Difference between revisions of "Python - ctypes"

From PeformIQ Upgrade
Jump to navigation Jump to search
Line 2: Line 2:


Extracted from http://starship.python.net/crew/theller/ctypes/tutorial.html
Extracted from http://starship.python.net/crew/theller/ctypes/tutorial.html
<html>
<h1 class="title">ctypes tutorial</h1>
<div class="contents topic">
<p class="topic-title first"><a id="contents" name="contents">Contents</a></p>
<ul class="simple">
<li><a class="reference" href="#loading-dynamic-link-libraries" id="id1" name="id1">Loading dynamic link libraries</a></li>
<li><a class="reference" href="#accessing-functions-from-loaded-dlls" id="id2" name="id2">Accessing functions from loaded dlls</a></li>
<li><a class="reference" href="#calling-functions" id="id3" name="id3">Calling functions</a></li>
<li><a class="reference" href="#fundamental-data-types" id="id4" name="id4">Fundamental data types</a></li>
<li><a class="reference" href="#calling-functions-continued" id="id5" name="id5">Calling functions, continued</a></li>
<li><a class="reference" href="#calling-functions-with-your-own-custom-data-types" id="id6" name="id6">Calling functions with your own custom data types</a></li>
<li><a class="reference" href="#specifying-the-required-argument-types-function-prototypes" id="id7" name="id7">Specifying the required argument types (function prototypes)</a></li>
<li><a class="reference" href="#return-types" id="id8" name="id8">Return types</a></li>
<li><a class="reference" href="#passing-pointers-or-passing-parameters-by-reference" id="id9" name="id9">Passing pointers (or: passing parameters by reference)</a></li>
<li><a class="reference" href="#structures-and-unions" id="id10" name="id10">Structures and unions</a></li>
<li><a class="reference" href="#structure-union-alignment-and-byte-order" id="id11" name="id11">Structure/union alignment and byte order</a></li>
<li><a class="reference" href="#bit-fields-in-structures-and-unions" id="id12" name="id12">Bit fields in structures and unions</a></li>
<li><a class="reference" href="#arrays" id="id13" name="id13">Arrays</a></li>
<li><a class="reference" href="#pointers" id="id14" name="id14">Pointers</a></li>
<li><a class="reference" href="#type-conversions" id="id15" name="id15">Type conversions</a></li>
<li><a class="reference" href="#incomplete-types" id="id16" name="id16">Incomplete Types</a></li>
<li><a class="reference" href="#callback-functions" id="id17" name="id17">Callback functions</a></li>
<li><a class="reference" href="#accessing-values-exported-from-dlls" id="id18" name="id18">Accessing values exported from dlls</a></li>
<li><a class="reference" href="#surprises" id="id19" name="id19">Surprises</a></li>
<li><a class="reference" href="#variable-sized-data-types" id="id20" name="id20">Variable-sized data types</a></li>
<li><a class="reference" href="#bugs-todo-and-non-implemented-things" id="id21" name="id21">Bugs, ToDo and non-implemented things</a></li>
</ul>
</div>
<p>Note: The code samples in this tutorial uses <tt class="docutils literal"><span class="pre">doctest</span></tt> to make sure
that they actually work.  Since some code samples behave differently
under Linux, Windows, or Mac OS X, they contain doctest directives in
comments.</p>
<p>Note: Quite some code samples references the ctypes <tt class="docutils literal"><span class="pre">c_int</span></tt> type.
This type is an alias to the <tt class="docutils literal"><span class="pre">c_long</span></tt> type on 32-bit systems.  So,
you should not be confused if <tt class="docutils literal"><span class="pre">c_long</span></tt> is printed if you would
expect <tt class="docutils literal"><span class="pre">c_int</span></tt> - they are actually the same type.</p>
<div class="section">
<h1><a class="toc-backref" href="#id1" id="loading-dynamic-link-libraries" name="loading-dynamic-link-libraries">Loading dynamic link libraries</a></h1>
<p><tt class="docutils literal"><span class="pre">ctypes</span></tt> exports the <tt class="docutils literal"><span class="pre">cdll</span></tt>, and on Windows also <tt class="docutils literal"><span class="pre">windll</span></tt> and
<tt class="docutils literal"><span class="pre">oledll</span></tt> objects to load dynamic link libraries.</p>
<p>You load libraries by accessing them as attributes of these objects.
<tt class="docutils literal"><span class="pre">cdll</span></tt> loads libraries which export functions using the standard
<tt class="docutils literal"><span class="pre">cdecl</span></tt> calling convention, while <tt class="docutils literal"><span class="pre">windll</span></tt> libraries call
functions using the <tt class="docutils literal"><span class="pre">stdcall</span></tt> calling convention. <tt class="docutils literal"><span class="pre">oledll</span></tt> also
uses the <tt class="docutils literal"><span class="pre">stdcall</span></tt> calling convention, and assumes the functions
return a Windows <tt class="docutils literal"><span class="pre">HRESULT</span></tt> error code. The error code is used to
automatically raise <tt class="docutils literal"><span class="pre">WindowsError</span></tt> Python exceptions when the
function call fails.</p>
<p>Here are some examples for Windows, note that <tt class="docutils literal"><span class="pre">msvcrt</span></tt> is the MS
standard C library containing most standard C functions, and uses the
cdecl calling convention:</p>
<pre class="literal-block">
&gt;&gt;&gt; from ctypes import *
&gt;&gt;&gt; print windll.kernel32 # doctest: +WINDOWS
&lt;WinDLL 'kernel32', handle ... at ...&gt;
&gt;&gt;&gt; print cdll.msvcrt # doctest: +WINDOWS
&lt;CDLL 'msvcrt', handle ... at ...&gt;
&gt;&gt;&gt; libc = cdll.msvcrt # doctest: +WINDOWS
&gt;&gt;&gt;
</pre>
<p>Windows appends the usual '.dll' file suffix automatically.</p>
<p>On Linux, it is required to specify the filename <em>including</em> the
extension to load a library, so attribute access does not work.
Either the <tt class="docutils literal"><span class="pre">LoadLibrary</span></tt> method of the dll loaders should be used,
or you should load the library by creating an instance of CDLL by
calling the constructor:</p>
<pre class="literal-block">
&gt;&gt;&gt; cdll.LoadLibrary(&quot;libc.so.6&quot;) # doctest: +LINUX
&lt;CDLL 'libc.so.6', handle ... at ...&gt;
&gt;&gt;&gt; libc = CDLL(&quot;libc.so.6&quot;)    # doctest: +LINUX
&gt;&gt;&gt; libc                        # doctest: +LINUX
&lt;CDLL 'libc.so.6', handle ... at ...&gt;
&gt;&gt;&gt;
</pre>
<!-- XXX Add section for Mac OS X. -->
</div>
<div class="section">
<h1><a class="toc-backref" href="#id2" id="accessing-functions-from-loaded-dlls" name="accessing-functions-from-loaded-dlls">Accessing functions from loaded dlls</a></h1>
<p>Functions are accessed as attributes of dll objects:</p>
<pre class="literal-block">
&gt;&gt;&gt; from ctypes import *
&gt;&gt;&gt; libc.printf
&lt;_FuncPtr object at 0x...&gt;
&gt;&gt;&gt; print windll.kernel32.GetModuleHandleA # doctest: +WINDOWS
&lt;_FuncPtr object at 0x...&gt;
&gt;&gt;&gt; print windll.kernel32.MyOwnFunction # doctest: +WINDOWS
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
  File &quot;ctypes.py&quot;, line 239, in __getattr__
    func = _StdcallFuncPtr(name, self)
AttributeError: function 'MyOwnFunction' not found
&gt;&gt;&gt;
</pre>
<p>Note that win32 system dlls like <tt class="docutils literal"><span class="pre">kernel32</span></tt> and <tt class="docutils literal"><span class="pre">user32</span></tt> often
export ANSI as well as UNICODE versions of a function. The UNICODE
version is exported with an <tt class="docutils literal"><span class="pre">W</span></tt> appended to the name, while the ANSI
version is exported with an <tt class="docutils literal"><span class="pre">A</span></tt> appended to the name. The win32
<tt class="docutils literal"><span class="pre">GetModuleHandle</span></tt> function, which returns a <em>module handle</em> for a
given module name, has the following C prototype, and a macro is used
to expose one of them as <tt class="docutils literal"><span class="pre">GetModuleHandle</span></tt> depending on whether
UNICODE is defined or not:</p>
<pre class="literal-block">
/* ANSI version */
HMODULE GetModuleHandleA(LPCSTR lpModuleName);
/* UNICODE version */
HMODULE GetModuleHandleW(LPCWSTR lpModuleName);
</pre>
<p><tt class="docutils literal"><span class="pre">windll</span></tt> does not try to select one of them by magic, you must
access the version you need by specifying <tt class="docutils literal"><span class="pre">GetModuleHandleA</span></tt> or
<tt class="docutils literal"><span class="pre">GetModuleHandleW</span></tt> explicitely, and then call it with normal strings
or unicode strings respectively.</p>
<p>Sometimes, dlls export functions with names which aren't valid Python
identifiers, like <tt class="docutils literal"><span class="pre">&quot;??2&#64;YAPAXI&#64;Z&quot;</span></tt>. In this case you have to use
<tt class="docutils literal"><span class="pre">getattr</span></tt> to retrieve the function:</p>
<pre class="literal-block">
&gt;&gt;&gt; getattr(cdll.msvcrt, &quot;??2&#64;YAPAXI&#64;Z&quot;) # doctest: +WINDOWS
&lt;_FuncPtr object at 0x...&gt;
&gt;&gt;&gt;
</pre>
<p>On Windows, some dlls export functions not by name but by ordinal.
These functions can be accessed by indexing the dll object with the
ordinal number:</p>
<pre class="literal-block">
&gt;&gt;&gt; cdll.kernel32[1] # doctest: +WINDOWS
&lt;_FuncPtr object at 0x...&gt;
&gt;&gt;&gt; cdll.kernel32[0] # doctest: +WINDOWS
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
  File &quot;ctypes.py&quot;, line 310, in __getitem__
    func = _StdcallFuncPtr(name, self)
AttributeError: function ordinal 0 not found
&gt;&gt;&gt;
</pre>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id3" id="calling-functions" name="calling-functions">Calling functions</a></h1>
<p>You can call these functions like any other Python callable. This
example uses the <tt class="docutils literal"><span class="pre">time()</span></tt> function, which returns system time in
seconds since the UNIX epoch, and the <tt class="docutils literal"><span class="pre">GetModuleHandleA()</span></tt> function,
which returns a win32 module handle.</p>
<p>This example calls both functions with a NULL pointer (<tt class="docutils literal"><span class="pre">None</span></tt> should
be used as the NULL pointer):</p>
<pre class="literal-block">
&gt;&gt;&gt; print libc.time(None) # doctest: +SKIP
1150640792
&gt;&gt;&gt; print hex(windll.kernel32.GetModuleHandleA(None)) # doctest: +WINDOWS
0x1d000000
&gt;&gt;&gt;
</pre>
<p><tt class="docutils literal"><span class="pre">ctypes</span></tt> tries to protect you from calling functions with the wrong
number of arguments or the wrong calling convention.  Unfortunately
this only works on Windows.  It does this by examining the stack after
the function returns, so although an error is raised the function
<em>has</em> been called:</p>
<pre class="literal-block">
&gt;&gt;&gt; windll.kernel32.GetModuleHandleA() # doctest: +WINDOWS
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
ValueError: Procedure probably called with not enough arguments (4 bytes missing)
&gt;&gt;&gt; windll.kernel32.GetModuleHandleA(0, 0) # doctest: +WINDOWS
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
ValueError: Procedure probably called with too many arguments (4 bytes in excess)
&gt;&gt;&gt;
</pre>
<p>The same exception is raised when you call an <tt class="docutils literal"><span class="pre">stdcall</span></tt> function
with the <tt class="docutils literal"><span class="pre">cdecl</span></tt> calling convention, or vice versa:</p>
<pre class="literal-block">
&gt;&gt;&gt; cdll.kernel32.GetModuleHandleA(None) # doctest: +WINDOWS
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
ValueError: Procedure probably called with not enough arguments (4 bytes missing)
&gt;&gt;&gt;
&gt;&gt;&gt; windll.msvcrt.printf(&quot;spam&quot;) # doctest: +WINDOWS
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
ValueError: Procedure probably called with too many arguments (4 bytes in excess)
&gt;&gt;&gt;
</pre>
<p>To find out the correct calling convention you have to look into the C
header file or the documentation for the function you want to call.</p>
<p>On Windows, <tt class="docutils literal"><span class="pre">ctypes</span></tt> uses win32 structured exception handling to
prevent crashes from general protection faults when functions are
called with invalid argument values:</p>
<pre class="literal-block">
&gt;&gt;&gt; windll.kernel32.GetModuleHandleA(32) # doctest: +WINDOWS
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
WindowsError: exception: access violation reading 0x00000020
&gt;&gt;&gt;
</pre>
<p>There are, however, enough ways to crash Python with <tt class="docutils literal"><span class="pre">ctypes</span></tt>, so
you should be careful anyway.</p>
<p><tt class="docutils literal"><span class="pre">None</span></tt>, integers, longs, byte strings and unicode strings are the
only native Python objects that can directly be used as parameters in
these function calls.  <tt class="docutils literal"><span class="pre">None</span></tt> is passed as a C <tt class="docutils literal"><span class="pre">NULL</span></tt> pointer,
byte strings and unicode strings are passed as pointer to the memory
block that contains their data (<tt class="docutils literal"><span class="pre">char</span> <span class="pre">*</span></tt> or <tt class="docutils literal"><span class="pre">wchar_t</span> <span class="pre">*</span></tt>).  Python
integers and Python longs are passed as the platforms default C
<tt class="docutils literal"><span class="pre">int</span></tt> type, their value is masked to fit into the C type.</p>
<p>Before we move on calling functions with other parameter types, we
have to learn more about <tt class="docutils literal"><span class="pre">ctypes</span></tt> data types.</p>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id4" id="fundamental-data-types" name="fundamental-data-types">Fundamental data types</a></h1>
<p><tt class="docutils literal"><span class="pre">ctypes</span></tt> defines a number of primitive C compatible data types :</p>
<blockquote>
<table border="1" class="docutils">
<colgroup>
<col width="29%" />
<col width="44%" />
<col width="27%" />
</colgroup>
<thead valign="bottom">
<tr><th class="head">ctypes type</th>
<th class="head">C type</th>
<th class="head">Python type</th>
</tr>
</thead>
<tbody valign="top">
<tr><td><tt class="docutils literal"><span class="pre">c_char</span></tt></td>
<td><tt class="docutils literal"><span class="pre">char</span></tt></td>
<td>1-character
string</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_wchar</span></tt></td>
<td><tt class="docutils literal"><span class="pre">wchar_t</span></tt></td>
<td>1-character
unicode string</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_byte</span></tt></td>
<td><tt class="docutils literal"><span class="pre">char</span></tt></td>
<td>int/long</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_ubyte</span></tt></td>
<td><tt class="docutils literal"><span class="pre">unsigned</span> <span class="pre">char</span></tt></td>
<td>int/long</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_short</span></tt></td>
<td><tt class="docutils literal"><span class="pre">short</span></tt></td>
<td>int/long</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_ushort</span></tt></td>
<td><tt class="docutils literal"><span class="pre">unsigned</span> <span class="pre">short</span></tt></td>
<td>int/long</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_int</span></tt></td>
<td><tt class="docutils literal"><span class="pre">int</span></tt></td>
<td>int/long</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_uint</span></tt></td>
<td><tt class="docutils literal"><span class="pre">unsigned</span> <span class="pre">int</span></tt></td>
<td>int/long</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_long</span></tt></td>
<td><tt class="docutils literal"><span class="pre">long</span></tt></td>
<td>int/long</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_ulong</span></tt></td>
<td><tt class="docutils literal"><span class="pre">unsigned</span> <span class="pre">long</span></tt></td>
<td>int/long</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_longlong</span></tt></td>
<td><tt class="docutils literal"><span class="pre">__int64</span></tt> or
<tt class="docutils literal"><span class="pre">long</span> <span class="pre">long</span></tt></td>
<td>int/long</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_ulonglong</span></tt></td>
<td><tt class="docutils literal"><span class="pre">unsigned</span> <span class="pre">__int64</span></tt> or
<tt class="docutils literal"><span class="pre">unsigned</span> <span class="pre">long</span> <span class="pre">long</span></tt></td>
<td>int/long</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_float</span></tt></td>
<td><tt class="docutils literal"><span class="pre">float</span></tt></td>
<td>float</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_double</span></tt></td>
<td><tt class="docutils literal"><span class="pre">double</span></tt></td>
<td>float</td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_char_p</span></tt></td>
<td><tt class="docutils literal"><span class="pre">char</span> <span class="pre">*</span></tt>
(NUL terminated)</td>
<td>string or
<tt class="docutils literal"><span class="pre">None</span></tt></td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_wchar_p</span></tt></td>
<td><tt class="docutils literal"><span class="pre">wchar_t</span> <span class="pre">*</span></tt>
(NUL terminated)</td>
<td>unicode or
<tt class="docutils literal"><span class="pre">None</span></tt></td>
</tr>
<tr><td><tt class="docutils literal"><span class="pre">c_void_p</span></tt></td>
<td><tt class="docutils literal"><span class="pre">void</span> <span class="pre">*</span></tt></td>
<td>int/long
or <tt class="docutils literal"><span class="pre">None</span></tt></td>
</tr>
</tbody>
</table>
</blockquote>
<p>All these types can be created by calling them with an optional
initializer of the correct type and value:</p>
<pre class="literal-block">
&gt;&gt;&gt; c_int()
c_long(0)
&gt;&gt;&gt; c_char_p(&quot;Hello, World&quot;)
c_char_p('Hello, World')
&gt;&gt;&gt; c_ushort(-3)
c_ushort(65533)
&gt;&gt;&gt;
</pre>
<p>Since these types are mutable, their value can also be changed
afterwards:</p>
<pre class="literal-block">
&gt;&gt;&gt; i = c_int(42)
&gt;&gt;&gt; print i
c_long(42)
&gt;&gt;&gt; print i.value
42
&gt;&gt;&gt; i.value = -99
&gt;&gt;&gt; print i.value
-99
&gt;&gt;&gt;
</pre>
<p>Assigning a new value to instances of the pointer types <tt class="docutils literal"><span class="pre">c_char_p</span></tt>,
<tt class="docutils literal"><span class="pre">c_wchar_p</span></tt>, and <tt class="docutils literal"><span class="pre">c_void_p</span></tt> changes the <em>memory location</em> they
point to, <em>not the contents</em> of the memory block (of course not,
because Python strings are immutable):</p>
<pre class="literal-block">
&gt;&gt;&gt; s = &quot;Hello, World&quot;
&gt;&gt;&gt; c_s = c_char_p(s)
&gt;&gt;&gt; print c_s
c_char_p('Hello, World')
&gt;&gt;&gt; c_s.value = &quot;Hi, there&quot;
&gt;&gt;&gt; print c_s
c_char_p('Hi, there')
&gt;&gt;&gt; print s                # first string is unchanged
Hello, World
&gt;&gt;&gt;
</pre>
<p>You should be careful, however, not to pass them to functions
expecting pointers to mutable memory. If you need mutable memory
blocks, ctypes has a <tt class="docutils literal"><span class="pre">create_string_buffer</span></tt> function which creates
these in various ways.  The current memory block contents can be
accessed (or changed) with the <tt class="docutils literal"><span class="pre">raw</span></tt> property, if you want to access
it as NUL terminated string, use the <tt class="docutils literal"><span class="pre">string</span></tt> property:</p>
<pre class="literal-block">
&gt;&gt;&gt; from ctypes import *
&gt;&gt;&gt; p = create_string_buffer(3)      # create a 3 byte buffer, initialized to NUL bytes
&gt;&gt;&gt; print sizeof(p), repr(p.raw)
3 '\x00\x00\x00'
&gt;&gt;&gt; p = create_string_buffer(&quot;Hello&quot;)      # create a buffer containing a NUL terminated string
&gt;&gt;&gt; print sizeof(p), repr(p.raw)
6 'Hello\x00'
&gt;&gt;&gt; print repr(p.value)
'Hello'
&gt;&gt;&gt; p = create_string_buffer(&quot;Hello&quot;, 10)  # create a 10 byte buffer
&gt;&gt;&gt; print sizeof(p), repr(p.raw)
10 'Hello\x00\x00\x00\x00\x00'
&gt;&gt;&gt; p.value = &quot;Hi&quot;
&gt;&gt;&gt; print sizeof(p), repr(p.raw)
10 'Hi\x00lo\x00\x00\x00\x00\x00'
&gt;&gt;&gt;
</pre>
<p>The <tt class="docutils literal"><span class="pre">create_string_buffer</span></tt> function replaces the <tt class="docutils literal"><span class="pre">c_buffer</span></tt>
function (which is still available as an alias), as well as the
<tt class="docutils literal"><span class="pre">c_string</span></tt> function from earlier ctypes releases.  To create a
mutable memory block containing unicode characters of the C type
<tt class="docutils literal"><span class="pre">wchar_t</span></tt> use the <tt class="docutils literal"><span class="pre">create_unicode_buffer</span></tt> function.</p>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id5" id="calling-functions-continued" name="calling-functions-continued">Calling functions, continued</a></h1>
<p>Note that printf prints to the real standard output channel, <em>not</em> to
<tt class="docutils literal"><span class="pre">sys.stdout</span></tt>, so these examples will only work at the console
prompt, not from within <em>IDLE</em> or <em>PythonWin</em>:</p>
<pre class="literal-block">
&gt;&gt;&gt; printf = libc.printf
&gt;&gt;&gt; printf(&quot;Hello, %s\n&quot;, &quot;World!&quot;)
Hello, World!
14
&gt;&gt;&gt; printf(&quot;Hello, %S&quot;, u&quot;World!&quot;)
Hello, World!
13
&gt;&gt;&gt; printf(&quot;%d bottles of beer\n&quot;, 42)
42 bottles of beer
19
&gt;&gt;&gt; printf(&quot;%f bottles of beer\n&quot;, 42.5)
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
ArgumentError: argument 2: exceptions.TypeError: Don't know how to convert parameter 2
&gt;&gt;&gt;
</pre>
<p>As has been mentioned before, all Python types except integers,
strings, and unicode strings have to be wrapped in their corresponding
<tt class="docutils literal"><span class="pre">ctypes</span></tt> type, so that they can be converted to the required C data
type:</p>
<pre class="literal-block">
&gt;&gt;&gt; printf(&quot;An int %d, a double %f\n&quot;, 1234, c_double(3.14))
Integer 1234, double 3.1400001049
31
&gt;&gt;&gt;
</pre>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id6" id="calling-functions-with-your-own-custom-data-types" name="calling-functions-with-your-own-custom-data-types">Calling functions with your own custom data types</a></h1>
<p>You can also customize <tt class="docutils literal"><span class="pre">ctypes</span></tt> argument conversion to allow
instances of your own classes be used as function arguments.
<tt class="docutils literal"><span class="pre">ctypes</span></tt> looks for an <tt class="docutils literal"><span class="pre">_as_parameter_</span></tt> attribute and uses this as
the function argument. Of course, it must be one of integer, string,
or unicode:</p>
<pre class="literal-block">
&gt;&gt;&gt; class Bottles(object):
...    def __init__(self, number):
...        self._as_parameter_ = number
...
&gt;&gt;&gt; bottles = Bottles(42)
&gt;&gt;&gt; printf(&quot;%d bottles of beer\n&quot;, bottles)
42 bottles of beer
19
&gt;&gt;&gt;
</pre>
<p>If you don't want to store the instance's data in the
<tt class="docutils literal"><span class="pre">_as_parameter_</span></tt> instance variable, you could define a <tt class="docutils literal"><span class="pre">property</span></tt>
which makes the data avaiblable.</p>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id7" id="specifying-the-required-argument-types-function-prototypes" name="specifying-the-required-argument-types-function-prototypes">Specifying the required argument types (function prototypes)</a></h1>
<p>It is possible to specify the required argument types of functions
exported from DLLs by setting the <tt class="docutils literal"><span class="pre">argtypes</span></tt> attribute.</p>
<p><tt class="docutils literal"><span class="pre">argtypes</span></tt> must be a sequence of C data types (the <tt class="docutils literal"><span class="pre">printf</span></tt>
function is probably not a good example here, because it takes a
variable number and different types of parameters depending on the
format string, on the other hand this is quite handy to experiment
with this feature):</p>
<pre class="literal-block">
&gt;&gt;&gt; printf.argtypes = [c_char_p, c_char_p, c_int, c_double]
&gt;&gt;&gt; printf(&quot;String '%s', Int %d, Double %f\n&quot;, &quot;Hi&quot;, 10, 2.2)
String 'Hi', Int 10, Double 2.200000
37
&gt;&gt;&gt;
</pre>
<p>Specifying a format protects against incompatible argument types (just
as a prototype for a C function), and tries to convert the arguments
to valid types:</p>
<pre class="literal-block">
&gt;&gt;&gt; printf(&quot;%d %d %d&quot;, 1, 2, 3)
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
ArgumentError: argument 2: exceptions.TypeError: wrong type
&gt;&gt;&gt; printf(&quot;%s %d %f&quot;, &quot;X&quot;, 2, 3)
X 2 3.00000012
12
&gt;&gt;&gt;
</pre>
<p>If you have defined your own classes which you pass to function calls,
you have to implement a <tt class="docutils literal"><span class="pre">from_param</span></tt> class method for them to be
able to use them in the <tt class="docutils literal"><span class="pre">argtypes</span></tt> sequence. The <tt class="docutils literal"><span class="pre">from_param</span></tt>
class method receives the Python object passed to the function call,
it should do a typecheck or whatever is needed to make sure this
object is acceptable, and then return the object itself, it's
<tt class="docutils literal"><span class="pre">_as_parameter_</span></tt> attribute, or whatever you want to pass as the C
function argument in this case. Again, the result should be an
integer, string, unicode, a <tt class="docutils literal"><span class="pre">ctypes</span></tt> instance, or something having
the <tt class="docutils literal"><span class="pre">_as_parameter_</span></tt> attribute.</p>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id8" id="return-types" name="return-types">Return types</a></h1>
<p>By default functions are assumed to return the C <tt class="docutils literal"><span class="pre">int</span></tt> type.  Other
return types can be specified by setting the <tt class="docutils literal"><span class="pre">restype</span></tt> attribute of
the function object.</p>
<p>Here is a more advanced example, it uses the <tt class="docutils literal"><span class="pre">strchr</span></tt> function, which
expects a string pointer and a char, and returns a pointer to a
string:</p>
<pre class="literal-block">
&gt;&gt;&gt; strchr = libc.strchr
&gt;&gt;&gt; strchr(&quot;abcdef&quot;, ord(&quot;d&quot;)) # doctest: +SKIP
8059983
&gt;&gt;&gt; strchr.restype = c_char_p # c_char_p is a pointer to a string
&gt;&gt;&gt; strchr(&quot;abcdef&quot;, ord(&quot;d&quot;))
'def'
&gt;&gt;&gt; print strchr(&quot;abcdef&quot;, ord(&quot;x&quot;))
None
&gt;&gt;&gt;
</pre>
<p>If you want to avoid the <tt class="docutils literal"><span class="pre">ord(&quot;x&quot;)</span></tt> calls above, you can set the
<tt class="docutils literal"><span class="pre">argtypes</span></tt> attribute, and the second argument will be converted from
a single character Python string into a C char:</p>
<pre class="literal-block">
&gt;&gt;&gt; strchr.restype = c_char_p
&gt;&gt;&gt; strchr.argtypes = [c_char_p, c_char]
&gt;&gt;&gt; strchr(&quot;abcdef&quot;, &quot;d&quot;)
'def'
&gt;&gt;&gt; strchr(&quot;abcdef&quot;, &quot;def&quot;)
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
ArgumentError: argument 2: exceptions.TypeError: one character string expected
&gt;&gt;&gt; print strchr(&quot;abcdef&quot;, &quot;x&quot;)
None
&gt;&gt;&gt; strchr(&quot;abcdef&quot;, &quot;d&quot;)
'def'
&gt;&gt;&gt;
</pre>
<p>You can also use a callable Python object (a function or a class for
example) as the <tt class="docutils literal"><span class="pre">restype</span></tt> attribute, if the foreign function returns
an integer.  The callable will be called with the <tt class="docutils literal"><span class="pre">integer</span></tt> the C
function returns, and the result of this call will be used as the
result of your function call. This is useful to check for error return
values and automatically raise an exception:</p>
<pre class="literal-block">
&gt;&gt;&gt; GetModuleHandle = windll.kernel32.GetModuleHandleA # doctest: +WINDOWS
&gt;&gt;&gt; def ValidHandle(value):
...    if value == 0:
...        raise WinError()
...    return value
...
&gt;&gt;&gt;
&gt;&gt;&gt; GetModuleHandle.restype = ValidHandle # doctest: +WINDOWS
&gt;&gt;&gt; GetModuleHandle(None) # doctest: +WINDOWS
486539264
&gt;&gt;&gt; GetModuleHandle(&quot;something silly&quot;) # doctest: +WINDOWS
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
  File &quot;&lt;stdin&gt;&quot;, line 3, in ValidHandle
WindowsError: [Errno 126] The specified module could not be found.
&gt;&gt;&gt;
</pre>
<p><tt class="docutils literal"><span class="pre">WinError</span></tt> is a function which will call Windows <tt class="docutils literal"><span class="pre">FormatMessage()</span></tt>
api to get the string representation of an error code, and <em>returns</em>
an exception.  <tt class="docutils literal"><span class="pre">WinError</span></tt> takes an optional error code parameter, if
no one is used, it calls <tt class="docutils literal"><span class="pre">GetLastError()</span></tt> to retrieve it.</p>
<p>Please note that a much more powerful error checking mechanism is
available through the <tt class="docutils literal"><span class="pre">errcheck</span></tt> attribute; see the reference manual
for details.</p>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id9" id="passing-pointers-or-passing-parameters-by-reference" name="passing-pointers-or-passing-parameters-by-reference">Passing pointers (or: passing parameters by reference)</a></h1>
<p>Sometimes a C api function expects a <em>pointer</em> to a data type as
parameter, probably to write into the corresponding location, or if
the data is too large to be passed by value. This is also known as
<em>passing parameters by reference</em>.</p>
<p><tt class="docutils literal"><span class="pre">ctypes</span></tt> exports the <tt class="docutils literal"><span class="pre">byref</span></tt> function which is used to pass
parameters by reference.  The same effect can be achieved with the
<tt class="docutils literal"><span class="pre">pointer</span></tt> function, although <tt class="docutils literal"><span class="pre">pointer</span></tt> does a lot more work since
it constructs a real pointer object, so it is faster to use <tt class="docutils literal"><span class="pre">byref</span></tt>
if you don't need the pointer object in Python itself:</p>
<pre class="literal-block">
&gt;&gt;&gt; i = c_int()
&gt;&gt;&gt; f = c_float()
&gt;&gt;&gt; s = create_string_buffer('\000' * 32)
&gt;&gt;&gt; print i.value, f.value, repr(s.value)
0 0.0 ''
&gt;&gt;&gt; libc.sscanf(&quot;1 3.14 Hello&quot;, &quot;%d %f %s&quot;,
...            byref(i), byref(f), s)
3
&gt;&gt;&gt; print i.value, f.value, repr(s.value)
1 3.1400001049 'Hello'
&gt;&gt;&gt;
</pre>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id10" id="structures-and-unions" name="structures-and-unions">Structures and unions</a></h1>
<p>Structures and unions must derive from the <tt class="docutils literal"><span class="pre">Structure</span></tt> and <tt class="docutils literal"><span class="pre">Union</span></tt>
base classes which are defined in the <tt class="docutils literal"><span class="pre">ctypes</span></tt> module. Each subclass
must define a <tt class="docutils literal"><span class="pre">_fields_</span></tt> attribute.  <tt class="docutils literal"><span class="pre">_fields_</span></tt> must be a list of
<em>2-tuples</em>, containing a <em>field name</em> and a <em>field type</em>.</p>
<p>The field type must be a <tt class="docutils literal"><span class="pre">ctypes</span></tt> type like <tt class="docutils literal"><span class="pre">c_int</span></tt>, or any other
derived <tt class="docutils literal"><span class="pre">ctypes</span></tt> type: structure, union, array, pointer.</p>
<p>Here is a simple example of a POINT structure, which contains two
integers named <tt class="docutils literal"><span class="pre">x</span></tt> and <tt class="docutils literal"><span class="pre">y</span></tt>, and also shows how to initialize a
structure in the constructor:</p>
<pre class="literal-block">
&gt;&gt;&gt; from ctypes import *
&gt;&gt;&gt; class POINT(Structure):
...    _fields_ = [(&quot;x&quot;, c_int),
...                (&quot;y&quot;, c_int)]
...
&gt;&gt;&gt; point = POINT(10, 20)
&gt;&gt;&gt; print point.x, point.y
10 20
&gt;&gt;&gt; point = POINT(y=5)
&gt;&gt;&gt; print point.x, point.y
0 5
&gt;&gt;&gt; POINT(1, 2, 3)
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
ValueError: too many initializers
&gt;&gt;&gt;
</pre>
<p>You can, however, build much more complicated structures. Structures
can itself contain other structures by using a structure as a field
type.</p>
<p>Here is a RECT structure which contains two POINTs named <tt class="docutils literal"><span class="pre">upperleft</span></tt>
and <tt class="docutils literal"><span class="pre">lowerright</span></tt></p>
<pre class="literal-block">
&gt;&gt;&gt; class RECT(Structure):
...    _fields_ = [(&quot;upperleft&quot;, POINT),
...                (&quot;lowerright&quot;, POINT)]
...
&gt;&gt;&gt; rc = RECT(point)
&gt;&gt;&gt; print rc.upperleft.x, rc.upperleft.y
0 5
&gt;&gt;&gt; print rc.lowerright.x, rc.lowerright.y
0 0
&gt;&gt;&gt;
</pre>
<p>Nested structures can also be initialized in the constructor in
several ways:</p>
<pre class="literal-block">
&gt;&gt;&gt; r = RECT(POINT(1, 2), POINT(3, 4))
&gt;&gt;&gt; r = RECT((1, 2), (3, 4))
</pre>
<p>Fields descriptors can be retrieved from the <em>class</em>, they are useful
for debugging because they can provide useful information:</p>
<pre class="literal-block">
&gt;&gt;&gt; print POINT.x
&lt;Field type=c_long, ofs=0, size=4&gt;
&gt;&gt;&gt; print POINT.y
&lt;Field type=c_long, ofs=4, size=4&gt;
&gt;&gt;&gt;
</pre>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id11" id="structure-union-alignment-and-byte-order" name="structure-union-alignment-and-byte-order">Structure/union alignment and byte order</a></h1>
<p>By default, Structure and Union fields are aligned in the same way the
C compiler does it. It is possible to override this behaviour be
specifying a <tt class="docutils literal"><span class="pre">_pack_</span></tt> class attribute in the subclass
definition. This must be set to a positive integer and specifies the
maximum alignment for the fields. This is what <tt class="docutils literal"><span class="pre">#pragma</span> <span class="pre">pack(n)</span></tt>
also does in MSVC.</p>
<p><tt class="docutils literal"><span class="pre">ctypes</span></tt> uses the native byte order for Structures and Unions.  To
build structures with non-native byte order, you can use one of the
BigEndianStructure, LittleEndianStructure, BigEndianUnion, and
LittleEndianUnion base classes.  These classes cannot contain pointer
fields.</p>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id12" id="bit-fields-in-structures-and-unions" name="bit-fields-in-structures-and-unions">Bit fields in structures and unions</a></h1>
<p>It is possible to create structures and unions containing bit fields.
Bit fields are only possible for integer fields, the bit width is
specified as the third item in the <tt class="docutils literal"><span class="pre">_fields_</span></tt> tuples:</p>
<pre class="literal-block">
&gt;&gt;&gt; class Int(Structure):
...    _fields_ = [(&quot;first_16&quot;, c_int, 16),
...                (&quot;second_16&quot;, c_int, 16)]
...
&gt;&gt;&gt; print Int.first_16
&lt;Field type=c_long, ofs=0:0, bits=16&gt;
&gt;&gt;&gt; print Int.second_16
&lt;Field type=c_long, ofs=0:16, bits=16&gt;
&gt;&gt;&gt;
</pre>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id13" id="arrays" name="arrays">Arrays</a></h1>
<p>Arrays are sequences, containing a fixed number of instances of the
same type.</p>
<p>The recommended way to create array types is by multiplying a data
type with a positive integer:</p>
<pre class="literal-block">
TenPointsArrayType = POINT * 10
</pre>
<p>Here is an example of an somewhat artifical data type, a structure
containing 4 POINTs among other stuff:</p>
<pre class="literal-block">
&gt;&gt;&gt; from ctypes import *
&gt;&gt;&gt; class POINT(Structure):
...    _fields_ = (&quot;x&quot;, c_int), (&quot;y&quot;, c_int)
...
&gt;&gt;&gt; class MyStruct(Structure):
...    _fields_ = [(&quot;a&quot;, c_int),
...                (&quot;b&quot;, c_float),
...                (&quot;point_array&quot;, POINT * 4)]
&gt;&gt;&gt;
&gt;&gt;&gt; print len(MyStruct().point_array)
4
&gt;&gt;&gt;
</pre>
<p>Instances are created in the usual way, by calling the class:</p>
<pre class="literal-block">
arr = TenPointsArrayType()
for pt in arr:
    print pt.x, pt.y
</pre>
<p>The above code print a series of <tt class="docutils literal"><span class="pre">0</span> <span class="pre">0</span></tt> lines, because the array
contents is initialized to zeros.</p>
<p>Initializers of the correct type can also be specified:</p>
<pre class="literal-block">
&gt;&gt;&gt; from ctypes import *
&gt;&gt;&gt; TenIntegers = c_int * 10
&gt;&gt;&gt; ii = TenIntegers(1, 2, 3, 4, 5, 6, 7, 8, 9, 10)
&gt;&gt;&gt; print ii
&lt;c_long_Array_10 object at 0x...&gt;
&gt;&gt;&gt; for i in ii: print i,
...
1 2 3 4 5 6 7 8 9 10
&gt;&gt;&gt;
</pre>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id14" id="pointers" name="pointers">Pointers</a></h1>
<p>Pointer instances are created by calling the <tt class="docutils literal"><span class="pre">pointer</span></tt> function on a
<tt class="docutils literal"><span class="pre">ctypes</span></tt> type:</p>
<pre class="literal-block">
&gt;&gt;&gt; from ctypes import *
&gt;&gt;&gt; i = c_int(42)
&gt;&gt;&gt; pi = pointer(i)
&gt;&gt;&gt;
</pre>
<p>Pointer instances have a <tt class="docutils literal"><span class="pre">contents</span></tt> attribute which returns the
object to which the pointer points, the <tt class="docutils literal"><span class="pre">i</span></tt> object above:</p>
<pre class="literal-block">
&gt;&gt;&gt; pi.contents
c_long(42)
&gt;&gt;&gt;
</pre>
<p>Note that <tt class="docutils literal"><span class="pre">ctypes</span></tt> does not have OOR (original object return), it
constructs a new, equivalent object each time you retrieve an
attribute:</p>
<pre class="literal-block">
&gt;&gt;&gt; pi.contents is i
False
&gt;&gt;&gt; pi.contents is pi.contents
False
&gt;&gt;&gt;
</pre>
<p>Assigning another <tt class="docutils literal"><span class="pre">c_int</span></tt> instance to the pointer's contents
attribute would cause the pointer to point to the memory location
where this is stored:</p>
<pre class="literal-block">
&gt;&gt;&gt; i = c_int(99)
&gt;&gt;&gt; pi.contents = i
&gt;&gt;&gt; pi.contents
c_long(99)
&gt;&gt;&gt;
</pre>
<p>Pointer instances can also be indexed with integers:</p>
<pre class="literal-block">
&gt;&gt;&gt; pi[0]
99
&gt;&gt;&gt;
</pre>
<p>Assigning to an integer index changes the pointed to value:</p>
<pre class="literal-block">
&gt;&gt;&gt; print i
c_long(99)
&gt;&gt;&gt; pi[0] = 22
&gt;&gt;&gt; print i
c_long(22)
&gt;&gt;&gt;
</pre>
<p>It is also possible to use indexes different from 0, but you must know
what you're doing, just as in C: You can access or change arbitrary
memory locations. Generally you only use this feature if you receive a
pointer from a C function, and you <em>know</em> that the pointer actually
points to an array instead of a single item.</p>
<p>Behind the scenes, the <tt class="docutils literal"><span class="pre">pointer</span></tt> function does more than simply
create pointer instances, it has to create pointer <em>types</em> first.
This is done with the <tt class="docutils literal"><span class="pre">POINTER</span></tt> function, which accepts any
<tt class="docutils literal"><span class="pre">ctypes</span></tt> type, and returns a new type:</p>
<pre class="literal-block">
&gt;&gt;&gt; PI = POINTER(c_int)
&gt;&gt;&gt; PI
&lt;class 'ctypes.LP_c_long'&gt;
&gt;&gt;&gt; PI(42)
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
TypeError: expected c_long instead of int
&gt;&gt;&gt; PI(c_int(42))
&lt;ctypes.LP_c_long object at 0x...&gt;
&gt;&gt;&gt;
</pre>
<p>Calling the pointer type without an argument creates a <tt class="docutils literal"><span class="pre">NULL</span></tt>
pointer.  <tt class="docutils literal"><span class="pre">NULL</span></tt> pointers have a <tt class="docutils literal"><span class="pre">False</span></tt> boolean value:</p>
<pre class="literal-block">
&gt;&gt;&gt; null_ptr = POINTER(c_int)()
&gt;&gt;&gt; print bool(null_ptr)
False
&gt;&gt;&gt;
</pre>
<p><tt class="docutils literal"><span class="pre">ctypes</span></tt> checks for <tt class="docutils literal"><span class="pre">NULL</span></tt> when dereferencing pointers (but
dereferencing non-<tt class="docutils literal"><span class="pre">NULL</span></tt> pointers would crash Python):</p>
<pre class="literal-block">
&gt;&gt;&gt; null_ptr[0]
Traceback (most recent call last):
    ....
ValueError: NULL pointer access
&gt;&gt;&gt;
&gt;&gt;&gt; null_ptr[0] = 1234
Traceback (most recent call last):
    ....
ValueError: NULL pointer access
&gt;&gt;&gt;
</pre>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id15" id="type-conversions" name="type-conversions">Type conversions</a></h1>
<p>Usually, ctypes does strict type checking.  This means, if you have
<tt class="docutils literal"><span class="pre">POINTER(c_int)</span></tt> in the <tt class="docutils literal"><span class="pre">argtypes</span></tt> list of a function or as the
type of a member field in a structure definition, only instances of
exactly the same type are accepted.  There are some exceptions to this
rule, where ctypes accepts other objects.  For example, you can pass
compatible array instances instead of pointer types.  So, for
<tt class="docutils literal"><span class="pre">POINTER(c_int)</span></tt>, ctypes accepts an array of c_int:</p>
<pre class="literal-block">
&gt;&gt;&gt; class Bar(Structure):
...    _fields_ = [(&quot;count&quot;, c_int), (&quot;values&quot;, POINTER(c_int))]
...
&gt;&gt;&gt; bar = Bar()
&gt;&gt;&gt; bar.values = (c_int * 3)(1, 2, 3)
&gt;&gt;&gt; bar.count = 3
&gt;&gt;&gt; for i in range(bar.count):
...    print bar.values[i]
...
1
2
3
&gt;&gt;&gt;
</pre>
<p>To set a POINTER type field to <tt class="docutils literal"><span class="pre">NULL</span></tt>, you can assign <tt class="docutils literal"><span class="pre">None</span></tt>:</p>
<pre class="literal-block">
&gt;&gt;&gt; bar.values = None
&gt;&gt;&gt;
</pre>
<p>XXX list other conversions...</p>
<p>Sometimes you have instances of incompatible types.  In <tt class="docutils literal"><span class="pre">C</span></tt>, you can
cast one type into another type.  <tt class="docutils literal"><span class="pre">ctypes</span></tt> provides a <tt class="docutils literal"><span class="pre">cast</span></tt>
function which can be used in the same way.  The <tt class="docutils literal"><span class="pre">Bar</span></tt> structure
defined above accepts <tt class="docutils literal"><span class="pre">POINTER(c_int)</span></tt> pointers or <tt class="docutils literal"><span class="pre">c_int</span></tt> arrays
for its <tt class="docutils literal"><span class="pre">values</span></tt> field, but not instances of other types:</p>
<pre class="literal-block">
&gt;&gt;&gt; bar.values = (c_byte * 4)()
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
TypeError: incompatible types, c_byte_Array_4 instance instead of LP_c_long instance
&gt;&gt;&gt;
</pre>
<p>For these cases, the <tt class="docutils literal"><span class="pre">cast</span></tt> function is handy.</p>
<p>The <tt class="docutils literal"><span class="pre">cast</span></tt> function can be used to cast a ctypes instance into a
pointer to a different ctypes data type.  <tt class="docutils literal"><span class="pre">cast</span></tt> takes two
parameters, a ctypes object that is or can be converted to a pointer
of some kind, and a ctypes pointer type.  It returns an instance of
the second argument, which references the same memory block as the
first argument:</p>
<pre class="literal-block">
&gt;&gt;&gt; a = (c_byte * 4)()
&gt;&gt;&gt; cast(a, POINTER(c_int))
&lt;ctypes.LP_c_long object at ...&gt;
&gt;&gt;&gt;
</pre>
<p>So, <tt class="docutils literal"><span class="pre">cast</span></tt> can be used to assign to the <tt class="docutils literal"><span class="pre">values</span></tt> field of <tt class="docutils literal"><span class="pre">Bar</span></tt>
the structure:</p>
<pre class="literal-block">
&gt;&gt;&gt; bar = Bar()
&gt;&gt;&gt; bar.values = cast((c_byte * 4)(), POINTER(c_int))
&gt;&gt;&gt; print bar.values[0]
0
&gt;&gt;&gt;
</pre>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id16" id="incomplete-types" name="incomplete-types">Incomplete Types</a></h1>
<p><em>Incomplete Types</em> are structures, unions or arrays whose members are
not yet specified. In C, they are specified by forward declarations, which
are defined later:</p>
<pre class="literal-block">
struct cell; /* forward declaration */
struct {
    char *name;
    struct cell *next;
} cell;
</pre>
<p>The straightforward translation into ctypes code would be this, but it
does not work:</p>
<pre class="literal-block">
&gt;&gt;&gt; class cell(Structure):
...    _fields_ = [(&quot;name&quot;, c_char_p),
...                (&quot;next&quot;, POINTER(cell))]
...
Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in ?
  File &quot;&lt;stdin&gt;&quot;, line 2, in cell
NameError: name 'cell' is not defined
&gt;&gt;&gt;
</pre>
<p>because the new <tt class="docutils literal"><span class="pre">class</span> <span class="pre">cell</span></tt> is not available in the class statement
itself.  In <tt class="docutils literal"><span class="pre">ctypes</span></tt>, we can define the <tt class="docutils literal"><span class="pre">cell</span></tt> class and set the
<tt class="docutils literal"><span class="pre">_fields_</span></tt> attribute later, after the class statement:</p>
<pre class="literal-block">
&gt;&gt;&gt; from ctypes import *
&gt;&gt;&gt; class cell(Structure):
...    pass
...
&gt;&gt;&gt; cell._fields_ = [(&quot;name&quot;, c_char_p),
...                  (&quot;next&quot;, POINTER(cell))]
&gt;&gt;&gt;
</pre>
<p>Lets try it. We create two instances of <tt class="docutils literal"><span class="pre">cell</span></tt>, and let them point
to each other, and finally follow the pointer chain a few times:</p>
<pre class="literal-block">
&gt;&gt;&gt; c1 = cell()
&gt;&gt;&gt; c1.name = &quot;foo&quot;
&gt;&gt;&gt; c2 = cell()
&gt;&gt;&gt; c2.name = &quot;bar&quot;
&gt;&gt;&gt; c1.next = pointer(c2)
&gt;&gt;&gt; c2.next = pointer(c1)
&gt;&gt;&gt; p = c1
&gt;&gt;&gt; for i in range(8):
...    print p.name,
...    p = p.next[0]
...
foo bar foo bar foo bar foo bar
&gt;&gt;&gt;
</pre>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id17" id="callback-functions" name="callback-functions">Callback functions</a></h1>
<p><tt class="docutils literal"><span class="pre">ctypes</span></tt> allows to create C callable function pointers from Python
callables. These are sometimes called <em>callback functions</em>.</p>
<p>First, you must create a class for the callback function, the class
knows the calling convention, the return type, and the number and
types of arguments this function will receive.</p>
<p>The CFUNCTYPE factory function creates types for callback functions
using the normal cdecl calling convention, and, on Windows, the
WINFUNCTYPE factory function creates types for callback functions
using the stdcall calling convention.</p>
<p>Both of these factory functions are called with the result type as
first argument, and the callback functions expected argument types as
the remaining arguments.</p>
<p>I will present an example here which uses the standard C library's
<tt class="docutils literal"><span class="pre">qsort</span></tt> function, this is used to sort items with the help of a
callback function. <tt class="docutils literal"><span class="pre">qsort</span></tt> will be used to sort an array of
integers:</p>
<pre class="literal-block">
&gt;&gt;&gt; IntArray5 = c_int * 5
&gt;&gt;&gt; ia = IntArray5(5, 1, 7, 33, 99)
&gt;&gt;&gt; qsort = libc.qsort
&gt;&gt;&gt; qsort.restype = None
&gt;&gt;&gt;
</pre>
<p><tt class="docutils literal"><span class="pre">qsort</span></tt> must be called with a pointer to the data to sort, the
number of items in the data array, the size of one item, and a pointer
to the comparison function, the callback. The callback will then be
called with two pointers to items, and it must return a negative
integer if the first item is smaller than the second, a zero if they
are equal, and a positive integer else.</p>
<p>So our callback function receives pointers to integers, and must
return an integer. First we create the <tt class="docutils literal"><span class="pre">type</span></tt> for the callback
function:</p>
<pre class="literal-block">
&gt;&gt;&gt; CMPFUNC = CFUNCTYPE(c_int, POINTER(c_int), POINTER(c_int))
&gt;&gt;&gt;
</pre>
<p>For the first implementation of the callback function, we simply print
the arguments we get, and return 0 (incremental development ;-):</p>
<pre class="literal-block">
&gt;&gt;&gt; def py_cmp_func(a, b):
...    print &quot;py_cmp_func&quot;, a, b
...    return 0
...
&gt;&gt;&gt;
</pre>
<p>Create the C callable callback:</p>
<pre class="literal-block">
&gt;&gt;&gt; cmp_func = CMPFUNC(py_cmp_func)
&gt;&gt;&gt;
</pre>
<p>And we're ready to go:</p>
<pre class="literal-block">
&gt;&gt;&gt; qsort(ia, len(ia), sizeof(c_int), cmp_func) # doctest: +WINDOWS
py_cmp_func &lt;ctypes.LP_c_long object at 0x00...&gt; &lt;ctypes.LP_c_long object at 0x00...&gt;
py_cmp_func &lt;ctypes.LP_c_long object at 0x00...&gt; &lt;ctypes.LP_c_long object at 0x00...&gt;
py_cmp_func &lt;ctypes.LP_c_long object at 0x00...&gt; &lt;ctypes.LP_c_long object at 0x00...&gt;
py_cmp_func &lt;ctypes.LP_c_long object at 0x00...&gt; &lt;ctypes.LP_c_long object at 0x00...&gt;
py_cmp_func &lt;ctypes.LP_c_long object at 0x00...&gt; &lt;ctypes.LP_c_long object at 0x00...&gt;
py_cmp_func &lt;ctypes.LP_c_long object at 0x00...&gt; &lt;ctypes.LP_c_long object at 0x00...&gt;
py_cmp_func &lt;ctypes.LP_c_long object at 0x00...&gt; &lt;ctypes.LP_c_long object at 0x00...&gt;
py_cmp_func &lt;ctypes.LP_c_long object at 0x00...&gt; &lt;ctypes.LP_c_long object at 0x00...&gt;
py_cmp_func &lt;ctypes.LP_c_long object at 0x00...&gt; &lt;ctypes.LP_c_long object at 0x00...&gt;
py_cmp_func &lt;ctypes.LP_c_long object at 0x00...&gt; &lt;ctypes.LP_c_long object at 0x00...&gt;
&gt;&gt;&gt;
</pre>
<p>We know how to access the contents of a pointer, so lets redefine our callback:</p>
<pre class="literal-block">
&gt;&gt;&gt; def py_cmp_func(a, b):
...    print &quot;py_cmp_func&quot;, a[0], b[0]
...    return 0
...
&gt;&gt;&gt; cmp_func = CMPFUNC(py_cmp_func)
&gt;&gt;&gt;
</pre>
<p>Here is what we get on Windows:</p>
<pre class="literal-block">
&gt;&gt;&gt; qsort(ia, len(ia), sizeof(c_int), cmp_func) # doctest: +WINDOWS
py_cmp_func 7 1
py_cmp_func 33 1
py_cmp_func 99 1
py_cmp_func 5 1
py_cmp_func 7 5
py_cmp_func 33 5
py_cmp_func 99 5
py_cmp_func 7 99
py_cmp_func 33 99
py_cmp_func 7 33
&gt;&gt;&gt;
</pre>
<p>It is funny to see that on linux the sort function seems to work much
more efficient, it is doing less comparisons:</p>
<pre class="literal-block">
&gt;&gt;&gt; qsort(ia, len(ia), sizeof(c_int), cmp_func) # doctest: +LINUX
py_cmp_func 5 1
py_cmp_func 33 99
py_cmp_func 7 33
py_cmp_func 5 7
py_cmp_func 1 7
&gt;&gt;&gt;
</pre>
<p>Ah, we're nearly done! The last step is to actually compare the two
items and return a useful result:</p>
<pre class="literal-block">
&gt;&gt;&gt; def py_cmp_func(a, b):
...    print &quot;py_cmp_func&quot;, a[0], b[0]
...    return a[0] - b[0]
...
&gt;&gt;&gt;
</pre>
<p>Final run on Windows:</p>
<pre class="literal-block">
&gt;&gt;&gt; qsort(ia, len(ia), sizeof(c_int), CMPFUNC(py_cmp_func)) # doctest: +WINDOWS
py_cmp_func 33 7
py_cmp_func 99 33
py_cmp_func 5 99
py_cmp_func 1 99
py_cmp_func 33 7
py_cmp_func 1 33
py_cmp_func 5 33
py_cmp_func 5 7
py_cmp_func 1 7
py_cmp_func 5 1
&gt;&gt;&gt;
</pre>
<p>and on Linux:</p>
<pre class="literal-block">
&gt;&gt;&gt; qsort(ia, len(ia), sizeof(c_int), CMPFUNC(py_cmp_func)) # doctest: +LINUX
py_cmp_func 5 1
py_cmp_func 33 99
py_cmp_func 7 33
py_cmp_func 1 7
py_cmp_func 5 7
&gt;&gt;&gt;
</pre>
<p>It is quite interesting to see that the Windows <tt class="docutils literal"><span class="pre">qsort</span></tt> function
needs more comparisons than the linux version!</p>
<p>As we can easily check, our array sorted now:</p>
<pre class="literal-block">
&gt;&gt;&gt; for i in ia: print i,
...
1 5 7 33 99
&gt;&gt;&gt;
</pre>
<p><strong>Important note for callback functions:</strong></p>
<p>Make sure you keep references to CFUNCTYPE objects as long as they are
used from C code. <tt class="docutils literal"><span class="pre">ctypes</span></tt> doesn't, and if you don't, they may be
garbage collected, crashing your program when a callback is made.</p>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id18" id="accessing-values-exported-from-dlls" name="accessing-values-exported-from-dlls">Accessing values exported from dlls</a></h1>
<p>Sometimes, a dll not only exports functions, it also exports
variables. An example in the Python library itself is the
<tt class="docutils literal"><span class="pre">Py_OptimizeFlag</span></tt>, an integer set to 0, 1, or 2, depending on the
<tt class="docutils literal"><span class="pre">-O</span></tt> or <tt class="docutils literal"><span class="pre">-OO</span></tt> flag given on startup.</p>
<p><tt class="docutils literal"><span class="pre">ctypes</span></tt> can access values like this with the <tt class="docutils literal"><span class="pre">in_dll</span></tt> class
methods of the type.  <tt class="docutils literal"><span class="pre">pythonapi</span></tt> ìs a predefined symbol giving
access to the Python C api:</p>
<pre class="literal-block">
&gt;&gt;&gt; opt_flag = c_int.in_dll(pythonapi, &quot;Py_OptimizeFlag&quot;)
&gt;&gt;&gt; print opt_flag
c_long(0)
&gt;&gt;&gt;
</pre>
<p>If the interpreter would have been started with <tt class="docutils literal"><span class="pre">-O</span></tt>, the sample
would have printed <tt class="docutils literal"><span class="pre">c_long(1)</span></tt>, or <tt class="docutils literal"><span class="pre">c_long(2)</span></tt> if <tt class="docutils literal"><span class="pre">-OO</span></tt> would have
been specified.</p>
<p>An extended example which also demonstrates the use of pointers
accesses the <tt class="docutils literal"><span class="pre">PyImport_FrozenModules</span></tt> pointer exported by Python.</p>
<p>Quoting the Python docs: <em>This pointer is initialized to point to an
array of ``struct _frozen`` records, terminated by one whose members
are all NULL or zero. When a frozen module is imported, it is searched
in this table. Third-party code could play tricks with this to provide
a dynamically created collection of frozen modules.</em></p>
<p>So manipulating this pointer could even prove useful. To restrict the
example size, we show only how this table can be read with
<tt class="docutils literal"><span class="pre">ctypes</span></tt>:</p>
<pre class="literal-block">
&gt;&gt;&gt; from ctypes import *
&gt;&gt;&gt;
&gt;&gt;&gt; class struct_frozen(Structure):
...    _fields_ = [(&quot;name&quot;, c_char_p),
...                (&quot;code&quot;, POINTER(c_ubyte)),
...                (&quot;size&quot;, c_int)]
...
&gt;&gt;&gt;
</pre>
<p>We have defined the <tt class="docutils literal"><span class="pre">struct</span> <span class="pre">_frozen</span></tt> data type, so we can get the
pointer to the table:</p>
<pre class="literal-block">
&gt;&gt;&gt; FrozenTable = POINTER(struct_frozen)
&gt;&gt;&gt; table = FrozenTable.in_dll(pythonapi, &quot;PyImport_FrozenModules&quot;)
&gt;&gt;&gt;
</pre>
<p>Since <tt class="docutils literal"><span class="pre">table</span></tt> is a <tt class="docutils literal"><span class="pre">pointer</span></tt> to the array of <tt class="docutils literal"><span class="pre">struct_frozen</span></tt>
records, we can iterate over it, but we just have to make sure that
our loop terminates, because pointers have no size. Sooner or later it
would probably crash with an access violation or whatever, so it's
better to break out of the loop when we hit the NULL entry:</p>
<pre class="literal-block">
&gt;&gt;&gt; for item in table:
...    print item.name, item.size
...    if item.name is None:
...        break
...
__hello__ 104
__phello__ -104
__phello__.spam 104
None 0
&gt;&gt;&gt;
</pre>
<p>The fact that standard Python has a frozen module and a frozen package
(indicated by the negative size member) is not wellknown, it is only
used for testing. Try it out with <tt class="docutils literal"><span class="pre">import</span> <span class="pre">__hello__</span></tt> for example.</p>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id19" id="surprises" name="surprises">Surprises</a></h1>
<p>There are some edges in <tt class="docutils literal"><span class="pre">ctypes</span></tt> where you may be expect something
else than what actually happens.</p>
<p>Consider the following example:</p>
<pre class="literal-block">
&gt;&gt;&gt; from ctypes import *
&gt;&gt;&gt; class POINT(Structure):
...    _fields_ = (&quot;x&quot;, c_int), (&quot;y&quot;, c_int)
...
&gt;&gt;&gt; class RECT(Structure):
...    _fields_ = (&quot;a&quot;, POINT), (&quot;b&quot;, POINT)
...
&gt;&gt;&gt; p1 = POINT(1, 2)
&gt;&gt;&gt; p2 = POINT(3, 4)
&gt;&gt;&gt; rc = RECT(p1, p2)
&gt;&gt;&gt; print rc.a.x, rc.a.y, rc.b.x, rc.b.y
1 2 3 4
&gt;&gt;&gt; # now swap the two points
&gt;&gt;&gt; rc.a, rc.b = rc.b, rc.a
&gt;&gt;&gt; print rc.a.x, rc.a.y, rc.b.x, rc.b.y
3 4 3 4
&gt;&gt;&gt;
</pre>
<p>Hm. We certainly expected the last statement to print <tt class="docutils literal"><span class="pre">3</span> <span class="pre">4</span> <span class="pre">1</span> <span class="pre">2</span></tt>.
What happended? Here are the steps of the <tt class="docutils literal"><span class="pre">rc.a,</span> <span class="pre">rc.b</span> <span class="pre">=</span> <span class="pre">rc.b,</span> <span class="pre">rc.a</span></tt>
line above:</p>
<pre class="literal-block">
&gt;&gt;&gt; temp0, temp1 = rc.b, rc.a
&gt;&gt;&gt; rc.a = temp0
&gt;&gt;&gt; rc.b = temp1
&gt;&gt;&gt;
</pre>
<p>Note that <tt class="docutils literal"><span class="pre">temp0</span></tt> and <tt class="docutils literal"><span class="pre">temp1</span></tt> are objects still using the internal
buffer of the <tt class="docutils literal"><span class="pre">rc</span></tt> object above. So executing <tt class="docutils literal"><span class="pre">rc.a</span> <span class="pre">=</span> <span class="pre">temp0</span></tt>
copies the buffer contents of <tt class="docutils literal"><span class="pre">temp0</span></tt> into <tt class="docutils literal"><span class="pre">rc</span></tt> 's buffer.  This,
in turn, changes the contents of <tt class="docutils literal"><span class="pre">temp1</span></tt>. So, the last assignment
<tt class="docutils literal"><span class="pre">rc.b</span> <span class="pre">=</span> <span class="pre">temp1</span></tt>, doesn't have the expected effect.</p>
<p>Keep in mind that retrieving subobjects from Structure, Unions, and
Arrays doesn't <em>copy</em> the subobject, instead it retrieves a wrapper
object accessing the root-object's underlying buffer.</p>
<p>Another example that may behave different from what one would expect is this:</p>
<pre class="literal-block">
&gt;&gt;&gt; s = c_char_p()
&gt;&gt;&gt; s.value = &quot;abc def ghi&quot;
&gt;&gt;&gt; s.value
'abc def ghi'
&gt;&gt;&gt; s.value is s.value
False
&gt;&gt;&gt;
</pre>
<p>Why is it printing <tt class="docutils literal"><span class="pre">False</span></tt>?  ctypes instances are objects containing
a memory block plus some descriptors accessing the contents of the
memory.  Storing a Python object in the memory block does not store
the object itself, instead the <tt class="docutils literal"><span class="pre">contents</span></tt> of the object is stored.
Accessing the contents again constructs a new Python each time!</p>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id20" id="variable-sized-data-types" name="variable-sized-data-types">Variable-sized data types</a></h1>
<p><tt class="docutils literal"><span class="pre">ctypes</span></tt> provides some support for variable-sized arrays and
structures (this was added in version 0.9.9.7).</p>
<p>The <tt class="docutils literal"><span class="pre">resize</span></tt> function can be used to resize the memory buffer of an
existing ctypes object.  The function takes the object as first
argument, and the requested size in bytes as the second argument.  The
memory block cannot be made smaller than the natural memory block
specified by the objects type, a <tt class="docutils literal"><span class="pre">ValueError</span></tt> is raised if this is
tried:</p>
<pre class="literal-block">
&gt;&gt;&gt; short_array = (c_short * 4)()
&gt;&gt;&gt; print sizeof(short_array)
8
&gt;&gt;&gt; resize(short_array, 4)
Traceback (most recent call last):
    ...
ValueError: minimum size is 8
&gt;&gt;&gt; resize(short_array, 32)
&gt;&gt;&gt; sizeof(short_array)
32
&gt;&gt;&gt; sizeof(type(short_array))
8
&gt;&gt;&gt;
</pre>
<p>This is nice and fine, but how would one access the additional
elements contained in this array?  Since the type still only knows
about 4 elements, we get errors accessing other elements:</p>
<pre class="literal-block">
&gt;&gt;&gt; short_array[:]
[0, 0, 0, 0]
&gt;&gt;&gt; short_array[7]
Traceback (most recent call last):
    ...
IndexError: invalid index
&gt;&gt;&gt;
</pre>
<p>Another way to use variable-sized data types with <tt class="docutils literal"><span class="pre">ctypes</span></tt> is to use
the dynamic nature of Python, and (re-)define the data type after the
required size is already known, on a case by case basis.</p>
</div>
<div class="section">
<h1><a class="toc-backref" href="#id21" id="bugs-todo-and-non-implemented-things" name="bugs-todo-and-non-implemented-things">Bugs, ToDo and non-implemented things</a></h1>
<p>Enumeration types are not implemented. You can do it easily yourself,
using <tt class="docutils literal"><span class="pre">c_int</span></tt> as the base class.</p>
<p><tt class="docutils literal"><span class="pre">long</span> <span class="pre">double</span></tt> is not implemented.</p>
<!-- Local Variables:
compile-command: "make.bat"
End: -->
</div>
</div>
</html>


= Examples =
= Examples =

Revision as of 14:56, 9 November 2007

Tutorial

Extracted from http://starship.python.net/crew/theller/ctypes/tutorial.html

Examples

A scrap of sample code from a past SVT exercise implemented usiong Pyth and ctypes:

#!/usr/bin/env python

## import ctypes
from ctypes import *
from ctypes.wintypes import *

import win32con
import sys

import time

#dll = windll.MouseHook
# the dll is called hooksm.dll
dll = cdll.hooksm

#startfn = dll.setMyHook
#stopfn = dll.clearMyHook


# startfn()

dll.search_msg = c_int (win32con.WM_SETTEXT)

#windll.user32.DefWindowProcA(c_int(hwnd), c_int(message), c_int(wParam), c_int(lParam))

txt = "Analysis is complete for:"

startfn = dll.hksmSetHook
#startfn.argtypes = [c_ulong, c_int, c_uint, c_long, c_char_p, c_ulong]
startfn (
#    c_ulong(0x00000238),
    c_ulong(0x00000000),
    c_int (win32con.WM_SETTEXT),
    c_uint(win32con.NULL),
    c_char_p(txt),
    0x00000000
    )

startt = time.time()
print startt

#time.sleep(20)
print "sync pre"
syncfn = dll.Synchronize
print "sync start"
syncfn()
print "sync stop"
endt = time.time()
print endt
print endt - startt

dll.hksmUnsetHook ()

#stopfn()

#flog = dll.log