python/doc/v2/faq.html

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">

<!-- Copyright David Abrahams 2006. Distributed under the Boost -->
<!-- Software License, Version 1.0. (See accompanying -->
<!-- file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) -->
<html>
  <head>
    <meta name="generator" content=
    "HTML Tidy for Cygwin (vers 1st April 2002), see www.w3.org">
    <meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
    <link rel="stylesheet" type="text/css" href="../boost.css">

    <title>Boost.Python - FAQ</title>
  </head>

  <body link="#0000ff" vlink="#800080">
    <table border="0" cellpadding="7" cellspacing="0" width="100%" summary=
    "header">
      <tr>
        <td valign="top" width="300">
          <h3><a href="../../../../index.htm"><img height="86" width="277"
          alt="C++ Boost" src="../../../../boost.png" border="0"></a></h3>
        </td>

        <td valign="top">
          <h1 align="center"><a href="../index.html">Boost.Python</a></h1>

          <h2 align="center">Frequently Asked Questions (FAQs)</h2>
        </td>
      </tr>
    </table>
    <hr>

    <dl class="page-index">

      <dt><a href="#funcptr">How can I wrap a function which takes a
      function pointer as an argument?</a><dd>

      <dt><a href="#dangling">I'm getting the "attempt to return dangling
      reference" error. What am I doing wrong?</a></dt>

      <dt><a href="#question1">Is return_internal_reference
      efficient?</a></dt>

      <dt><a href="#question2">How can I wrap functions which take C++
      containers as arguments?</a></dt>

      <dt><a href="#c1204">fatal error C1204:Compiler limit:internal
      structure overflow</a></dt>

      <dt><a href="#debugging">How do I debug my Python extensions?</a></dt>

      <dt><a href="#imul">Why doesn't my <code>*=</code> operator
      work?</a></dt>

      <dt><a href="#macosx">Does Boost.Python work with Mac OS X?</a></dt>

      <dt><a href="#xref">How can I find the existing PyObject that holds a
      C++ object?</a></dt>

      <dt><a href="#ownership">How can I wrap a function which needs to take
      ownership of a raw pointer?</a></dt>

      <dt><a href="#slow_compilation">Compilation takes too much time and eats too much memory!
      What can I do to make it faster?</a></dt>

      <dt><a href="#packages">How do I create sub-packages using Boost.Python?</a></dt>

      <dt><a href="#msvcthrowbug"
      >error C2064: term does not evaluate to a function taking 2 arguments</a>
      </dt>

      <dt><a href="#custom_string"
      >How can I automatically convert my custom string type to
       and from a Python string?</a></dt>

      <dt><a href="#topythonconversionfailed">Why is my automatic to-python conversion not being
      found?</a></dt>

      <dt><a href="#threadsupport">Is Boost.Python thread-aware/compatible with multiple interpreters?</a></dt>
    </dl>
    <hr>

    <h2><a name="funcptr">How can I wrap a function which takes a
      function pointer as an argument?</a></h2>

    If what you're trying to do is something like this:
<pre>
typedef boost::function&lt;void (string s) &gt; funcptr;

void foo(funcptr fp)
{
    fp(&quot;hello,world!&quot;);
}

BOOST_PYTHON_MODULE(test)
{
    def(&quot;foo&quot;,foo) ;
}
</pre>

And then:

<pre>
&gt;&gt;&gt; def hello(s):
...    print s
...
&gt;&gt;&gt; foo(hello)
hello, world!
</pre>

    The short answer is: &quot;you can't&quot;.  This is not a
    Boost.Python limitation so much as a limitation of C++.  The
    problem is that a Python function is actually data, and the only
    way of associating data with a C++ function pointer is to store it
    in a static variable of the function.  The problem with that is
    that you can only associate one piece of data with every C++
    function, and we have no way of compiling a new C++ function
    on-the-fly for every Python function you decide to pass
    to <code>foo</code>.  In other words, this could work if the C++
    function is always going to invoke the <em>same</em> Python
    function, but you probably don't want that.

    <p>If you have the luxury of changing the C++ code you're
    wrapping, pass it an <code>object</code> instead and call that;
    the overloaded function call operator will invoke the Python
    function you pass it behind the <code>object</code>.

    <p>For more perspective on the issue, see <a
    href="http://aspn.activestate.com/ASPN/Mail/Message/1554837">this
    posting</a>.

    <hr>

    <h2><a name="dangling">I'm getting the "attempt to return dangling
    reference" error. What am I doing wrong?</a></h2>
    That exception is protecting you from causing a nasty crash. It usually
    happens in response to some code like this:
<pre>
period const&amp; get_floating_frequency() const
{
  return boost::python::call_method&lt;period const&amp;&gt;(
      m_self,"get_floating_frequency");
}
</pre>
    And you get:
<pre>
ReferenceError: Attempt to return dangling reference to object of type:
class period
</pre>

    <p>In this case, the Python method invoked by <code>call_method</code>
    constructs a new Python object. You're trying to return a reference to a
    C++ object (an instance of <code>class period</code>) contained within
    and owned by that Python object. Because the called method handed back a
    brand new object, the only reference to it is held for the duration of
    <code>get_floating_frequency()</code> above. When the function returns,
    the Python object will be destroyed, destroying the instance of
    <code>class period</code>, and leaving the returned reference dangling.
    That's already undefined behavior, and if you try to do anything with
    that reference you're likely to cause a crash. Boost.Python detects this
    situation at runtime and helpfully throws an exception instead of letting
    you do that.<br>
     &nbsp;</p>
    <hr>

    <h2><a name="question1"></a>Is return_internal_reference efficient?</h2>

    <blockquote>
      <b>Q:</b> <i>I have an object composed of 12 doubles. A const&amp; to
      this object is returned by a member function of another class. From the
      viewpoint of using the returned object in Python I do not care if I get
      a copy or a reference to the returned object. In Boost.Python Version 2
      I have the choice of using copy_const_reference or
      return_internal_reference. Are there considerations that would lead me
      to prefer one over the other, such as size of generated code or memory
      overhead?</i>

      <p><b>A:</b> copy_const_reference will make an instance with storage
      for one of your objects, size = base_size + 12 * sizeof(double).
      return_internal_reference will make an instance with storage for a
      pointer to one of your objects, size = base_size + sizeof(void*).
      However, it will also create a weak reference object which goes in the
      source object's weakreflist and a special callback object to manage the
      lifetime of the internally-referenced object. My guess?
      copy_const_reference is your friend here, resulting in less overall
      memory use and less fragmentation, also probably fewer total
      cycles.</p>
    </blockquote>
    <hr>

    <h2><a name="question2"></a>How can I wrap functions which take C++
    containers as arguments?</h2>

    <p>Ralf W. Grosse-Kunstleve provides these notes:</p>

    <ol>
      <li>
        Using the regular <code>class_&lt;&gt;</code> wrapper:
<pre>
class_&lt;std::vector&lt;double&gt; &gt;("std_vector_double")
  .def(...)
  ...
  ;
</pre>
        This can be moved to a template so that several types (double, int,
        long, etc.) can be wrapped with the same code. This technique is used
        in the file

        <blockquote>
          scitbx/include/scitbx/array_family/boost_python/flex_wrapper.h
        </blockquote>
        in the "scitbx" package. The file could easily be modified for
        wrapping std::vector&lt;&gt; instantiations.

        <p>This type of C++/Python binding is most suitable for containers
        that may contain a large number of elements (&gt;10000).</p>
      </li>

      <li>
        Using custom rvalue converters. Boost.Python "rvalue converters"
        match function signatures such as:
<pre>
void foo(std::vector&lt;double&gt; const&amp; array); // pass by const-reference
void foo(std::vector&lt;double&gt; array); // pass by value
</pre>
        Some custom rvalue converters are implemented in the file

        <blockquote>
          scitbx/include/scitbx/boost_python/container_conversions.h
        </blockquote>
        This code can be used to convert from C++ container types such as
        std::vector&lt;&gt; or std::list&lt;&gt; to Python tuples and vice
        versa. A few simple examples can be found in the file

        <blockquote>
          scitbx/array_family/boost_python/regression_test_module.cpp
        </blockquote>
        Automatic C++ container &lt;-&gt; Python tuple conversions are most
        suitable for containers of moderate size. These converters generate
        significantly less object code compared to alternative 1 above.
      </li>
    </ol>
    A disadvantage of using alternative 2 is that operators such as
    arithmetic +,-,*,/,% are not available. It would be useful to have custom
    rvalue converters that convert to a "math_array" type instead of tuples.
    This is currently not implemented but is possible within the framework of
    Boost.Python V2 as it will be released in the next couple of weeks. [ed.:
    this was posted on 2002/03/10]

    <p>It would also be useful to also have "custom lvalue converters" such
    as std::vector&lt;&gt; &lt;-&gt; Python list. These converters would
    support the modification of the Python list from C++. For example:</p>

    <p>C++:</p>
<pre>
void foo(std::vector&lt;double&gt;&amp; array)
{
  for(std::size_t i=0;i&lt;array.size();i++) {
    array[i] *= 2;
  }
}
</pre>
    Python:
<pre>
&gt;&gt;&gt; l = [1, 2, 3]
&gt;&gt;&gt; foo(l)
&gt;&gt;&gt; print l
[2, 4, 6]
</pre>
    Custom lvalue converters require changes to the Boost.Python core library
    and are currently not available.

    <p>P.S.:</p>

    <p>The "scitbx" files referenced above are available via anonymous
    CVS:</p>
<pre>
cvs -d:pserver:anonymous@cvs.cctbx.sourceforge.net:/cvsroot/cctbx login
cvs -d:pserver:anonymous@cvs.cctbx.sourceforge.net:/cvsroot/cctbx co scitbx
</pre>
    <hr>

    <h2><a name="c1204"></a>fatal error C1204:Compiler limit:internal
    structure overflow</h2>

    <blockquote>
      <b>Q:</b> <i>I get this error message when compiling a large source
      file. What can I do?</i>

      <p><b>A:</b> You have two choices:</p>

      <ol>
        <li>Upgrade your compiler (preferred)</li>

        <li>
          Break your source file up into multiple translation units.

          <p><code><b>my_module.cpp</b></code>:</p>
<pre>
...
void more_of_my_module();
BOOST_PYTHON_MODULE(my_module)
{
   def("foo", foo);
   def("bar", bar);
   ...
   more_of_my_module();
}
</pre>
          <code><b>more_of_my_module.cpp</b></code>:
<pre>
void more_of_my_module()
{
   def("baz", baz);
   ...
}
</pre>
          If you find that a <code><a href=
          "class.html#class_-spec">class_</a>&lt;...&gt;</code> declaration
          can't fit in a single source file without triggering the error, you
          can always pass a reference to the <code>class_</code> object to a
          function in another source file, and call some of its member
          functions (e.g. <code>.def(...)</code>) in the auxilliary source
          file:

          <p><code><b>more_of_my_class.cpp</b></code>:</p>
<pre>
void more_of_my_class(class&lt;my_class&gt;&amp; x)
{
   x
     .def("baz", baz)
     .add_property("xx", &amp;my_class::get_xx, &amp;my_class::set_xx)
     ;

   ...
}
</pre>
        </li>
      </ol>
    </blockquote>
    <hr>

    <h2><a name="debugging"></a>How do I debug my Python extensions?</h2>

    <p>Greg Burley gives the following answer for Unix GCC users:</p>

    <blockquote>
      Once you have created a boost python extension for your c++ library or
      class, you may need to debug the code. Afterall this is one of the
      reasons for wrapping the library in python. An expected side-effect or
      benefit of using BPL is that debugging should be isolated to the c++
      library that is under test, given that python code is minimal and
      boost::python either works or it doesn't. (ie. While errors can occur
      when the wrapping method is invalid, most errors are caught by the
      compiler ;-).

      <p>The basic steps required to initiate a gdb session to debug a c++
      library via python are shown here. Note, however that you should start
      the gdb session in the directory that contains your BPL my_ext.so
      module.</p>
<pre>
(gdb) target exec python
(gdb) run
 &gt;&gt;&gt; from my_ext import *
 &gt;&gt;&gt; [C-c]
(gdb) break MyClass::MyBuggyFunction
(gdb) cont
 &gt;&gt;&gt; pyobj = MyClass()
 &gt;&gt;&gt; pyobj.MyBuggyFunction()
Breakpoint 1, MyClass::MyBuggyFunction ...
Current language:  auto; currently c++
(gdb) do debugging stuff
</pre>
    </blockquote>

    <p>Greg's approach works even better using Emacs' "<code>gdb</code>"
    command, since it will show you each line of source as you step through
    it.</p>

    <p>On <b>Windows</b>, my favorite debugging solution is the debugger that
    comes with Microsoft Visual C++ 7. This debugger seems to work with code
    generated by all versions of Microsoft and Metrowerks toolsets; it's rock
    solid and "just works" without requiring any special tricks from the
    user.</p>

    <p>Raoul Gough has provided the following for gdb on Windows:</p>

    <blockquote>

      <p>gdb support for Windows DLLs has improved lately, so it is
      now possible to debug Python extensions using a few
      tricks. Firstly, you will need an up-to-date gdb with support
      for minimal symbol extraction from a DLL. Any gdb from version 6
      onwards, or Cygwin gdb-20030214-1 and onwards should do. A
      suitable release will have a section in the gdb.info file under
      Configuration &ndash; Native &ndash; Cygwin Native &ndash;
      Non-debug DLL symbols. Refer to that info section for more
      details of the procedures outlined here.</p>

      <p>Secondly, it seems necessary to set a breakpoint in the
      Python interpreter, rather than using ^C to break execution. A
      good place to set this breakpoint is PyOS_Readline, which will
      stop execution immediately before reading each interactive
      Python command. You have to let Python start once under the
      debugger, so that it loads its own DLL, before you can set the
      breakpoint:</p>

<p>
<pre>
$ gdb python
GNU gdb 2003-09-02-cvs (cygwin-special)
[...]

(gdb) run
Starting program: /cygdrive/c/Python22/python.exe
Python 2.2.2 (#37, Oct 14 2002, 17:02:34) [MSC 32 bit (Intel)] on win32
Type "help", "copyright", "credits" or "license" for more information.
&gt;&gt;&gt; ^Z


Program exited normally.
(gdb) break *&amp;PyOS_Readline
Breakpoint 1 at 0x1e04eff0
(gdb) run
Starting program: /cygdrive/c/Python22/python.exe
Python 2.2.2 (#37, Oct 14 2002, 17:02:34) [MSC 32 bit (Intel)] on win32
Type "help", "copyright", "credits" or "license" for more information.

Breakpoint 1, 0x1e04eff0 in python22!PyOS_Readline ()
   from /cygdrive/c/WINNT/system32/python22.dll
(gdb) cont
Continuing.
&gt;&gt;&gt; from my_ext import *

Breakpoint 1, 0x1e04eff0 in python22!PyOS_Readline ()
   from /cygdrive/c/WINNT/system32/python22.dll
(gdb) # my_ext now loaded (with any debugging symbols it contains)
</pre>
    </blockquote>

    <h3>Debugging extensions through Boost.Build</h3>
    If you are launching your extension module tests with <a href=
    "../../../../tools/build/v1/build_system.htm">Boost.Build</a> using the
    <code>boost-python-runtest</code> rule, you can ask it to launch your
    debugger for you by adding "--debugger=<i>debugger</i>" to your bjam
    command-line:
<pre>
bjam -sTOOLS=vc7.1 "--debugger=devenv /debugexe" test
bjam -sTOOLS=gcc -sPYTHON_LAUNCH=gdb test
</pre>
    It can also be extremely useful to add the <code>-d+2</code> option when
    you run your test, because Boost.Build will then show you the exact
    commands it uses to invoke it. This will invariably involve setting up
    PYTHONPATH and other important environment variables such as
    LD_LIBRARY_PATH which may be needed by your debugger in order to get
    things to work right.
    <hr>

    <h2><a name="imul"></a>Why doesn't my <code>*=</code> operator work?</h2>

    <blockquote>
      <b>Q:</b> <i>I have exported my class to python, with many overloaded
      operators. it works fine for me except the</i> <code>*=</code>
      <i>operator. It always tells me "can't multiply sequence with non int
      type". If I use</i> <code>p1.__imul__(p2)</code> <i>instead of</i>
      <code>p1 *= p2</code><i>, it successfully executes my code. What's
      wrong with me?</i>

      <p><b>A:</b> There's nothing wrong with you. This is a bug in Python
      2.2. You can see the same effect in Pure Python (you can learn a lot
      about what's happening in Boost.Python by playing with new-style
      classes in Pure Python).</p>
<pre>
&gt;&gt;&gt; class X(object):
...     def __imul__(self, x):
...         print 'imul'
...
&gt;&gt;&gt; x = X()
&gt;&gt;&gt; x *= 1
</pre>
      To cure this problem, all you need to do is upgrade your Python to
      version 2.2.1 or later.
    </blockquote>
    <hr>

    <h2><a name="macosx"></a>Does Boost.Python work with Mac OS X?</h2>

    It is known to work under 10.2.8 and 10.3 using
    Apple's gcc 3.3 compiler:
    <pre>gcc (GCC) 3.3 20030304 (Apple Computer, Inc. build 1493)</pre>
    Under 10.2.8 get the August 2003 gcc update (free at
    <a href="http://connect.apple.com/">http://connect.apple.com/</a>).
    Under 10.3 get the Xcode Tools v1.0 (also free).
    <p>
    Python 2.3 is required. The Python that ships with 10.3 is
    fine. Under 10.2.8 use these commands to install Python
    as a framework:
    <pre>./configure --enable-framework
make
make frameworkinstall</pre>
    The last command requires root privileges because the target
    directory is
    <tt>/Library/Frameworks/Python.framework/Versions/2.3</tt>.
    However, the installation does not interfere with the Python
    version that ships with 10.2.8.
    <p>
    It is also crucial to increase the <tt>stacksize</tt> before
    starting compilations, e.g.:
    <pre>limit stacksize 8192k</pre>
    If the <tt>stacksize</tt> is too small the build might crash with
    internal compiler errors.
    <p>
    Sometimes Apple's compiler exhibits a bug by printing an error
    like the following while compiling a
    <tt>boost::python::class_&lt;your_type&gt;</tt>
    template instantiation:
    <pre>.../inheritance.hpp:44: error: cannot
 dynamic_cast `p' (of type `struct cctbx::boost_python::&lt;unnamed&gt;::add_pair*
   ') to type `void*' (source type is not polymorphic)</pre>

    We do not know a general workaround, but if the definition of
    <tt>your_type</tt> can be modified the following was found
    to work in all cases encountered so far:<pre>struct your_type
{
  // before defining any member data
#if defined(__MACH__) &amp;&amp; defined(__APPLE_CC__) &amp;&amp; __APPLE_CC__ == 1493
  bool dummy_;
#endif
  // now your member data, e.g.
  double x;
  int j;
  // etc.
};</pre>

    <hr>
    <h2><a name="xref">How can I find the existing PyObject that holds a C++
    object?</a></h2>

    <blockquote>
      "I am wrapping a function that always returns a pointer to an
      already-held C++ object."
    </blockquote>
    One way to do that is to hijack the mechanisms used for wrapping a class
    with virtual functions. If you make a wrapper class with an initial
    PyObject* constructor argument and store that PyObject* as "self", you
    can get back to it by casting down to that wrapper type in a thin wrapper
    function. For example:
<pre>
class X { X(int); virtual ~X(); ... };
X* f();  // known to return Xs that are managed by Python objects


// wrapping code

struct X_wrap : X
{
    X_wrap(PyObject* self, int v) : self(self), X(v) {}
    PyObject* self;
};

handle&lt;&gt; f_wrap()
{
    X_wrap* xw = dynamic_cast&lt;X_wrap*&gt;(f());
    assert(xw != 0);
    return handle&lt;&gt;(borrowed(xw-&gt;self));
}

...

def("f", f_wrap());
class_&lt;X,X_wrap,boost::noncopyable&gt;("X", init&lt;int&gt;())
   ...
   ;
</pre>
    Of course, if X has no virtual functions you'll have to use
    <code>static_cast</code> instead of <code>dynamic_cast</code> with no
    runtime check that it's valid. This approach also only works if the
    <code>X</code> object was constructed from Python, because
    <code>X</code>s constructed from C++ are of course never
    <code>X_wrap</code> objects.

    <p>Another approach to this requires you to change your C++ code a bit;
    if that's an option for you it might be a better way to go. work we've
    been meaning to get to anyway. When a <code>shared_ptr&lt;X&gt;</code> is
    converted from Python, the shared_ptr actually manages a reference to the
    containing Python object. When a shared_ptr&lt;X&gt; is converted back to
    Python, the library checks to see if it's one of those "Python object
    managers" and if so just returns the original Python object. So you could
    just write <code>object(p)</code> to get the Python object back. To
    exploit this you'd have to be able to change the C++ code you're wrapping
    so that it deals with shared_ptr instead of raw pointers.</p>

    <p>There are other approaches too. The functions that receive the Python
    object that you eventually want to return could be wrapped with a thin
    wrapper that records the correspondence between the object address and
    its containing Python object, and you could have your f_wrap function
    look in that mapping to get the Python object out.</p>

    <hr>

    <h2><a name="ownership">How can I wrap a function which needs to take
    ownership of a raw pointer?</a></h2>

    <blockquote>
      <i>Part of an API that I'm wrapping goes something like this:</i>
<pre>
struct A {}; struct B { void add( A* ); }
where B::add() takes ownership of the pointer passed to it.
</pre>

      <p><i>However:</i></p>
<pre>
a = mod.A()
b = mod.B()
b.add( a )
del a
del b
# python interpreter crashes
# later due to memory corruption.
</pre>

      <p><i>Even binding the lifetime of a</i> to b via
      with_custodian_and_ward doesn't prevent the python object a from
      ultimately trying to delete the object it's pointing to. Is there a way
      to accomplish a 'transfer-of-ownership' of a wrapped C++ object?</p>

      <p><i>--Bruce Lowery</i></p>
    </blockquote>
    Yes: Make sure the C++ object is held by auto_ptr:
<pre>
class_&lt;A, std::auto_ptr&lt;A&gt; &gt;("A")
    ...
    ;
</pre>
    Then make a thin wrapper function which takes an auto_ptr parameter:
<pre>
void b_insert(B&amp; b, std::auto_ptr&lt;A&gt; a)
{
    b.insert(a.get());
    a.release();
}
</pre>
    Wrap that as B.add. Note that pointers returned via <code><a href=
    "manage_new_object.html#manage_new_object-spec">manage_new_object</a></code>
    will also be held by <code>auto_ptr</code>, so this transfer-of-ownership
    will also work correctly.

    <hr>
    <h2><a name="slow_compilation">Compilation takes too much time and eats too
            much memory!  What can I do to make it faster?</a></h2>
    <p>
    Please refer to the <a href="../tutorial/doc/html/python/techniques.html#python.reducing_compiling_time"
    >Reducing Compiling Time</a> section in the tutorial.
    </p>

    <hr>
    <h2><a name="packages">How do I create sub-packages using Boost.Python?</a></h2>
    <p>
    Please refer to the <a href="../tutorial/doc/html/python/techniques.html#python.creating_packages"
    >Creating Packages</a> section in the tutorial.
    </p>

    <hr>
    <h2><a name="msvcthrowbug"></a>error C2064: term does
        not evaluate to a function taking 2 arguments</h2>
    <font size="-1"><i>Niall Douglas provides these notes:</i></font><p>
    If you see Microsoft Visual C++ 7.1 (MS Visual Studio .NET 2003) issue
    an error message like the following it is most likely due to a bug
    in the compiler:
    <pre>boost\boost\python\detail\invoke.hpp(76):
error C2064: term does not evaluate to a function taking 2 arguments"</pre>
    This message is triggered by code like the following:
<pre>#include &lt;boost/python.hpp&gt;

using namespace boost::python;

class FXThread
{
public:
    bool setAutoDelete(bool doso) throw();
};

void Export_FXThread()
{
    class_< FXThread >("FXThread")
        .def("setAutoDelete", &amp;FXThread::setAutoDelete)
    ;
}
    </pre>
    The bug is related to the <code>throw()</code> modifier.
    As a workaround cast off the modifier. E.g.:
<pre>
        .def("setAutoDelete", (bool (FXThread::*)(bool)) &amp;FXThread::setAutoDelete)</pre>
    <p>(The bug has been reported to Microsoft.)</p>

    <hr>
    <h2><a name="custom_string"></a>How can I automatically
    convert my custom string type to and from a Python string?</h2>
    <font size="-1"><i>Ralf W. Grosse-Kunstleve provides these
    notes:</i></font><p>
    Below is a small, self-contained demo extension module that shows
    how to do this. Here is the corresponding trivial test:
    <pre>import custom_string
assert custom_string.hello() == "Hello world."
assert custom_string.size("california") == 10</pre>

    If you look at the code you will find:

    <ul>
    <li>A custom <tt>to_python</tt> converter (easy):
        <tt>custom_string_to_python_str</tt>

    <li>A custom lvalue converter (needs more code):
        <tt>custom_string_from_python_str</tt>
    </ul>

    The custom converters are registered in the global Boost.Python
    registry near the top of the module initialization function. Once
    flow control has passed through the registration code the automatic
    conversions from and to Python strings will work in any module
    imported in the same process.

<pre>#include &lt;boost/python/module.hpp&gt;
#include &lt;boost/python/def.hpp&gt;
#include &lt;boost/python/to_python_converter.hpp&gt;

namespace sandbox { namespace {

  class custom_string
  {
    public:
      custom_string() {}
      custom_string(std::string const&amp; value) : value_(value) {}
      std::string const&amp; value() const { return value_; }
    private:
      std::string value_;
  };

  struct custom_string_to_python_str
  {
    static PyObject* convert(custom_string const&amp; s)
    {
      return boost::python::incref(boost::python::object(s.value()).ptr());
    }
  };

  struct custom_string_from_python_str
  {
    custom_string_from_python_str()
    {
      boost::python::converter::registry::push_back(
        &amp;convertible,
        &amp;construct,
        boost::python::type_id&lt;custom_string&gt;());
    }

    static void* convertible(PyObject* obj_ptr)
    {
      if (!PyString_Check(obj_ptr)) return 0;
      return obj_ptr;
    }

    static void construct(
      PyObject* obj_ptr,
      boost::python::converter::rvalue_from_python_stage1_data* data)
    {
      const char* value = PyString_AsString(obj_ptr);
      if (value == 0) boost::python::throw_error_already_set();
      void* storage = (
        (boost::python::converter::rvalue_from_python_storage&lt;custom_string&gt;*)
          data)-&gt;storage.bytes;
      new (storage) custom_string(value);
      data-&gt;convertible = storage;
    }
  };

  custom_string hello() { return custom_string(&quot;Hello world.&quot;); }

  std::size_t size(custom_string const&amp; s) { return s.value().size(); }

  void init_module()
  {
    using namespace boost::python;

    boost::python::to_python_converter&lt;
      custom_string,
      custom_string_to_python_str&gt;();

    custom_string_from_python_str();

    def(&quot;hello&quot;, hello);
    def(&quot;size&quot;, size);
  }

}} // namespace sandbox::&lt;anonymous&gt;

BOOST_PYTHON_MODULE(custom_string)
{
  sandbox::init_module();
}</pre>

    <hr>
    <h2><a name="topythonconversionfailed"></a
    >Why is my automatic to-python conversion not being found?</h2>
    <font size="-1"><i>Niall Douglas provides these notes:</i></font><p>
    If you define custom converters similar to the ones
    shown above the <tt>def_readonly()</tt> and <tt>def_readwrite()</tt>
    member functions provided by <tt>boost::python::class_</tt> for
    direct access to your member data will not work as expected.
    This is because <tt>def_readonly("bar",&nbsp;&amp;foo::bar)</tt> is
    equivalent to:

<pre>.add_property("bar", make_getter(&amp;foo::bar, return_internal_reference()))</pre>

    Similarly, <tt>def_readwrite("bar",&nbsp;&amp;foo::bar)</tt> is
    equivalent to:

<pre>.add_property("bar", make_getter(&amp;foo::bar, return_internal_reference()),
                     make_setter(&amp;foo::bar, return_internal_reference())</pre>

    In order to define return value policies compatible with the
    custom conversions replace <tt>def_readonly()</tt> and
    <tt>def_readwrite()</tt> by <tt>add_property()</tt>. E.g.:

<pre>.add_property("bar", make_getter(&amp;foo::bar, return_value_policy&lt;return_by_value&gt;()),
                     make_setter(&amp;foo::bar, return_value_policy&lt;return_by_value&gt;()))</pre>

    <hr>
    <h2><a name="threadsupport"></a
    >Is Boost.Python thread-aware/compatible with multiple interpreters?</h2>
    <font size="-1"><i>Niall Douglas provides these notes:</i></font><p>
    The quick answer to this is: no.</p>
    <p>
    The longer answer is that it can be patched to be so, but it's
    complex. You will need to add custom lock/unlock wrapping of every
    time your code enters Boost.Python (particularly every virtual
    function override) plus heavily modify
    <tt>boost/python/detail/invoke.hpp</tt> with custom unlock/lock
    wrapping of every time Boost.Python enters your code. You must
    furthermore take care to <i>not</i> unlock/lock when Boost.Python
    is invoking iterator changes via <tt>invoke.hpp</tt>.</p>
    <p>
    There is a patched <tt>invoke.hpp</tt> posted on the C++-SIG
    mailing list archives and you can find a real implementation of all
    the machinery necessary to fully implement this in the TnFOX
    project at <a href="http://sourceforge.net/projects/tnfox/"> this
    SourceForge project location</a>.</p>

    <hr>

    <p>Revised
    <!--webbot bot="Timestamp" S-Type="EDITED" S-Format="%d %B, %Y" startspan -->
     12 March, 2006
    <!--webbot bot="Timestamp" endspan i-checksum="39359" -->
    </p>

    <p><i>&copy; Copyright <a href=
    "http://www.boost.org/people/dave_abrahams.htm">Dave Abrahams</a> 2002-2006.</i></p>
  </body>
</html>