xref: /netbsd-src/external/gpl3/gcc.old/dist/libstdc++-v3/doc/html/manual/strings.html (revision a2dc1f3faca890bc62c61c70cbcb4657d1fe6044)
136ac495dSmrg<?xml version="1.0" encoding="UTF-8" standalone="no"?>
236ac495dSmrg<!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>Chapter 7.  Strings</title><meta name="generator" content="DocBook XSL Stylesheets Vsnapshot" /><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="std_contents.html" title="Part II.  Standard Contents" /><link rel="prev" href="traits.html" title="Traits" /><link rel="next" href="localization.html" title="Chapter 8.  Localization" /></head><body><div class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="3" align="center">Chapter 7. 
336ac495dSmrg  Strings
436ac495dSmrg
536ac495dSmrg</th></tr><tr><td width="20%" align="left"><a accesskey="p" href="traits.html">Prev</a> </td><th width="60%" align="center">Part II. 
636ac495dSmrg    Standard Contents
736ac495dSmrg  </th><td width="20%" align="right"> <a accesskey="n" href="localization.html">Next</a></td></tr></table><hr /></div><div class="chapter"><div class="titlepage"><div><div><h2 class="title"><a id="std.strings"></a>Chapter 7. 
836ac495dSmrg  Strings
936ac495dSmrg  <a id="id-1.3.4.5.1.1.1" class="indexterm"></a>
1036ac495dSmrg</h2></div></div></div><div class="toc"><p><strong>Table of Contents</strong></p><dl class="toc"><dt><span class="section"><a href="strings.html#std.strings.string">String Classes</a></span></dt><dd><dl><dt><span class="section"><a href="strings.html#strings.string.simple">Simple Transformations</a></span></dt><dt><span class="section"><a href="strings.html#strings.string.case">Case Sensitivity</a></span></dt><dt><span class="section"><a href="strings.html#strings.string.character_types">Arbitrary Character Types</a></span></dt><dt><span class="section"><a href="strings.html#strings.string.token">Tokenizing</a></span></dt><dt><span class="section"><a href="strings.html#strings.string.shrink">Shrink to Fit</a></span></dt><dt><span class="section"><a href="strings.html#strings.string.Cstring">CString (MFC)</a></span></dt></dl></dd></dl></div><div class="section"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="std.strings.string"></a>String Classes</h2></div></div></div><div class="section"><div class="titlepage"><div><div><h3 class="title"><a id="strings.string.simple"></a>Simple Transformations</h3></div></div></div><p>
1136ac495dSmrg      Here are Standard, simple, and portable ways to perform common
1236ac495dSmrg      transformations on a <code class="code">string</code> instance, such as
1336ac495dSmrg      "convert to all upper case." The word transformations
1436ac495dSmrg      is especially apt, because the standard template function
1536ac495dSmrg      <code class="code">transform&lt;&gt;</code> is used.
1636ac495dSmrg   </p><p>
1736ac495dSmrg     This code will go through some iterations.  Here's a simple
1836ac495dSmrg     version:
1936ac495dSmrg   </p><pre class="programlisting">
2036ac495dSmrg   #include &lt;string&gt;
2136ac495dSmrg   #include &lt;algorithm&gt;
2236ac495dSmrg   #include &lt;cctype&gt;      // old &lt;ctype.h&gt;
2336ac495dSmrg
2436ac495dSmrg   struct ToLower
2536ac495dSmrg   {
2636ac495dSmrg     char operator() (char c) const  { return std::tolower(c); }
2736ac495dSmrg   };
2836ac495dSmrg
2936ac495dSmrg   struct ToUpper
3036ac495dSmrg   {
3136ac495dSmrg     char operator() (char c) const  { return std::toupper(c); }
3236ac495dSmrg   };
3336ac495dSmrg
3436ac495dSmrg   int main()
3536ac495dSmrg   {
3636ac495dSmrg     std::string  s ("Some Kind Of Initial Input Goes Here");
3736ac495dSmrg
3836ac495dSmrg     // Change everything into upper case
3936ac495dSmrg     std::transform (s.begin(), s.end(), s.begin(), ToUpper());
4036ac495dSmrg
4136ac495dSmrg     // Change everything into lower case
4236ac495dSmrg     std::transform (s.begin(), s.end(), s.begin(), ToLower());
4336ac495dSmrg
4436ac495dSmrg     // Change everything back into upper case, but store the
4536ac495dSmrg     // result in a different string
4636ac495dSmrg     std::string  capital_s;
4736ac495dSmrg     capital_s.resize(s.size());
4836ac495dSmrg     std::transform (s.begin(), s.end(), capital_s.begin(), ToUpper());
4936ac495dSmrg   }
5036ac495dSmrg   </pre><p>
5136ac495dSmrg     <span class="emphasis"><em>Note</em></span> that these calls all
5236ac495dSmrg      involve the global C locale through the use of the C functions
5336ac495dSmrg      <code class="code">toupper/tolower</code>.  This is absolutely guaranteed to work --
5436ac495dSmrg      but <span class="emphasis"><em>only</em></span> if the string contains <span class="emphasis"><em>only</em></span> characters
5536ac495dSmrg      from the basic source character set, and there are <span class="emphasis"><em>only</em></span>
5636ac495dSmrg      96 of those.  Which means that not even all English text can be
5736ac495dSmrg      represented (certain British spellings, proper names, and so forth).
5836ac495dSmrg      So, if all your input forevermore consists of only those 96
5936ac495dSmrg      characters (hahahahahaha), then you're done.
6036ac495dSmrg   </p><p><span class="emphasis"><em>Note</em></span> that the
6136ac495dSmrg      <code class="code">ToUpper</code> and <code class="code">ToLower</code> function objects
6236ac495dSmrg      are needed because <code class="code">toupper</code> and <code class="code">tolower</code>
6336ac495dSmrg      are overloaded names (declared in <code class="code">&lt;cctype&gt;</code> and
6436ac495dSmrg      <code class="code">&lt;locale&gt;</code>) so the template-arguments for
6536ac495dSmrg      <code class="code">transform&lt;&gt;</code> cannot be deduced, as explained in
6636ac495dSmrg      <a class="link" href="http://gcc.gnu.org/ml/libstdc++/2002-11/msg00180.html" target="_top">this
6736ac495dSmrg      message</a>.
6836ac495dSmrg
6936ac495dSmrg      At minimum, you can write short wrappers like
7036ac495dSmrg   </p><pre class="programlisting">
7136ac495dSmrg   char toLower (char c)
7236ac495dSmrg   {
73*a2dc1f3fSmrg      // std::tolower(c) is undefined if c &lt; 0 so cast to unsigned char.
74*a2dc1f3fSmrg      return std::tolower((unsigned char)c);
7536ac495dSmrg   } </pre><p>(Thanks to James Kanze for assistance and suggestions on all of this.)
7636ac495dSmrg   </p><p>Another common operation is trimming off excess whitespace.  Much
7736ac495dSmrg      like transformations, this task is trivial with the use of string's
7836ac495dSmrg      <code class="code">find</code> family.  These examples are broken into multiple
7936ac495dSmrg      statements for readability:
8036ac495dSmrg   </p><pre class="programlisting">
8136ac495dSmrg   std::string  str (" \t blah blah blah    \n ");
8236ac495dSmrg
8336ac495dSmrg   // trim leading whitespace
8436ac495dSmrg   string::size_type  notwhite = str.find_first_not_of(" \t\n");
8536ac495dSmrg   str.erase(0,notwhite);
8636ac495dSmrg
8736ac495dSmrg   // trim trailing whitespace
8836ac495dSmrg   notwhite = str.find_last_not_of(" \t\n");
8936ac495dSmrg   str.erase(notwhite+1); </pre><p>Obviously, the calls to <code class="code">find</code> could be inserted directly
9036ac495dSmrg      into the calls to <code class="code">erase</code>, in case your compiler does not
9136ac495dSmrg      optimize named temporaries out of existence.
9236ac495dSmrg   </p></div><div class="section"><div class="titlepage"><div><div><h3 class="title"><a id="strings.string.case"></a>Case Sensitivity</h3></div></div></div><p>
9336ac495dSmrg    </p><p>The well-known-and-if-it-isn't-well-known-it-ought-to-be
9436ac495dSmrg      <a class="link" href="http://www.gotw.ca/gotw/" target="_top">Guru of the Week</a>
9536ac495dSmrg      discussions held on Usenet covered this topic in January of 1998.
9636ac495dSmrg      Briefly, the challenge was, <span class="quote">“<span class="quote">write a 'ci_string' class which
9736ac495dSmrg      is identical to the standard 'string' class, but is
9836ac495dSmrg      case-insensitive in the same way as the (common but nonstandard)
9936ac495dSmrg      C function stricmp()</span>”</span>.
10036ac495dSmrg   </p><pre class="programlisting">
10136ac495dSmrg   ci_string s( "AbCdE" );
10236ac495dSmrg
10336ac495dSmrg   // case insensitive
10436ac495dSmrg   assert( s == "abcde" );
10536ac495dSmrg   assert( s == "ABCDE" );
10636ac495dSmrg
10736ac495dSmrg   // still case-preserving, of course
10836ac495dSmrg   assert( strcmp( s.c_str(), "AbCdE" ) == 0 );
10936ac495dSmrg   assert( strcmp( s.c_str(), "abcde" ) != 0 ); </pre><p>The solution is surprisingly easy.  The original answer was
11036ac495dSmrg   posted on Usenet, and a revised version appears in Herb Sutter's
11136ac495dSmrg   book <span class="emphasis"><em>Exceptional C++</em></span> and on his website as <a class="link" href="http://www.gotw.ca/gotw/029.htm" target="_top">GotW 29</a>.
11236ac495dSmrg   </p><p>See?  Told you it was easy!</p><p>
11336ac495dSmrg     <span class="emphasis"><em>Added June 2000:</em></span> The May 2000 issue of C++
11436ac495dSmrg     Report contains a fascinating <a class="link" href="http://lafstern.org/matt/col2_new.pdf" target="_top"> article</a> by
11536ac495dSmrg     Matt Austern (yes, <span class="emphasis"><em>the</em></span> Matt Austern) on why
11636ac495dSmrg     case-insensitive comparisons are not as easy as they seem, and
11736ac495dSmrg     why creating a class is the <span class="emphasis"><em>wrong</em></span> way to go
11836ac495dSmrg     about it in production code.  (The GotW answer mentions one of
11936ac495dSmrg     the principle difficulties; his article mentions more.)
12036ac495dSmrg   </p><p>Basically, this is "easy" only if you ignore some things,
12136ac495dSmrg      things which may be too important to your program to ignore.  (I chose
12236ac495dSmrg      to ignore them when originally writing this entry, and am surprised
12336ac495dSmrg      that nobody ever called me on it...)  The GotW question and answer
12436ac495dSmrg      remain useful instructional tools, however.
12536ac495dSmrg   </p><p><span class="emphasis"><em>Added September 2000:</em></span>  James Kanze provided a link to a
12636ac495dSmrg      <a class="link" href="http://www.unicode.org/reports/tr21/tr21-5.html" target="_top">Unicode
12736ac495dSmrg      Technical Report discussing case handling</a>, which provides some
12836ac495dSmrg      very good information.
12936ac495dSmrg   </p></div><div class="section"><div class="titlepage"><div><div><h3 class="title"><a id="strings.string.character_types"></a>Arbitrary Character Types</h3></div></div></div><p>
13036ac495dSmrg    </p><p>The <code class="code">std::basic_string</code> is tantalizingly general, in that
13136ac495dSmrg      it is parameterized on the type of the characters which it holds.
13236ac495dSmrg      In theory, you could whip up a Unicode character class and instantiate
13336ac495dSmrg      <code class="code">std::basic_string&lt;my_unicode_char&gt;</code>, or assuming
13436ac495dSmrg      that integers are wider than characters on your platform, maybe just
13536ac495dSmrg      declare variables of type <code class="code">std::basic_string&lt;int&gt;</code>.
13636ac495dSmrg   </p><p>That's the theory.  Remember however that basic_string has additional
13736ac495dSmrg      type parameters, which take default arguments based on the character
13836ac495dSmrg      type (called <code class="code">CharT</code> here):
13936ac495dSmrg   </p><pre class="programlisting">
14036ac495dSmrg      template &lt;typename CharT,
14136ac495dSmrg		typename Traits = char_traits&lt;CharT&gt;,
14236ac495dSmrg		typename Alloc = allocator&lt;CharT&gt; &gt;
14336ac495dSmrg      class basic_string { .... };</pre><p>Now, <code class="code">allocator&lt;CharT&gt;</code> will probably Do The Right
14436ac495dSmrg      Thing by default, unless you need to implement your own allocator
14536ac495dSmrg      for your characters.
14636ac495dSmrg   </p><p>But <code class="code">char_traits</code> takes more work.  The char_traits
14736ac495dSmrg      template is <span class="emphasis"><em>declared</em></span> but not <span class="emphasis"><em>defined</em></span>.
14836ac495dSmrg      That means there is only
14936ac495dSmrg   </p><pre class="programlisting">
15036ac495dSmrg      template &lt;typename CharT&gt;
15136ac495dSmrg	struct char_traits
15236ac495dSmrg	{
15336ac495dSmrg	    static void foo (type1 x, type2 y);
15436ac495dSmrg	    ...
15536ac495dSmrg	};</pre><p>and functions such as char_traits&lt;CharT&gt;::foo() are not
15636ac495dSmrg      actually defined anywhere for the general case.  The C++ standard
15736ac495dSmrg      permits this, because writing such a definition to fit all possible
15836ac495dSmrg      CharT's cannot be done.
15936ac495dSmrg   </p><p>The C++ standard also requires that char_traits be specialized for
16036ac495dSmrg      instantiations of <code class="code">char</code> and <code class="code">wchar_t</code>, and it
16136ac495dSmrg      is these template specializations that permit entities like
16236ac495dSmrg      <code class="code">basic_string&lt;char,char_traits&lt;char&gt;&gt;</code> to work.
16336ac495dSmrg   </p><p>If you want to use character types other than char and wchar_t,
16436ac495dSmrg      such as <code class="code">unsigned char</code> and <code class="code">int</code>, you will
16536ac495dSmrg      need suitable specializations for them.  For a time, in earlier
16636ac495dSmrg      versions of GCC, there was a mostly-correct implementation that
16736ac495dSmrg      let programmers be lazy but it broke under many situations, so it
16836ac495dSmrg      was removed.  GCC 3.4 introduced a new implementation that mostly
16936ac495dSmrg      works and can be specialized even for <code class="code">int</code> and other
17036ac495dSmrg      built-in types.
17136ac495dSmrg   </p><p>If you want to use your own special character class, then you have
17236ac495dSmrg      <a class="link" href="http://gcc.gnu.org/ml/libstdc++/2002-08/msg00163.html" target="_top">a lot
17336ac495dSmrg      of work to do</a>, especially if you with to use i18n features
17436ac495dSmrg      (facets require traits information but don't have a traits argument).
17536ac495dSmrg   </p><p>Another example of how to specialize char_traits was given <a class="link" href="http://gcc.gnu.org/ml/libstdc++/2002-08/msg00260.html" target="_top">on the
17636ac495dSmrg      mailing list</a> and at a later date was put into the file <code class="code">
17736ac495dSmrg      include/ext/pod_char_traits.h</code>.  We agree
17836ac495dSmrg      that the way it's used with basic_string (scroll down to main())
17936ac495dSmrg      doesn't look nice, but that's because <a class="link" href="http://gcc.gnu.org/ml/libstdc++/2002-08/msg00236.html" target="_top">the
18036ac495dSmrg      nice-looking first attempt</a> turned out to <a class="link" href="http://gcc.gnu.org/ml/libstdc++/2002-08/msg00242.html" target="_top">not
18136ac495dSmrg      be conforming C++</a>, due to the rule that CharT must be a POD.
18236ac495dSmrg      (See how tricky this is?)
18336ac495dSmrg   </p></div><div class="section"><div class="titlepage"><div><div><h3 class="title"><a id="strings.string.token"></a>Tokenizing</h3></div></div></div><p>
18436ac495dSmrg    </p><p>The Standard C (and C++) function <code class="code">strtok()</code> leaves a lot to
18536ac495dSmrg      be desired in terms of user-friendliness.  It's unintuitive, it
18636ac495dSmrg      destroys the character string on which it operates, and it requires
18736ac495dSmrg      you to handle all the memory problems.  But it does let the client
18836ac495dSmrg      code decide what to use to break the string into pieces; it allows
18936ac495dSmrg      you to choose the "whitespace," so to speak.
19036ac495dSmrg   </p><p>A C++ implementation lets us keep the good things and fix those
19136ac495dSmrg      annoyances.  The implementation here is more intuitive (you only
19236ac495dSmrg      call it once, not in a loop with varying argument), it does not
19336ac495dSmrg      affect the original string at all, and all the memory allocation
19436ac495dSmrg      is handled for you.
19536ac495dSmrg   </p><p>It's called stringtok, and it's a template function. Sources are
19636ac495dSmrg   as below, in a less-portable form than it could be, to keep this
19736ac495dSmrg   example simple (for example, see the comments on what kind of
19836ac495dSmrg   string it will accept).
19936ac495dSmrg   </p><pre class="programlisting">
20036ac495dSmrg#include &lt;string&gt;
20136ac495dSmrgtemplate &lt;typename Container&gt;
20236ac495dSmrgvoid
20336ac495dSmrgstringtok(Container &amp;container, string const &amp;in,
20436ac495dSmrg	  const char * const delimiters = " \t\n")
20536ac495dSmrg{
20636ac495dSmrg    const string::size_type len = in.length();
20736ac495dSmrg	  string::size_type i = 0;
20836ac495dSmrg
20936ac495dSmrg    while (i &lt; len)
21036ac495dSmrg    {
21136ac495dSmrg	// Eat leading whitespace
21236ac495dSmrg	i = in.find_first_not_of(delimiters, i);
21336ac495dSmrg	if (i == string::npos)
21436ac495dSmrg	  return;   // Nothing left but white space
21536ac495dSmrg
21636ac495dSmrg	// Find the end of the token
21736ac495dSmrg	string::size_type j = in.find_first_of(delimiters, i);
21836ac495dSmrg
21936ac495dSmrg	// Push token
22036ac495dSmrg	if (j == string::npos)
22136ac495dSmrg	{
22236ac495dSmrg	  container.push_back(in.substr(i));
22336ac495dSmrg	  return;
22436ac495dSmrg	}
22536ac495dSmrg	else
22636ac495dSmrg	  container.push_back(in.substr(i, j-i));
22736ac495dSmrg
22836ac495dSmrg	// Set up for next loop
22936ac495dSmrg	i = j + 1;
23036ac495dSmrg    }
23136ac495dSmrg}
23236ac495dSmrg</pre><p>
23336ac495dSmrg     The author uses a more general (but less readable) form of it for
23436ac495dSmrg     parsing command strings and the like.  If you compiled and ran this
23536ac495dSmrg     code using it:
23636ac495dSmrg   </p><pre class="programlisting">
23736ac495dSmrg   std::list&lt;string&gt;  ls;
23836ac495dSmrg   stringtok (ls, " this  \t is\t\n  a test  ");
23936ac495dSmrg   for (std::list&lt;string&gt;const_iterator i = ls.begin();
24036ac495dSmrg	i != ls.end(); ++i)
24136ac495dSmrg   {
24236ac495dSmrg       std::cerr &lt;&lt; ':' &lt;&lt; (*i) &lt;&lt; ":\n";
24336ac495dSmrg   } </pre><p>You would see this as output:
24436ac495dSmrg   </p><pre class="programlisting">
24536ac495dSmrg   :this:
24636ac495dSmrg   :is:
24736ac495dSmrg   :a:
24836ac495dSmrg   :test: </pre><p>with all the whitespace removed.  The original <code class="code">s</code> is still
24936ac495dSmrg      available for use, <code class="code">ls</code> will clean up after itself, and
25036ac495dSmrg      <code class="code">ls.size()</code> will return how many tokens there were.
25136ac495dSmrg   </p><p>As always, there is a price paid here, in that stringtok is not
25236ac495dSmrg      as fast as strtok.  The other benefits usually outweigh that, however.
25336ac495dSmrg   </p><p><span class="emphasis"><em>Added February 2001:</em></span>  Mark Wilden pointed out that the
25436ac495dSmrg      standard <code class="code">std::getline()</code> function can be used with standard
25536ac495dSmrg      <code class="code">istringstreams</code> to perform
25636ac495dSmrg      tokenizing as well.  Build an istringstream from the input text,
25736ac495dSmrg      and then use std::getline with varying delimiters (the three-argument
25836ac495dSmrg      signature) to extract tokens into a string.
25936ac495dSmrg   </p></div><div class="section"><div class="titlepage"><div><div><h3 class="title"><a id="strings.string.shrink"></a>Shrink to Fit</h3></div></div></div><p>
26036ac495dSmrg    </p><p>From GCC 3.4 calling <code class="code">s.reserve(res)</code> on a
26136ac495dSmrg      <code class="code">string s</code> with <code class="code">res &lt; s.capacity()</code> will
26236ac495dSmrg      reduce the string's capacity to <code class="code">std::max(s.size(), res)</code>.
26336ac495dSmrg   </p><p>This behaviour is suggested, but not required by the standard. Prior
26436ac495dSmrg      to GCC 3.4 the following alternative can be used instead
26536ac495dSmrg   </p><pre class="programlisting">
26636ac495dSmrg      std::string(str.data(), str.size()).swap(str);
26736ac495dSmrg   </pre><p>This is similar to the idiom for reducing
26836ac495dSmrg      a <code class="code">vector</code>'s memory usage
26936ac495dSmrg      (see <a class="link" href="../faq.html#faq.size_equals_capacity" title="7.8.">this FAQ
27036ac495dSmrg      entry</a>) but the regular copy constructor cannot be used
27136ac495dSmrg      because libstdc++'s <code class="code">string</code> is Copy-On-Write in GCC 3.
27236ac495dSmrg   </p><p>In <a class="link" href="status.html#status.iso.2011" title="C++ 2011">C++11</a> mode you can call
27336ac495dSmrg      <code class="code">s.shrink_to_fit()</code> to achieve the same effect as
27436ac495dSmrg      <code class="code">s.reserve(s.size())</code>.
27536ac495dSmrg   </p></div><div class="section"><div class="titlepage"><div><div><h3 class="title"><a id="strings.string.Cstring"></a>CString (MFC)</h3></div></div></div><p>
27636ac495dSmrg    </p><p>A common lament seen in various newsgroups deals with the Standard
27736ac495dSmrg      string class as opposed to the Microsoft Foundation Class called
27836ac495dSmrg      CString.  Often programmers realize that a standard portable
27936ac495dSmrg      answer is better than a proprietary nonportable one, but in porting
28036ac495dSmrg      their application from a Win32 platform, they discover that they
28136ac495dSmrg      are relying on special functions offered by the CString class.
28236ac495dSmrg   </p><p>Things are not as bad as they seem.  In
28336ac495dSmrg      <a class="link" href="http://gcc.gnu.org/ml/gcc/1999-04n/msg00236.html" target="_top">this
28436ac495dSmrg      message</a>, Joe Buck points out a few very important things:
28536ac495dSmrg   </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>The Standard <code class="code">string</code> supports all the operations
28636ac495dSmrg	     that CString does, with three exceptions.
28736ac495dSmrg	 </p></li><li class="listitem"><p>Two of those exceptions (whitespace trimming and case
28836ac495dSmrg	     conversion) are trivial to implement.  In fact, we do so
28936ac495dSmrg	     on this page.
29036ac495dSmrg	 </p></li><li class="listitem"><p>The third is <code class="code">CString::Format</code>, which allows formatting
29136ac495dSmrg	     in the style of <code class="code">sprintf</code>.  This deserves some mention:
29236ac495dSmrg	 </p></li></ul></div><p>
29336ac495dSmrg      The old libg++ library had a function called form(), which did much
29436ac495dSmrg      the same thing.  But for a Standard solution, you should use the
29536ac495dSmrg      stringstream classes.  These are the bridge between the iostream
29636ac495dSmrg      hierarchy and the string class, and they operate with regular
29736ac495dSmrg      streams seamlessly because they inherit from the iostream
29836ac495dSmrg      hierarchy.  An quick example:
29936ac495dSmrg   </p><pre class="programlisting">
30036ac495dSmrg   #include &lt;iostream&gt;
30136ac495dSmrg   #include &lt;string&gt;
30236ac495dSmrg   #include &lt;sstream&gt;
30336ac495dSmrg
30436ac495dSmrg   string f (string&amp; incoming)     // incoming is "foo  N"
30536ac495dSmrg   {
30636ac495dSmrg       istringstream   incoming_stream(incoming);
30736ac495dSmrg       string          the_word;
30836ac495dSmrg       int             the_number;
30936ac495dSmrg
31036ac495dSmrg       incoming_stream &gt;&gt; the_word        // extract "foo"
31136ac495dSmrg		       &gt;&gt; the_number;     // extract N
31236ac495dSmrg
31336ac495dSmrg       ostringstream   output_stream;
31436ac495dSmrg       output_stream &lt;&lt; "The word was " &lt;&lt; the_word
31536ac495dSmrg		     &lt;&lt; " and 3*N was " &lt;&lt; (3*the_number);
31636ac495dSmrg
31736ac495dSmrg       return output_stream.str();
31836ac495dSmrg   } </pre><p>A serious problem with CString is a design bug in its memory
31936ac495dSmrg      allocation.  Specifically, quoting from that same message:
32036ac495dSmrg   </p><pre class="programlisting">
32136ac495dSmrg   CString suffers from a common programming error that results in
32236ac495dSmrg   poor performance.  Consider the following code:
32336ac495dSmrg
32436ac495dSmrg   CString n_copies_of (const CString&amp; foo, unsigned n)
32536ac495dSmrg   {
32636ac495dSmrg	   CString tmp;
32736ac495dSmrg	   for (unsigned i = 0; i &lt; n; i++)
32836ac495dSmrg		   tmp += foo;
32936ac495dSmrg	   return tmp;
33036ac495dSmrg   }
33136ac495dSmrg
33236ac495dSmrg   This function is O(n^2), not O(n).  The reason is that each +=
33336ac495dSmrg   causes a reallocation and copy of the existing string.  Microsoft
33436ac495dSmrg   applications are full of this kind of thing (quadratic performance
33536ac495dSmrg   on tasks that can be done in linear time) -- on the other hand,
33636ac495dSmrg   we should be thankful, as it's created such a big market for high-end
33736ac495dSmrg   ix86 hardware. :-)
33836ac495dSmrg
33936ac495dSmrg   If you replace CString with string in the above function, the
34036ac495dSmrg   performance is O(n).
34136ac495dSmrg   </pre><p>Joe Buck also pointed out some other things to keep in mind when
34236ac495dSmrg      comparing CString and the Standard string class:
34336ac495dSmrg   </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>CString permits access to its internal representation; coders
34436ac495dSmrg	     who exploited that may have problems moving to <code class="code">string</code>.
34536ac495dSmrg	 </p></li><li class="listitem"><p>Microsoft ships the source to CString (in the files
34636ac495dSmrg	     MFC\SRC\Str{core,ex}.cpp), so you could fix the allocation
34736ac495dSmrg	     bug and rebuild your MFC libraries.
34836ac495dSmrg	     <span class="emphasis"><em><span class="emphasis"><em>Note:</em></span> It looks like the CString shipped
34936ac495dSmrg	     with VC++6.0 has fixed this, although it may in fact have been
35036ac495dSmrg	     one of the VC++ SPs that did it.</em></span>
35136ac495dSmrg	 </p></li><li class="listitem"><p><code class="code">string</code> operations like this have O(n) complexity
35236ac495dSmrg	     <span class="emphasis"><em>if the implementors do it correctly</em></span>.  The libstdc++
35336ac495dSmrg	     implementors did it correctly.  Other vendors might not.
35436ac495dSmrg	 </p></li><li class="listitem"><p>While parts of the SGI STL are used in libstdc++, their
35536ac495dSmrg	     string class is not.  The SGI <code class="code">string</code> is essentially
35636ac495dSmrg	     <code class="code">vector&lt;char&gt;</code> and does not do any reference
35736ac495dSmrg	     counting like libstdc++'s does.  (It is O(n), though.)
35836ac495dSmrg	     So if you're thinking about SGI's string or rope classes,
35936ac495dSmrg	     you're now looking at four possibilities:  CString, the
36036ac495dSmrg	     libstdc++ string, the SGI string, and the SGI rope, and this
36136ac495dSmrg	     is all before any allocator or traits customizations!  (More
36236ac495dSmrg	     choices than you can shake a stick at -- want fries with that?)
36336ac495dSmrg	 </p></li></ul></div></div></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="traits.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="std_contents.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="localization.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">Traits </td><td width="20%" align="center"><a accesskey="h" href="../index.html">Home</a></td><td width="40%" align="right" valign="top"> Chapter 8. 
36436ac495dSmrg  Localization
36536ac495dSmrg
36636ac495dSmrg</td></tr></table></div></body></html>