123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186 |
- <?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>
|