3rdparty/boost_1_81_0/libs/python/doc/internals.html

<?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_&lt;boring&gt;(&quot;boring&quot;)
    ;
</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 &quot;off-the-top-of-my-head&quot; 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">
&gt;&gt;&gt; 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&lt;boring,</span> <span class="pre">non_copyable,</span>
<span class="pre">...&gt;</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&lt;boring</span> <span class="pre">...&gt;(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&lt;boring&gt;</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&lt;&gt;</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 -&gt;
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&lt;&gt;(...)</span></tt>, <tt class="literal"><span class="pre">call_method&lt;&gt;(...)</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, &quot;return value&quot; conversion
(for Python-&gt;C++ calls) and &quot;argument&quot; conversion (for
C++-&gt;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 &quot;default&quot; 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 &quot;return value&quot;
and &quot;argument&quot; 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>