Files
optional/doc/optional.html
Fernando Cacciola 929847570c Merged to RC
[SVN r17957]
2003-03-17 13:42:06 +00:00

983 lines
38 KiB
HTML

<!DOCTYPE HTML PUBLIC "-//SoftQuad Software//DTD HoTMetaL PRO 5.0::19981217::extensions to HTML 4.0//EN" "hmpro5.dtd">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<LINK REL="stylesheet" TYPE="text/css" HREF="../../../boost.css">
<TITLE>Header </TITLE>
</HEAD>
<BODY BGCOLOR="#FFFFFF" TEXT="#000000" LINK="#0000FF" VLINK="#800080">
<H2><IMG SRC="../../../c++boost.gif" WIDTH="276" HEIGHT="86">Header &lt;<A
HREF="../../../boost/optional.hpp">boost/optional.hpp</A>&gt; </H2>
<H2>Contents</H2>
<DL CLASS="page-index">
<DT><A HREF="#mot">Motivation</A></DT>
<DT><A HREF="#dev">Development</A></DT>
<DT><A HREF="#synopsis">Synopsis</A></DT>
<DT><A HREF="#semantics">Semantics</A></DT>
<DT><A HREF="#examples">Examples</A></DT>
<DT><A HREF="#bool">A note about optional&lt;bool&gt;</A></DT>
<DT><A HREF="#exsafety">Exception Safety Guarantees</A></DT>
<DT><A HREF="#requirements">Type requirements</A></DT>
<DT><A HREF="#impl">Implementation Notes</A></DT>
<DT><A HREF="#porta">Dependencies and Portability</A></DT>
<DT><A HREF="#credits">Acknowledgment</A></DT>
</DL>
<HR>
<H2><A NAME="mot"></A>Motivation</H2>
<P>Consider these functions which should return a value but which might not have
a value to return:</P>
<pre>(A) double sqrt(double n );
(B) char get_async_input();
(C) point polygon::get_any_point_effectively_inside();</pre>
<P>There are different approaches to the issue of not having a value to return.</P>
<P>A typical approach is to consider the existence of a valid return value as
a postcondition, so that if the function cannot compute the value to return,
it has either undefined behavior (and can use asssert in a debug build)
or uses a runtime check and throws an exception if the postcondition is violated.
This is a reasonable choice for example, for function (A), because the
lack of a proper return value is directly related to an invalid parameter (out
of domain argument), so it is appropriate to require the callee to supply only
parameters in a valid domain for execution to continue normally.</P>
<P>However, function (B), because of its asynchronous nature, does not fail just
because it can't find a value to return; so it is incorrect to consider
such a situation an error and assert or throw an exception. This function must
return, and somehow, must tell the callee that it is not returning a meaningful
value.</P>
<P>A similar situation occurs with function (C): it is conceptually an error to
ask a <i>null-area</i> polygon to return a point inside itself, but in many
applications, it is just impractical for performance reasons to treat this as
an error (because detecting that the polygon has no area might be too expensive
to be required to be tested previously), and either an arbitrary point (typically
at infinity) is returned, or some efficient way to tell the callee that there
is no such point is used.</P>
<P>There are various mechanisms to let functions communicate that the returned
value is not valid. One such mechanism, which is quite common since it has zero
or negligible overhead, is to use a special value which is reserved to communicate
this. Classical examples of such special values are EOF, string::npos, points
at infinity, etc...</P>
<P>When those values exist, i.e. the return type can hold all meaningful values
<i>plus</i> the <i>signal</i> value, this mechanism is quite appropriate and
well known. Unfortunately, there are cases when such values do not exist. In
these cases, the usual alternative is either to use a wider type, such as 'int'
in place of 'char'; or a compound type, such as std::pair&lt;point,bool&gt;.
</P>
<P>Returning a std::pair&lt;T,bool&gt;, thus attaching a boolean flag to the result
which indicates if the result is meaningful, has the advantage that can be turned
into a consistent idiom since the first element of the pair can be whatever
the function would conceptually return. For example, the last two functions
could have the following interface:</P>
<pre>std::pair&lt;char,bool&gt; get_async_input();
std::pair&lt;point,bool&gt; polygon::get_any_point_effectively_inside();</pre>
<p>These functions use a consistent interface for dealing with possibly inexistent
results:</p>
<pre>std::pair&lt;point,bool&gt; p = poly.get_any_point_effectively_inside();
if ( p.second )
flood_fill(p.first);
</pre>
<P>However, not only is this quite a burden syntactically, it is also error
prone since the user can easily use the function result (first element of the
pair) without ever checking if it has a valid value.</P>
<P>Clearly, we need a better idiom.</P>
<H2><A NAME="dev"></A>Development</H2>
<h3>The model</h3>
<P>In C++, we can <i>declare</i> an object (a variable) of type T, and we can give this variable
an <i>initial value</i> (through an <i>initializer</i>. (c.f. 8.5)).<br>
When a declaration includes a non-empty initializer (an initial value is given), it is said that
the object has been <i><b>initialized</b></i>.<br>
If the declaration uses an empty initializer (no initial value is given),
and neither default nor value initialization applies, it is said that the object is
<i><b>uninitialized</b></i>. Its actual value exist but has an
<i>indeterminate inital value</i> (c.f. 8.5.9).<br>
<code>optional&lt;T&gt;</code> intends to formalize the notion of initialization/no-initialization
allowing a program to test whether an object has been initialized and stating that access to
the value of an uninitialized object is undefined behaviour. That is,
when a variable is declared as optional&lt;T&gt; and no initial value is given,
the variable is formally uninitialized. A formally uninitialized optional object has conceptually
no value at all and this situation can be tested at runtime. It is formally <i>undefined behaviour</i>
to try to access the value of an uninitialized optional. An uninitialized optional can be <i>assigned</i> a value, in which case its
initialization state changes to initialized. And given the formal treatment of initialization
states in optional objects, it is even possible to
reset an optional to <i>uninitialized</i>.</P>
<P>In C++ there is no formal notion of uninitialized objects, which
means that objects always have an initial value even if indeterminate.<br>
As discussed on the previous section, this has a drawback because you need additional
information to tell if an object has been effectively initialized.<br>
One of the typical ways in which this has been historically
dealt with is via a special value: EOF,npos,-1, etc... This is equivalent to adding
the special value to the set of possible values of a given type.
On modern languages, this can be modeled with a <b>discriminated
union</b> of T and something else such as a trivial POD or enum.
Discriminated unions are often called <i>variants</i>.
A variant has a <i>current type</i>, which in our case is either T or something else. In C++,
such a variant would be typically implemented as a template class of the form: <code>variant&lt;T,nil_t&gt;</code>
</P>
<P>There is precedence for a discriminated union as a model for an optional
value: the <a href="http://www.haskell.org/"><u>Haskell</u></a> <b>Maybe</b> builtin type constructor.</p>
<p>A discriminated union, which can be seen as a <b>container</b> which has an object of either
type T or something else, has <i>exactly</i> the semantics required for a wrapper of optional values:</p>
<li><b>deep-copy</b> semantics: copies of the variant implies copies of the contained value.</li>
<li><b>deep-relational</b> semantics: comparisons between variants matches both current types and values</li>
<li>If the variant's current type is T, it is modeling an <i>initialized</i> optional.</li>
<li>If the variant's current type is not T, it is modeling an <i>uninitialized</i> optional.</li>
<li>Testing if the variant's current type is T models testing if the optional is initialized</li>
<li>Trying to extract a T from a variant when its current type is not T, models the undefined behaviour
of trying to access the value of an uninitialized optional</li>
<P>Because of the way a discriminated union is used for this purpose, it only matters
whether its current type is T or not. We can put a layer on top of the variant hidding the other type
transforming a container of fixed size 1 into a variable size container which either has
a T or has nothing. Thus, the variant&lt;T,nil_t&gt; can be seen as if it were a variable-size
fixed-capacity stack-based container with the following optional-oriented interface:</P>
<pre>// Uninitialized (internally, current type is nil_t)
optional&lt;T&gt;::optional();
// Initialized with 'v' (internally, current type is T)
optional&lt;T&gt;::optional( T const&amp; v ) ;
// Back to uninitialized (current type is set to nil_t)
void optional&lt;T&gt;::reset();
// Assigns 'v' whether previously initialized or not (current type is set to T)
void optional&lt;T&gt;::reset( T const&amp; v ) ;
// Returns 'true' if the optional is initialized, 'false' otherwise.
bool optional&lt;T&gt;::initialized() ;
// If the optional is initialized (current type is T), returns a reference to its value.
// Otherwise (current type is nil_t), the result is undefined.
T const&amp; optional&lt;T&gt;::ref() const ;
T&amp; optional&lt;T&gt;::ref() ;
// If both are initialized, calls swap(T&amp;,T&amp;);
// If only one is initialized, calls reset(T const&amp;) and reset().
// If both are uninitalized, do nothing.
void swap ( optional&lt;T&gt;&amp; lhs, optional&lt;T&gt;&amp; rhs ) ;
// If both are initialized, compare values.
// If only one is initialized, they are not equal.
// If both are uninitalized, they are equal.
bool operator == ( optional&lt;T&gt; const&amp; lhs, optional&lt;T&gt; const&amp; rhs ) ;
bool operator != ( optional&lt;T&gt; const&amp; lhs, optional&lt;T&gt; const&amp; rhs ) ;
</pre>
<h3>Pointers and optional objects</h3>
<P>In C++, unlike many other languages, objects can be referenced <i>indirectly</i>
by means of a <b>pointer</b> (or a reference). Pointers have several nice features,
two of which are relevant to this development.</p>
<p>One is that a pointer has its own <i>pointer value</i>, which in effect
references the object being pointed to: the <b>pointee</b>. Consequently,
copies of pointers do not involve copies of pointees. This effect results in <i>aliasing</i>:
different pointers can refer to the same object.
The particular semantic that a copy of a pointer does not involve
a copy of the pointee is called <b>shallow-copy</b>, which is oppossed to <b>deep-copy</b> were
a copy of a wrapper involves a copy of the wrapped object (as with optional&lt;&gt;)<br>
Since this is the semantic followed by pointers (and references), shallow-copy
(and therefore aliasing) is implied in <b>pointer semantics</b>.</p>
<p>The other relevant feature of a pointer is that a pointer can have a <b>null
pointer value</b>. This is a <i>special</i> value which is used to indicate that the
pointer is not referring to any object at all. In other words, null pointer
values convey the notion of inexistent objects.</P>
<P>This meaning of the null pointer value allowed pointers to became a defacto standard
for handling optional objects because all you have to do to refer to a value which you
don't really have is to use a null pointer value of the appropriate type.
Pointers have been used for decades&mdash;from the days of C APIs to modern C++ libraries&mdash;to
<i>refer</i> to optional (that is, possibly inexistent) objects; particularly
as optional arguments to a function, but also quite often as optional data members.</P>
<P>The possible presence of a null pointer value makes the operations that access the
pointee's value possibly undefined, therefore, expressions which use dereference
and access operators, such as: <code>( *p = 2 )</code> and <code>( p-&gt;foo())</code>,
implicitly convey the notion of optionality, and this information is tied to
the <i>syntax</i> of the expressions. That is, the presence of operators * and -&gt; tell by
themselves&mdash;without any additional context&mdash;that the expression will be undefined unless
the implied pointee actually exist.<br>
Furthermore, the existence of the pointee can be tested by a comparison against
the null pointer value or via a conversion to bool, which allows expressions of
the form: if ( p != 0 ), or if ( p ) to be used to test for the existence of the pointee.</P>
<p>Such a defacto idiom for referring to optional objects can be formalized in the form of a
concept: the <a href="../../utility/OptionalPointee.html">OptionalPointee</a> concept.<br>
This concept captures the syntactic usage of operatos *, -> and conversion to bool to convey
the notion of optionality.</p>
<P>However, pointers are good to <u>refer</u> to optional objects, but not particularly good
to handle the optional objects in all other respects, such as initializing or moving/copying
them. The problem resides in the shallow-copy of pointer semantics: if you need to
effectively move or copy the object, pointers alone are not enough. The problem
is that copies of pointers do not imply copies of pointees. For example, as
was discussed in the motivation, pointers alone cannot be used to return optional
objects from a function because the object must move outside from the function and
into the caller's context.<br>
A solution to the shallow-copy problem that is often used is to resort to dynamic
allocation and use a smart pointer to automatically handle the details of this.
For example, if a function is to optionally return an object X, it can use shared_ptr&lt;X&gt;
as the return value. However, this requires dynamic allocation of X. If X is
a builtin or small POD, this technique is very poor in terms of required resources.
Optional objects are essentially values so it is very convenient to be able to use automatic
storage and deep-copy semantics to manipulate optional values just as we do with ordinary
values. Pointers do not have this semantics, so are unappropriate for the initialization and
transport of optional values, yet are quite convenient for handling the access to the
possible undefined value because of the idiomatic aid present in the OptionalPointee
concept incarnated by pointers. <br>
Therefore, the final solution which is presented in this library is to shape the
previously discussed optional&mdash;which is a value-based container&mdash;as a model
of the OptionalPointee concept.
</p>
<h3>Optional&lt;T&gt; as a model of OptionalPointee</h3>
<P>The optional&lt;&gt; template class presented with this library is a variation of the
sketch shown before (as a layer on top of a variant). It features <b>deep-copy</b> and
<b>deep relational operators</b>, but also models the OptionalPointee concept.
Instead of the member function 'initialized()' it has a safe conversion to bool,
and instead of the 'ref()' member function, it has operators*() and ->().<br>
However, it is particularly important that optional<> objects are not mistaken by pointers,
they are not. <u><b>optional&lt;&gt; does not model a pointer</b></u>.
For instance, optional&lt;&gt; has not shallow-copy so does not alias: two different optionals
never refer to the <i>same</i> value (but my have <i>equivalent</i> values).<br>
The difference between an optional&lt;T&gt; and a pointer must be kept in mind, particularly
because the semantics of relational operators are different: since optional&lt;T&gt;
is a value-wrapper, relational operators are deep: they compare optional values;
but relational operators for pointers are shallow: they do not compare pointee values.<br>
As a result, you might be able to replace optional&lt;T&gt; by T* on some situations but
not always. Specifically, on generic code written for both, you cannot use relational
operators directly, and must use the template function
<a href="../../utility/OptionalPointee.html#equal">equal_pointees()</a> instead.
<HR>
<H2><A NAME="synopsis">Synopsis</A></H2>
<PRE>namespace boost {
template&lt;class T>
class optional
{
public :
optional () ;
explicit optional ( T const&amp; v ) ;
optional ( optional const&amp; rhs ) ;
template&lt;class U&gt; explicit optional ( optional&lt;U&gt; const&amp; rhs ) ;
optional&amp; operator = ( optional const&amp; rhs ) ;
template&lt;class U&gt; optional&amp; operator = ( optional&lt;U&gt; const&amp rhs ) ;
T const* get() const ;
T* get() ;
T const* operator -&gt;() const ;
T* operator -&gt;() ;
T const&amp; operator *() const ;
T&amp; operator *() ;
void reset();
void reset ( T const&amp; ) ;
operator <i>unspecified-bool-type</i>() const ;
bool operator!() const ;
} ;
template&lt;class T&gt; inline bool operator == ( optional&lt;T&gt; const& x, optional&lt;T&gt; const& y ) ;
template&lt;class T&gt; inline bool operator != ( optional&lt;T&gt; const& x, optional&lt;T&gt; const& y ) ;
template&lt;class T&gt; inline T* get_pointer ( optional&lt;T&gt; const& opt ) ;
template&lt;class T&gt; inline void swap( optional&lt;T&gt;& x, optional&lt;T&gt;&amp; y ) ;
} // namespace boost
</PRE>
<HR>
<h2><A NAME="semantics">Semantics</a></h2>
<p><i>Note: the following section contains various assert() which are used only to
show the postconditions as sample code.
It is not implied that the type T must support each particular expression
but that if the expression is supported, the implied condition holds.</i></p>
<hr>
<pre>optional&lt;T&gt;::optional();</pre>
<blockquote>
<p><b>Effect:</b> Default-Constructs an <b>optional</b>.</p>
<p><b>Postconditions:</b> <b>*this</b> is <u>uninitialized</u>.</p>
<p><b>Throws:</b> Nothing.</p>
<p><b>Notes:</b> T's default constructor <u><i>is not</i></u> called.</p>
<p><b>Example:</b></p>
<blockquote>
<pre>optional&lt;T&gt; def ;
assert ( !def ) ;</pre>
</blockquote>
</blockquote>
<HR>
<pre>explicit optional&lt;T&gt;::optional( T const&amp; v )</pre>
<blockquote>
<p><b>Effect:</b> Directly-Constructs an <b>optional</b>.</p>
<p><b>Postconditions:</b> <b>*this</b> is <u>initialized</u> and its value is a <i>copy</i> of 'v'.</p>
<p><b>Throws:</b> Whatever T::T( T const&amp; ) throws.</p>
<p><b>Notes:</b> T::T( T const&amp; ) is called.</p>
<p><b>Exception Safety:</b> Exceptions can only be thrown during T::T( T const&amp; );
in that case, this constructor has no effect.
</p>
<p><b>Example:</b></p>
<blockquote>
<pre>
T v;
optional&lt;T&gt; opt(v);
assert ( *opt == v ) ;
</pre>
</blockquote>
</blockquote>
<HR>
<pre>optional&lt;T&gt;::optional( optional const&amp; rhs );</pre>
<blockquote>
<p><b>Effect:</b> Copy-Constructs an <b>optional</b>.</p>
<p><b>Postconditions:</b> If <b>rhs</b> is initialized, <b>*this</b> is initialized
and its value is a <i>copy</i> of the value of <b>rhs</b>; else <b>*this</b>
is uninitialized.</p>
<p><b>Throws:</b> Whatever T::T( T const& ) throws.</p>
<p><b>Notes:</b> T::T( T const& ) is called if <b>rhs</b> is initialized.</p>
<p><b>Exception Safety:</b> Exceptions can only be thrown during T::T( T const& );
in that case, this constructor has no effect.
</p>
<p><b>Example:</b></p>
<blockquote>
<pre>optional&lt;T&gt; uninit ;
assert (!uninit);
optional&lt;T&gt; uinit2 ( uninit ) ;
assert ( uninit2 == uninit );
optional&lt;T&gt; init( T(2) );
assert ( *init == T(2) ) ;
optional&lt;T&gt; init2 ( init ) ;
assert ( init2 == init ) ;
</pre>
</blockquote>
</blockquote>
<HR>
<pre>explicit template&lt;U&gt; optional&lt;T&gt;::optional( optional&lt;U&gt; const&amp; rhs );</pre>
<blockquote>
<p><b>Effect:</b> Copy-Constructs an <b>optional</b>.</p>
<p><b>Postconditions:</b> If <b>rhs</b> is initialized, <b>*this</b> is initialized
and its value is a <i>copy</i> of the value of <b>rhs</b> <i>converted</i>
to type T; else <b>*this</b> is uninitialized.
</p>
<p><b>Throws:</b> Whatever T::T( U const& ) throws.</p>
<p><b>Notes:</b> T::T( U const& ) is called if <b>rhs</b> is initialized, which requires
a valid conversion from U to T.
</p>
<p><b>Exception Safety:</b> Exceptions can only be thrown during T::T( U const& );
in that case, this constructor has no effect.
</p>
<p><b>Example:</b></p>
<blockquote>
<pre>optional&lt;double&gt; x(123.4);
assert ( *x == 123.4 ) ;
optional&lt;int&gt; y(x) ;
assert( *y == 123 ) ;
</pre>
</blockquote>
</blockquote>
<HR>
<pre>optional&amp; optional&lt;T&gt;::operator= ( optional const&amp; rhs ) ;</pre>
<blockquote>
<p><b>Effect:</b> Assigns another <b>optional</b> to an <b>optional</b>.</p>
<p><b>Postconditions:</b> If <b>rhs</b> is initialized, <b>*this</b> is initialized
and its value is a <i>copy</i> of the value of <b>rhs</b>; else <b>*this</b>
is uninitialized.
</p>
<p><b>Throws:</b> Whatever T::T( T const& ) throws.</p>
<p><b>Notes:</b> If <b>*this</b> was initialized, it is first reset to uninitialized
using T::~T(), then T::T( T const& ) is called if <b>rhs</b> is initialized.
</p>
<p><b>Exception Safety:</b> <u>Basic:</u> Exceptions can only be thrown during T::T( T const& );
in that case, <b>*this</b> is left <u>uninitialized</u>.
</p>
<p><b>Example:</b></p>
<blockquote>
<pre>
T v;
optional&lt;T&gt; opt(v);
optional&lt;T&gt; uninit ;
opt = uninit ;
assert ( !opt ) ;
// previous value (copy of 'v') destroyed from within 'opt'.
</pre>
</blockquote>
</blockquote>
<HR>
<pre>template&ltU&gt; optional&amp; optional&lt;T&gt;::operator= ( optional&lt;U&gt; const&amp; rhs ) ;</pre>
<blockquote>
<p><b>Effect:</b> Assigns another <i>convertible</i> <b>optional</b> to an <b>optional</b>.</p>
<p><b>Postconditions:</b> If <b>rhs</b> is initialized, <b>*this</b> is initialized
and its value is a <i>copy</i> of the value of <b>rhs</b> <i>converted</i>
to type T; else <b>*this</b> is uninitialized.
</p>
<p><b>Throws:</b> Whatever T::T( U const& ) throws.</p>
<p><b>Notes:</b> If <b>*this</b> was initialized, it is first reset to uninitialized
using T::~T(), then T::T( U const& ) is called if <b>rhs</b> is initialized,
which requires a valid conversion from U to T.
</p>
<p><b>Exception Safety:</b> <u>Basic:</u> Exceptions can only be thrown during T::T( U const& );
in that case, <b>*this</b> is left <u>uninitialized</u>.
</p>
<p><b>Example:</b></p>
<blockquote>
<pre>
T v;
optional&lt;T&gt; opt0(v);
optional&lt;U&gt; opt1;
opt1 = opt0 ;
assert ( *opt1 == static_cast&lt;U&gt;(v) ) ;
</pre>
</blockquote>
</blockquote>
<HR>
<pre>void optional&lt;T&gt;::reset( T const&amp v ) ;</pre>
<blockquote>
<p><b>Effect:</b> Resets the current value.</p>
<p><b>Postconditions: </b><b>*this</b> is <u>initialized</u> and its value is
a <i>copy</i> of 'v'.
</p>
<p><b>Throws:</b> Whatever T::T( T const& ) throws.</p>
<p><b>Notes:</b> If <b>*this</b> was initialized, it is first reset to uninitialized
using T::~T(), then T::T( T const& ) is called.
</p>
<p><b>Exception Safety:</b> <u>Basic:</u> Exceptions can only be thrown during T::T( T const& );
in that case, <b>*this</b> is left <u>uninitialized</u>.
</p>
<p><b>Example:</b></p>
<blockquote>
<pre>optional&lt;T&gt; opt ( some_T ) ;
assert( *opt == some_T );
opt.reset ( some_other_T ) ;
assert( *opt == some_other_T );
</pre>
</blockquote>
</blockquote>
<HR>
<pre>void optional&lt;T&gt;::reset() ;</pre>
<blockquote>
<p><b>Effect:</b> Destroys the current value.</p>
<p><b>Postconditions: *this</b> is uninitialized.</p>
<p><b>Throws:</b> Nothing.</p>
<p><b>Notes:</b> T::~T() is called.</p>
<p><b>Example:</b></p>
<blockquote>
<pre>optional&lt;T&gt; opt ( some_T ) ;
assert( *opt == some_T );
opt.reset();
assert( !opt );
</pre>
</blockquote>
</blockquote>
<HR>
<pre>
T const* optional&lt;T&gt;::get() const ;
T* optional&lt;T&gt;::get() ;
inline T const* get_pointer ( optional&lt;T&gt; const&amp; ) ;
inline T* get_pointer ( optional&lt;T&gt;&amp;) ;
</pre>
<blockquote>
<p><b>Returns:</b> If <b>*this</b> is initialized, a pointer to the contained
value; else 0 (<i>null</i>).
</p>
<p><b>Throws:</b> Nothing.</p>
<p><b>Notes:</b> The contained value is permanently stored within *this, so
you should not hold nor delete this pointer
</p>
<p><b>Example:</b></p>
<blockquote>
<pre>
T v;
optional&lt;T&gt; opt(v);
optional&lt;T&gt; const copt(v);
T* p = opt.get() ;
T const* cp = copt.get();
assert ( p == get_pointer(opt) );
assert ( cp == get_pointer(copt) ) ;
</pre>
</blockquote>
</blockquote>
<HR>
<pre>
T const* optional&lt;T&gt;::operator -&gt;() const ;
T* optional&lt;T&gt;::operator -&gt;() ;
</pre>
<blockquote>
<p><b>Requirements: *this</b> is initialized.</p>
<p><b>Returns:</b> A pointer to the contained value.</p>
<p><b>Throws:</b> Nothing.</p>
<p><b>Notes:</b> The requirement is asserted via BOOST_ASSERT().</p>
<p><b>Example:</b></p>
<blockquote>
<pre>
struct X { int mdata ; } ;
X x ;
optional&lt;X&gt; opt (x);
opt-&gt;mdata = 2 ;
</pre>
</blockquote>
</blockquote>
<HR>
<pre>T const&amp; optional&lt;T&gt;::operator*() const ;
T&amp; optional&lt;T&gt;::operator*();</pre>
<blockquote>
<p><b>Requirements: *this</b> is initialized</p>
<p><b>Returns:</b> A reference to the contained value</p>
<p><b>Throws:</b> Nothing.</p>
<p><b>Notes:</b> The requirement is asserted via BOOST_ASSERT().</p>
<p><b>Example:</b></p>
<blockquote>
<pre>T v ;
optional&lt;T&gt; opt ( v );
T const&amp; u = *opt;
assert ( u == v ) ;
T w ;
*opt = w ;
assert ( *opt == w ) ;
</pre>
</blockquote>
</blockquote>
<HR>
<pre>optional&lt;T&gt;::operator <i>unspecified-bool-type</i>() const ;</pre>
<blockquote>
<p><b>Returns:</b> An unspecified value which if used on a boolean context is equivalent to (get() != 0)</p>
<p><b>Throws:</b> Nothing.</p>
<blockquote>
<pre>optional&lt;T&gt; def ;
assert ( def == 0 );
optional&lt;T&gt; opt ( v ) ;
assert ( opt );
assert ( opt != 0 );
</pre>
</blockquote>
</blockquote>
<HR>
<pre> bool optional&lt;T&gt;::operator!() ;</pre>
<blockquote>
<p><b>Returns:</b> If <b>*this</b> is uninitialized, <code>true</code>; else <code>false.</code></p>
<p><b>Throws:</b> Nothing.</p>
<p><b>Notes:</b> This operator is provided for those compilers which can't use
the <i>unspecified-bool-type</i> operator in certain boolean contexts.
</p>
<p><b>Example:</b></p>
<blockquote>
<pre>optional&lt;T&gt; opt ;
assert ( !opt );
*opt = some_T ;
// Notice the &quot;double-bang&quot; idiom here.
assert ( !!opt ) ;
</pre>
</blockquote>
</blockquote>
<HR>
<pre>bool operator == ( optional&lt;T&gt; const&amp x, optional&lt;T&gt const&amp y );</pre>
<blockquote>
<p><b>Returns:</b> If both <b>x</b> and <b>y</b> are initialied, <code>(*x == *y)</code>.
If only x or y is initialized, <code>false</code>. If both are uninitialized, <code>true</code>.
</p>
<p><b>Throws:</b> Nothing.</p>
<p><b>Notes:</b> Pointers have shallow relational operators while <b>optional</b> has
deep relational operators. Do not use operator == directly in generic code
which expect to be given either an optional&lt;T&gt; or a pointer;
use <a href="../../utility/OptionalPointee.html#equal">equal_pointees()</a> instead
</p>
<p><b>Example:</b></p>
<blockquote>
<pre>
T x(12);
T y(12);
T z(21);
optional&lt;T&gt; def0 ;
optional&lt;T&gt; def1 ;
optional&lt;T&gt; optX(x);
optional&lt;T&gt; optY(y);
optional&lt;T&gt; optZ(z);
// Identity always hold
assert ( def0 == def0 );
assert ( optX == optX );
// Both uninitialized compare equal
assert ( def0 == def1 );
// Only one initialized compare unequal.
assert ( def0 != optX );
// Both initialized compare as (*lhs == *rhs)
assert ( optX == optY ) ;
assert ( optX != optZ ) ;
</pre>
</blockquote>
</blockquote>
<HR>
<pre>bool operator != ( optional&lt;T&gt; const&amp x, optional&lt;T&gt const&amp y );
</pre>
<blockquote>
<p><b>Returns:</b> !( x == y );</p>
<p><b>Throws:</b> Nothing.</p>
</blockquote>
<HR>
<pre>void swap ( optional&lt;T&gt;&amp x, optional&lt;T&gt&amp y );</pre>
<blockquote>
<p><b>Effect:</b> If both <b>x</b> and <b>y</b> are initialized, calls <code>swap(*x,*y)</code>
using std::swap.<br>
If only one is initialized, say x, calls: <code>y.reset(*x); x.reset();</code><br>
If none is initialized, does nothing.
</p>
<p><b>Postconditions:</b> The states of x and y interchanged.</p>
<p><b>Throws:</b> If both are initialized, whatever swap(T&amp;,T&amp;) throws.
If only one is initialized, whatever T::T ( T const&amp; ) throws.
</p>
<p><b>Notes:</b> If both are initialized, swap(T&amp;,T&amp;) is used <i>unqualified</i>
but with std::swap introduced in scope.<br>
If only one is initialized, T::~T() and T::T( T const& ) is called.
</p>
<p><b>Exception Safety:</b> If both are initialized, this operation has the exception
safety guarantees of swap(T&,T&).<br>
If only one is initialized, it has the same <b>basic</b> guarantee as optional&lt;T&gt;::reset( T const& ).
</p>
<p><b>Example:</b></p>
<blockquote>
<pre>
T x(12);
T y(21);
optional&lt;T&gt; def0 ;
optional&lt;T&gt; def1 ;
optional&lt;T&gt; optX(x);
optional&lt;T&gt; optY(y);
boost::swap(def0,def1); // no-op
boost::swap(def0,optX);
assert ( *def0 == x );
assert ( !optX );
boost::swap(def0,optX); // Get back to original values
boost::swap(optX,optY);
assert ( *optX == y );
assert ( *optY == x );
</pre>
</blockquote>
</blockquote>
<HR>
<H2><A NAME="examples">Examples</A></H2>
<h3>Optional return values</h3>
<PRE>optional&lt;char&gt; get_async_input()
{
if ( !queue.empty() )
return optional&lt;char&gt;(queue.top());
else return optional&lt;char&gt;(); // uninitialized
}
void recieve_async_message()
{
optional&lt;char&gt; rcv ;
// The safe boolean conversion from 'rcv' is used here.
while ( (rcv = get_async_input()) &amp;&amp; !timeout() )
output(*rcv);
}
</pre>
<h3>Optional local variables</h3>
<pre>optional&lt;string&gt; name ;
if ( database.open() )
{
name.reset ( database.lookup(employer_name) ) ;
}
else
{
if ( can_ask_user )
name.reset ( user.ask(employer_name) ) ;
}
if ( name )
print(*name);
else print(&quot;employer's name not found!&quot;);
</pre>
<h3>Optional data members</h3>
<pre>class figure
{
public:
figure()
{
// data member 'm_clipping_rect' is uninitialized at this point.
}
void clip_in_rect ( rect const&amp; rect )
{
....
m_clipping_rect.reset ( rect ) ; // initialized here.
}
void draw ( canvas& cvs )
{
if ( m_clipping_rect )
do_clipping(*m_clipping_rect);
cvs.drawXXX(..);
}
// this can return NULL.
rect const* get_clipping_rect() { return get_pointer(m_clipping_rect); }
private :
optional&lt;rect&gt; m_clipping_rect ;
};
</pre>
<h3>Bypassing expensive unnecesary default construction</h3>
<pre>class ExpensiveCtor { ... } ;
class Fred
{
Fred() : mLargeVector(10000) {}
std::vector< optional&lt;ExpensiveCtor&gt; > mLargeVector ;
} ;
</pre>
<HR>
<H2><A NAME="bool">A note about optional&lt;bool&gt;</A></H2>
<p><code>optional&lt;bool&gt;</code> should be used with special caution and consideration.</p>
<p>First, it is functionally similar to a tristate boolean (false,maybe,true) &mdash;such as
<u>boost::tribool</u> (not yet formally in boost)&mdash;except that in a tristate boolean,
the <i>maybe</i> state <u>represents a valid value</u>, unlike the corresponding state
of an uninitialized optional&lt;bool&gt;.<br>
It should be carefully considered if an optional bool instead of a tribool is really needed</p>
<p>Second, optional&lt;&gt; provides an implicit conversion to bool. This conversion
refers to the initialization state and not to the contained value.<br>
Using optional&lt;bool&gt; can lead to subtle errors due to the implicit bool conversion:</p>
<pre>
void foo ( bool v ) ;
void bar()
{
optional&lt;bool&gt; v = try();
// The following intended to pass the <b>value</b> of 'v' to foo():
foo(v);
// But instead, the <i>initialization state</i> is passed
// due to a typo: it should have been foo(<b>*</b>v).
}
</pre>
<p>The only implicit conversion is to bool, and it is <i>safe</i> in the sense that typical
integral promotions don't apply (i.e. if foo() takes an 'int' instead, it won't compile).
<HR>
<H2><A NAME="exsafety">Exception Safety Guarantees</A></H2>
<H3><u>Assignment and Reset:</u></H3>
<p>Because of the current implementation (see <A HREF="#impl">Implementation Notes</A>),
<code> optional&lt;T&gt;::operator=( optional&lt;T&gt; const&amp; )
and optional&lt;T&gt;::reset( T const&amp; )</code>
can only <i>guarantee</i> the <u>basic exception safety</u>: The lvalue optional is left
<u>uninitialized</u> if an exception is thrown (any previous value is <i>first</i>
destroyed using T::~T())</p>
<p><code>optional&lt;T&gt;::reset()</code> provides the no-throw guarantee (assuming a no-throw T::~T())</p>
<p>However, since <code>optional&lt&gt</code> itself doesn't throw any exceptions,
the only source for exceptions here is T's copy constructor, so if you know the exception guarantees
for T::T ( T const&amp; ), you know that optional's assignment and reset has the same guarantees.</p>
<pre>//
// Case 1: Exception thrown during assignment.
//
T v0(123);
optional&ltT&gt opt0(v0);
try
{
T v1(456);
optional&ltT&gt opt1(v1);
opt0 = opt1 ;
// If no exception was thrown, assignment succeeded.
assert( *opt0 == v1 ) ;
}
catch(...)
{
// If any exception was thrown, 'opt0' is reset to uninitialized.
assert( !opt0 ) ;
}
//
// Case 2: Exception thrown during reset(v)
//
T v0(123);
optional&ltT&gt opt(v0);
try
{
T v1(456);
opt.reset ( v1 ) ;
// If no exception was thrown, reset succeeded.
assert( *opt == v1 ) ;
}
catch(...)
{
// If any exception was thrown, 'opt' is reset to uninitialized.
assert( !opt ) ;
}
</pre>
<H3><u>Swap:</u></H3>
<p><code>void swap( optional&lt;T&gt;&amp;, optional&lt;T&gt;&amp; )</code>
has the same exception guarantee as <code>swap(T&amp;,T&amp;)</code> when both optionals are initialized.<br>
If only one of the optionals is initialized, it gives the same
<i>basic</i> exception guarantee as <code>optional&lt;T&gt;::reset( T const&amp; )</code>
(since <code>optional&lt;T&gt;::reset()</code> doesn't throw).<br>
If none of the optionals is initialized, it has no-throw guarantee since it is a no-op.
</p>
<HR>
<H2><A NAME="requirements">Type requirements</A></H2>
<p>T must be <a href="../../utility/CopyConstructible.html">Copy Constructible</a>
and have a no-throw destructor.<br>
T <u>is not</u> required to be
<a href="http://www.sgi.com/tech/stl/DefaultConstructible.html">Default Constructible</a>
</p>
<HR>
<H2><A NAME="impl">Implementation Notes</A></H2>
<p>optional&lt;T&gt; is currently implemented
using a custom aligned storage facility built from <code>alignment_of</code> and
<code>type_with_alignment</code> (both from Type Traits).
It uses a separate boolean flag to indicate the initialization state.<br>
Placement new with T's copy constructor and T's destructor
are explicitly used to initialize,copy and destroy optional values.<br>
As a result, T's default constructor is effectively by-passed, but the exception
guarantees are basic.<br>
It is planned to replace the current implementation with another with
stronger exception safety, such as a future boost::variant<T,nil_t>.
</p>
<HR>
<H2><A NAME="porta">Dependencies and Portability</A></H2>
<p>The implementation uses <code>type_traits/alignment_of.hpp</code>
and <code>type_traits/type_with_alignment.hpp</code></p>
<p>It has been tested on bcc5.5.1, vc6.0 and gcc2.95.2</p>
<HR>
<H2><A NAME="credits">Acknowledgments</A></H2>
<p>Pre-formal review:</p>
<blockquote>
<p>Peter Dimov suggested the name 'optional', and was the first to point out the
need for aligned storage<br>
Douglas Gregor developed 'type_with_alignment', and later Eric Friedman coded
'aligned_storage', which are the core of the optional class implementation.<br>
Andrei Alexandrescu and Brian Parker also worked with aligned storage techniques
and their work influenced the current implementation.<br>
Gennadiy Rozental made extensive and important comments which shaped the design.<br>
Vesa Karvonen and Douglas Gregor made quite useful comparisons between optional,
variant and any; and made other relevant comments. Douglas Gregor and Peter
Dimov commented on comparisons and evaluation in boolean contexts.<br>
Eric Friedman helped understand the issues involved with aligned storage, move/copy
operations and exception safety.<br>
Many others have participated with useful comments: Aleksey Gurotov, Kevlin
Henney, David Abrahams, and others I can't recall.
</p>
</blockquote>
<p>Post-formal review:</p>
<blockquote>
<p>William Kempf carefully considered the originally proposed interface
and suggested the new interface which is currently used. He also started
and fueled the discussion about the analogy optional&lt;&gt;/smart pointer
and about relational operators.<br>
Peter Dimov, Joel de Guzman, David Abrahams, Tanton Gibbs and Ian Hanson
focused on the relational semantics of optional (originally undefined);
concluding with the fact that the pointer-like interface doesn't make it
a pointer so it shall have deep relational operators.<br>
Augustus Saunders also explored the different relational
semantics between optional&lt;&gt; and a pointer and developed the
OptionalPointee concept as an aid against potential conflicts on generic code.<br>
Joel de Guzman noticed that optional&lt;&gt; can be seen as an
API on top of variant&lt;T,nil_t&gt;.<br>
Dave Gomboc explained the meaning and usage of the Haskell analog to optional&lt;&gt;:
the Maybe type constructor (analogy originally pointed out by David Sankel).<br>
Other comments were posted by Vincent Finn, Anthony Williams, Ed Brey, Rob Stewart,
and others.
</p>
</blockquote>
<HR>
<P>Revised January 20, 2003</P>
<P>&copy; Copyright boost.org 2003. 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>Developed by <A HREF="mailto:fernando_cacciola@hotmail.com">Fernando Cacciola</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.yahoogroups.com/list/boost">www.yahoogroups.com/list/boost</A>.
</P>
</BODY>
</HTML>