADS}, and to re-acquire it using |
| 1173 | \code{Py_END_ALLOW_THREADS}. This is common around blocking I/O |
| 1174 | calls, to let other threads use the CPU while waiting for the I/O to |
| 1175 | complete. Obviously, the following function has the same problem as |
| 1176 | the previous one: |
| 1177 | |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1178 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1179 | bug(PyObject *list) { |
| 1180 | PyObject *item = PyList_GetItem(list, 0); |
| 1181 | Py_BEGIN_ALLOW_THREADS |
| 1182 | ...some blocking I/O call... |
| 1183 | Py_END_ALLOW_THREADS |
| 1184 | PyObject_Print(item, stdout, 0); /* BUG! */ |
| 1185 | } |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1186 | \end{verbatim}\ecode |
| 1187 | % |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1188 | \subsection{NULL Pointers} |
| 1189 | |
| 1190 | In general, functions that take object references as arguments don't |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1191 | expect you to pass them \NULL{} pointers, and will dump core (or |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1192 | cause later core dumps) if you do so. Functions that return object |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1193 | references generally return \NULL{} only to indicate that an |
| 1194 | exception occurred. The reason for not testing for \NULL{} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1195 | arguments is that functions often pass the objects they receive on to |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1196 | other function --- if each function were to test for \NULL{}, |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1197 | there would be a lot of redundant tests and the code would run slower. |
| 1198 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1199 | It is better to test for \NULL{} only at the ``source'', i.e.\ |
| 1200 | when a pointer that may be \NULL{} is received, e.g.\ from |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1201 | \code{malloc()} or from a function that may raise an exception. |
| 1202 | |
| 1203 | The macros \code{Py_INCREF()} and \code{Py_DECREF()} |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1204 | don't check for \NULL{} pointers --- however, their variants |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1205 | \code{Py_XINCREF()} and \code{Py_XDECREF()} do. |
| 1206 | |
| 1207 | The macros for checking for a particular object type |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1208 | (\code{Py\var{type}_Check()}) don't check for \NULL{} pointers --- |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1209 | again, there is much code that calls several of these in a row to test |
| 1210 | an object against various different expected types, and this would |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1211 | generate redundant tests. There are no variants with \NULL{} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1212 | checking. |
| 1213 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1214 | The \C{} function calling mechanism guarantees that the argument list |
| 1215 | passed to \C{} functions (\code{args} in the examples) is never |
| 1216 | \NULL{} --- in fact it guarantees that it is always a tuple.% |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1217 | \footnote{These guarantees don't hold when you use the ``old'' style |
| 1218 | calling convention --- this is still found in much existing code.} |
| 1219 | |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1220 | It is a severe error to ever let a \NULL{} pointer ``escape'' to |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1221 | the Python user. |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1222 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1223 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1224 | \section{Writing Extensions in \Cpp{}} |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1225 | |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 1226 | It is possible to write extension modules in \Cpp{}. Some restrictions |
Guido van Rossum | ed39cd0 | 1995-10-08 00:17:19 +0000 | [diff] [blame] | 1227 | apply. If the main program (the Python interpreter) is compiled and |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1228 | linked by the \C{} compiler, global or static objects with constructors |
Guido van Rossum | ed39cd0 | 1995-10-08 00:17:19 +0000 | [diff] [blame] | 1229 | cannot be used. This is not a problem if the main program is linked |
| 1230 | by the \Cpp{} compiler. All functions that will be called directly or |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1231 | indirectly (i.e. via function pointers) by the Python interpreter will |
| 1232 | have to be declared using \code{extern "C"}; this applies to all |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 1233 | ``methods'' as well as to the module's initialization function. |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1234 | It is unnecessary to enclose the Python header files in |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1235 | \code{extern "C" \{...\}} --- they use this form already if the symbol |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1236 | \samp{__cplusplus} is defined (all recent \Cpp{} compilers define this |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1237 | symbol). |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1238 | |
| 1239 | \chapter{Embedding Python in another application} |
| 1240 | |
| 1241 | Embedding Python is similar to extending it, but not quite. The |
| 1242 | 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] | 1243 | application is still the Python interpreter, while if you embed |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1244 | Python, the main program may have nothing to do with Python --- |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1245 | instead, some parts of the application occasionally call the Python |
| 1246 | interpreter to run some Python code. |
| 1247 | |
| 1248 | So if you are embedding Python, you are providing your own main |
| 1249 | program. One of the things this main program has to do is initialize |
| 1250 | 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] | 1251 | function \code{Py_Initialize()}. There are optional calls to pass command |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1252 | line arguments to Python. Then later you can call the interpreter |
| 1253 | from any part of the application. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1254 | |
| 1255 | 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] | 1256 | a string containing Python statements to \code{PyRun_SimpleString()}, |
| 1257 | or you can pass a stdio file pointer and a file name (for |
| 1258 | identification in error messages only) to \code{PyRun_SimpleFile()}. You |
| 1259 | can also call the lower-level operations described in the previous |
| 1260 | chapters to construct and use Python objects. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1261 | |
| 1262 | 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] | 1263 | \file{Demo/embed}. |
Guido van Rossum | db65a6c | 1993-11-05 17:11:16 +0000 | [diff] [blame] | 1264 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1265 | |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 1266 | \section{Embedding Python in \Cpp{}} |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1267 | |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 1268 | It is also possible to embed Python in a \Cpp{} program; precisely how this |
| 1269 | is done will depend on the details of the \Cpp{} system used; in general you |
| 1270 | will need to write the main program in \Cpp{}, and use the \Cpp{} compiler |
| 1271 | to compile and link your program. There is no need to recompile Python |
| 1272 | itself using \Cpp{}. |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1273 | |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1274 | |
| 1275 | \chapter{Dynamic Loading} |
| 1276 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1277 | On most modern systems it is possible to configure Python to support |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1278 | dynamic loading of extension modules implemented in \C{}. When shared |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1279 | libraries are used dynamic loading is configured automatically; |
| 1280 | otherwise you have to select it as a build option (see below). Once |
| 1281 | configured, dynamic loading is trivial to use: when a Python program |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1282 | executes \code{import spam}, the search for modules tries to find a |
| 1283 | file \file{spammodule.o} (\file{spammodule.so} when using shared |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1284 | libraries) in the module search path, and if one is found, it is |
| 1285 | loaded into the executing binary and executed. Once loaded, the |
| 1286 | module acts just like a built-in extension module. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1287 | |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 1288 | The advantages of dynamic loading are twofold: the ``core'' Python |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1289 | binary gets smaller, and users can extend Python with their own |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1290 | modules implemented in \C{} without having to build and maintain their |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1291 | own copy of the Python interpreter. There are also disadvantages: |
| 1292 | dynamic loading isn't available on all systems (this just means that |
| 1293 | on some systems you have to use static loading), and dynamically |
| 1294 | 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] | 1295 | (e.g. with a different representation of objects) may dump core. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1296 | |
| 1297 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1298 | \section{Configuring and Building the Interpreter for Dynamic Loading} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1299 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1300 | There are three styles of dynamic loading: one using shared libraries, |
| 1301 | one using SGI IRIX 4 dynamic loading, and one using GNU dynamic |
| 1302 | loading. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1303 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1304 | \subsection{Shared Libraries} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1305 | |
Guido van Rossum | 16d6e71 | 1994-08-08 12:30:22 +0000 | [diff] [blame] | 1306 | The following systems support dynamic loading using shared libraries: |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1307 | SunOS 4; Solaris 2; SGI IRIX 5 (but not SGI IRIX 4!); and probably all |
| 1308 | systems derived from SVR4, or at least those SVR4 derivatives that |
| 1309 | support shared libraries (are there any that don't?). |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1310 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1311 | You don't need to do anything to configure dynamic loading on these |
| 1312 | systems --- the \file{configure} detects the presence of the |
| 1313 | \file{<dlfcn.h>} header file and automatically configures dynamic |
| 1314 | loading. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1315 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1316 | \subsection{SGI IRIX 4 Dynamic Loading} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1317 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1318 | Only SGI IRIX 4 supports dynamic loading of modules using SGI dynamic |
| 1319 | loading. (SGI IRIX 5 might also support it but it is inferior to |
| 1320 | using shared libraries so there is no reason to; a small test didn't |
| 1321 | work right away so I gave up trying to support it.) |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1322 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1323 | Before you build Python, you first need to fetch and build the \code{dl} |
| 1324 | package written by Jack Jansen. This is available by anonymous ftp |
Fred Drake | ca6567f | 1998-01-22 20:44:18 +0000 | [diff] [blame] | 1325 | from \url{ftp://ftp.cwi.nl/pub/dynload}, file |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1326 | \file{dl-1.6.tar.Z}. (The version number may change.) Follow the |
| 1327 | instructions in the package's \file{README} file to build it. |
| 1328 | |
| 1329 | Once you have built \code{dl}, you can configure Python to use it. To |
| 1330 | this end, you run the \file{configure} script with the option |
| 1331 | \code{--with-dl=\var{directory}} where \var{directory} is the absolute |
| 1332 | pathname of the \code{dl} directory. |
| 1333 | |
| 1334 | Now build and install Python as you normally would (see the |
| 1335 | \file{README} file in the toplevel Python directory.) |
| 1336 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1337 | \subsection{GNU Dynamic Loading} |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1338 | |
| 1339 | GNU dynamic loading supports (according to its \file{README} file) the |
| 1340 | following hardware and software combinations: VAX (Ultrix), Sun 3 |
| 1341 | (SunOS 3.4 and 4.0), Sparc (SunOS 4.0), Sequent Symmetry (Dynix), and |
| 1342 | Atari ST. There is no reason to use it on a Sparc; I haven't seen a |
| 1343 | Sun 3 for years so I don't know if these have shared libraries or not. |
| 1344 | |
Guido van Rossum | 7e924dd | 1997-02-10 16:51:52 +0000 | [diff] [blame] | 1345 | You need to fetch and build two packages. |
| 1346 | One is GNU DLD. All development of this code has been done with DLD |
Fred Drake | ca6567f | 1998-01-22 20:44:18 +0000 | [diff] [blame] | 1347 | version 3.2.3, which is available by anonymous ftp from |
| 1348 | \url{ftp://ftp.cwi.nl/pub/dynload}, file |
Guido van Rossum | 7e924dd | 1997-02-10 16:51:52 +0000 | [diff] [blame] | 1349 | \file{dld-3.2.3.tar.Z}. (A more recent version of DLD is available |
Fred Drake | ca6567f | 1998-01-22 20:44:18 +0000 | [diff] [blame] | 1350 | via \url{http://www-swiss.ai.mit.edu/~jaffer/DLD.html} but this has |
Guido van Rossum | 7e924dd | 1997-02-10 16:51:52 +0000 | [diff] [blame] | 1351 | not been tested.) |
| 1352 | The other package needed is an |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1353 | emulation of Jack Jansen's \code{dl} package that I wrote on top of |
| 1354 | GNU DLD 3.2.3. This is available from the same host and directory, |
Guido van Rossum | 98046b9 | 1997-08-14 19:50:18 +0000 | [diff] [blame] | 1355 | file \file{dl-dld-1.1.tar.Z}. (The version number may change --- but I doubt |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1356 | it will.) Follow the instructions in each package's \file{README} |
Guido van Rossum | 98046b9 | 1997-08-14 19:50:18 +0000 | [diff] [blame] | 1357 | file to configure and build them. |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1358 | |
| 1359 | Now configure Python. Run the \file{configure} script with the option |
| 1360 | \code{--with-dl-dld=\var{dl-directory},\var{dld-directory}} where |
| 1361 | \var{dl-directory} is the absolute pathname of the directory where you |
| 1362 | have built the \file{dl-dld} package, and \var{dld-directory} is that |
| 1363 | of the GNU DLD package. The Python interpreter you build hereafter |
| 1364 | will support GNU dynamic loading. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1365 | |
| 1366 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1367 | \section{Building a Dynamically Loadable Module} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1368 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1369 | Since there are three styles of dynamic loading, there are also three |
| 1370 | groups of instructions for building a dynamically loadable module. |
| 1371 | Instructions common for all three styles are given first. Assuming |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1372 | your module is called \code{spam}, the source filename must be |
| 1373 | \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] | 1374 | module must be written as a normal Python extension module (as |
| 1375 | described earlier). |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1376 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1377 | Note that in all cases you will have to create your own Makefile that |
| 1378 | compiles your module file(s). This Makefile will have to pass two |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1379 | \samp{-I} arguments to the \C{} compiler which will make it find the |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1380 | Python header files. If the Make variable \var{PYTHONTOP} points to |
| 1381 | the toplevel Python directory, your \var{CFLAGS} Make variable should |
| 1382 | contain the options \samp{-I\$(PYTHONTOP) -I\$(PYTHONTOP)/Include}. |
| 1383 | (Most header files are in the \file{Include} subdirectory, but the |
Guido van Rossum | 305ed11 | 1996-08-19 22:59:46 +0000 | [diff] [blame] | 1384 | \file{config.h} header lives in the toplevel directory.) |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1385 | |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1386 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1387 | \subsection{Shared Libraries} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1388 | |
Fred Drake | af8a015 | 1998-01-14 14:51:31 +0000 | [diff] [blame] | 1389 | You must link the \file{.o} file to produce a shared library. This is |
| 1390 | done using a special invocation of the \UNIX{} loader/linker, |
| 1391 | \emph{ld}(1). Unfortunately the invocation differs slightly per |
| 1392 | system. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1393 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1394 | On SunOS 4, use |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1395 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1396 | ld spammodule.o -o spammodule.so |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1397 | \end{verbatim}\ecode |
| 1398 | % |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1399 | On Solaris 2, use |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1400 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1401 | ld -G spammodule.o -o spammodule.so |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1402 | \end{verbatim}\ecode |
| 1403 | % |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1404 | On SGI IRIX 5, use |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1405 | \bcode\begin{verbatim} |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1406 | ld -shared spammodule.o -o spammodule.so |
Guido van Rossum | e47da0a | 1997-07-17 16:34:52 +0000 | [diff] [blame] | 1407 | \end{verbatim}\ecode |
| 1408 | % |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 1409 | On other systems, consult the manual page for \code{ld}(1) to find what |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1410 | flags, if any, must be used. |
| 1411 | |
| 1412 | If your extension module uses system libraries that haven't already |
| 1413 | been linked with Python (e.g. a windowing system), these must be |
Guido van Rossum | b92112d | 1995-03-20 14:24:09 +0000 | [diff] [blame] | 1414 | passed to the \code{ld} command as \samp{-l} options after the |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1415 | \samp{.o} file. |
| 1416 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1417 | 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] | 1418 | along the Python module search path. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1419 | |
| 1420 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1421 | \subsection{SGI IRIX 4 Dynamic Loading} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1422 | |
Fred Drake | af8a015 | 1998-01-14 14:51:31 +0000 | [diff] [blame] | 1423 | \strong{IMPORTANT:} You must compile your extension module with the |
Fred Drake | 0fd8268 | 1998-01-09 05:39:38 +0000 | [diff] [blame] | 1424 | additional \C{} flag \samp{-G0} (or \samp{-G 0}). This instruct the |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1425 | assembler to generate position-independent code. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1426 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1427 | 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] | 1428 | copy it into a directory along the Python module search path. |
| 1429 | |
| 1430 | The first time your extension is loaded, it takes some extra time and |
| 1431 | a few messages may be printed. This creates a file |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1432 | \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] | 1433 | the Python interpreter process. When a new Python interpreter is |
| 1434 | installed, the \code{dl} package detects this and rebuilds |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1435 | \file{spammodule.ld}. The file \file{spammodule.ld} is placed in the |
| 1436 | directory where \file{spammodule.o} was found, unless this directory is |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1437 | unwritable; in that case it is placed in a temporary |
| 1438 | directory.\footnote{Check the manual page of the \code{dl} package for |
| 1439 | details.} |
| 1440 | |
| 1441 | If your extension modules uses additional system libraries, you must |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1442 | create a file \file{spammodule.libs} in the same directory as the |
| 1443 | \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] | 1444 | whitespace-separated options that will be passed to the linker --- |
| 1445 | normally only \samp{-l} options or absolute pathnames of libraries |
| 1446 | (\samp{.a} files) should be used. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1447 | |
| 1448 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1449 | \subsection{GNU Dynamic Loading} |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1450 | |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1451 | 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] | 1452 | search path. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1453 | |
Guido van Rossum | 6938f06 | 1994-08-01 12:22:53 +0000 | [diff] [blame] | 1454 | If your extension modules uses additional system libraries, you must |
Guido van Rossum | 5049bcb | 1995-03-13 16:55:23 +0000 | [diff] [blame] | 1455 | create a file \file{spammodule.libs} in the same directory as the |
| 1456 | \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] | 1457 | whitespace-separated absolute pathnames of libraries (\samp{.a} |
| 1458 | files). No \samp{-l} options can be used. |
Guido van Rossum | 6f0132f | 1993-11-19 13:13:22 +0000 | [diff] [blame] | 1459 | |
| 1460 | |
Guido van Rossum | 9231c8f | 1997-05-15 21:43:21 +0000 | [diff] [blame] | 1461 | %\input{extref} |
Guido van Rossum | 267e80d | 1996-08-09 21:01:07 +0000 | [diff] [blame] | 1462 | |
Guido van Rossum | 7a2dba2 | 1993-11-05 14:45:11 +0000 | [diff] [blame] | 1463 | \input{ext.ind} |
| 1464 | |
| 1465 | \end{document} |