blob: 423905c110c342114be27bf7c0abdf75c5f5f80d [file] [log] [blame]
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0.1 Transitional//EN">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
<title>Boost.MultiIndex Documentation - Release notes</title>
<link rel="stylesheet" href="style.css" type="text/css">
<link rel="start" href="index.html">
<link rel="prev" href="future_work.html">
<link rel="up" href="index.html">
<link rel="next" href="acknowledgements.html">
</head>
<body>
<h1><img src="../../../boost.png" alt="boost.png (6897 bytes)" align=
"middle" width="277" height="86">Boost.MultiIndex Release notes</h1>
<div class="prev_link"><a href="future_work.html"><img src="prev.gif" alt="future work" border="0"><br>
Future work
</a></div>
<div class="up_link"><a href="index.html"><img src="up.gif" alt="index" border="0"><br>
Index
</a></div>
<div class="next_link"><a href="acknowledgements.html"><img src="next.gif" alt="acknowledgements" border="0"><br>
Acknowledgements
</a></div><br clear="all" style="clear: all;">
<hr>
<h2>Contents</h2>
<ul>
<li><a href="#boost_1_58">Boost 1.58 release</a></li>
<li><a href="#boost_1_57">Boost 1.57 release</a></li>
<li><a href="#boost_1_56">Boost 1.56 release</a></li>
<li><a href="#boost_1_55">Boost 1.55 release</a></li>
<li><a href="#boost_1_54">Boost 1.54 release</a></li>
<li><a href="#boost_1_49">Boost 1.49 release</a></li>
<li><a href="#boost_1_48">Boost 1.48 release</a></li>
<li><a href="#boost_1_47">Boost 1.47 release</a></li>
<li><a href="#boost_1_44">Boost 1.44 release</a></li>
<li><a href="#boost_1_43">Boost 1.43 release</a></li>
<li><a href="#boost_1_42">Boost 1.42 release</a></li>
<li><a href="#boost_1_41">Boost 1.41 release</a></li>
<li><a href="#boost_1_38">Boost 1.38 release</a></li>
<li><a href="#boost_1_37">Boost 1.37 release</a></li>
<li><a href="#boost_1_36">Boost 1.36 release</a></li>
<li><a href="#boost_1_35">Boost 1.35 release</a></li>
<li><a href="#boost_1_34">Boost 1.34 release</a></li>
<li><a href="#boost_1_33_1">Boost 1.33.1 release</a></li>
<li><a href="#boost_1_33">Boost 1.33 release</a></li>
</ul>
<h2><a name="boost_1_58">Boost 1.58 release</a></h2>
<p>
<ul>
<li>The efficiency of lookup operations has improved in situations where they involve
the generation of temporary values of <code>key_type</code>. Consider for instance
the following code:
<blockquote><pre>
<span class=keyword>typedef</span> <span class=identifier>multi_index_container</span><span class=special>&lt;</span>
<span class=identifier>std</span><span class=special>::</span><span class=identifier>string</span><span class=special>,</span>
<span class=identifier>indexed_by</span><span class=special>&lt;</span>
<span class=identifier>ordered_unique</span><span class=special>&lt;</span><span class=identifier>identity</span><span class=special>&lt;</span><span class=identifier>std</span><span class=special>::</span><span class=identifier>string</span><span class=special>&gt;</span> <span class=special>&gt;</span>
<span class=special>&gt;</span>
<span class=special>&gt;</span> <span class=identifier>multi_t</span><span class=special>;</span>
<span class=special>...</span>
<span class=identifier>multi_t</span> <span class=identifier>m</span><span class=special>=...;</span>
<span class=identifier>m</span><span class=special>.</span><span class=identifier>find</span><span class=special>(</span><span class=string>&quot;boost&quot;</span><span class=special>);</span> <span class=comment>// passed a const char*, not a std::string</span>
</pre></blockquote>
In previous versions of the library, the <code>find</code> operation generates
several temporary <code>std::string</code>s (one every time an internal comparison is made).
In sufficiently advanced compilers, this is now avoided so that only one temporary is
created.
</li>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_57">Boost 1.57 release</a></h2>
<p>
<ul>
<li>When <code>std::tuple</code>s are available, these can be used for lookup operations
in indices equipped with composite keys. <code>boost::tuple</code>s are also supported
for backwards compatibility.
</li>
</ul>
</p>
<h2><a name="boost_1_56">Boost 1.56 release</a></h2>
<p>
<ul>
<li>The <code>erase(iterator)</code> member function of hashed indices
used to have poor performance under low load conditions due to the requirement
that an iterator to the next element must be returned (see ticket
<a href="https://svn.boost.org/trac/boost/ticket/4264">#4264</a>). In accordance with
the resolution of <a href="http://lwg.github.io/issues/lwg-closed.html#579">LWG
issue #579</a>, this problem has been fixed while maintaining the interface of
<code>erase</code>, at the expense of using one more
word of memory per element. In fact, C++ complexity requirements on unordered
associative containers have been improved for hashed indices so that
<ul>
<li>deletion of a given element is unconditionally constant-time,</li>
<li>worst-case performance is not <code>O(n)</code> but <code>O(n<sub>dist</sub>)</code>,
where <code>n<sub>dist</sub></code> is the number of non-equivalent elements in the index.
</li>
</ul>
Due to the fact that hashed indices rely on a new data structure, the internal representation of
their iterators and local iterators have changed, which affects serialization: their corresponding
serialization <a href="../../serialization/doc/tutorial.html#versioning">class version</a> has been
bumped from 0 to 1. Old archives involving hashed index (local) iterators can be loaded
by Boost 1.56 version of Boost.MultiIndex, but not the other way around.
</li>
<li>Hashed indices now provide <code>reserve</code>.</li>
<li>Hashed indices can now be checked for equality and inequality following the
(suitably adapted) C++ standard specification in <b>[unord.req]</b>.</li>
<li>The public interface of Boost.MultiIndex provide <code>noexcept</code> specifications
where appropriate (for compliant compilers).
</li>
<li>Improved performance of failed insertions into a <code>multi_index_container</code>.</li>
<li>Much internal code aimed at supporting MSVC++ 7.0 and prior has been removed.
Compilation times without this legacy code might be slightly faster.
</li>
<li>Fixed a bug with insertion via iterators dereferencing to rvalues
(ticket <a href="https://svn.boost.org/trac/boost/ticket/9665">#9665</a>).
</li>
<li>Made Boost.MultiIndex compatible with <code>BOOST_BIND_NO_PLACEHOLDERS</code>
(ticket <a href="https://svn.boost.org/trac/boost/ticket/9798">#9798</a>).
</li>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_55">Boost 1.55 release</a></h2>
<p>
<ul>
<li>Boost.MultiIndex has been brought to a higher level of compliance
with C++11.
<ul>
<li><code>multi_index_container</code> is now efficiently movable.</li>
<li>Initializer lists supported.</li>
<li>Emplace functions provided.</li>
<li>Non-copyable elements (such as <code>std::unique_ptr&lt;T&gt;</code>) supported. This includes
insertion of a range [<code>first</code>,<code>last</code>) where the iterators point to a type that is
convertible to that of the element: no copy construction happens in the process.
</li>
<li>Random access indices provide <code>shrink_to_fit()</code>.</li>
</ul>
Refer to the <a href="compiler_specifics.html">compiler specifics</a> section for limitations
on pre-C++11 compilers.
</li>
<li>The following classes are deprecated:
<ul>
<li><a href="reference/key_extraction.html#member_offset"><code>member_offset</code></a>,</li>
<li><a href="reference/key_extraction.html#const_mem_fun_explicit"><code>const_mem_fun_explicit</code></a>,</li>
<li><a href="reference/key_extraction.html#const_mem_fun_explicit"><code>mem_fun_explicit</code></a>,</li>
<li><a href="reference/key_extraction.html#composite_key_result_equal_to"><code>composite_key_result_equal_to</code></a>,</li>
<li><a href="reference/key_extraction.html#composite_key_result_less"><code>composite_key_result_less</code></a>,</li>
<li><a href="reference/key_extraction.html#composite_key_result_greater"><code>composite_key_result_greater</code></a>,</li>
<li><a href="reference/key_extraction.html#composite_key_result_hash"><code>composite_key_result_hash</code></a>.</li>
</ul>
</li>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_54">Boost 1.54 release</a></h2>
<p>
<ul>
<li>Suppressed some potential warnings described in
tickets <a href="https://svn.boost.org/trac/boost/ticket/8034">#8034</a> and <a href="https://svn.boost.org/trac/boost/ticket/8129">#8129</a>.
</li>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_49">Boost 1.49 release</a></h2>
<p>
<ul>
<li>Suppressed a potential narrow conversion warning described in
ticket <a href="https://svn.boost.org/trac/boost/ticket/3365">#3365</a>.
</li>
</ul>
</p>
<h2><a name="boost_1_48">Boost 1.48 release</a></h2>
<p>
<ul>
<li>Fixed a compatibility problem with
<a href="../../../doc/html/foreach.html">Boost.Foreach</a>
(ticket <a href="https://svn.boost.org/trac/boost/ticket/5741">#5741</a>).
</li>
</ul>
</p>
<h2><a name="boost_1_47">Boost 1.47 release</a></h2>
<p>
<ul>
<li>Fixed an ADL problem in
<a href="reference/key_extraction.html#composite_key_compare"><code>composite_key_compare</code></a>
related with <a href="../../../doc/html/ref.html">Boost.Ref</a>.
</li>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_44">Boost 1.44 release</a></h2>
<p>
<ul>
<li>
Fixed a bug preventing the use of <code>modify_key</code> with
rollback in <a href="reference/ord_indices.html#modify_key">ordered</a> and
<a href="reference/hash_indices.html#modify_key">hashed</a> indices when
<code>Modifier</code> and <code>Rollback</code> are different types
(ticket <a href="https://svn.boost.org/trac/boost/ticket/4130">#4130</a>).
</li>
</ul>
</p>
<h2><a name="boost_1_43">Boost 1.43 release</a></h2>
<p>
<ul>
<li>
<a href="../../serialization/doc/serialization.html#constructors">Serialization
of non default constructible values</a> is now properly supported
through user-provided facilities <code>save_construct_data</code> and
<code>load_construct_data</code>.
<code>multi_index_container</code> serialization
<a href="../../serialization/doc/tutorial.html#versioning">class version</a> has been
bumped from 1 to 2.
</li>
</ul>
</p>
<h2><a name="boost_1_42">Boost 1.42 release</a></h2>
<p>
<ul>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_41">Boost 1.41 release</a></h2>
<p>
<ul>
<li>Serialization now uses the portable
<a href="../../serialization/doc/wrappers.html#collection_size_type"><code>collection_size_type</code></a>
type instead of the original <code>std::size_t</code> (ticket
<a href="https://svn.boost.org/trac/boost/ticket/3365">#3365</a>).
<code>multi_index_container</code> serialization
<a href="../../serialization/doc/tutorial.html#versioning">class version</a> has been
bumped from 0 to 1.
</li>
<li>Fixed a concurrency bug in the implementation of
<a href="tutorial/debug.html#safe_mode">safe mode</a>
(ticket <a href="https://svn.boost.org/trac/boost/ticket/3462">#3462</a>).
</li>
</ul>
</p>
<h2><a name="boost_1_38">Boost 1.38 release</a></h2>
<p>
<ul>
<li>These constructs are deprecated:
<ul>
<li><code>nth_index_iterator&lt;MultiIndexContainer,N&gt;::type</code>,</li>
<li><code>multi_index_container&lt;...&gt;::nth_index_iterator&lt;N&gt;::type</code>,</li>
<li><code>nth_index_const_iterator&lt;MultiIndexContainer,N&gt;::type</code>,</li>
<li><code>multi_index_container&lt;...&gt;::nth_index_const_iterator&lt;N&gt;::type</code>,</li>
<li><code>index_iterator&lt;MultiIndexContainer,Tag&gt;::type</code>,</li>
<li><code>multi_index_container&lt;...&gt;::index_iterator&lt;Tag&gt;::type</code>,</li>
<li><code>index_const_iterator&lt;MultiIndexContainer,Tag&gt;::type</code>,</li>
<li><code>multi_index_container&lt;...&gt;::index_const_iterator&lt;Tag&gt;::type</code>.</li>
</ul>
Use the following instead:
<ul>
<li><code>nth_index&lt;MultiIndexContainer,N&gt;::type::iterator</code>,</li>
<li><code>multi_index_container&lt;...&gt;::nth_index&lt;N&gt;::type::iterator</code>,</li>
<li><code>nth_index&lt;MultiIndexContainer,N&gt;::type::const_iterator</code>,</li>
<li><code>multi_index_container&lt;...&gt;::nth_index&lt;N&gt;::type::const_iterator</code>,</li>
<li><code>index&lt;MultiIndexContainer,Tag&gt;::type::iterator</code>,</li>
<li><code>multi_index_container&lt;...&gt;::index&lt;Tag&gt;::type::iterator</code>,</li>
<li><code>index&lt;MultiIndexContainer,Tag&gt;::type::const_iterator</code>,</li>
<li><code>multi_index_container&lt;...&gt;::index&lt;Tag&gt;::type::const_iterator</code>.</li>
</ul>
</li>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_37">Boost 1.37 release</a></h2>
<p>
<ul>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_36">Boost 1.36 release</a></h2>
<p>
<ul>
<li><a name="stable_update">On prior versions of the library, the <a href="tutorial/indices.html#hash_updating">update
member functions</a> of hashed indices could alter the position of an element even if the
associated key did not change with the update. This is legal but probably unexpected behavior.
The functions have been rewritten to provide the additional guarantee that elements with
unmodified key will not change position in hashed indices, just as always was the case with
ordered indices. These guarantees are now documented in the reference.</a></li>
<li>Added the constructor <code>multi_index_container::multi_index_container(const allocator_type&amp;)</code>
to mimic the equivalent interface in STL sequence containers.
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_35">Boost 1.35 release</a></h2>
<p>
<ul>
<li>New <a href="tutorial/key_extraction.html#global_fun"><code>global_fun</code></a>
predefined key extractor.
</li>
<li>Added <a href="tutorial/indices.html#iterator_to"><code>iterator_to</code></a>
facility.
</li>
<li>Included <a href="tutorial/creation.html#special_allocator">support for
non-standard allocators</a> such as those of
<a href="../../interprocess/index.html">Boost.Interprocess</a>, which makes
<code>multi_index_container</code>s placeable in shared memory.
</li>
<li>New versions of <code>modify</code> and <code>modify_key</code> with
rollback, as described in the
<a href="tutorial/basics.html#ord_updating">tutorial</a>.
</li>
<li>Indices provide the new <code>cbegin</code>, <code>cend</code> and,
when applicable, <code>crbegin</code> and <code>crend</code>
member functions, in accordance with the latest drafts of the next
revision of the C++ standard.
</li>
<li>Hinted insertion in ordered indices fully conforms to the resolutions of
C++ Standard Library
<a href="http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-defects.html#233">Defect
Report 233</a>. The new requirement that the point of insertion
be always as close as possible to the hint induces a different behavior than
exhibited in former releases of Boost.MultiIndex, which can potentially cause
backwards compatibility problems; in any case, the likelihood of these
compatibility issues arising in a real scenario is very low.
</li>
<li>Sequenced and random access indices now follow the requirements of the
C++ standard for sequence containers with respect to the operations
<code>assign(f,l)</code> and <code>insert(p,f,l)</code> (23.1.1/9): if
<code>f</code> and <code>l</code> are of the same integral type, the
iterator-based overloads of these member functions are avoided:
<blockquote><pre>
<span class=keyword>typedef</span> <span class=identifier>multi_index_container</span><span class=special>&lt;</span>
<span class=keyword>int</span><span class=special>,</span><span class=identifier>indexed_by</span><span class=special>&lt;</span><span class=identifier>sequenced</span><span class=special>&lt;&gt;</span> <span class=special>&gt;</span>
<span class=special>&gt;</span> <span class=identifier>sequenced_container</span><span class=special>;</span>
<span class=identifier>std</span><span class=special>::</span><span class=identifier>list</span><span class=special>&lt;</span><span class=keyword>int</span><span class=special>&gt;</span> <span class=identifier>l</span><span class=special>(...);</span>
<span class=identifier>sequenced_container</span> <span class=identifier>c</span><span class=special>;</span>
<span class=comment>// iterator-based overload of assign</span>
<span class=identifier>c</span><span class=special>.</span><span class=identifier>assign</span><span class=special>(</span><span class=identifier>l</span><span class=special>.</span><span class=identifier>begin</span><span class=special>(),</span><span class=identifier>l</span><span class=special>.</span><span class=identifier>end</span><span class=special>());</span>
<span class=comment>// The following is equivalent to
// c.assign(
// static_cast&lt;sequenced_container::size_type&gt;(10),100);
// that is, &quot;10&quot; and &quot;100&quot; are not taken to be iterators as
// in the previous expression.</span>
<span class=identifier>c</span><span class=special>.</span><span class=identifier>assign</span><span class=special>(</span><span class=number>10</span><span class=special>,</span><span class=number>100</span><span class=special>);</span>
</pre></blockquote>
</li>
<li>The performance of ordered indices <code>range</code> and
<code>equal_range</code> has been improved.
</li>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_34">Boost 1.34 release</a></h2>
<p>
<ul>
<li>Added <a href="tutorial/indices.html#rnd_indices">random access
indices</a>.
</li>
<li>Non key-based indices provide new
<a href="tutorial/indices.html#rearrange">rearrange facilities</a>
allowing for interaction with external mutating algorithms.
</li>
<li>All predefined Boost.MultiIndex key extractors
instantiated for a given type <code>T</code> can handle objects of types
derived from or convertible to <code>T</code> (and
<a href="reference/key_extraction.html#chained_pointers">chained pointers</a>
to those). Previously, only objects of the exact type specified (along with
<code>reference_wrapper</code>s and chained pointers to them) were accepted.
</li>
<li><a href="reference/key_extraction.html#composite_key_compare"><code>composite_key_compare</code></a>
and related classes accept operands not included in tuples as if they were passed
in a tuple of length 1; this allows the user to omit tuple enclosing in
lookup operations involving composite keys when only the first key is provided.
</li>
<li>The core algorithms of ordered indices have been optimized, yielding
an estimated reduction of about 5% in insertion times.
</li>
<li>Size of ordered indices node headers have been reduced by 25% on
most platforms, using a well known
<a href="tutorial/indices.html#ordered_node_compression">optimization
technique</a>.
</li>
<li>The tutorial has been restructured, new examples added.</li>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_33_1">Boost 1.33.1 release</a></h2>
<p>
<ul>
<li>For ordered and hashed indices, <code>erase(it)</code> and
<code>erase(first,last)</code> now return an iterator to the element
following those being deleted (previously nothing was returned), in
accordance with the C++ Standard Library
<a href="http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-defects.html#130">Defect
Report 130</a> and issue 6.19 of TR1
<a href="http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1837.pdf">Issues
List</a>.
</li>
<li>Boost.MultiIndex offers the usual guarantees with respect to
multithreading code provided by most STL implementations:
<ol>
<li>Concurrent access to different containers is safe.</li>
<li>Concurrent read-only access to the same container is safe.</li>
</ol>
In previous versions of the library, the latter guarantee was not properly
maintained if the <a href="tutorial/debug.html#safe_mode">safe
mode</a> was set. This problem has been fixed now.
</li>
<li>Maintenance fixes.</li>
</ul>
</p>
<h2><a name="boost_1_33">Boost 1.33 release</a></h2>
<p>
<ul>
<li>Added <a href="tutorial/indices.html#hashed_indices">hashed indices</a>,
whose interface is based on the specification for unordered associative
containers by the C++ Standard Library Technical Report (TR1).
</li>
<li>Added <a href="tutorial/creation.html#serialization">serialization support</a>
for <a href="../../serialization/index.html">Boost.Serialization</a>.
</li>
<li>Destruction of <code>multi_index_container</code>s and <code>clear</code>
memfuns now perform faster.
</li>
<li>Internal changes aimed at reducing the length of symbol names generated
by the compiler; cuts of up to a 50% can be achieved with respect to the
Boost 1.32 release. This results in much shorter and more readable error
messages and has also a beneficial impact on compilers with strict limits on
symbol name lengths. Additionally, a section on further
<a href="compiler_specifics.html#symbol_reduction">reduction of symbol name
lengths</a> has been added.
</li>
<li>Restructured some parts of the documentation, new examples.</li>
<li>Maintenance fixes.</li>
</ul>
</p>
<hr>
<div class="prev_link"><a href="future_work.html"><img src="prev.gif" alt="future work" border="0"><br>
Future work
</a></div>
<div class="up_link"><a href="index.html"><img src="up.gif" alt="index" border="0"><br>
Index
</a></div>
<div class="next_link"><a href="acknowledgements.html"><img src="next.gif" alt="acknowledgements" border="0"><br>
Acknowledgements
</a></div><br clear="all" style="clear: all;">
<br>
<p>Revised November 12th 2014</p>
<p>&copy; Copyright 2003-2014 Joaqu&iacute;n M L&oacute;pez Mu&ntilde;oz.
Distributed under the Boost Software
License, Version 1.0. (See accompanying file <a href="../../../LICENSE_1_0.txt">
LICENSE_1_0.txt</a> or copy at <a href="http://www.boost.org/LICENSE_1_0.txt">
http://www.boost.org/LICENSE_1_0.txt</a>)
</p>
</body>
</html>