<?xml version="1.0" encoding="utf-8" ?> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en"> <head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8" /> <meta name="generator" content="Docutils 0.3.0: http://docutils.sourceforge.net/" /> <title>Boost.Python Internals Boost</title> <link rel="stylesheet" href="../../../rst.css" type="text/css" /> </head> <body> <div class="document" id="boost-python-internals-logo"> <h1 class="title"><a class="reference" href="index.html">Boost.Python</a> Internals <a class="reference" href="../../../index.htm"><img alt="Boost" src="../../../boost.png" /></a></h1> <div class="section" id="a-conversation-between-brett-calcott-and-david-abrahams"> <h1><a name="a-conversation-between-brett-calcott-and-david-abrahams">A conversation between Brett Calcott and David Abrahams</a></h1> <table class="field-list" frame="void" rules="none"> <col class="field-name" /> <col class="field-body" /> <tbody valign="top"> <tr class="field"><th class="field-name">copyright:</th><td class="field-body">Copyright David Abrahams and Brett Calcott 2003. See accompanying <a class="reference" href="../../../LICENSE_1_0.txt">license</a> for terms of use.</td> </tr> </tbody> </table> <p>In both of these cases, I'm quite capable of reading code - but the thing I don't get from scanning the source is a sense of the architecture, both structurally, and temporally (er, I mean in what order things go on).</p> <ol class="arabic"> <li><p class="first">What happens when you do the following:</p> <pre class="literal-block"> struct boring {}; ...etc... class_<boring>("boring") ; </pre> </li> </ol> <p>There seems to be a fair bit going on.</p> <blockquote> <ul class="simple"> <li>Python needs a new ClassType to be registered.</li> <li>We need to construct a new type that can hold our boring struct.</li> <li>Inward and outward converters need to be registered for the type.</li> </ul> </blockquote> <p>Can you gesture in the general direction where these things are done?</p> <blockquote> <p>I only have time for a "off-the-top-of-my-head" answer at the moment; I suggest you step through the code with a debugger after reading this to see how it works, fill in details, and make sure I didn't forget anything.</p> <blockquote> <p>A new (Python) subclass of Boost.Python.Instance (see libs/python/src/object/class.cpp) is created by invoking Boost.Python.class, the metatype:</p> <pre class="literal-block"> >>> boring = Boost.Python.class( ... 'boring' ... , bases_tuple # in this case, just () ... , { ... '__module__' : module_name ... , '__doc__' : doc_string # optional ... } ... ) </pre> <p>A handle to this object is stuck in the m_class_object field of the registration associated with <tt class="literal"><span class="pre">typeid(boring)</span></tt>. The registry will keep that object alive forever, even if you wipe out the 'boring' attribute of the extension module (probably not a good thing).</p> <p>Because you didn't specify <tt class="literal"><span class="pre">class<boring,</span> <span class="pre">non_copyable,</span> <span class="pre">...></span></tt>, a to-python converter for boring is registered which copies its argument into a value_holder held by the the Python boring object.</p> <p>Because you didn't specify <tt class="literal"><span class="pre">class<boring</span> <span class="pre">...>(no_init)</span></tt>, an <tt class="literal"><span class="pre">__init__</span></tt> function object is added to the class dictionary which default-constructs a boring in a value_holder (because you didn't specify some smart pointer or derived wrapper class as a holder) held by the Python boring object.</p> <p><tt class="literal"><span class="pre">register_class_from_python</span></tt> is used to register a from-python converter for <tt class="literal"><span class="pre">shared_ptr<boring></span></tt>. <tt class="literal"><span class="pre">boost::shared_ptr</span></tt>s are special among smart pointers because their Deleter argument can be made to manage the whole Python object, not just the C++ object it contains, no matter how the C++ object is held.</p> <p>If there were any <tt class="literal"><span class="pre">bases<></span></tt>, we'd also be registering the relationship between these base classes and boring in the up/down cast graph (<tt class="literal"><span class="pre">inheritance.[hpp/cpp]</span></tt>).</p> <p>In earlier versions of the code, we'd be registering lvalue from-python converters for the class here, but now from-python conversion for wrapped classes is handled as a special case, before consulting the registry, if the source Python object's metaclass is the Boost.Python metaclass.</p> <p>Hmm, that from-python converter probably ought to be handled the way class converters are, with no explicit conversions registered.</p> </blockquote> </blockquote> <ol class="arabic" start="2"> <li><p class="first">Can you give a brief overview of the data structures that are present in the registry</p> <blockquote> <p>The registry is simple: it's just a map from typeid -> registration (see boost/python/converter/registrations.hpp). <tt class="literal"><span class="pre">lvalue_chain</span></tt> and <tt class="literal"><span class="pre">rvalue_chain</span></tt> are simple endogenous linked lists.</p> <p>If you want to know more, just ask.</p> <p>If you want to know about the cast graph, ask me something specific in a separate message.</p> </blockquote> <p>and an overview of the process that happens as a type makes its way from c++ to python and back again.</p> </li> </ol> <blockquote> <p>Big subject. I suggest some background reading: look for relevant info in the LLNL progress reports and the messages they link to. Also,</p> <blockquote> <p><a class="reference" href="http://mail.python.org/pipermail/c++-sig/2002-May/001023.html">http://mail.python.org/pipermail/c++-sig/2002-May/001023.html</a></p> <p><a class="reference" href="http://mail.python.org/pipermail/c++-sig/2002-December/003115.html">http://mail.python.org/pipermail/c++-sig/2002-December/003115.html</a></p> <p><a class="reference" href="http://aspn.activestate.com/ASPN/Mail/Message/1280898">http://aspn.activestate.com/ASPN/Mail/Message/1280898</a></p> <p><a class="reference" href="http://mail.python.org/pipermail/c++-sig/2002-July/001755.html">http://mail.python.org/pipermail/c++-sig/2002-July/001755.html</a></p> </blockquote> <p>from c++ to python:</p> <blockquote> <p>It depends on the type and the call policies in use or, for <tt class="literal"><span class="pre">call<>(...)</span></tt>, <tt class="literal"><span class="pre">call_method<>(...)</span></tt>, or <tt class="literal"><span class="pre">object(...)</span></tt>, if <tt class="literal"><span class="pre">ref</span></tt> or <tt class="literal"><span class="pre">ptr</span></tt> is used. There are also two basic categories to to-python conversion, "return value" conversion (for Python->C++ calls) and "argument" conversion (for C++->Python calls and explicit <tt class="literal"><span class="pre">object()</span></tt> conversions). The behavior of these two categories differs subtly in various ways whose details I forget at the moment. You can probably find the answers in the above references, and certainly in the code.</p> <p>The "default" case is by-value (copying) conversion, which uses to_python_value as a to-python converter.</p> <blockquote> <p>Since there can sensibly be only one way to convert any type to python (disregarding the idea of scoped registries for the moment), it makes sense that to-python conversions can be handled by specializing a template. If the type is one of the types handled by a built-in conversion (builtin_converters.hpp), the corresponding template specialization of to_python_value gets used.</p> <p>Otherwise, to_python_value uses the <tt class="literal"><span class="pre">m_to_python</span></tt> function in the registration for the C++ type.</p> </blockquote> <p>Other conversions, like by-reference conversions, are only available for wrapped classes, and are requested explicitly by using <tt class="literal"><span class="pre">ref(...)</span></tt>, <tt class="literal"><span class="pre">ptr(...)</span></tt>, or by specifying different CallPolicies for a call, which can cause a different to-python converter to be used. These conversions are never registered anywhere, though they do need to use the registration to find the Python class corresponding to the C++ type being referred to. They just build a new Python instance and stick the appropriate Holder instance in it.</p> </blockquote> <p>from python to C++:</p> <blockquote> <p>Once again I think there is a distinction between "return value" and "argument" conversions, and I forget exactly what that is.</p> <p>What happens depends on whether an lvalue conversion is needed (see <a class="reference" href="http://mail.python.org/pipermail/c++-sig/2002-May/001023.html">http://mail.python.org/pipermail/c++-sig/2002-May/001023.html</a>) All lvalue conversions are also registered in a type's rvalue conversion chain, since when an rvalue will do, an lvalue is certainly good enough.</p> <p>An lvalue conversion can be done in one step (just get me the pointer to the object - it can be <tt class="literal"><span class="pre">NULL</span></tt> if no conversion is possible) while an rvalue conversion requires two steps to support wrapped function overloading and multiple converters for a given C++ target type: first tell me if a conversion is possible, then construct the converted object as a second step.</p> </blockquote> </blockquote> </div> </div> <hr class="footer"/> <div class="footer"> <a class="reference" href="internals.rst">View document source</a>. Generated on: 2003-09-12 14:51 UTC. Generated by <a class="reference" href="http://docutils.sourceforge.net/">Docutils</a> from <a class="reference" href="http://docutils.sourceforge.net/rst.html">reStructuredText</a> source. </div> </body> </html>