Compare commits

..

3 Commits

Author SHA1 Message Date
Ralf W. Grosse-Kunstleve
7ea60a96f8 Join ralf_grosse_kunstleve with HEAD
[SVN r9444]
2001-03-05 20:01:01 +00:00
nobody
a9adec170a This commit was manufactured by cvs2svn to create branch
'unlabeled-1.2.2'.

[SVN r9442]
2001-03-05 12:03:56 +00:00
nobody
8808f55435 This commit was manufactured by cvs2svn to create branch
'unlabeled-1.2.2'.

[SVN r9019]
2001-02-07 23:58:46 +00:00
6 changed files with 1858 additions and 458 deletions

762
call_traits.htm Normal file
View File

@@ -0,0 +1,762 @@
<html>
<head>
<meta http-equiv="Content-Type"
content="text/html; charset=iso-8859-1">
<meta name="Template"
content="C:\PROGRAM FILES\MICROSOFT OFFICE\OFFICE\html.dot">
<meta name="GENERATOR" content="Microsoft FrontPage Express 2.0">
<title>Call Traits</title>
</head>
<body bgcolor="#FFFFFF" text="#000000" link="#0000FF"
vlink="#800080">
<h1><img src="../../c++boost.gif" width="276" height="86">Header
&lt;<a href="../../boost/detail/call_traits.hpp">boost/call_traits.hpp</a>&gt;</h1>
<p>All of the contents of &lt;boost/call_traits.hpp&gt; are
defined inside namespace boost.</p>
<p>The template class call_traits&lt;T&gt; encapsulates the
&quot;best&quot; method to pass a parameter of some type T to or
from a function, and consists of a collection of typedefs defined
as in the table below. The purpose of call_traits is to ensure
that problems like &quot;<a href="#refs">references to references</a>&quot;
never occur, and that parameters are passed in the most efficient
manner possible (see <a href="#examples">examples</a>). In each
case if your existing practice is to use the type defined on the
left, then replace it with the call_traits defined type on the
right. </p>
<p>Note that for compilers that do not support either partial
specialization or member templates, no benefit will occur from
using call_traits: the call_traits defined types will always be
the same as the existing practice in this case. In addition if
only member templates and not partial template specialisation is
support by the compiler (for example Visual C++ 6) then
call_traits can not be used with array types (although it can be
used to solve the reference to reference problem).</p>
<table border="0" cellpadding="7" cellspacing="1" width="797">
<tr>
<td valign="top" width="17%" bgcolor="#008080"><p
align="center">Existing practice</p>
</td>
<td valign="top" width="35%" bgcolor="#008080"><p
align="center">call_traits equivalent</p>
</td>
<td valign="top" width="32%" bgcolor="#008080"><p
align="center">Description</p>
</td>
<td valign="top" width="16%" bgcolor="#008080"><p
align="center">Notes</p>
</td>
</tr>
<tr>
<td valign="top" width="17%"><p align="center">T<br>
(return by value)</p>
</td>
<td valign="top" width="35%"><p align="center"><code>call_traits&lt;T&gt;::value_type</code></p>
</td>
<td valign="top" width="32%">Defines a type that
represents the &quot;value&quot; of type T. Use this for
functions that return by value, or possibly for stored
values of type T.</td>
<td valign="top" width="16%"><p align="center">2</p>
</td>
</tr>
<tr>
<td valign="top" width="17%"><p align="center">T&amp;<br>
(return value)</p>
</td>
<td valign="top" width="35%"><p align="center"><code>call_traits&lt;T&gt;::reference</code></p>
</td>
<td valign="top" width="32%">Defines a type that
represents a reference to type T. Use for functions that
would normally return a T&amp;.</td>
<td valign="top" width="16%"><p align="center">1</p>
</td>
</tr>
<tr>
<td valign="top" width="17%"><p align="center">const
T&amp;<br>
(return value)</p>
</td>
<td valign="top" width="35%"><p align="center"><code>call_traits&lt;T&gt;::const_reference</code></p>
</td>
<td valign="top" width="32%">Defines a type that
represents a constant reference to type T. Use for
functions that would normally return a const T&amp;.</td>
<td valign="top" width="16%"><p align="center">1</p>
</td>
</tr>
<tr>
<td valign="top" width="17%"><p align="center">const
T&amp;<br>
(function parameter)</p>
</td>
<td valign="top" width="35%"><p align="center"><code>call_traits&lt;T&gt;::param_type</code></p>
</td>
<td valign="top" width="32%">Defines a type that
represents the &quot;best&quot; way to pass a parameter
of type T to a function.</td>
<td valign="top" width="16%"><p align="center">1,3</p>
</td>
</tr>
</table>
<p>Notes:</p>
<ol>
<li>If T is already reference type, then call_traits is
defined such that <a href="#refs">references to
references</a> do not occur (requires partial
specialization).</li>
<li>If T is an array type, then call_traits defines <code>value_type</code>
as a &quot;constant pointer to type&quot; rather than an
&quot;array of type&quot; (requires partial
specialization). Note that if you are using value_type as
a stored value then this will result in storing a &quot;constant
pointer to an array&quot; rather than the array itself.
This may or may not be a good thing depending upon what
you actually need (in other words take care!).</li>
<li>If T is a small built in type or a pointer, then <code>param_type</code>
is defined as <code>T const</code>, instead of <code>T
const&amp;</code>. This can improve the ability of the
compiler to optimize loops in the body of the function if
they depend upon the passed parameter, the semantics of
the passed parameter is otherwise unchanged (requires
partial specialization).</li>
</ol>
<p>&nbsp;</p>
<h3>Copy constructibility</h3>
<p>The following table defines which call_traits types can always
be copy-constructed from which other types, those entries marked
with a '?' are true only if and only if T is copy constructible:</p>
<table border="0" cellpadding="7" cellspacing="1" width="766">
<tr>
<td valign="top" width="17%">&nbsp;</td>
<td valign="top" colspan="5" width="85%"
bgcolor="#008080"><p align="center">To:</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#008080">From:</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">T</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">value_type</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">reference</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">const_reference</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">param_type</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0">T</td>
<td valign="top" width="17%"><p align="center">?</p>
</td>
<td valign="top" width="17%"><p align="center">?</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0">value_type</td>
<td valign="top" width="17%"><p align="center">?</p>
</td>
<td valign="top" width="17%"><p align="center">?</p>
</td>
<td valign="top" width="17%"><p align="center">N</p>
</td>
<td valign="top" width="17%"><p align="center">N</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0">reference</td>
<td valign="top" width="17%"><p align="center">?</p>
</td>
<td valign="top" width="17%"><p align="center">?</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0">const_reference</td>
<td valign="top" width="17%"><p align="center">?</p>
</td>
<td valign="top" width="17%"><p align="center">N</p>
</td>
<td valign="top" width="17%"><p align="center">N</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0">param_type</td>
<td valign="top" width="17%"><p align="center">?</p>
</td>
<td valign="top" width="17%"><p align="center">?</p>
</td>
<td valign="top" width="17%"><p align="center">N</p>
</td>
<td valign="top" width="17%"><p align="center">N</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
</tr>
</table>
<p>&nbsp;</p>
<p>If T is an assignable type the following assignments are
possible:</p>
<table border="0" cellpadding="7" cellspacing="1" width="766">
<tr>
<td valign="top" width="17%">&nbsp;</td>
<td valign="top" colspan="5" width="85%"
bgcolor="#008080"><p align="center">To:</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#008080">From:</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">T</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">value_type</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">reference</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">const_reference</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">param_type</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0">T</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0">value_type</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0">reference</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0">const_reference</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0">param_type</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">Y</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
<td valign="top" width="17%"><p align="center">-</p>
</td>
</tr>
</table>
<p>&nbsp;</p>
<h3><a name="examples"></a>Examples</h3>
<p>The following table shows the effect that call_traits has on
various types, the table assumes that the compiler supports
partial specialization: if it doesn't then all types behave in
the same way as the entry for &quot;myclass&quot;, and
call_traits can not be used with reference or array types.</p>
<table border="0" cellpadding="7" cellspacing="1" width="766">
<tr>
<td valign="top" width="17%">&nbsp;</td>
<td valign="top" colspan="5" width="85%"
bgcolor="#008080"><p align="center">Call_traits type:</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#008080"><p
align="center">Original type T</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">value_type</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">reference</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">const_reference</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">param_type</p>
</td>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">Applies to:</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">myclass</p>
</td>
<td valign="top" width="17%"><p align="center">myclass</p>
</td>
<td valign="top" width="17%"><p align="center">myclass&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">const
myclass&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">myclass
const&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">All user
defined types.</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">int</p>
</td>
<td valign="top" width="17%"><p align="center">int</p>
</td>
<td valign="top" width="17%"><p align="center">int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">const
int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">int const</p>
</td>
<td valign="top" width="17%"><p align="center">All small
built-in types.</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">int*</p>
</td>
<td valign="top" width="17%"><p align="center">int*</p>
</td>
<td valign="top" width="17%"><p align="center">int*&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">int*const&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">int* const</p>
</td>
<td valign="top" width="17%"><p align="center">All
pointer types.</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">const
int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">All
reference types.</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">const int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">const
int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">const
int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">const
int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">const
int&amp;</p>
</td>
<td valign="top" width="17%"><p align="center">All
constant-references.</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">int[3]</p>
</td>
<td valign="top" width="17%"><p align="center">const int*</p>
</td>
<td valign="top" width="17%"><p align="center">int(&amp;)[3]</p>
</td>
<td valign="top" width="17%"><p align="center">const int(&amp;)[3]</p>
</td>
<td valign="top" width="17%"><p align="center">const int*
const</p>
</td>
<td valign="top" width="17%"><p align="center">All array
types.</p>
</td>
</tr>
<tr>
<td valign="top" width="17%" bgcolor="#C0C0C0"><p
align="center">const int[3]</p>
</td>
<td valign="top" width="17%"><p align="center">const int*</p>
</td>
<td valign="top" width="17%"><p align="center">const int(&amp;)[3]</p>
</td>
<td valign="top" width="17%"><p align="center">const int(&amp;)[3]</p>
</td>
<td valign="top" width="17%"><p align="center">const int*
const</p>
</td>
<td valign="top" width="17%"><p align="center">All
constant-array types.</p>
</td>
</tr>
</table>
<p>&nbsp;</p>
<h4>Example 1:</h4>
<p>The following class is a trivial class that stores some type T
by value (see the <a href="call_traits_test.cpp">call_traits_test.cpp</a>
file), the aim is to illustrate how each of the available
call_traits typedefs may be used:</p>
<pre>template &lt;class T&gt;
struct contained
{
// define our typedefs first, arrays are stored by value
// so value_type is not the same as result_type:
typedef typename boost::call_traits&lt;T&gt;::param_type param_type;
typedef typename boost::call_traits&lt;T&gt;::reference reference;
typedef typename boost::call_traits&lt;T&gt;::const_reference const_reference;
typedef T value_type;
typedef typename boost::call_traits&lt;T&gt;::value_type result_type;
// stored value:
value_type v_;
// constructors:
contained() {}
contained(param_type p) : v_(p){}
// return byval:
result_type value() { return v_; }
// return by_ref:
reference get() { return v_; }
const_reference const_get()const { return v_; }
// pass value:
void call(param_type p){}
};</pre>
<h4><a name="refs"></a>Example 2 (the reference to reference
problem):</h4>
<p>Consider the definition of std::binder1st:</p>
<pre>template &lt;class Operation&gt;
class binder1st :
public unary_function&lt;typename Operation::second_argument_type, typename Operation::result_type&gt;
{
protected:
Operation op;
typename Operation::first_argument_type value;
public:
binder1st(const Operation&amp; x, const typename Operation::first_argument_type&amp; y);
typename Operation::result_type operator()(const typename Operation::second_argument_type&amp; x) const;
}; </pre>
<p>Now consider what happens in the relatively common case that
the functor takes its second argument as a reference, that
implies that <code>Operation::second_argument_type</code> is a
reference type, <code>operator()</code> will now end up taking a
reference to a reference as an argument, and that is not
currently legal. The solution here is to modify <code>operator()</code>
to use call_traits:</p>
<pre>typename Operation::result_type operator()(typename call_traits&lt;typename Operation::second_argument_type&gt;::param_type x) const;</pre>
<p>Now in the case that <code>Operation::second_argument_type</code>
is a reference type, the argument is passed as a reference, and
the no &quot;reference to reference&quot; occurs.</p>
<h4><a name="ex3"></a>Example 3 (the make_pair problem):</h4>
<p>If we pass the name of an array as one (or both) arguments to <code>std::make_pair</code>,
then template argument deduction deduces the passed parameter as
&quot;const reference to array of T&quot;, this also applies to
string literals (which are really array literals). Consequently
instead of returning a pair of pointers, it tries to return a
pair of arrays, and since an array type is not copy-constructible
the code fails to compile. One solution is to explicitly cast the
arguments to make_pair to pointers, but call_traits provides a
better (i.e. automatic) solution (and one that works safely even
in generic code where the cast might do the wrong thing):</p>
<pre>template &lt;class T1, class T2&gt;
std::pair&lt;
typename boost::call_traits&lt;T1&gt;::value_type,
typename boost::call_traits&lt;T2&gt;::value_type&gt;
make_pair(const T1&amp; t1, const T2&amp; t2)
{
return std::pair&lt;
typename boost::call_traits&lt;T1&gt;::value_type,
typename boost::call_traits&lt;T2&gt;::value_type&gt;(t1, t2);
}</pre>
<p>Here, the deduced argument types will be automatically
degraded to pointers if the deduced types are arrays, similar
situations occur in the standard binders and adapters: in
principle in any function that &quot;wraps&quot; a temporary
whose type is deduced. Note that the function arguments to
make_pair are not expressed in terms of call_traits: doing so
would prevent template argument deduction from functioning.</p>
<h4><a name="ex4"></a>Example 4 (optimising fill):</h4>
<p>The call_traits template will &quot;optimize&quot; the passing
of a small built-in type as a function parameter, this mainly has
an effect when the parameter is used within a loop body. In the
following example (see <a href="algo_opt_examples.cpp">algo_opt_examples.cpp</a>),
a version of std::fill is optimized in two ways: if the type
passed is a single byte built-in type then std::memset is used to
effect the fill, otherwise a conventional C++ implemention is
used, but with the passed parameter &quot;optimized&quot; using
call_traits:</p>
<pre>namespace detail{
template &lt;bool opt&gt;
struct filler
{
template &lt;typename I, typename T&gt;
static void do_fill(I first, I last, typename boost::call_traits&lt;T&gt;::param_type val);
{
while(first != last)
{
*first = val;
++first;
}
}
};
template &lt;&gt;
struct filler&lt;true&gt;
{
template &lt;typename I, typename T&gt;
static void do_fill(I first, I last, T val)
{
memset(first, val, last-first);
}
};
}
template &lt;class I, class T&gt;
inline void fill(I first, I last, const T&amp; val)
{
enum{ can_opt = boost::is_pointer&lt;I&gt;::value
&amp;&amp; boost::is_arithmetic&lt;T&gt;::value
&amp;&amp; (sizeof(T) == 1) };
typedef detail::filler&lt;can_opt&gt; filler_t;
filler_t::template do_fill&lt;I,T&gt;(first, last, val);
}</pre>
<p>Footnote: the reason that this is &quot;optimal&quot; for
small built-in types is that with the value passed as &quot;T
const&quot; instead of &quot;const T&amp;&quot; the compiler is
able to tell both that the value is constant and that it is free
of aliases. With this information the compiler is able to cache
the passed value in a register, unroll the loop, or use
explicitly parallel instructions: if any of these are supported.
Exactly how much mileage you will get from this depends upon your
compiler - we could really use some accurate benchmarking
software as part of boost for cases like this.</p>
<p>Note that the function arguments to fill are not expressed in
terms of call_traits: doing so would prevent template argument
deduction from functioning. Instead fill acts as a &quot;thin
wrapper&quot; that is there to perform template argument
deduction, the compiler will optimise away the call to fill all
together, replacing it with the call to filler&lt;&gt;::do_fill,
which does use call_traits.</p>
<h3>Rationale</h3>
<p>The following notes are intended to briefly describe the
rational behind choices made in call_traits.</p>
<p>All user-defined types follow &quot;existing practice&quot;
and need no comment.</p>
<p>Small built-in types (what the standard calls fundamental
types [3.9.1]) differ from existing practice only in the <i>param_type</i>
typedef. In this case passing &quot;T const&quot; is compatible
with existing practice, but may improve performance in some cases
(see <a href="#ex4">Example 4</a>), in any case this should never
be any worse than existing practice.</p>
<p>Pointers follow the same rational as small built-in types.</p>
<p>For reference types the rational follows <a href="#refs">Example
2</a> - references to references are not allowed, so the
call_traits members must be defined such that these problems do
not occur. There is a proposal to modify the language such that
&quot;a reference to a reference is a reference&quot; (issue #106,
submitted by Bjarne Stroustrup), call_traits&lt;T&gt;::value_type
and call_traits&lt;T&gt;::param_type both provide the same effect
as that proposal, without the need for a language change (in
other words it's a workaround).</p>
<p>For array types, a function that takes an array as an argument
will degrade the array type to a pointer type: this means that
the type of the actual parameter is different from its declared
type, something that can cause endless problems in template code
that relies on the declared type of a parameter. For example:</p>
<pre>template &lt;class T&gt;
struct A
{
void foo(T t);
};</pre>
<p><font face="Times New Roman">In this case if we instantiate
A&lt;int[2]&gt; then the declared type of the parameter passed to
member function foo is int[2], but it's actual type is const int*,
if we try to use the type T within the function body, then there
is a strong likelyhood that our code will not compile:</font></p>
<pre>template &lt;class T&gt;
void A&lt;T&gt;::foo(T t)
{
T dup(t); // doesn't compile for case that T is an array.
}</pre>
<p>By using call_traits the degradation from array to pointer is
explicit, and the type of the parameter is the same as it's
declared type:</p>
<pre>template &lt;class T&gt;
struct A
{
void foo(typename call_traits&lt;T&gt;::value_type t);
};
template &lt;class T&gt;
void A&lt;T&gt;::foo(typename call_traits&lt;T&gt;::value_type t)
{
typename call_traits&lt;T&gt;::value_type dup(t); // OK even if T is an array type.
}</pre>
<p>For value_type (return by value), again only a pointer may be
returned, not a copy of the whole array, and again call_traits
makes the degradation explicit. The value_type member is useful
whenever an array must be explicitly degraded to a pointer - <a
href="#ex3">Example 3</a> provides the test case (Footnote: the
array specialisation for call_traits is the least well understood
of all the call_traits specialisations, if the given semantics
cause specific problems for you, or don't solve a particular
array-related problem, then I would be interested to hear about
it. Most people though will probably never need to use this
specialisation).</p>
<hr>
<p>Revised 01 September 2000</p>
<p><EFBFBD> Copyright boost.org 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
&quot;as is&quot; without express or implied warranty, and with
no claim as to its suitability for any purpose.</p>
<p>Based on contributions by Steve Cleary, Beman Dawes, Howard
Hinnant and John Maddock.</p>
<p>Maintained by <a href="mailto:John_Maddock@compuserve.com">John
Maddock</a>, the latest version of this file can be found at <a
href="http://www.boost.org/">www.boost.org</a>, and the boost
discussion list at <a href="http://www.egroups.com/list/boost">www.egroups.com/list/boost</a>.</p>
<p>.</p>
<p>&nbsp;</p>
<p>&nbsp;</p>
</body>
</html>

366
half_open_range_test.cpp Normal file
View File

@@ -0,0 +1,366 @@
// (C) Copyright David Abrahams 2001. Permission to copy, use, modify, sell and
// distribute this software is granted provided this copyright notice appears in
// all copies. This software is provided "as is" without express or implied
// warranty, and with no claim as to its suitability for any purpose.
//
// See http://www.boost.org for most recent version including documentation.
//
// Revision History
// 11 Feb 2001 Compile with Borland, re-enable failing tests (David Abrahams)
// 29 Jan 2001 Initial revision (David Abrahams)
#include <boost/half_open_range.hpp>
#include <boost/utility.hpp>
#include <iterator>
#include <stdlib.h>
#include <vector>
#include <list>
#include <cassert>
#include <stdexcept>
#ifndef BOOST_NO_LIMITS
# include <limits>
#endif
#ifndef BOOST_NO_SLIST
# include <slist>
#endif
inline unsigned unsigned_random(unsigned max)
{
return (max > 0) ? (unsigned)rand() % max : 0;
}
// Special tests for ranges supporting random access
template <class T>
void category_test_1(
const boost::half_open_range<T>& r, std::random_access_iterator_tag)
{
typedef boost::half_open_range<T> range;
typedef typename range::size_type size_type;
size_type size = r.size();
// pick a random offset
size_type offset = unsigned_random(size);
typename range::value_type x = *(r.begin() + offset);
// test contains(value_type)
assert(r.contains(r.start()) == !r.empty());
assert(!r.contains(r.finish()));
assert(r.contains(x) == (offset != size));
range::const_iterator p = r.find(x);
assert((p == r.end()) == (x == r.finish()));
assert(r.find(r.finish()) == r.end());
if (offset != size)
{
assert(x == r[offset]);
assert(x == r.at(offset));
}
bool caught_out_of_range = false;
try {
bool never_initialized = x == r.at(size);
(void)never_initialized;
}
catch(std::out_of_range&)
{
caught_out_of_range = true;
}
catch(...)
{
}
assert(caught_out_of_range);
}
// Those tests must be skipped for other ranges
template <class T>
void category_test_1(
const boost::half_open_range<T>&, std::forward_iterator_tag)
{
}
unsigned indices[][2] = { {0,0},{0,1},{0,2},{0,3},
{1,1},{1,2},{1,3},
{2,2},{2,3},
{3,3}};
template <class Range>
void category_test_2(
const std::vector<Range>& ranges, unsigned i, unsigned j, std::random_access_iterator_tag)
{
typedef Range range;
const range& ri = ranges[i];
const range& rj = ranges[j];
if (indices[i][0] <= indices[j][0] && indices[i][1] >= indices[j][1])
assert(ri.contains(rj));
if (ri.contains(rj))
assert((ri & rj) == rj);
assert(boost::intersects(ri, rj) == !(ri & rj).empty());
range t1(ri);
t1 &= rj;
assert(t1 == range(indices[i][0] > indices[j][0] ? ri.start() : rj.start(),
indices[i][1] < indices[j][1] ? ri.finish() : rj.finish()));
assert(t1 == (ri & rj));
range t2(ri);
t2 |= rj;
if (ri.empty())
assert(t2 == rj);
else if (rj.empty())
assert(t2 == ri);
else
assert(t2 == range(indices[i][0] < indices[j][0] ? ri.start() : rj.start(),
indices[i][1] > indices[j][1] ? ri.finish() : rj.finish()));
assert(t2 == (ri | rj));
if (i == j)
assert(ri == rj);
if (ri.empty() || rj.empty())
assert((ri == rj) == (ri.empty() && rj.empty()));
else
assert((ri == rj) == (ri.start() == rj.start() && ri.finish() == rj.finish()));
assert((ri == rj) == !(ri != rj));
bool same = ri == rj;
bool one_empty = ri.empty() != rj.empty();
std::less<range> less;
std::less_equal<range> less_equal;
std::greater<range> greater;
std::greater_equal<range> greater_equal;
if (same)
{
assert(greater_equal(ri,rj));
assert(less_equal(ri,rj));
assert(!greater(ri,rj));
assert(!less(ri,rj));
}
else if (one_empty)
{
const range& empty = ri.empty() ? ri : rj;
const range& non_empty = rj.empty() ? ri : rj;
assert(less(empty,non_empty));
assert(less_equal(empty,non_empty));
assert(!greater(empty,non_empty));
assert(!greater_equal(empty,non_empty));
assert(!less(non_empty,empty));
assert(!less_equal(non_empty,empty));
assert(greater(non_empty,empty));
assert(greater_equal(non_empty,empty));
}
else {
if (indices[i][0] < indices[j][0] ||
indices[i][0] == indices[j][0] && indices[i][1] < indices[j][1])
{
assert(!greater_equal(ri,rj));
assert(less(ri,rj));
}
if (indices[i][0] < indices[j][0] ||
indices[i][0] == indices[j][0] && indices[i][1] <= indices[j][1])
{
assert(!greater(ri,rj));
assert(less_equal(ri,rj));
}
if (indices[i][0] > indices[j][0] ||
indices[i][0] == indices[j][0] && indices[i][1] > indices[j][1])
{
assert(!less_equal(ri,rj));
assert(greater(ri,rj));
}
if (indices[i][0] > indices[j][0] ||
indices[i][0] == indices[j][0] && indices[i][1] >= indices[j][1])
{
assert(!less(ri,rj));
assert(greater_equal(ri,rj));
}
}
}
template <class Range>
void category_test_2(
const std::vector<Range>&, unsigned, unsigned, std::forward_iterator_tag)
{
}
template <class T>
void category_test_2(
const std::vector<boost::half_open_range<T> >&, unsigned, unsigned, std::bidirectional_iterator_tag)
{
}
template <class Range>
void test_back(Range& x, std::bidirectional_iterator_tag)
{
assert(x.back() == boost::prior(x.finish()));
}
template <class Range>
void test_back(Range& x, std::forward_iterator_tag)
{
}
template <class T>
boost::half_open_range<T> range_identity(const boost::half_open_range<T>& x)
{
return x;
}
template <class T>
void test(T x0, T x1, T x2, T x3)
{
std::vector<boost::half_open_range<T> > ranges;
typedef boost::half_open_range<T> range;
T bounds[4] = { x0, x1, x2, x3 };
const std::size_t num_ranges = sizeof(indices)/sizeof(*indices);
// test construction
for (std::size_t n = 0; n < num_ranges;++n)
{
T start = bounds[indices[n][0]];
T finish = bounds[indices[n][1]];
boost::half_open_range<T> r(start, finish);
ranges.push_back(r);
}
// test implicit conversion from std::pair<T,T>
range converted = std::pair<T,T>(x0,x0);
(void)converted;
// test assignment, equality and inequality
range r00 = range(x0, x0);
assert(r00 == range(x0,x0));
assert(r00 == range(x1,x1)); // empty ranges are all equal
if (x3 != x0)
assert(r00 != range(x0, x3));
r00 = range(x0, x3);
assert(r00 == range(x0, x3));
if (x3 != x0)
assert(r00 != range(x0, x0));
typedef typename range::iterator iterator;
typedef typename iterator::iterator_category category;
for (unsigned i = 0; i < num_ranges; ++i)
{
const range& r = ranges[i];
// test begin(), end(), basic iteration.
unsigned count = 0;
for (range::const_iterator p = r.begin(), finish = r.end();
p != finish;
++p, ++count)
{
assert(count < 2100);
}
// test size(), empty(), front(), back()
assert((unsigned)r.size() == count);
if (indices[i][0] == indices[i][1])
assert(r.empty());
if (r.empty())
assert(r.size() == 0);
if (!r.empty())
{
assert(r.front() == r.start());
test_back(r, category());
}
// test swap
range r1(r);
range r2(x0,x3);
const bool same = r1 == r2;
r1.swap(r2);
assert(r1 == range(x0,x3));
assert(r2 == r);
if (!same) {
assert(r1 != r);
assert(r2 != range(x0,x3));
}
// do individual tests for random-access iterators
category_test_1(r, category());
}
for (unsigned j = 0; j < num_ranges; ++j) {
for (unsigned k = 0; k < num_ranges; ++k) {
category_test_2(ranges, j, k, category());
}
}
}
template <class Integer>
void test_integer(Integer* = 0) // default arg works around MSVC bug
{
Integer a = 0;
Integer b = a + unsigned_random(128 - a);
Integer c = b + unsigned_random(128 - b);
Integer d = c + unsigned_random(128 - c);
test(a, b, c, d);
}
template <class Container>
void test_container(Container* = 0) // default arg works around MSVC bug
{
Container c(unsigned_random(1673));
const typename Container::size_type offset1 = unsigned_random(c.size());
const typename Container::size_type offset2 = unsigned_random(c.size() - offset1);
typename Container::iterator internal1 = c.begin();
std::advance(internal1, offset1);
typename Container::iterator internal2 = internal1;
std::advance(internal2, offset2);
test(c.begin(), internal1, internal2, c.end());
typedef typename Container::const_iterator const_iterator;
test(const_iterator(c.begin()),
const_iterator(internal1),
const_iterator(internal2),
const_iterator(c.end()));
}
int main()
{
// Test the built-in integer types.
test_integer<char>();
test_integer<unsigned char>();
test_integer<signed char>();
test_integer<wchar_t>();
test_integer<short>();
test_integer<unsigned short>();
test_integer<int>();
test_integer<unsigned int>();
test_integer<long>();
test_integer<unsigned long>();
#if defined(ULLONG_MAX) || defined(ULONG_LONG_MAX)
test_integer<long long>();
test_integer<unsigned long long>();
#endif
// Some tests on container iterators, to prove we handle a few different categories
test_container<std::vector<int> >();
test_container<std::list<int> >();
#ifndef BOOST_NO_SLIST
test_container<BOOST_STD_EXTENSION_NAMESPACE::slist<int> >();
#endif
// Also prove that we can handle raw pointers.
int array[2000];
const std::size_t a = 0;
const std::size_t b = a + unsigned_random(2000 - a);
const std::size_t c = b + unsigned_random(2000 - b);
test(array, array+b, array+c, array+2000);
return 0;
}

565
include/boost/operators.hpp Normal file
View File

@@ -0,0 +1,565 @@
// Boost operators.hpp header file ----------------------------------------//
// (C) Copyright David Abrahams 1999. Permission to copy, use,
// modify, sell and distribute this software is granted provided this
// copyright notice appears in all copies. This software is provided
// "as is" without express or implied warranty, and with no claim as
// to its suitability for any purpose.
// (C) Copyright Jeremy Siek 1999. Permission to copy, use, modify,
// sell and distribute this software is granted provided this
// copyright notice appears in all copies. This software is provided
// "as is" without express or implied warranty, and with no claim as
// to its suitability for any purpose.
// See http://www.boost.org for most recent version including documentation.
// Revision History
// 11 Feb 01 Fixed bugs in the iterator helpers which prevented explicitly
// supplied arguments from actually being used (Dave Abrahams)
// 04 Jul 00 Fixed NO_OPERATORS_IN_NAMESPACE bugs, major cleanup and
// refactoring of compiler workarounds, additional documentation
// (Alexy Gurtovoy and Mark Rodgers with some help and prompting from
// Dave Abrahams)
// 28 Jun 00 General cleanup and integration of bugfixes from Mark Rodgers and
// Jeremy Siek (Dave Abrahams)
// 20 Jun 00 Changes to accommodate Borland C++Builder 4 and Borland C++ 5.5
// (Mark Rodgers)
// 20 Jun 00 Minor fixes to the prior revision (Aleksey Gurtovoy)
// 10 Jun 00 Support for the base class chaining technique was added
// (Aleksey Gurtovoy). See documentation and the comments below
// for the details.
// 12 Dec 99 Initial version with iterator operators (Jeremy Siek)
// 18 Nov 99 Change name "divideable" to "dividable", remove unnecessary
// specializations of dividable, subtractable, modable (Ed Brey)
// 17 Nov 99 Add comments (Beman Dawes)
// Remove unnecessary specialization of operators<> (Ed Brey)
// 15 Nov 99 Fix less_than_comparable<T,U> second operand type for first two
// operators.(Beman Dawes)
// 12 Nov 99 Add operators templates (Ed Brey)
// 11 Nov 99 Add single template parameter version for compilers without
// partial specialization (Beman Dawes)
// 10 Nov 99 Initial version
// 10 Jun 00:
// An additional optional template parameter was added to most of
// operator templates to support the base class chaining technique (see
// documentation for the details). Unfortunately, a straightforward
// implementation of this change would have broken compatibility with the
// previous version of the library by making it impossible to use the same
// template name (e.g. 'addable') for both the 1- and 2-argument versions of
// an operator template. This implementation solves the backward-compatibility
// issue at the cost of some simplicity.
//
// One of the complications is an existence of special auxiliary class template
// 'is_chained_base<>' (see 'detail' namespace below), which is used
// to determine whether its template parameter is a library's operator template
// or not. You have to specialize 'is_chained_base<>' for each new
// operator template you add to the library.
//
// However, most of the non-trivial implementation details are hidden behind
// several local macros defined below, and as soon as you understand them,
// you understand the whole library implementation.
#ifndef BOOST_OPERATORS_HPP
#define BOOST_OPERATORS_HPP
#include <boost/config.hpp>
#include <boost/iterator.hpp>
#if defined(__sgi) && !defined(__GNUC__)
#pragma set woff 1234
#endif
#if defined(BOOST_MSVC)
# pragma warning( disable : 4284 ) // complaint about return type of
#endif // operator-> not begin a UDT
namespace boost {
namespace detail {
class empty_base {};
} // namespace detail
} // namespace boost
// In this section we supply the xxxx1 and xxxx2 forms of the operator
// templates, which are explicitly targeted at the 1-type-argument and
// 2-type-argument operator forms, respectively. Some compilers get confused
// when inline friend functions are overloaded in namespaces other than the
// global namespace. When BOOST_NO_OPERATORS_IN_NAMESPACE is defined, all of
// these templates must go in the global namespace.
#ifndef BOOST_NO_OPERATORS_IN_NAMESPACE
namespace boost
{
#endif
// Basic operator classes (contributed by Dave Abrahams) ------------------//
// Note that friend functions defined in a class are implicitly inline.
// See the C++ std, 11.4 [class.friend] paragraph 5
template <class T, class U, class B = ::boost::detail::empty_base>
struct less_than_comparable2 : B
{
friend bool operator<=(const T& x, const U& y) { return !(x > y); }
friend bool operator>=(const T& x, const U& y) { return !(x < y); }
friend bool operator>(const U& x, const T& y) { return y < x; }
friend bool operator<(const U& x, const T& y) { return y > x; }
friend bool operator<=(const U& x, const T& y) { return !(y < x); }
friend bool operator>=(const U& x, const T& y) { return !(y > x); }
};
template <class T, class B = ::boost::detail::empty_base>
struct less_than_comparable1 : B
{
friend bool operator>(const T& x, const T& y) { return y < x; }
friend bool operator<=(const T& x, const T& y) { return !(y < x); }
friend bool operator>=(const T& x, const T& y) { return !(x < y); }
};
template <class T, class U, class B = ::boost::detail::empty_base>
struct equality_comparable2 : B
{
friend bool operator==(const U& y, const T& x) { return x == y; }
friend bool operator!=(const U& y, const T& x) { return !(x == y); }
friend bool operator!=(const T& y, const U& x) { return !(y == x); }
};
template <class T, class B = ::boost::detail::empty_base>
struct equality_comparable1 : B
{
friend bool operator!=(const T& x, const T& y) { return !(x == y); }
};
template <class T, class U, class B = ::boost::detail::empty_base>
struct multipliable2 : B
{
friend T operator*(T x, const U& y) { return x *= y; }
friend T operator*(const U& y, T x) { return x *= y; }
};
template <class T, class B = ::boost::detail::empty_base>
struct multipliable1 : B
{
friend T operator*(T x, const T& y) { return x *= y; }
};
template <class T, class U, class B = ::boost::detail::empty_base>
struct addable2 : B
{
friend T operator+(T x, const U& y) { return x += y; }
friend T operator+(const U& y, T x) { return x += y; }
};
template <class T, class B = ::boost::detail::empty_base>
struct addable1 : B
{
friend T operator+(T x, const T& y) { return x += y; }
};
template <class T, class U, class B = ::boost::detail::empty_base>
struct subtractable2 : B
{
friend T operator-(T x, const U& y) { return x -= y; }
};
template <class T, class B = ::boost::detail::empty_base>
struct subtractable1 : B
{
friend T operator-(T x, const T& y) { return x -= y; }
};
template <class T, class U, class B = ::boost::detail::empty_base>
struct dividable2 : B
{
friend T operator/(T x, const U& y) { return x /= y; }
};
template <class T, class B = ::boost::detail::empty_base>
struct dividable1 : B
{
friend T operator/(T x, const T& y) { return x /= y; }
};
template <class T, class U, class B = ::boost::detail::empty_base>
struct modable2 : B
{
friend T operator%(T x, const U& y) { return x %= y; }
};
template <class T, class B = ::boost::detail::empty_base>
struct modable1 : B
{
friend T operator%(T x, const T& y) { return x %= y; }
};
template <class T, class U, class B = ::boost::detail::empty_base>
struct xorable2 : B
{
friend T operator^(T x, const U& y) { return x ^= y; }
friend T operator^(const U& y, T x) { return x ^= y; }
};
template <class T, class B = ::boost::detail::empty_base>
struct xorable1 : B
{
friend T operator^(T x, const T& y) { return x ^= y; }
};
template <class T, class U, class B = ::boost::detail::empty_base>
struct andable2 : B
{
friend T operator&(T x, const U& y) { return x &= y; }
friend T operator&(const U& y, T x) { return x &= y; }
};
template <class T, class B = ::boost::detail::empty_base>
struct andable1 : B
{
friend T operator&(T x, const T& y) { return x &= y; }
};
template <class T, class U, class B = ::boost::detail::empty_base>
struct orable2 : B
{
friend T operator|(T x, const U& y) { return x |= y; }
friend T operator|(const U& y, T x) { return x |= y; }
};
template <class T, class B = ::boost::detail::empty_base>
struct orable1 : B
{
friend T operator|(T x, const T& y) { return x |= y; }
};
// incrementable and decrementable contributed by Jeremy Siek
template <class T, class B = ::boost::detail::empty_base>
struct incrementable : B
{
friend T operator++(T& x, int)
{
incrementable_type tmp(x);
++x;
return tmp;
}
private: // The use of this typedef works around a Borland bug
typedef T incrementable_type;
};
template <class T, class B = ::boost::detail::empty_base>
struct decrementable : B
{
friend T operator--(T& x, int)
{
decrementable_type tmp(x);
--x;
return tmp;
}
private: // The use of this typedef works around a Borland bug
typedef T decrementable_type;
};
// Iterator operator classes (contributed by Jeremy Siek) ------------------//
template <class T, class P, class B = ::boost::detail::empty_base>
struct dereferenceable : B
{
P operator->() const
{
return &*static_cast<const T&>(*this);
}
};
template <class T, class I, class R, class B = ::boost::detail::empty_base>
struct indexable : B
{
R operator[](I n) const
{
return *(static_cast<const T&>(*this) + n);
}
};
#ifndef BOOST_NO_OPERATORS_IN_NAMESPACE
} // namespace boost
#endif // BOOST_NO_OPERATORS_IN_NAMESPACE
// BOOST_IMPORT_TEMPLATE1/BOOST_IMPORT_TEMPLATE2 -
//
// When BOOST_NO_OPERATORS_IN_NAMESPACE is defined we need a way to import an
// operator template into the boost namespace. BOOST_IMPORT_TEMPLATE1 is used
// for one-argument forms of operator templates; BOOST_IMPORT_TEMPLATE2 for
// two-argument forms. Note that these macros expect to be invoked from within
// boost.
#if defined(BOOST_NO_OPERATORS_IN_NAMESPACE)
# if defined(BOOST_NO_USING_TEMPLATE)
// Because a Borland C++ 5.5 bug prevents a using declaration from working,
// we are forced to use inheritance for that compiler.
# define BOOST_IMPORT_TEMPLATE2(template_name) \
template <class T, class U, class B = ::boost::detail::empty_base> \
struct template_name : ::template_name<T, U, B> {};
# define BOOST_IMPORT_TEMPLATE1(template_name) \
template <class T, class B = ::boost::detail::empty_base> \
struct template_name : ::template_name<T, B> {};
# else
// Otherwise, bring the names in with a using-declaration to avoid
// stressing the compiler
# define BOOST_IMPORT_TEMPLATE2(template_name) using ::template_name;
# define BOOST_IMPORT_TEMPLATE1(template_name) using ::template_name;
# endif // BOOST_NO_USING_TEMPLATE
#else // !BOOST_NO_OPERATORS_IN_NAMESPACE
// The template is already in boost so we have nothing to do.
# define BOOST_IMPORT_TEMPLATE2(template_name)
# define BOOST_IMPORT_TEMPLATE1(template_name)
#endif // BOOST_NO_OPERATORS_IN_NAMESPACE
//
// Here's where we put it all together, defining the xxxx forms of the templates
// in namespace boost. We also define specializations of is_chained_base<> for
// the xxxx, xxxx1, and xxxx2 templates, importing them into boost:: as
// neccessary.
//
#if !defined(BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION)
// is_chained_base<> - a traits class used to distinguish whether an operator
// template argument is being used for base class chaining, or is specifying a
// 2nd argument type.
namespace boost {
// A type parameter is used instead of a plain bool because Borland's compiler
// didn't cope well with the more obvious non-type template parameter.
namespace detail {
struct true_t {};
struct false_t {};
} // namespace detail
// Unspecialized version assumes that most types are not being used for base
// class chaining. We specialize for the operator templates defined in this
// library.
template<class T> struct is_chained_base {
typedef ::boost::detail::false_t value;
};
} // namespace boost
// Import a 2-type-argument operator template into boost (if neccessary) and
// provide a specialization of 'is_chained_base<>' for it.
# define BOOST_OPERATOR_TEMPLATE2(template_name2) \
BOOST_IMPORT_TEMPLATE2(template_name2) \
template<class T, class U, class B> \
struct is_chained_base< ::boost::template_name2<T, U, B> > { \
typedef ::boost::detail::true_t value; \
};
// Import a 1-type-argument operator template into boost (if neccessary) and
// provide a specialization of 'is_chained_base<>' for it.
# define BOOST_OPERATOR_TEMPLATE1(template_name1) \
BOOST_IMPORT_TEMPLATE1(template_name1) \
template<class T, class B> \
struct is_chained_base< ::boost::template_name1<T, B> > { \
typedef ::boost::detail::true_t value; \
};
// BOOST_OPERATOR_TEMPLATE(template_name) defines template_name<> such that it
// can be used for specifying both 1-argument and 2-argument forms. Requires the
// existence of two previously defined class templates named '<template_name>1'
// and '<template_name>2' which must implement the corresponding 1- and 2-
// argument forms.
//
// The template type parameter O == is_chained_base<U>::value is used to
// distinguish whether the 2nd argument to <template_name> is being used for
// base class chaining from another boost operator template or is describing a
// 2nd operand type. O == true_t only when U is actually an another operator
// template from the library. Partial specialization is used to select an
// implementation in terms of either '<template_name>1' or '<template_name>2'.
//
# define BOOST_OPERATOR_TEMPLATE(template_name) \
template <class T \
,class U = T \
,class B = ::boost::detail::empty_base \
,class O = typename is_chained_base<U>::value \
> \
struct template_name : template_name##2<T, U, B> {}; \
\
template<class T, class U, class B> \
struct template_name<T, U, B, ::boost::detail::true_t> \
: template_name##1<T, U> {}; \
\
template <class T, class B> \
struct template_name<T, T, B, ::boost::detail::false_t> \
: template_name##1<T, B> {}; \
\
template<class T, class U, class B, class O> \
struct is_chained_base< ::boost::template_name<T, U, B, O> > { \
typedef ::boost::detail::true_t value; \
}; \
\
BOOST_OPERATOR_TEMPLATE2(template_name##2) \
BOOST_OPERATOR_TEMPLATE1(template_name##1)
#else // BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
# define BOOST_OPERATOR_TEMPLATE2(template_name2) \
BOOST_IMPORT_TEMPLATE2(template_name2)
# define BOOST_OPERATOR_TEMPLATE1(template_name1) \
BOOST_IMPORT_TEMPLATE1(template_name1)
// In this case we can only assume that template_name<> is equivalent to the
// more commonly needed template_name1<> form.
# define BOOST_OPERATOR_TEMPLATE(template_name) \
template <class T, class B = ::boost::detail::empty_base> \
struct template_name : template_name##1<T, B> {};
#endif // BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
namespace boost {
BOOST_OPERATOR_TEMPLATE(less_than_comparable)
BOOST_OPERATOR_TEMPLATE(equality_comparable)
BOOST_OPERATOR_TEMPLATE(multipliable)
BOOST_OPERATOR_TEMPLATE(addable)
BOOST_OPERATOR_TEMPLATE(subtractable)
BOOST_OPERATOR_TEMPLATE(dividable)
BOOST_OPERATOR_TEMPLATE(modable)
BOOST_OPERATOR_TEMPLATE(xorable)
BOOST_OPERATOR_TEMPLATE(andable)
BOOST_OPERATOR_TEMPLATE(orable)
BOOST_OPERATOR_TEMPLATE1(incrementable)
BOOST_OPERATOR_TEMPLATE1(decrementable)
BOOST_OPERATOR_TEMPLATE2(dereferenceable)
// indexable doesn't follow the patterns above (it has 4 template arguments), so
// we just write out the compiler hacks explicitly.
#ifdef BOOST_NO_OPERATORS_IN_NAMESPACE
# ifdef BOOST_NO_USING_TEMPLATE
template <class T, class I, class R, class B = ::boost::detail::empty_base>
struct indexable : ::indexable<T,I,R,B> {};
# else
using ::indexable;
# endif
#endif
#ifndef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
template <class T, class I, class R, class B>
struct is_chained_base< ::boost::indexable<T, I, R, B> > {
typedef ::boost::detail::true_t operator_template_type;
};
#endif
#undef BOOST_OPERATOR_TEMPLATE
#undef BOOST_OPERATOR_TEMPLATE2
#undef BOOST_OPERATOR_TEMPLATE1
#undef BOOST_IMPORT_TEMPLATE1
#undef BOOST_IMPORT_TEMPLATE2
// The following 'operators' classes can only be used portably if the derived class
// declares ALL of the required member operators.
template <class T, class U>
struct operators2
: less_than_comparable2<T,U
, equality_comparable2<T,U
, addable2<T,U
, subtractable2<T,U
, multipliable2<T,U
, dividable2<T,U
, modable2<T,U
, orable2<T,U
, andable2<T,U
, xorable2<T,U
> > > > > > > > > > {};
#ifndef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
template <class T, class U = T>
struct operators : operators2<T, U> {};
template <class T> struct operators<T, T>
#else
template <class T> struct operators
#endif
: less_than_comparable<T
, equality_comparable<T
, addable<T
, subtractable<T
, multipliable<T
, dividable<T
, modable<T
, orable<T
, andable<T
, xorable<T
, incrementable<T
, decrementable<T
> > > > > > > > > > > > {};
// Iterator helper classes (contributed by Jeremy Siek) -------------------//
template <class T,
class V,
class D = std::ptrdiff_t,
class P = V*,
class R = V&>
struct forward_iterator_helper
: equality_comparable<T
, incrementable<T
, dereferenceable<T,P
, boost::iterator<std::forward_iterator_tag,V,D,P,R
> > > > {};
template <class T,
class V,
class D = std::ptrdiff_t,
class P = V*,
class R = V&>
struct bidirectional_iterator_helper
: equality_comparable<T
, incrementable<T
, decrementable<T
, dereferenceable<T,P
, boost::iterator<std::bidirectional_iterator_tag,V,D,P,R
> > > > > {};
template <class T,
class V,
class D = std::ptrdiff_t,
class P = V*,
class R = V&>
struct random_access_iterator_helper
: equality_comparable<T
, less_than_comparable<T
, incrementable<T
, decrementable<T
, dereferenceable<T,P
, addable2<T,D
, subtractable2<T,D
, indexable<T,D,R
, boost::iterator<std::random_access_iterator_tag,V,D,P,R
> > > > > > > > >
{
#ifndef __BORLANDC__
friend D requires_difference_operator(const T& x, const T& y) {
return x - y;
}
#endif
}; // random_access_iterator_helper
} // namespace boost
#if defined(__sgi) && !defined(__GNUC__)
#pragma reset woff 1234
#endif
#endif // BOOST_OPERATORS_HPP

View File

@@ -0,0 +1,61 @@
// (C) Copyright Jeremy Siek 2000. Permission to copy, use, modify, sell and
// distribute this software is granted provided this copyright notice appears
// in all copies. This software is provided "as is" without express or implied
// warranty, and with no claim as to its suitability for any purpose.
#include <boost/config.hpp>
#include <vector>
#include <iostream>
#include <iterator>
#include <functional>
#include <boost/iterator_adaptors.hpp>
int main(int, char*[])
{
char characters[] = "abcdefg";
const int N = sizeof(characters)/sizeof(char) - 1; // -1 since characters has a null char
char* pointers_to_chars[N]; // at the end.
for (int i = 0; i < N; ++i)
pointers_to_chars[i] = &characters[i];
// Example of using indirect_iterator_generator
boost::indirect_iterator_generator<char**, char>::type
indirect_first(pointers_to_chars), indirect_last(pointers_to_chars + N);
std::copy(indirect_first, indirect_last, std::ostream_iterator<char>(std::cout, ","));
std::cout << std::endl;
// Example of using indirect_iterator_pair_generator
typedef boost::indirect_iterator_pair_generator<char**,
char, char*, char&, const char*, const char&> PairGen;
char mutable_characters[N];
char* pointers_to_mutable_chars[N];
for (int i = 0; i < N; ++i)
pointers_to_mutable_chars[i] = &mutable_characters[i];
PairGen::iterator mutable_indirect_first(pointers_to_mutable_chars),
mutable_indirect_last(pointers_to_mutable_chars + N);
PairGen::const_iterator const_indirect_first(pointers_to_chars),
const_indirect_last(pointers_to_chars + N);
std::transform(const_indirect_first, const_indirect_last,
mutable_indirect_first, std::bind1st(std::plus<char>(), 1));
std::copy(mutable_indirect_first, mutable_indirect_last,
std::ostream_iterator<char>(std::cout, ","));
std::cout << std::endl;
// Example of using make_indirect_iterator()
std::copy(boost::make_indirect_iterator(pointers_to_chars),
boost::make_indirect_iterator(pointers_to_chars + N),
std::ostream_iterator<char>(std::cout, ","));
std::cout << std::endl;
return 0;
}

View File

@@ -1,458 +0,0 @@
// Demonstrate and test boost/operators.hpp on std::iterators -------------//
// (C) Copyright Jeremy Siek 1999. Permission to copy, use, modify,
// sell and distribute this software is granted provided this
// copyright notice appears in all copies. This software is provided
// "as is" without express or implied warranty, and with no claim as
// to its suitability for any purpose.
// See http://www.boost.org for most recent version including documentation.
// Revision History
// 19 Feb 01 Take adavantage of improved iterator_traits to do more tests
// on MSVC. Hack around an MSVC-with-STLport internal compiler
// error. (David Abrahams)
// 11 Feb 01 Added test of operator-> for forward and input iterators.
// (Jeremy Siek)
// 11 Feb 01 Borland fixes (David Abrahams)
// 10 Feb 01 Use new adaptors interface. (David Abrahams)
// 10 Feb 01 Use new filter_ interface. (David Abrahams)
// 09 Feb 01 Use new reverse_ and indirect_ interfaces. Replace
// BOOST_NO_STD_ITERATOR_TRAITS with
// BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION to prove we've
// normalized to core compiler capabilities (David Abrahams)
// 08 Feb 01 Use Jeremy's new make_reverse_iterator form; add more
// comprehensive testing. Force-decay array function arguments to
// pointers.
// 07 Feb 01 Added tests for the make_xxx_iterator() helper functions.
// (Jeremy Siek)
// 07 Feb 01 Replaced use of xxx_pair_generator with xxx_generator where
// possible (which was all but the projection iterator).
// (Jeremy Siek)
// 06 Feb 01 Removed now-defaulted template arguments where possible
// Updated names to correspond to new generator naming convention.
// Added a trivial test for make_transform_iterator().
// Gave traits for const iterators a mutable value_type, per std.
// Resurrected my original tests for indirect iterators.
// (David Abrahams)
// 04 Feb 01 Fix for compilers without standard iterator_traits
// (David Abrahams)
// 13 Jun 00 Added const version of the iterator tests (Jeremy Siek)
// 12 Dec 99 Initial version with iterator operators (Jeremy Siek)
#include <boost/config.hpp>
#include <iostream>
#include <algorithm>
#include <functional>
#include <boost/iterator_adaptors.hpp>
#include <boost/pending/iterator_tests.hpp>
#include <boost/pending/integer_range.hpp>
#include <boost/concept_archetype.hpp>
#include <stdlib.h>
#include <vector>
#include <deque>
#include <set>
struct my_iterator_tag : public std::random_access_iterator_tag { };
using boost::dummyT;
struct my_iter_traits {
typedef dummyT value_type;
typedef dummyT* pointer;
typedef dummyT& reference;
typedef my_iterator_tag iterator_category;
typedef std::ptrdiff_t difference_type;
};
struct my_const_iter_traits {
typedef dummyT value_type;
typedef const dummyT* pointer;
typedef const dummyT& reference;
typedef my_iterator_tag iterator_category;
typedef std::ptrdiff_t difference_type;
};
typedef boost::iterator_adaptor<dummyT*,
boost::default_iterator_policies, dummyT> my_iterator;
typedef boost::iterator_adaptor<const dummyT*,
boost::default_iterator_policies, const dummyT> const_my_iterator;
struct mult_functor {
typedef int result_type;
typedef int argument_type;
// Functors used with transform_iterator must be
// DefaultConstructible, as the transform_iterator must be
// DefaultConstructible to satisfy the requirements for
// TrivialIterator.
mult_functor() { }
mult_functor(int aa) : a(aa) { }
int operator()(int b) const { return a * b; }
int a;
};
template <class Pair>
struct select1st_
: public std::unary_function<Pair, typename Pair::first_type>
{
const typename Pair::first_type& operator()(const Pair& x) const {
return x.first;
}
typename Pair::first_type& operator()(Pair& x) const {
return x.first;
}
};
struct one_or_four {
bool operator()(dummyT x) const {
return x.foo() == 1 || x.foo() == 4;
}
};
typedef std::deque<int> storage;
typedef std::deque<int*> pointer_deque;
typedef std::set<storage::iterator> iterator_set;
void more_indirect_iterator_tests()
{
// For some reason all heck breaks loose in the compiler under these conditions.
#if !defined(BOOST_MSVC) || !defined(__STL_DEBUG)
storage store(1000);
std::generate(store.begin(), store.end(), rand);
pointer_deque ptr_deque;
iterator_set iter_set;
for (storage::iterator p = store.begin(); p != store.end(); ++p)
{
ptr_deque.push_back(&*p);
iter_set.insert(p);
}
typedef boost::indirect_iterator_pair_generator<
pointer_deque::iterator
#ifdef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
, int
#endif
> IndirectDeque;
IndirectDeque::iterator db(ptr_deque.begin());
IndirectDeque::iterator de(ptr_deque.end());
assert(static_cast<std::size_t>(de - db) == store.size());
assert(db + store.size() == de);
IndirectDeque::const_iterator dci(db);
assert(db == dci);
assert(dci == db);
assert(dci != de);
assert(dci < de);
assert(dci <= de);
assert(de >= dci);
assert(de > dci);
dci = de;
assert(dci == de);
boost::random_access_iterator_test(db + 1, store.size() - 1, boost::next(store.begin()));
*db = 999;
assert(store.front() == 999);
typedef boost::indirect_iterator_generator<
iterator_set::iterator
#ifdef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
, int
#endif
>::type indirect_set_iterator;
typedef boost::indirect_iterator_generator<
iterator_set::iterator,
const int
>::type const_indirect_set_iterator;
indirect_set_iterator sb(iter_set.begin());
indirect_set_iterator se(iter_set.end());
const_indirect_set_iterator sci(iter_set.begin());
assert(sci == sb);
assert(sci != se);
sci = se;
assert(sci == se);
*boost::prior(se) = 888;
assert(store.back() == 888);
assert(std::equal(sb, se, store.begin()));
boost::bidirectional_iterator_test(boost::next(sb), store[1], store[2]);
assert(std::equal(db, de, store.begin()));
#endif
}
int
main()
{
dummyT array[] = { dummyT(0), dummyT(1), dummyT(2),
dummyT(3), dummyT(4), dummyT(5) };
const int N = sizeof(array)/sizeof(dummyT);
// sanity check, if this doesn't pass the test is buggy
boost::random_access_iterator_test(array,N,array);
// Check that the policy concept checks and the default policy
// implementation match up.
boost::function_requires<
boost::RandomAccessIteratorPoliciesConcept<
boost::default_iterator_policies, int*,
boost::iterator<std::random_access_iterator_tag, int, std::ptrdiff_t,
int*, int&>
> >();
// Test the iterator_adaptor
{
my_iterator i(array);
boost::random_access_iterator_test(i, N, array);
const_my_iterator j(array);
boost::random_access_iterator_test(j, N, array);
boost::const_nonconst_iterator_test(i, ++j);
}
// Test transform_iterator
{
int x[N], y[N];
for (int k = 0; k < N; ++k)
x[k] = k;
std::copy(x, x + N, y);
for (int k2 = 0; k2 < N; ++k2)
x[k2] = x[k2] * 2;
boost::transform_iterator_generator<mult_functor, int*>::type
i(y, mult_functor(2));
boost::input_iterator_test(i, x[0], x[1]);
boost::input_iterator_test(boost::make_transform_iterator(&y[0], mult_functor(2)), x[0], x[1]);
}
// Test indirect_iterator_generator
{
dummyT* ptr[N];
for (int k = 0; k < N; ++k)
ptr[k] = array + k;
typedef boost::indirect_iterator_generator<dummyT**
#ifdef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
, dummyT
#endif
>::type indirect_iterator;
typedef boost::indirect_iterator_generator<dummyT**, const dummyT>::type const_indirect_iterator;
indirect_iterator i(ptr);
boost::random_access_iterator_test(i, N, array);
#ifndef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
boost::random_access_iterator_test(boost::make_indirect_iterator(ptr), N, array);
#endif
// check operator->
assert((*i).m_x == i->foo());
const_indirect_iterator j(ptr);
boost::random_access_iterator_test(j, N, array);
dummyT*const* const_ptr = ptr;
#ifndef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
boost::random_access_iterator_test(boost::make_indirect_iterator(const_ptr), N, array);
#endif
boost::const_nonconst_iterator_test(i, ++j);
more_indirect_iterator_tests();
}
// Test projection_iterator_pair_generator
{
typedef std::pair<dummyT,dummyT> Pair;
Pair pair_array[N];
for (int k = 0; k < N; ++k)
pair_array[k].first = array[k];
typedef boost::projection_iterator_pair_generator<select1st_<Pair>,
Pair*, const Pair*
> Projection;
Projection::iterator i(pair_array);
boost::random_access_iterator_test(i, N, array);
boost::random_access_iterator_test(boost::make_projection_iterator(pair_array, select1st_<Pair>()), N, array);
boost::random_access_iterator_test(boost::make_projection_iterator< select1st_<Pair> >(pair_array), N, array);
Projection::const_iterator j(pair_array);
boost::random_access_iterator_test(j, N, array);
boost::random_access_iterator_test(boost::make_const_projection_iterator(pair_array, select1st_<Pair>()), N, array);
boost::random_access_iterator_test(boost::make_const_projection_iterator<select1st_<Pair> >(pair_array), N, array);
boost::const_nonconst_iterator_test(i, ++j);
}
// Test reverse_iterator_generator
{
dummyT reversed[N];
std::copy(array, array + N, reversed);
std::reverse(reversed, reversed + N);
typedef boost::reverse_iterator_generator<dummyT*
#ifdef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
, dummyT
#endif
>::type reverse_iterator;
reverse_iterator i(reversed + N);
boost::random_access_iterator_test(i, N, array);
#ifndef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
boost::random_access_iterator_test(boost::make_reverse_iterator(reversed + N), N, array);
#endif
typedef boost::reverse_iterator_generator<const dummyT*
#ifdef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
, const dummyT
#endif
>::type const_reverse_iterator;
const_reverse_iterator j(reversed + N);
boost::random_access_iterator_test(j, N, array);
const dummyT* const_reversed = reversed;
#ifndef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
boost::random_access_iterator_test(boost::make_reverse_iterator(const_reversed + N), N, array);
#endif
boost::const_nonconst_iterator_test(i, ++j);
}
// Test reverse_iterator_generator again, with traits fully deducible on all platforms
{
std::deque<dummyT> reversed_container;
std::reverse_copy(array, array + N, std::back_inserter(reversed_container));
const std::deque<dummyT>::iterator reversed = reversed_container.begin();
typedef boost::reverse_iterator_generator<
std::deque<dummyT>::iterator>::type reverse_iterator;
typedef boost::reverse_iterator_generator<
std::deque<dummyT>::const_iterator, const dummyT>::type const_reverse_iterator;
// MSVC/STLport gives an INTERNAL COMPILER ERROR when any computation
// (e.g. "reversed + N") is used in the constructor below.
const std::deque<dummyT>::iterator finish = reversed_container.end();
reverse_iterator i(finish);
boost::random_access_iterator_test(i, N, array);
boost::random_access_iterator_test(boost::make_reverse_iterator(reversed + N), N, array);
const_reverse_iterator j = reverse_iterator(finish);
boost::random_access_iterator_test(j, N, array);
const std::deque<dummyT>::const_iterator const_reversed = reversed;
boost::random_access_iterator_test(boost::make_reverse_iterator(const_reversed + N), N, array);
// Many compilers' builtin deque iterators don't interoperate well, though
// STLport fixes that problem.
#if defined(__SGI_STL_PORT) || !defined(__GNUC__) && !defined(__BORLANDC__) && !defined(BOOST_MSVC)
boost::const_nonconst_iterator_test(i, ++j);
#endif
}
// Test integer_range's iterators
{
int int_array[] = { 0, 1, 2, 3, 4, 5 };
boost::integer_range<int> r(0, 5);
boost::random_access_iterator_test(r.begin(), r.size(), int_array);
}
// Test filter iterator
{
// Using typedefs for filter_gen::type and filter_gen::policies_type
// confused Borland terribly.
typedef boost::detail::non_bidirectional_category<dummyT*>::type category;
typedef ::boost::filter_iterator_generator<one_or_four, dummyT*
#ifdef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
, dummyT
#endif
>::type filter_iter;
filter_iter i(array, filter_iter::policies_type(one_or_four(), array + N));
boost::forward_iterator_test(i, dummyT(1), dummyT(4));
enum { is_forward = boost::is_same<
filter_iter::iterator_category,
std::forward_iterator_tag>::value };
BOOST_STATIC_ASSERT(is_forward);
// On compilers not supporting partial specialization, we can do more type
// deduction with deque iterators than with pointers... unless the library
// is broken ;-(
#if !defined(BOOST_MSVC) || defined(__SGI_STL_PORT)
std::deque<dummyT> array2;
std::copy(array+0, array+N, std::back_inserter(array2));
boost::forward_iterator_test(
boost::make_filter_iterator(array2.begin(), array2.end(), one_or_four()),
dummyT(1), dummyT(4));
boost::forward_iterator_test(
boost::make_filter_iterator<one_or_four>(array2.begin(), array2.end()),
dummyT(1), dummyT(4));
#endif
#if !defined(BOOST_MSVC) // This just freaks MSVC out completely
boost::forward_iterator_test(
boost::make_filter_iterator<one_or_four>(
boost::make_reverse_iterator(array2.end()),
boost::make_reverse_iterator(array2.begin())
),
dummyT(4), dummyT(1));
#endif
#ifndef BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
boost::forward_iterator_test(
boost::make_filter_iterator(array+0, array+N, one_or_four()),
dummyT(1), dummyT(4));
boost::forward_iterator_test(
boost::make_filter_iterator<one_or_four>(array, array + N),
dummyT(1), dummyT(4));
#endif
}
// check operator-> with a forward iterator
{
boost::forward_iterator_archetype<dummyT> forward_iter;
typedef boost::iterator_adaptor<boost::forward_iterator_archetype<dummyT>,
boost::default_iterator_policies,
dummyT, const dummyT&, const dummyT*,
std::forward_iterator_tag, std::ptrdiff_t> adaptor_type;
adaptor_type i(forward_iter);
if (0) // don't do this, just make sure it compiles
assert((*i).m_x == i->foo());
}
// check operator-> with an input iterator
{
boost::input_iterator_archetype<dummyT> input_iter;
typedef boost::iterator_adaptor<boost::input_iterator_archetype<dummyT>,
boost::default_iterator_policies,
dummyT, const dummyT&, const dummyT*,
std::input_iterator_tag, std::ptrdiff_t> adaptor_type;
adaptor_type i(input_iter);
if (0) // don't do this, just make sure it compiles
assert((*i).m_x == i->foo());
}
std::cout << "test successful " << std::endl;
return 0;
}

104
utility.htm Normal file
View File

@@ -0,0 +1,104 @@
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<title>Header boost/utility.hpp Documentation</title>
</head>
<body bgcolor="#FFFFFF" text="#000000">
<h1><img src="../../c++boost.gif" alt="c++boost.gif (8819 bytes)" align="center" WIDTH="277" HEIGHT="86">Header
<a href="../../boost/utility.hpp">boost/utility.hpp</a></h1>
<p>The entire contents of the header <code><a href="../../boost/utility.hpp">&lt;boost/utility.hpp&gt;</a></code>
are in <code>namespace boost</code>.</p>
<h2>Contents</h2>
<ul>
<li>Function templates <a href="#functions next">next() and prior()</a></li>
<li>Class <a href="#Class noncopyable">noncopyable</a></li>
<li>Function template <a href="tie.html">tie()</a> and supporting class tied.</li>
</ul>
<h2> <a name="functions next">Function</a> templates next() and prior()</h2>
<p>Certain data types, such as the C++ Standard Library's forward and
bidirectional iterators, do not provide addition and subtraction via operator+()
or operator-().&nbsp; This means that non-modifying computation of the next or
prior value requires a temporary, even though operator++() or operator--() is
provided.&nbsp; It also means that writing code like <code>itr+1</code> inside a
template restricts the iterator category to random access iterators.</p>
<p>The next() and prior() functions provide a simple way around these problems:</p>
<blockquote>
<pre>template &lt;class T&gt;
T next(T x) { return ++x; }
template &lt;class X&gt;
T prior(T x) { return --x; }</pre>
</blockquote>
<p>Usage is simple:</p>
<blockquote>
<pre>const std::list&lt;T&gt;::iterator p = get_some_iterator();
const std::list&lt;T&gt;::iterator prev = boost::prior(p);</pre>
</blockquote>
<p>Contributed by <a href="../../people/dave_abrahams.htm">Dave Abrahams</a>.</p>
<h2><a name="Class noncopyable">Class noncopyable</a></h2>
<p>Class <strong>noncopyable</strong> is a base class.&nbsp; Derive your own class from <strong>noncopyable</strong>
when you want to prohibit copy construction and copy assignment.</p>
<p>Some objects, particularly those which hold complex resources like files or
network connections, have no sensible copy semantics.&nbsp; Sometimes there are
possible copy semantics, but these would be of very limited usefulness and be
very difficult to implement correctly.&nbsp; Sometimes you're implementing a class that doesn't need to be copied
just yet and you don't want to take the time to write the appropriate functions.&nbsp;
Deriving from <b> noncopyable</b> will prevent the otherwise implicitly-generated
functions (which don't have the proper semantics) from becoming a trap for other programmers.</p>
<p>The traditional way to deal with these is to declare a private copy constructor and copy assignment, and then
document why this is done.&nbsp; But deriving from <b>noncopyable</b> is simpler
and clearer, and doesn't require additional documentation.</p>
<p>The program <a href="noncopyable_test.cpp">noncopyable_test.cpp</a> can be
used to verify class <b>noncopyable</b> works as expected. It has have been run successfully under
GCC 2.95, Metrowerks
CodeWarrior 5.0, and Microsoft Visual C++ 6.0 sp 3.</p>
<p>Contributed by <a href="../../people/dave_abrahams.htm">Dave Abrahams</a>.</p>
<h3>Example</h3>
<blockquote>
<pre>// inside one of your own headers ...
#include &lt;boost/utility.hpp&gt;
class ResourceLadenFileSystem : boost::noncopyable {
...</pre>
</blockquote>
<h3>Rationale</h3>
<p>Class noncopyable has protected constructor and destructor members to
emphasize that it is to be used only as a base class.&nbsp; Dave Abrahams notes
concern about the effect on compiler optimization of adding (even trivial inline)
destructor declarations. He says &quot;Probably this concern is misplaced, because
noncopyable will be used mostly for classes which own resources and thus have non-trivial destruction semantics.&quot;</p>
<hr>
<p>Revised&nbsp; <!--webbot bot="Timestamp" S-Type="EDITED" S-Format="%d %B, %Y" startspan
-->16 February, 2001<!--webbot bot="Timestamp" endspan i-checksum="40407"
-->
</p>
<p><EFBFBD> Copyright boost.org 1999. Permission to copy, use, modify, sell and
distribute this document is granted provided this copyright notice appears in
all copies. This document is provided &quot;as is&quot; without express or
implied warranty, and with no claim as to its suitability for any purpose.</p>
</body>
</html>