2
0
mirror of https://github.com/boostorg/python.git synced 2026-01-21 17:12:22 +00:00
Files
python/special.html
2000-11-22 18:15:05 +00:00

817 lines
30 KiB
HTML
Raw Blame History

This file contains invisible Unicode characters
This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.0//EN"
"http://www.w3.org/TR/REC-html40/strict.dtd">
<title>
Special Method and Operator Support
</title>
<div>
<h1>
<img width="277" height="86" id="_x0000_i1025" align="center" src=
"c++boost.gif" alt="c++boost.gif (8819 bytes)">Special Method and
Operator Support
</h1>
<h2>
Overview
</h2>
<p>
Py_cpp is able to wrap suitable C++ functions and C++ operators into
Python operators. It supports all of the standard <a href=
"http://www.pythonlabs.com/pub/www.python.org/doc/current/ref/specialnames.html">
special method names</a> supported by real Python class instances <em>
except</em> <code>__complex__</code> (more on the reasons <a href=
"#reasons">below</a>). Supported operators include <a href="#general">
general</a>, <a href="#numeric">numeric</a>, and <a href=
"#sequence_and_mapping">sequence and mapping</a> operators. In
addition, py_cpp provides a simple way to export member variables and
define attributes by means of <a href="#getter_setter">getters and
setters</a>.
<h2>
<a name="general">General Operators</a>
</h2>
Python provides a number of special operatos for basic customization of a
class:
<dl>
<dt>
<b><tt class='method'>__repr__:</tt></b>
<dd>
create a string representation from which the object can be
reconstructed
<dt>
<b><tt class='method'>__str__:</tt></b>
<dd>
create a string representation which is suitable for printing
<dt>
<b><tt class='method'>__cmp__:</tt></b>
<dd>
three-way compare function, used to implement comparison operators
(&lt; etc.)
<dt>
<b><tt class='method'>__hash__:</tt></b>
<dd>
needed to use the object as a dictionary key (only allowed if __cmp__
is also defined)
<dt>
<b><tt class='method'>__nonzero__:</tt></b>
<dd>
called if the object is used as a truth value (e.g. in an if
statement)
<dt>
<b><tt class='method'>__call__:</tt></b>
<dd>
make instances of the class callable like a function
</dl>
If we have a suitable C++ function that supports any of these features,
we can export it like any other function, using its Python special name.
For example, suppose that class <code>Foo</code> provides a string
conversion function:
<pre>
std::string to_string(Foo const &amp; f)
{
std::ostringstream s;
s &lt;&lt; f;
return s.str();
}
</pre>
This function would be wrapped like this:
<pre>
python::class_builder&lt;Foo&gt; foo_class(my_module, "Foo");
foo_class.def(&amp;to_string, "__str__");
</pre>
Note that py_cpp also supports <em>automatic wrapping</em> of
"__str__" and "__cmp__". This is explained in the <a href="#numeric">next
section</a> and the <a href="#numeric_table">table of numeric
operators</a>.
<h2>
<a name="numeric">Numeric Operators</a>
</h2>
There are two fundamental ways to define numeric operators within py_cpp:
manual wrapping (as is done with <a href="#general">general
operators</a>) and automatic wrapping. Lets start with the second
possibility. Suppose, C++ defines a class <code>Int</code> (which might
represent an infinite-precision integer) which supports addition, so that
we can write (in C++):
<pre>
Int a, b, c;
...
c = a + b;
</pre>
To enable the same functionality in Python, we first wrap the <code>
Int</code> class as usual:
<pre>
python::class_builder&lt;Int&gt; int_class(my_module, "Int");
int_class.def(python::constructor&lt;&gt;());
...
</pre>
Then we export the addition operator like this:
<pre>
int_class.def(python::operators&lt;python::op_add&gt;());
</pre>
Since Int also supports subtraction, multiplication, adn division, we
want to export those also. This can be done in a single command by
'or'ing the operator identifiers together (a complete list of these
identifiers and the corresponding operators can be found in the <a href=
"#numeric_table">table</a>):
<pre>
int_class.def(python::operators&lt;(python::op_sub | python::op_mul | python::op_div)&gt;());
</pre>
Note that the or-expression must be enclosed in parentheses. This form of
operator definition will wrap homogeneous operators, i.e. operators whose
left and right operand have the same type. Now, suppose that our C++
library also supports addition of Ints and plain integers:
<pre>
Int a, b;
int i;
...
a = b + i;
a = i + b;
</pre>
To wrap these heterogeneous operators (left and right hand side have
different types), we need a possibility to specify a different type for
one of the operands. This is done using the <code>right_operand</code>
and <code>left_operand</code> templates:
<pre>
int_class.def(python::operators&lt;python::op_add&gt;(), python::right_operand&lt;int&gt;());
int_class.def(python::operators&lt;python::op_add&gt;(), python::left_operand&lt;int&gt;());
</pre>
Py_cpp uses overloading to register several variants of the same
operation (more on this in the context of <a href="#coercion">
coercion</a>). Again, several operators can be exported at once:
<pre>
int_class.def(python::operators&lt;(python::op_sub | python::op_mul | python::op_div)&gt;(),
python::right_operand&lt;int&gt;());
int_class.def(python::operators&lt;(python::op_sub | python::op_mul | python::op_div)&gt;(),
python::left_operand&lt;int&gt;());
</pre>
The type of the operand not mentioned is taken from the class object. In
our example, the class object is <code>int_class</code>, and thus the
other operand's type is `<code>Int const &amp;</code>'. You can override
this default by explicitly specifying a type in the <code>
operators</code> template:
<pre>
int_class.def(python::operators&lt;python::op_add, Int&gt;(), python::right_operand&lt;int&gt;());
</pre>
Here, `<code>Int</code>' would be used instead of `<code>Int const
&amp;</code>'.
<p>
Note that automatic wrapping doesn't need any specific form of <code>
operator+()</code> (or one of the other operators), but rather wraps
the <em>expression</em> `<code>left + right</code>'. That is, this
mechanism can be used for any definition of <code>operator+()</code>,
such as a free function `<code>Int operator+(Int, Int)</code>' or a
member function `<code>Int Int::operator+(Int)</code>'.
<p>
For the Python operators <code>pow()</code> and <code>abs()</code>,
there is no corresponding C++ operator. Instead, automatic wrapping
attempts to wrap C++ functions of the same name. This only works if
those functions are known in namespace <code>python::detail</code>.
Thus it might be necessary to add a using declaration prior to
wrapping:
<pre>
namespace python {
namespace detail {
using my_namespace::pow;
using my_namespace::abs;
}}
</pre>
<p>
In some cases, automatic wrapping of operators is not possible or not
desirable. Suppose, for example, that the modulo operation for Ints is
defined by a set of functions <code>mod()</code> (for automatic
wrapping, we would need <code>operator%()</code>):
<pre>
Int mod(Int const &amp; left, Int const &amp; right);
Int mod(Int const &amp; left, int right);
Int mod(int left, Int const &amp; right);
</pre>
In order to create the Python operator "__mod__" from these functions, we
have to wrap them manually:
<pre>
int_class.def((Int (*)(Int const &amp;, Int const &amp;))&amp;mod, "__mod__");
int_class.def((Int (*)(Int const &amp;, int))&amp;mod, "__mod__");
</pre>
The third form (with <code>int</code> as left operand) cannot be wrapped
this way. We must first create a function <code>rmod()</code> with the
operands reversed:
<pre>
Int rmod(Int const &amp; right, int left)
{
return mod(left, right);
}
</pre>
This function must be wrapped under the name "__rmod__":
<pre>
int_class.def(&amp;rmod, "__rmod__");
</pre>
A list of the possible operator names is also found in the <a href=
"#numeric_table">table</a>. Special treatment is necessary to export the
<a href="#ternary_pow">ternary pow</a> operator.
<p>
Automatic and manual wrapping can be mixed arbitrarily. Note that you
cannot overload the same operator for a given extension class on both
`<code>int</code>' and `<code>float</code>', because Python implicitly
converts these types into each other. Thus, the overloaded variant
found first (be it `<code>int</code>' or `<code>float</code>') will be
used for either of the two types.
<h4>
<a name="coercion">Coercion</a>
</h4>
Plain Python can only execute operators with identical types on the left
and right hand side. If it encounters an expression where the types of
the left and right operand differ, it tries to coerce these type to a
common type before invoking the actual operator. Implementing good
coercion functions can be difficult if many type combinations must be
supported.
<p>
In contrast, py_cpp provides <em><a href="overloading.html">
overloading</a></em>. By means of overloading, operator calling can be
simplyfied drastically: you just register operators for all desired
type combinations, and py_cpp automatically ensures that the correct
function is called in each case. User defined coercion functions are
<em>not necessary</em>. To enable operator overloading, py_cpp provides
a standard coercion which is <em>implicitly registered</em> whenever
automatic operator wrapping is used.
<p>
If you wrap all operator functions manually, but still want to use
operator overloading, you have to register the standard coercion
function explicitly:
<pre>
// this is not necessary if automatic operator wrapping is used
int_class.def_standard_coerce();
</pre>
In case you encounter a situation where you absolutely need a customized
coercion, you can overload the "__coerce__" operator itself. The
signature of a coercion function must look like this:
<pre>
python::tuple custom_coerce(PyObject * left, PyObject * right);
</pre>
The resulting <code>tuple</code> must contain two elements which
represent the values of <code>left</code> and <code>right</code>
converted to the same type. Such a function is wrapped as usual:
<pre>
some_class.def(&amp;custom_coerce, "__coerce__");
</pre>
Note that the custom coercion function is only used if it is defined <em>
before</em> any automatic operator wrapping on the given class or a call
to `<code>some_class.def_standard_coerce()</code>'.
<h4>
<a name="ternary_pow">The Ternary <code>pow()</code> Operator</a>
</h4>
In addition to the usual binary <code>pow()</code>-operator (meaning
<code>x^y</code>), Python also provides a ternary variant that implements
<code>(x^y) % z</code> (presumably using a more efficient algorithm than
concatenation of power and modulo operators). Automatic operator wrapping
can only be used with the binary variant. Ternary <code>pow()</code> must
always be wrapped manually. For a homgeneous ternary <code>pow()</code>,
this is done as usual:
<pre>
Int power(Int const &amp; first, Int const &amp; second, Int const &amp; module);
typedef Int (ternary_function1)(const Int&amp;, const Int&amp;, const Int&amp;);
...
int_class.def((ternary_function1)&amp;power, "__pow__");
</pre>
In case you want to support this function with non-uniform argument
types, wrapping is a little more involved. Suppose, you have to wrap:
<pre>
Int power(Int const &amp; first, int second, int module);
Int power(int first, Int const &amp; second, int module);
Int power(int first, int second, Int const &amp; module);
</pre>
The first variant can be wrapped as usual:
<pre>
typedef Int (ternary_function2)(const Int&amp;, int, int);
int_class.def((ternary_function2)&amp;power, "__pow__");
</pre>
In the second variant, however, <code>Int</code> appears only as second
argument, and in the last one it is the third argument. Therefor we must
first provide functions where the argumant order is changed so that
<code>Int</code> appears in first place:
<pre>
Int rpower(Int const &amp; second, int first, int module)
{
return power(first, second, third);
}
Int rrpower(Int const &amp; third, int first, int second)
{
return power(first, second, third);
}
</pre>
These functions must be wrapped under the names "__rpow__" and
"__rrpow__" respectively:
<pre>
int_class.def((ternary_function2)&amp;rpower, "__rpow__");
int_class.def((ternary_function2)&amp;rrpower, "__rrpow__");
</pre>
Note that "__rrpow__" is an extension not present in plain Python.
<h4>
<a name="numeric_table">Table of Numeric Operators</a>
</h4>
<p>
Py_cpp supports the <a href=
"http://www.pythonlabs.com/pub/www.python.org/doc/current/ref/numeric-types.html">
Python operators</a> listed in the following table. Note that
comparison (__cmp__) and string conversion (__str__) operators are
included in the list, although they are not strictly "numeric".
<p>
<table summary="special numeric methods" cellpadding="5" border="1"
width="100%">
<tr>
<td align="center">
<b>Python Operator Name</b>
<td align="center">
<b>Python Expression</b>
<td align="center">
<b>C++ Operator Id</b>
<td align="center">
<b>C++ Expression Used For Automatic Wrapping</b><br>
with <code>cpp_left = from_python(left,
type&lt;Left&gt;())</code>,<br>
<code>cpp_right = from_python(right,
type&lt;Right&gt;())</code>,<br>
and <code>cpp_oper = from_python(oper, type&lt;Oper&gt;())</code>
<tr>
<td>
<code>__add__, __radd__</code>
<td>
<code>left + right</code>
<td>
<code>python::op_add</code>
<td>
<code>cpp_left + cpp_right</code>
<tr>
<td>
<code>__sub__, __rsub__</code>
<td>
<code>left - right</code>
<td>
<code>python::op_sub</code>
<td>
<code>cpp_left - cpp_right</code>
<tr>
<td>
<code>__mul__, __rmul__</code>
<td>
<code>left * right</code>
<td>
<code>python::op_mul</code>
<td>
<code>cpp_left * cpp_right</code>
<tr>
<td>
<code>__div__, __rdiv__</code>
<td>
<code>left / right</code>
<td>
<code>python::op_div</code>
<td>
<code>cpp_left / cpp_right</code>
<tr>
<td>
<code>__mod__, __rmod__</code>
<td>
<code>left % right</code>
<td>
<code>python::op_mod</code>
<td>
<code>cpp_left % cpp_right</code>
<tr>
<td>
<code>__divmod__, __rdivmod__</code>
<td>
<code>(quotient, remainder)<br>
= divmod(left, right)</code>
<td>
<code>python::op_divmod</code>
<td>
<code>cpp_left / cpp_right </code> and <code> cpp_left %
cpp_right</code>
<tr>
<td>
<code>__pow__, __rpow__</code>
<td>
<code>pow(left, right)</code><br>
(binary power)
<td>
<code>python::op_pow</code>
<td>
<code>pow(cpp_left, cpp_right)</code>
<tr>
<td>
<code>__pow__</code>
<td>
<code>pow(left, right, modulo)</code><br>
(ternary power modulo)
<td colspan="2">
no automatic wrapping, <a href="#ternary_pow">special treatment</a>
required
<tr>
<td>
<code>__lshift__, __rlshift__</code>
<td>
<code>left &lt;&lt; right</code>
<td>
<code>python::op_lshift</code>
<td>
<code>cpp_left &lt;&lt; cpp_right</code>
<tr>
<td>
<code>__rshift__, __rrshift__</code>
<td>
<code>left &gt;&gt; right</code>
<td>
<code>python::op_rshift</code>
<td>
<code>cpp_left &gt;&gt; cpp_right</code>
<tr>
<td>
<code>__and__, __rand__</code>
<td>
<code>left &amp; right</code>
<td>
<code>python::op_and</code>
<td>
<code>cpp_left &amp; cpp_right</code>
<tr>
<td>
<code>__xor__, __rxor__</code>
<td>
<code>left ^ right</code>
<td>
<code>python::op_xor</code>
<td>
<code>cpp_left ^ cpp_right</code>
<tr>
<td>
<code>__or__, __ror__</code>
<td>
<code>left | right</code>
<td>
<code>python::op_or</code>
<td>
<code>cpp_left | cpp_right</code>
<tr>
<td>
<code>__cmp__, __rcmp__</code>
<td>
<code>cmp(left, right)</code> (3-way compare)<br>
<code>left &lt; right</code><br>
<code>left &lt;= right</code><br>
<code>left &gt; right</code><br>
<code>left &gt;= right</code><br>
<code>left == right</code><br>
<code>left != right</code>
<td>
<code>python::op_cmp</code>
<td>
<code>cpp_left &lt; cpp_right </code> and <code> cpp_right &lt;
cpp_left</code>
<tr>
<td>
<code>__neg__</code>
<td>
<code>-oper </code> (unary negation)
<td>
<code>python::op_neg</code>
<td>
<code>-cpp_oper</code>
<tr>
<td>
<code>__pos__</code>
<td>
<code>+oper </code> (identity)
<td>
<code>python::op_pos</code>
<td>
<code>+cpp_oper</code>
<tr>
<td>
<code>__abs__</code>
<td>
<code>abs(oper) </code> (absolute value)
<td>
<code>python::op_abs</code>
<td>
<code>abs(cpp_oper)</code>
<tr>
<td>
<code>__invert__</code>
<td>
<code>~oper </code> (bitwise inversion)
<td>
<code>python::op_invert</code>
<td>
<code>~cpp_oper</code>
<tr>
<td>
<code>__int__</code>
<td>
<code>int(oper) </code> (integer conversion)
<td>
<code>python::op_int</code>
<td>
<code>long(cpp_oper)</code>
<tr>
<td>
<code>__long__</code>
<td>
<code>long(oper) </code><br>
(infinite precision integer conversion)
<td>
<code>python::op_long</code>
<td>
<code>PyLong_FromLong(cpp_oper)</code>
<tr>
<td>
<code>__float__</code>
<td>
<code>float(oper) </code> (float conversion)
<td>
<code>python::op_float</code>
<td>
<code>double(cpp_oper)</code>
<tr>
<td>
<code>__oct__</code>
<td>
<code>oct(oper) </code> (octal conversion)
<td colspan="2">
must be wrapped manually (wrapped function should return a string)
<tr>
<td>
<code>__hex__</code>
<td>
<code>hex(oper) </code> (hex conversion)
<td colspan="2">
must be wrapped manually (wrapped function should return a string)
<tr>
<td>
<code>__str__</code>
<td>
<code>str(oper) </code> (string conversion)
<td>
<code>python::op_str</code>
<td>
<code>std::ostringstream s; s &lt;&lt; oper;</code>
<tr>
<td>
<code>__coerce__</code>
<td>
<code>coerce(left, right)</code>
<td colspan="2">
usually defined automatically, otherwise <a href="#coercion">
special treatment</a> required
</table>
<h2>
<a name="sequence_and_mapping">Sequence and Mapping Operators</a>
</h2>
Sequence and mapping operators let wrapped objects behave in accordance
to Python's iteration and access protocols. These protocols differ
considerably from the ones found in C++. For example, Python's typically
iteration idiom looks like  "<code>for i in S:</code>" , while in C++ one
uses  "<code>for(iterator i = S.begin(); i != S.end(); ++i)</code>". One
could try to wrap C++ iterators in order to carry the C++ idiom into
Python. However, this does not work very well because (1) it leads to
non-uniform Python code (wrapped types must be used in a different way
than Python built-in types) and (2) iterators are often implemented as
plain C++ pointers which cannot be wrapped easily because py_cpp is
designed to handle objects only.
<p>
Thus, it is a good idea to provide sequence and mapping operators for
your wrapped containers. These operators have to be wrapped manually
because there are no corresponding C++ operators that could be used for
automatic wrapping. The Python documentation lists the relevant <a
href=
"http://www.pythonlabs.com/pub/www.python.org/doc/current/ref/sequence-types.html">
container operators</a>. In particular, expose __getitem__, __setitem__
and remember to throw the <code>PyExc_IndexError</code> when the index
is out-of-range in order to enable the  "<code>for i in S:</code>" 
idiom.
<p>
Here is an example. Suppose, we want to wrap a <code>
std::map&lt;std::size_t,std::string&gt;</code>. This is done as follows
as follows:
<blockquote>
<pre>
typedef std::map&lt;std::size_t, std::string&gt; StringMap;
// A helper function for dealing with errors. Throw a Python exception
// if p == m.end().
void throw_key_error_if_end(
const StringMap&amp; m,
StringMap::const_iterator p,
std::size_t key)
{
if (p == m.end())
{
PyErr_SetObject(PyExc_KeyError, python::converters::to_python(key));
throw python::error_already_set();
}
}
// Define some simple wrapper functions which match the Python protocol
// for __getitem__, __setitem__, and __delitem__. Just as in Python, a
// free function with a "self" first parameter makes a fine class method.
const std::string&amp; get_item(const StringMap&amp; self, std::size_t key)
{
const StringMap::const_iterator p = self.find(key);
throw_key_error_if_end(self, p, key);
return p-&gt;second;
}
// Sets the item corresponding to key in the map.
void StringMapPythonClass::set_item(StringMap&amp; self, std::size_t key, const std::string&amp; value)
{
self[key] = value;
}
// Deletes the item corresponding to key from the map.
void StringMapPythonClass::del_item(StringMap&amp; self, std::size_t key)
{
const StringMap::iterator p = self.find(key);
throw_key_error_if_end(self, p, key);
self.erase(p);
}
class_builder&lt;StringMap&gt; string_map(my_module, "StringMap");
string_map.def(python::constructor&lt;&gt;());
string_map.def(&amp;StringMap::size, "__len__");
string_map.def(get_item, "__getitem__");
string_map.def(set_item, "__setitem__");
string_map.def(del_item, "__delitem__");
</pre>
</blockquote>
<p>
Then in Python:
<blockquote>
<pre>
&gt;&gt;&gt; m = StringMap()
&gt;&gt;&gt; m[1]
Traceback (innermost last):
File "&lt;stdin&gt;", line 1, in ?
KeyError: 1
&gt;&gt;&gt; m[1] = 'hello'
&gt;&gt;&gt; m[1]
'hello'
&gt;&gt;&gt; del m[1]
&gt;&gt;&gt; m[1] # prove that it's gone
Traceback (innermost last):
File "&lt;stdin&gt;", line 1, in ?
KeyError: 1
&gt;&gt;&gt; del m[2]
Traceback (innermost last):
File "&lt;stdin&gt;", line 1, in ?
KeyError: 2
&gt;&gt;&gt; len(m)
0
&gt;&gt;&gt; m[0] = 'zero'
&gt;&gt;&gt; m[1] = 'one'
&gt;&gt;&gt; m[2] = 'two'
&gt;&gt;&gt; m[3] = 'three'
&gt;&gt;&gt; len(m)
4
&gt;&gt;&gt; for i in m:
... print i
...
zero
one
two
three
</pre>
</blockquote>
<h2>
<a name="getter_setter">Getters and Setters</a>
</h2>
<p>
Py_cpp extension classes support some additional "special method"
protocols not supported by built-in Python classes. Because writing
<code>__getattr__</code>, <code> __setattr__</code>, and <code>
__delattr__</code> functions can be tedious in the common case where
the attributes being accessed are known statically, py_cpp checks the
special names
<ul>
<li>
<code>__getattr__<em>&lt;name&gt;</em>__</code>
<li>
<code>__setattr__<em>&lt;name&gt;</em>__</code>
<li>
<code>__delattr__<em>&lt;name&gt;</em>__</code>
</ul>
to provide functional access to the attribute <em>&lt;name&gt;</em>. This
facility can be used from C++ or entirely from Python. For example, the
following shows how we can implement a "computed attribute" in Python:
<blockquote>
<pre>
&gt;&gt;&gt; class Range(AnyPy_cppExtensionClass):
... def __init__(self, start, end):
... self.start = start
... self.end = end
... def __getattr__length__(self):
... return self.end - self.start
...
&gt;&gt;&gt; x = Range(3, 9)
&gt;&gt;&gt; x.length
6
</pre>
</blockquote>
<h4>
Direct Access to Data Members
</h4>
<p>
Py_cpp uses the special <code>
__xxxattr__<em>&lt;name&gt;</em>__</code> functionality described above
to allow direct access to data members through the following special
functions on <code>class_builder&lt;&gt;</code> and <code>
extension_class&lt;&gt;</code>:
<ul>
<li>
<code>def_getter(<em>pointer-to-member</em>, <em>name</em>)</code> //
read access to the member via attribute <em>name</em>
<li>
<code>def_setter(<em>pointer-to-member</em>, <em>name</em>)</code> //
write access to the member via attribute <em>name</em>
<li>
<code>def_readonly(<em>pointer-to-member</em>, <em>name</em>)</code>
// read-only access to the member via attribute <em>name</em>
<li>
<code>def_read_write(<em>pointer-to-member</em>, <em>
name</em>)</code> // read/write access to the member via attribute
<em>name</em>
</ul>
<p>
Note that the first two functions, used alone, may produce surprising
behavior. For example, when <code>def_getter()</code> is used, the
default functionality for <code>setattr()</code> and <code>
delattr()</code> remains in effect, operating on items in the extension
instance's name-space (i.e., its <code>__dict__</code>). For that
reason, you'll usually want to stick with <code>def_readonly</code> and
<code>def_read_write</code>.
<p>
For example, to expose a <code>std::pair&lt;int,long&gt;</code> we
might write:
<blockquote>
<pre>
typedef std::pair&lt;int,long&gt; Pil;
int first(const Pil&amp; x) { return x.first; }
long second(const Pil&amp; x) { return x.second; }
...
my_module.def(first, "first");
my_module.def(second, "second");
class_builder&lt;Pil&gt; pair_int_long(my_module, "Pair");
pair_int_long.def(python::constructor&lt;&gt;());
pair_int_long.def(python::constructor&lt;int,long&gt;());
pair_int_long.def_read_write(&amp;Pil::first, "first");
pair_int_long.def_read_write(&amp;Pil::second, "second");
</pre>
</blockquote>
<p>
Now your Python class has attributes <code>first</code> and <code>
second</code> which, when accessed, actually modify or reflect the
values of corresponding data members of the underlying C++ object. Now
in Python:
<blockquote>
<pre>
&gt;&gt;&gt; x = Pair(3,5)
&gt;&gt;&gt; x.first
3
&gt;&gt;&gt; x.second
5
&gt;&gt;&gt; x.second = 8
&gt;&gt;&gt; x.second
8
&gt;&gt;&gt; second(x) # Prove that we're not just changing the instance __dict__
8
</pre>
</blockquote>
<h2>
<a name="reasons">And what about <code>__complex__</code>?</a>
</h2>
<p>
That, dear reader, is one problem we don't know how to solve. The
Python source contains the following fragment, indicating the
special-case code really is hardwired:
<blockquote>
<pre>
/* XXX Hack to support classes with __complex__ method */
if (PyInstance_Check(r)) { ...
</pre>
</blockquote>
<p>
Previous: <a href="inheritance.html">Inheritance</a> Next: <a href=
"under-the-hood.html">A Peek Under the Hood</a> Up: <a href=
"py_cpp.html">Top</a>
<p>
&copy; Copyright David Abrahams and Ullrich K&ouml;the 2000.
Permission to copy, use, modify, sell and distribute this document is
granted provided this copyright notice appears in all copies. This
document is provided "as is" without express or implied warranty, and
with no claim as to its suitability for any purpose.
<p>
Updated: Nov 21, 2000
</div>