associative.html revision 1.5
1<?xml version="1.0" encoding="UTF-8" standalone="no"?> 2<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"><html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>Associative</title><meta name="generator" content="DocBook XSL-NS Stylesheets V1.78.1" /><meta name="keywords" content="ISO C++, library" /><meta name="keywords" content="ISO C++, runtime, library" /><link rel="home" href="../index.html" title="The GNU C++ Library" /><link rel="up" href="containers.html" title="Chapter��9.�� Containers" /><link rel="prev" href="containers.html" title="Chapter��9.�� Containers" /><link rel="next" href="unordered_associative.html" title="Unordered Associative" /></head><body><div class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="3" align="center">Associative</th></tr><tr><td width="20%" align="left"><a accesskey="p" href="containers.html">Prev</a>��</td><th width="60%" align="center">Chapter��9.�� 3 Containers 4 5</th><td width="20%" align="right">��<a accesskey="n" href="unordered_associative.html">Next</a></td></tr></table><hr /></div><div class="section"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="std.containers.associative"></a>Associative</h2></div></div></div><div class="section"><div class="titlepage"><div><div><h3 class="title"><a id="containers.associative.insert_hints"></a>Insertion Hints</h3></div></div></div><p> 6 Section [23.1.2], Table 69, of the C++ standard lists this 7 function for all of the associative containers (map, set, etc): 8 </p><pre class="programlisting"> 9 a.insert(p,t); 10 </pre><p> 11 where 'p' is an iterator into the container 'a', and 't' is the 12 item to insert. The standard says that <span class="quote">���<span class="quote"><code class="code">t</code> is 13 inserted as close as possible to the position just prior to 14 <code class="code">p</code>.</span>���</span> (Library DR #233 addresses this topic, 15 referring to <a class="link" href="http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1780.html" target="_top">N1780</a>. 16 Since version 4.2 GCC implements the resolution to DR 233, so 17 that insertions happen as close as possible to the hint. For 18 earlier releases the hint was only used as described below. 19 </p><p> 20 Here we'll describe how the hinting works in the libstdc++ 21 implementation, and what you need to do in order to take 22 advantage of it. (Insertions can change from logarithmic 23 complexity to amortized constant time, if the hint is properly 24 used.) Also, since the current implementation is based on the 25 SGI STL one, these points may hold true for other library 26 implementations also, since the HP/SGI code is used in a lot of 27 places. 28 </p><p> 29 In the following text, the phrases <span class="emphasis"><em>greater 30 than</em></span> and <span class="emphasis"><em>less than</em></span> refer to the 31 results of the strict weak ordering imposed on the container by 32 its comparison object, which defaults to (basically) 33 <span class="quote">���<span class="quote"><</span>���</span>. Using those phrases is semantically sloppy, 34 but I didn't want to get bogged down in syntax. I assume that if 35 you are intelligent enough to use your own comparison objects, 36 you are also intelligent enough to assign <span class="quote">���<span class="quote">greater</span>���</span> 37 and <span class="quote">���<span class="quote">lesser</span>���</span> their new meanings in the next 38 paragraph. *grin* 39 </p><p> 40 If the <code class="code">hint</code> parameter ('p' above) is equivalent to: 41 </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p> 42 <code class="code">begin()</code>, then the item being inserted should 43 have a key less than all the other keys in the container. 44 The item will be inserted at the beginning of the container, 45 becoming the new entry at <code class="code">begin()</code>. 46 </p></li><li class="listitem"><p> 47 <code class="code">end()</code>, then the item being inserted should have 48 a key greater than all the other keys in the container. The 49 item will be inserted at the end of the container, becoming 50 the new entry before <code class="code">end()</code>. 51 </p></li><li class="listitem"><p> 52 neither <code class="code">begin()</code> nor <code class="code">end()</code>, then: 53 Let <code class="code">h</code> be the entry in the container pointed to 54 by <code class="code">hint</code>, that is, <code class="code">h = *hint</code>. Then 55 the item being inserted should have a key less than that of 56 <code class="code">h</code>, and greater than that of the item preceding 57 <code class="code">h</code>. The new item will be inserted between 58 <code class="code">h</code> and <code class="code">h</code>'s predecessor. 59 </p></li></ul></div><p> 60 For <code class="code">multimap</code> and <code class="code">multiset</code>, the 61 restrictions are slightly looser: <span class="quote">���<span class="quote">greater than</span>���</span> 62 should be replaced by <span class="quote">���<span class="quote">not less than</span>���</span>and <span class="quote">���<span class="quote">less 63 than</span>���</span> should be replaced by <span class="quote">���<span class="quote">not greater 64 than.</span>���</span> (Why not replace greater with 65 greater-than-or-equal-to? You probably could in your head, but 66 the mathematicians will tell you that it isn't the same thing.) 67 </p><p> 68 If the conditions are not met, then the hint is not used, and the 69 insertion proceeds as if you had called <code class="code"> a.insert(t) 70 </code> instead. (<span class="emphasis"><em>Note </em></span> that GCC releases 71 prior to 3.0.2 had a bug in the case with <code class="code">hint == 72 begin()</code> for the <code class="code">map</code> and <code class="code">set</code> 73 classes. You should not use a hint argument in those releases.) 74 </p><p> 75 This behavior goes well with other containers' 76 <code class="code">insert()</code> functions which take an iterator: if used, 77 the new item will be inserted before the iterator passed as an 78 argument, same as the other containers. 79 </p><p> 80 <span class="emphasis"><em>Note </em></span> also that the hint in this 81 implementation is a one-shot. The older insertion-with-hint 82 routines check the immediately surrounding entries to ensure that 83 the new item would in fact belong there. If the hint does not 84 point to the correct place, then no further local searching is 85 done; the search begins from scratch in logarithmic time. 86 </p></div><div class="section"><div class="titlepage"><div><div><h3 class="title"><a id="containers.associative.bitset"></a>bitset</h3></div></div></div><div class="section"><div class="titlepage"><div><div><h4 class="title"><a id="associative.bitset.size_variable"></a>Size Variable</h4></div></div></div><p> 87 No, you cannot write code of the form 88 </p><pre class="programlisting"> 89 #include <bitset> 90 91 void foo (size_t n) 92 { 93 std::bitset<n> bits; 94 .... 95 } 96 </pre><p> 97 because <code class="code">n</code> must be known at compile time. Your 98 compiler is correct; it is not a bug. That's the way templates 99 work. (Yes, it <span class="emphasis"><em>is</em></span> a feature.) 100 </p><p> 101 There are a couple of ways to handle this kind of thing. Please 102 consider all of them before passing judgement. They include, in 103 no particular order: 104 </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>A very large N in <code class="code">bitset<N></code>.</p></li><li class="listitem"><p>A container<bool>.</p></li><li class="listitem"><p>Extremely weird solutions.</p></li></ul></div><p> 105 <span class="emphasis"><em>A very large N in 106 <code class="code">bitset<N></code>.����</em></span> It has been 107 pointed out a few times in newsgroups that N bits only takes up 108 (N/8) bytes on most systems, and division by a factor of eight is 109 pretty impressive when speaking of memory. Half a megabyte given 110 over to a bitset (recall that there is zero space overhead for 111 housekeeping info; it is known at compile time exactly how large 112 the set is) will hold over four million bits. If you're using 113 those bits as status flags (e.g., 114 <span class="quote">���<span class="quote">changed</span>���</span>/<span class="quote">���<span class="quote">unchanged</span>���</span> flags), that's a 115 <span class="emphasis"><em>lot</em></span> of state. 116 </p><p> 117 You can then keep track of the <span class="quote">���<span class="quote">maximum bit used</span>���</span> 118 during some testing runs on representative data, make note of how 119 many of those bits really need to be there, and then reduce N to 120 a smaller number. Leave some extra space, of course. (If you 121 plan to write code like the incorrect example above, where the 122 bitset is a local variable, then you may have to talk your 123 compiler into allowing that much stack space; there may be zero 124 space overhead, but it's all allocated inside the object.) 125 </p><p> 126 <span class="emphasis"><em>A container<bool>.����</em></span> The 127 Committee made provision for the space savings possible with that 128 (N/8) usage previously mentioned, so that you don't have to do 129 wasteful things like <code class="code">Container<char></code> or 130 <code class="code">Container<short int></code>. Specifically, 131 <code class="code">vector<bool></code> is required to be specialized for 132 that space savings. 133 </p><p> 134 The problem is that <code class="code">vector<bool></code> doesn't 135 behave like a normal vector anymore. There have been 136 journal articles which discuss the problems (the ones by Herb 137 Sutter in the May and July/August 1999 issues of C++ Report cover 138 it well). Future revisions of the ISO C++ Standard will change 139 the requirement for <code class="code">vector<bool></code> 140 specialization. In the meantime, <code class="code">deque<bool></code> 141 is recommended (although its behavior is sane, you probably will 142 not get the space savings, but the allocation scheme is different 143 than that of vector). 144 </p><p> 145 <span class="emphasis"><em>Extremely weird solutions.����</em></span> If 146 you have access to the compiler and linker at runtime, you can do 147 something insane, like figuring out just how many bits you need, 148 then writing a temporary source code file. That file contains an 149 instantiation of <code class="code">bitset</code> for the required number of 150 bits, inside some wrapper functions with unchanging signatures. 151 Have your program then call the compiler on that file using 152 Position Independent Code, then open the newly-created object 153 file and load those wrapper functions. You'll have an 154 instantiation of <code class="code">bitset<N></code> for the exact 155 <code class="code">N</code> that you need at the time. Don't forget to delete 156 the temporary files. (Yes, this <span class="emphasis"><em>can</em></span> be, and 157 <span class="emphasis"><em>has been</em></span>, done.) 158 </p><p> 159 This would be the approach of either a visionary genius or a 160 raving lunatic, depending on your programming and management 161 style. Probably the latter. 162 </p><p> 163 Which of the above techniques you use, if any, are up to you and 164 your intended application. Some time/space profiling is 165 indicated if it really matters (don't just guess). And, if you 166 manage to do anything along the lines of the third category, the 167 author would love to hear from you... 168 </p><p> 169 Also note that the implementation of bitset used in libstdc++ has 170 <a class="link" href="ext_containers.html#manual.ext.containers.sgi" title="Backwards Compatibility">some extensions</a>. 171 </p></div><div class="section"><div class="titlepage"><div><div><h4 class="title"><a id="associative.bitset.type_string"></a>Type String</h4></div></div></div><p> 172 </p><p> 173 Bitmasks do not take char* nor const char* arguments in their 174 constructors. This is something of an accident, but you can read 175 about the problem: follow the library's <span class="quote">���<span class="quote">Links</span>���</span> from 176 the homepage, and from the C++ information <span class="quote">���<span class="quote">defect 177 reflector</span>���</span> link, select the library issues list. Issue 178 number 116 describes the problem. 179 </p><p> 180 For now you can simply make a temporary string object using the 181 constructor expression: 182 </p><pre class="programlisting"> 183 std::bitset<5> b ( std::string("10110") ); 184 </pre><p> 185 instead of 186 </p><pre class="programlisting"> 187 std::bitset<5> b ( "10110" ); // invalid 188 </pre></div></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="containers.html">Prev</a>��</td><td width="20%" align="center"><a accesskey="u" href="containers.html">Up</a></td><td width="40%" align="right">��<a accesskey="n" href="unordered_associative.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">Chapter��9.�� 189 Containers 190 191��</td><td width="20%" align="center"><a accesskey="h" href="../index.html">Home</a></td><td width="40%" align="right" valign="top">��Unordered Associative</td></tr></table></div></body></html>