| [29] | 1 | <?xml version="1.0" encoding="utf-8" ?> |
|---|
| 2 | <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> |
|---|
| 3 | <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en"> |
|---|
| 4 | <head> |
|---|
| 5 | <meta http-equiv="Content-Type" content="text/html; charset=utf-8" /> |
|---|
| 6 | <meta name="generator" content="Docutils 0.3.0: http://docutils.sourceforge.net/" /> |
|---|
| 7 | <title>Boost.Python Internals Boost</title> |
|---|
| 8 | <link rel="stylesheet" href="../../../rst.css" type="text/css" /> |
|---|
| 9 | </head> |
|---|
| 10 | <body> |
|---|
| 11 | <div class="document" id="boost-python-internals-logo"> |
|---|
| 12 | <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> |
|---|
| 13 | <div class="section" id="a-conversation-between-brett-calcott-and-david-abrahams"> |
|---|
| 14 | <h1><a name="a-conversation-between-brett-calcott-and-david-abrahams">A conversation between Brett Calcott and David Abrahams</a></h1> |
|---|
| 15 | <table class="field-list" frame="void" rules="none"> |
|---|
| 16 | <col class="field-name" /> |
|---|
| 17 | <col class="field-body" /> |
|---|
| 18 | <tbody valign="top"> |
|---|
| 19 | <tr class="field"><th class="field-name">copyright:</th><td class="field-body">Copyright David Abrahams and Brett Calcott 2003. See |
|---|
| 20 | accompanying <a class="reference" href="../../../LICENSE_1_0.txt">license</a> for terms of use.</td> |
|---|
| 21 | </tr> |
|---|
| 22 | </tbody> |
|---|
| 23 | </table> |
|---|
| 24 | <p>In both of these cases, I'm quite capable of reading code - but the |
|---|
| 25 | thing I don't get from scanning the source is a sense of the |
|---|
| 26 | architecture, both structurally, and temporally (er, I mean in what |
|---|
| 27 | order things go on).</p> |
|---|
| 28 | <ol class="arabic"> |
|---|
| 29 | <li><p class="first">What happens when you do the following:</p> |
|---|
| 30 | <pre class="literal-block"> |
|---|
| 31 | struct boring {}; |
|---|
| 32 | ...etc... |
|---|
| 33 | class_<boring>("boring") |
|---|
| 34 | ; |
|---|
| 35 | </pre> |
|---|
| 36 | </li> |
|---|
| 37 | </ol> |
|---|
| 38 | <p>There seems to be a fair bit going on.</p> |
|---|
| 39 | <blockquote> |
|---|
| 40 | <ul class="simple"> |
|---|
| 41 | <li>Python needs a new ClassType to be registered.</li> |
|---|
| 42 | <li>We need to construct a new type that can hold our boring struct.</li> |
|---|
| 43 | <li>Inward and outward converters need to be registered for the type.</li> |
|---|
| 44 | </ul> |
|---|
| 45 | </blockquote> |
|---|
| 46 | <p>Can you gesture in the general direction where these things are done?</p> |
|---|
| 47 | <blockquote> |
|---|
| 48 | <p>I only have time for a "off-the-top-of-my-head" answer at the moment; |
|---|
| 49 | I suggest you step through the code with a debugger after reading this |
|---|
| 50 | to see how it works, fill in details, and make sure I didn't forget |
|---|
| 51 | anything.</p> |
|---|
| 52 | <blockquote> |
|---|
| 53 | <p>A new (Python) subclass of Boost.Python.Instance (see |
|---|
| 54 | libs/python/src/object/class.cpp) is created by invoking |
|---|
| 55 | Boost.Python.class, the metatype:</p> |
|---|
| 56 | <pre class="literal-block"> |
|---|
| 57 | >>> boring = Boost.Python.class( |
|---|
| 58 | ... 'boring' |
|---|
| 59 | ... , bases_tuple # in this case, just () |
|---|
| 60 | ... , { |
|---|
| 61 | ... '__module__' : module_name |
|---|
| 62 | ... , '__doc__' : doc_string # optional |
|---|
| 63 | ... } |
|---|
| 64 | ... ) |
|---|
| 65 | </pre> |
|---|
| 66 | <p>A handle to this object is stuck in the m_class_object field |
|---|
| 67 | of the registration associated with <tt class="literal"><span class="pre">typeid(boring)</span></tt>. The |
|---|
| 68 | registry will keep that object alive forever, even if you |
|---|
| 69 | wipe out the 'boring' attribute of the extension module |
|---|
| 70 | (probably not a good thing).</p> |
|---|
| 71 | <p>Because you didn't specify <tt class="literal"><span class="pre">class<boring,</span> <span class="pre">non_copyable,</span> |
|---|
| 72 | <span class="pre">...></span></tt>, a to-python converter for boring is registered which |
|---|
| 73 | copies its argument into a value_holder held by the the |
|---|
| 74 | Python boring object.</p> |
|---|
| 75 | <p>Because you didn't specify <tt class="literal"><span class="pre">class<boring</span> <span class="pre">...>(no_init)</span></tt>, |
|---|
| 76 | an <tt class="literal"><span class="pre">__init__</span></tt> function object is added to the class |
|---|
| 77 | dictionary which default-constructs a boring in a |
|---|
| 78 | value_holder (because you didn't specify some smart pointer |
|---|
| 79 | or derived wrapper class as a holder) held by the Python |
|---|
| 80 | boring object.</p> |
|---|
| 81 | <p><tt class="literal"><span class="pre">register_class_from_python</span></tt> is used to register a |
|---|
| 82 | from-python converter for <tt class="literal"><span class="pre">shared_ptr<boring></span></tt>. |
|---|
| 83 | <tt class="literal"><span class="pre">boost::shared_ptr</span></tt>s are special among smart pointers |
|---|
| 84 | because their Deleter argument can be made to manage the |
|---|
| 85 | whole Python object, not just the C++ object it contains, no |
|---|
| 86 | matter how the C++ object is held.</p> |
|---|
| 87 | <p>If there were any <tt class="literal"><span class="pre">bases<></span></tt>, we'd also be registering the |
|---|
| 88 | relationship between these base classes and boring in the |
|---|
| 89 | up/down cast graph (<tt class="literal"><span class="pre">inheritance.[hpp/cpp]</span></tt>).</p> |
|---|
| 90 | <p>In earlier versions of the code, we'd be registering lvalue |
|---|
| 91 | from-python converters for the class here, but now |
|---|
| 92 | from-python conversion for wrapped classes is handled as a |
|---|
| 93 | special case, before consulting the registry, if the source |
|---|
| 94 | Python object's metaclass is the Boost.Python metaclass.</p> |
|---|
| 95 | <p>Hmm, that from-python converter probably ought to be handled |
|---|
| 96 | the way class converters are, with no explicit conversions |
|---|
| 97 | registered.</p> |
|---|
| 98 | </blockquote> |
|---|
| 99 | </blockquote> |
|---|
| 100 | <ol class="arabic" start="2"> |
|---|
| 101 | <li><p class="first">Can you give a brief overview of the data structures that are |
|---|
| 102 | present in the registry</p> |
|---|
| 103 | <blockquote> |
|---|
| 104 | <p>The registry is simple: it's just a map from typeid -> |
|---|
| 105 | registration (see boost/python/converter/registrations.hpp). |
|---|
| 106 | <tt class="literal"><span class="pre">lvalue_chain</span></tt> and <tt class="literal"><span class="pre">rvalue_chain</span></tt> are simple endogenous |
|---|
| 107 | linked lists.</p> |
|---|
| 108 | <p>If you want to know more, just ask.</p> |
|---|
| 109 | <p>If you want to know about the cast graph, ask me something specific in |
|---|
| 110 | a separate message.</p> |
|---|
| 111 | </blockquote> |
|---|
| 112 | <p>and an overview of the process that happens as a type makes its |
|---|
| 113 | way from c++ to python and back again.</p> |
|---|
| 114 | </li> |
|---|
| 115 | </ol> |
|---|
| 116 | <blockquote> |
|---|
| 117 | <p>Big subject. I suggest some background reading: look for relevant |
|---|
| 118 | info in the LLNL progress reports and the messages they link to. |
|---|
| 119 | Also,</p> |
|---|
| 120 | <blockquote> |
|---|
| 121 | <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> |
|---|
| 122 | <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> |
|---|
| 123 | <p><a class="reference" href="http://aspn.activestate.com/ASPN/Mail/Message/1280898">http://aspn.activestate.com/ASPN/Mail/Message/1280898</a></p> |
|---|
| 124 | <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> |
|---|
| 125 | </blockquote> |
|---|
| 126 | <p>from c++ to python:</p> |
|---|
| 127 | <blockquote> |
|---|
| 128 | <p>It depends on the type and the call policies in use or, for |
|---|
| 129 | <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 |
|---|
| 130 | <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 |
|---|
| 131 | categories to to-python conversion, "return value" conversion |
|---|
| 132 | (for Python->C++ calls) and "argument" conversion (for |
|---|
| 133 | C++->Python calls and explicit <tt class="literal"><span class="pre">object()</span></tt> conversions). The |
|---|
| 134 | behavior of these two categories differs subtly in various ways |
|---|
| 135 | whose details I forget at the moment. You can probably find |
|---|
| 136 | the answers in the above references, and certainly in the code.</p> |
|---|
| 137 | <p>The "default" case is by-value (copying) conversion, which uses |
|---|
| 138 | to_python_value as a to-python converter.</p> |
|---|
| 139 | <blockquote> |
|---|
| 140 | <p>Since there can sensibly be only one way to convert any type |
|---|
| 141 | to python (disregarding the idea of scoped registries for the |
|---|
| 142 | moment), it makes sense that to-python conversions can be |
|---|
| 143 | handled by specializing a template. If the type is one of |
|---|
| 144 | the types handled by a built-in conversion |
|---|
| 145 | (builtin_converters.hpp), the corresponding template |
|---|
| 146 | specialization of to_python_value gets used.</p> |
|---|
| 147 | <p>Otherwise, to_python_value uses the <tt class="literal"><span class="pre">m_to_python</span></tt> |
|---|
| 148 | function in the registration for the C++ type.</p> |
|---|
| 149 | </blockquote> |
|---|
| 150 | <p>Other conversions, like by-reference conversions, are only |
|---|
| 151 | available for wrapped classes, and are requested explicitly by |
|---|
| 152 | using <tt class="literal"><span class="pre">ref(...)</span></tt>, <tt class="literal"><span class="pre">ptr(...)</span></tt>, or by specifying different |
|---|
| 153 | CallPolicies for a call, which can cause a different to-python |
|---|
| 154 | converter to be used. These conversions are never registered |
|---|
| 155 | anywhere, though they do need to use the registration to find |
|---|
| 156 | the Python class corresponding to the C++ type being referred |
|---|
| 157 | to. They just build a new Python instance and stick the |
|---|
| 158 | appropriate Holder instance in it.</p> |
|---|
| 159 | </blockquote> |
|---|
| 160 | <p>from python to C++:</p> |
|---|
| 161 | <blockquote> |
|---|
| 162 | <p>Once again I think there is a distinction between "return value" |
|---|
| 163 | and "argument" conversions, and I forget exactly what that is.</p> |
|---|
| 164 | <p>What happens depends on whether an lvalue conversion is needed |
|---|
| 165 | (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>) |
|---|
| 166 | All lvalue conversions are also registered in a type's rvalue |
|---|
| 167 | conversion chain, since when an rvalue will do, an lvalue is |
|---|
| 168 | certainly good enough.</p> |
|---|
| 169 | <p>An lvalue conversion can be done in one step (just get me the |
|---|
| 170 | pointer to the object - it can be <tt class="literal"><span class="pre">NULL</span></tt> if no conversion is |
|---|
| 171 | possible) while an rvalue conversion requires two steps to |
|---|
| 172 | support wrapped function overloading and multiple converters for |
|---|
| 173 | a given C++ target type: first tell me if a conversion is |
|---|
| 174 | possible, then construct the converted object as a second step.</p> |
|---|
| 175 | </blockquote> |
|---|
| 176 | </blockquote> |
|---|
| 177 | </div> |
|---|
| 178 | </div> |
|---|
| 179 | <hr class="footer"/> |
|---|
| 180 | <div class="footer"> |
|---|
| 181 | <a class="reference" href="internals.rst">View document source</a>. |
|---|
| 182 | Generated on: 2003-09-12 14:51 UTC. |
|---|
| 183 | 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. |
|---|
| 184 | </div> |
|---|
| 185 | </body> |
|---|
| 186 | </html> |
|---|
