libstdc++-v3/doc/html/manual/strings.html

   1 <html><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"><title>Chapter 7.  Strings</title><meta name="generator" content="DocBook XSL-NS Stylesheets V1.76.1"><meta name="keywords" content="
   2       ISO C++
   3     ,
   4       library
   5     "><meta name="keywords" content="
   6       ISO C++
   7     ,
   8       runtime
   9     ,
  10       library
  11     "><link rel="home" href="../index.html" title="The GNU C++ Library"><link rel="up" href="bk01pt02.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 bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF"><div class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="3" align="center">Chapter 7.
  12   Strings
  13
  14 </th></tr><tr><td width="20%" align="left"><a accesskey="p" href="traits.html">Prev</a> </td><th width="60%" align="center">Part II.
  15     Standard Contents
  16   </th><td width="20%" align="right"> <a accesskey="n" href="localization.html">Next</a></td></tr></table><hr></div><div class="chapter" title="Chapter 7.  Strings"><div class="titlepage"><div><div><h2 class="title"><a name="std.strings"></a>Chapter 7.
  17   Strings
  18   <a class="indexterm" name="id623691"></a>
  19 </h2></div></div></div><div class="toc"><p><b>Table of Contents</b></p><dl><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" title="String Classes"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="std.strings.string"></a>String Classes</h2></div></div></div><div class="section" title="Simple Transformations"><div class="titlepage"><div><div><h3 class="title"><a name="strings.string.simple"></a>Simple Transformations</h3></div></div></div><p>
  20       Here are Standard, simple, and portable ways to perform common
  21       transformations on a <code class="code">string</code> instance, such as
  22       "convert to all upper case." The word transformations
  23       is especially apt, because the standard template function
  24       <code class="code">transform&lt;&gt;</code> is used.
  25    </p><p>
  26      This code will go through some iterations.  Here's a simple
  27      version:
  28    </p><pre class="programlisting">
  29    #include &lt;string&gt;
  30    #include &lt;algorithm&gt;
  31    #include &lt;cctype&gt;      // old &lt;ctype.h&gt;
  32
  33    struct ToLower
  34    {
  35      char operator() (char c) const  { return std::tolower(c); }
  36    };
  37
  38    struct ToUpper
  39    {
  40      char operator() (char c) const  { return std::toupper(c); }
  41    };
  42
  43    int main()
  44    {
  45      std::string  s ("Some Kind Of Initial Input Goes Here");
  46
  47      // Change everything into upper case
  48      std::transform (s.begin(), s.end(), s.begin(), ToUpper());
  49
  50      // Change everything into lower case
  51      std::transform (s.begin(), s.end(), s.begin(), ToLower());
  52
  53      // Change everything back into upper case, but store the
  54      // result in a different string
  55      std::string  capital_s;
  56      capital_s.resize(s.size());
  57      std::transform (s.begin(), s.end(), capital_s.begin(), ToUpper());
  58    }
  59    </pre><p>
  60      <span class="emphasis"><em>Note</em></span> that these calls all
  61       involve the global C locale through the use of the C functions
  62       <code class="code">toupper/tolower</code>.  This is absolutely guaranteed to work --
  63       but <span class="emphasis"><em>only</em></span> if the string contains <span class="emphasis"><em>only</em></span> characters
  64       from the basic source character set, and there are <span class="emphasis"><em>only</em></span>
  65       96 of those.  Which means that not even all English text can be
  66       represented (certain British spellings, proper names, and so forth).
  67       So, if all your input forevermore consists of only those 96
  68       characters (hahahahahaha), then you're done.
  69    </p><p><span class="emphasis"><em>Note</em></span> that the
  70       <code class="code">ToUpper</code> and <code class="code">ToLower</code> function objects
  71       are needed because <code class="code">toupper</code> and <code class="code">tolower</code>
  72       are overloaded names (declared in <code class="code">&lt;cctype&gt;</code> and
  73       <code class="code">&lt;locale&gt;</code>) so the template-arguments for
  74       <code class="code">transform&lt;&gt;</code> cannot be deduced, as explained in
  75       <a class="link" href="http://gcc.gnu.org/ml/libstdc++/2002-11/msg00180.html" target="_top">this
  76       message</a>.
  77
  78       At minimum, you can write short wrappers like
  79    </p><pre class="programlisting">
  80    char toLower (char c)
  81    {
  82       return std::tolower(c);
  83    } </pre><p>(Thanks to James Kanze for assistance and suggestions on all of this.)
  84    </p><p>Another common operation is trimming off excess whitespace.  Much
  85       like transformations, this task is trivial with the use of string's
  86       <code class="code">find</code> family.  These examples are broken into multiple
  87       statements for readability:
  88    </p><pre class="programlisting">
  89    std::string  str (" \t blah blah blah    \n ");
  90
  91    // trim leading whitespace
  92    string::size_type  notwhite = str.find_first_not_of(" \t\n");
  93    str.erase(0,notwhite);
  94
  95    // trim trailing whitespace
  96    notwhite = str.find_last_not_of(" \t\n");
  97    str.erase(notwhite+1); </pre><p>Obviously, the calls to <code class="code">find</code> could be inserted directly
  98       into the calls to <code class="code">erase</code>, in case your compiler does not
  99       optimize named temporaries out of existence.
 100    </p></div><div class="section" title="Case Sensitivity"><div class="titlepage"><div><div><h3 class="title"><a name="strings.string.case"></a>Case Sensitivity</h3></div></div></div><p>
 101     </p><p>The well-known-and-if-it-isn't-well-known-it-ought-to-be
 102       <a class="link" href="http://www.gotw.ca/gotw/" target="_top">Guru of the Week</a>
 103       discussions held on Usenet covered this topic in January of 1998.
 104       Briefly, the challenge was, <span class="quote">“<span class="quote">write a 'ci_string' class which
 105       is identical to the standard 'string' class, but is
 106       case-insensitive in the same way as the (common but nonstandard)
 107       C function stricmp()</span>”</span>.
 108    </p><pre class="programlisting">
 109    ci_string s( "AbCdE" );
 110
 111    // case insensitive
 112    assert( s == "abcde" );
 113    assert( s == "ABCDE" );
 114
 115    // still case-preserving, of course
 116    assert( strcmp( s.c_str(), "AbCdE" ) == 0 );
 117    assert( strcmp( s.c_str(), "abcde" ) != 0 ); </pre><p>The solution is surprisingly easy.  The original answer was
 118    posted on Usenet, and a revised version appears in Herb Sutter's
 119    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>.
 120    </p><p>See?  Told you it was easy!</p><p>
 121      <span class="emphasis"><em>Added June 2000:</em></span> The May 2000 issue of C++
 122      Report contains a fascinating <a class="link" href="http://lafstern.org/matt/col2_new.pdf" target="_top"> article</a> by
 123      Matt Austern (yes, <span class="emphasis"><em>the</em></span> Matt Austern) on why
 124      case-insensitive comparisons are not as easy as they seem, and
 125      why creating a class is the <span class="emphasis"><em>wrong</em></span> way to go
 126      about it in production code.  (The GotW answer mentions one of
 127      the principle difficulties; his article mentions more.)
 128    </p><p>Basically, this is "easy" only if you ignore some things,
 129       things which may be too important to your program to ignore.  (I chose
 130       to ignore them when originally writing this entry, and am surprised
 131       that nobody ever called me on it...)  The GotW question and answer
 132       remain useful instructional tools, however.
 133    </p><p><span class="emphasis"><em>Added September 2000:</em></span>  James Kanze provided a link to a
 134       <a class="link" href="http://www.unicode.org/reports/tr21/tr21-5.html" target="_top">Unicode
 135       Technical Report discussing case handling</a>, which provides some
 136       very good information.
 137    </p></div><div class="section" title="Arbitrary Character Types"><div class="titlepage"><div><div><h3 class="title"><a name="strings.string.character_types"></a>Arbitrary Character Types</h3></div></div></div><p>
 138     </p><p>The <code class="code">std::basic_string</code> is tantalizingly general, in that
 139       it is parameterized on the type of the characters which it holds.
 140       In theory, you could whip up a Unicode character class and instantiate
 141       <code class="code">std::basic_string&lt;my_unicode_char&gt;</code>, or assuming
 142       that integers are wider than characters on your platform, maybe just
 143       declare variables of type <code class="code">std::basic_string&lt;int&gt;</code>.
 144    </p><p>That's the theory.  Remember however that basic_string has additional
 145       type parameters, which take default arguments based on the character
 146       type (called <code class="code">CharT</code> here):
 147    </p><pre class="programlisting">
 148       template &lt;typename CharT,
 149                 typename Traits = char_traits&lt;CharT&gt;,
 150                 typename Alloc = allocator&lt;CharT&gt; &gt;
 151       class basic_string { .... };</pre><p>Now, <code class="code">allocator&lt;CharT&gt;</code> will probably Do The Right
 152       Thing by default, unless you need to implement your own allocator
 153       for your characters.
 154    </p><p>But <code class="code">char_traits</code> takes more work.  The char_traits
 155       template is <span class="emphasis"><em>declared</em></span> but not <span class="emphasis"><em>defined</em></span>.
 156       That means there is only
 157    </p><pre class="programlisting">
 158       template &lt;typename CharT&gt;
 159         struct char_traits
 160         {
 161             static void foo (type1 x, type2 y);
 162             ...
 163         };</pre><p>and functions such as char_traits&lt;CharT&gt;::foo() are not
 164       actually defined anywhere for the general case.  The C++ standard
 165       permits this, because writing such a definition to fit all possible
 166       CharT's cannot be done.
 167    </p><p>The C++ standard also requires that char_traits be specialized for
 168       instantiations of <code class="code">char</code> and <code class="code">wchar_t</code>, and it
 169       is these template specializations that permit entities like
 170       <code class="code">basic_string&lt;char,char_traits&lt;char&gt;&gt;</code> to work.
 171    </p><p>If you want to use character types other than char and wchar_t,
 172       such as <code class="code">unsigned char</code> and <code class="code">int</code>, you will
 173       need suitable specializations for them.  For a time, in earlier
 174       versions of GCC, there was a mostly-correct implementation that
 175       let programmers be lazy but it broke under many situations, so it
 176       was removed.  GCC 3.4 introduced a new implementation that mostly
 177       works and can be specialized even for <code class="code">int</code> and other
 178       built-in types.
 179    </p><p>If you want to use your own special character class, then you have
 180       <a class="link" href="http://gcc.gnu.org/ml/libstdc++/2002-08/msg00163.html" target="_top">a lot
 181       of work to do</a>, especially if you with to use i18n features
 182       (facets require traits information but don't have a traits argument).
 183    </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
 184       mailing list</a> and at a later date was put into the file <code class="code">
 185       include/ext/pod_char_traits.h</code>.  We agree
 186       that the way it's used with basic_string (scroll down to main())
 187       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
 188       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
 189       be conforming C++</a>, due to the rule that CharT must be a POD.
 190       (See how tricky this is?)
 191    </p></div><div class="section" title="Tokenizing"><div class="titlepage"><div><div><h3 class="title"><a name="strings.string.token"></a>Tokenizing</h3></div></div></div><p>
 192     </p><p>The Standard C (and C++) function <code class="code">strtok()</code> leaves a lot to
 193       be desired in terms of user-friendliness.  It's unintuitive, it
 194       destroys the character string on which it operates, and it requires
 195       you to handle all the memory problems.  But it does let the client
 196       code decide what to use to break the string into pieces; it allows
 197       you to choose the "whitespace," so to speak.
 198    </p><p>A C++ implementation lets us keep the good things and fix those
 199       annoyances.  The implementation here is more intuitive (you only
 200       call it once, not in a loop with varying argument), it does not
 201       affect the original string at all, and all the memory allocation
 202       is handled for you.
 203    </p><p>It's called stringtok, and it's a template function. Sources are
 204    as below, in a less-portable form than it could be, to keep this
 205    example simple (for example, see the comments on what kind of
 206    string it will accept).
 207    </p><pre class="programlisting">
 208 #include &lt;string&gt;
 209 template &lt;typename Container&gt;
 210 void
 211 stringtok(Container &amp;container, string const &amp;in,
 212           const char * const delimiters = " \t\n")
 213 {
 214     const string::size_type len = in.length();
 215           string::size_type i = 0;
 216
 217     while (i &lt; len)
 218     {
 219         // Eat leading whitespace
 220         i = in.find_first_not_of(delimiters, i);
 221         if (i == string::npos)
 222           return;   // Nothing left but white space
 223
 224         // Find the end of the token
 225         string::size_type j = in.find_first_of(delimiters, i);
 226
 227         // Push token
 228         if (j == string::npos)
 229         {
 230           container.push_back(in.substr(i));
 231           return;
 232         }
 233         else
 234           container.push_back(in.substr(i, j-i));
 235
 236         // Set up for next loop
 237         i = j + 1;
 238     }
 239 }
 240 </pre><p>
 241      The author uses a more general (but less readable) form of it for
 242      parsing command strings and the like.  If you compiled and ran this
 243      code using it:
 244    </p><pre class="programlisting">
 245    std::list&lt;string&gt;  ls;
 246    stringtok (ls, " this  \t is\t\n  a test  ");
 247    for (std::list&lt;string&gt;const_iterator i = ls.begin();
 248         i != ls.end(); ++i)
 249    {
 250        std::cerr &lt;&lt; ':' &lt;&lt; (*i) &lt;&lt; ":\n";
 251    } </pre><p>You would see this as output:
 252    </p><pre class="programlisting">
 253    :this:
 254    :is:
 255    :a:
 256    :test: </pre><p>with all the whitespace removed.  The original <code class="code">s</code> is still
 257       available for use, <code class="code">ls</code> will clean up after itself, and
 258       <code class="code">ls.size()</code> will return how many tokens there were.
 259    </p><p>As always, there is a price paid here, in that stringtok is not
 260       as fast as strtok.  The other benefits usually outweigh that, however.
 261    </p><p><span class="emphasis"><em>Added February 2001:</em></span>  Mark Wilden pointed out that the
 262       standard <code class="code">std::getline()</code> function can be used with standard
 263       <code class="code">istringstreams</code> to perform
 264       tokenizing as well.  Build an istringstream from the input text,
 265       and then use std::getline with varying delimiters (the three-argument
 266       signature) to extract tokens into a string.
 267    </p></div><div class="section" title="Shrink to Fit"><div class="titlepage"><div><div><h3 class="title"><a name="strings.string.shrink"></a>Shrink to Fit</h3></div></div></div><p>
 268     </p><p>From GCC 3.4 calling <code class="code">s.reserve(res)</code> on a
 269       <code class="code">string s</code> with <code class="code">res &lt; s.capacity()</code> will
 270       reduce the string's capacity to <code class="code">std::max(s.size(), res)</code>.
 271    </p><p>This behaviour is suggested, but not required by the standard. Prior
 272       to GCC 3.4 the following alternative can be used instead
 273    </p><pre class="programlisting">
 274       std::string(str.data(), str.size()).swap(str);
 275    </pre><p>This is similar to the idiom for reducing
 276       a <code class="code">vector</code>'s memory usage
 277       (see <a class="link" href="../faq.html#faq.size_equals_capacity" title="7.8.">this FAQ
 278       entry</a>) but the regular copy constructor cannot be used
 279       because libstdc++'s <code class="code">string</code> is Copy-On-Write.
 280    </p><p>In <a class="link" href="status.html#status.iso.2011" title="C++ 2011">C++11</a> mode you can call
 281       <code class="code">s.shrink_to_fit()</code> to achieve the same effect as
 282       <code class="code">s.reserve(s.size())</code>.
 283    </p></div><div class="section" title="CString (MFC)"><div class="titlepage"><div><div><h3 class="title"><a name="strings.string.Cstring"></a>CString (MFC)</h3></div></div></div><p>
 284     </p><p>A common lament seen in various newsgroups deals with the Standard
 285       string class as opposed to the Microsoft Foundation Class called
 286       CString.  Often programmers realize that a standard portable
 287       answer is better than a proprietary nonportable one, but in porting
 288       their application from a Win32 platform, they discover that they
 289       are relying on special functions offered by the CString class.
 290    </p><p>Things are not as bad as they seem.  In
 291       <a class="link" href="http://gcc.gnu.org/ml/gcc/1999-04n/msg00236.html" target="_top">this
 292       message</a>, Joe Buck points out a few very important things:
 293    </p><div class="itemizedlist"><ul class="itemizedlist" type="disc"><li class="listitem"><p>The Standard <code class="code">string</code> supports all the operations
 294              that CString does, with three exceptions.
 295          </p></li><li class="listitem"><p>Two of those exceptions (whitespace trimming and case
 296              conversion) are trivial to implement.  In fact, we do so
 297              on this page.
 298          </p></li><li class="listitem"><p>The third is <code class="code">CString::Format</code>, which allows formatting
 299              in the style of <code class="code">sprintf</code>.  This deserves some mention:
 300          </p></li></ul></div><p>
 301       The old libg++ library had a function called form(), which did much
 302       the same thing.  But for a Standard solution, you should use the
 303       stringstream classes.  These are the bridge between the iostream
 304       hierarchy and the string class, and they operate with regular
 305       streams seamlessly because they inherit from the iostream
 306       hierarchy.  An quick example:
 307    </p><pre class="programlisting">
 308    #include &lt;iostream&gt;
 309    #include &lt;string&gt;
 310    #include &lt;sstream&gt;
 311
 312    string f (string&amp; incoming)     // incoming is "foo  N"
 313    {
 314        istringstream   incoming_stream(incoming);
 315        string          the_word;
 316        int             the_number;
 317
 318        incoming_stream &gt;&gt; the_word        // extract "foo"
 319                        &gt;&gt; the_number;     // extract N
 320
 321        ostringstream   output_stream;
 322        output_stream &lt;&lt; "The word was " &lt;&lt; the_word
 323                      &lt;&lt; " and 3*N was " &lt;&lt; (3*the_number);
 324
 325        return output_stream.str();
 326    } </pre><p>A serious problem with CString is a design bug in its memory
 327       allocation.  Specifically, quoting from that same message:
 328    </p><pre class="programlisting">
 329    CString suffers from a common programming error that results in
 330    poor performance.  Consider the following code:
 331
 332    CString n_copies_of (const CString&amp; foo, unsigned n)
 333    {
 334            CString tmp;
 335            for (unsigned i = 0; i &lt; n; i++)
 336                    tmp += foo;
 337            return tmp;
 338    }
 339
 340    This function is O(n^2), not O(n).  The reason is that each +=
 341    causes a reallocation and copy of the existing string.  Microsoft
 342    applications are full of this kind of thing (quadratic performance
 343    on tasks that can be done in linear time) -- on the other hand,
 344    we should be thankful, as it's created such a big market for high-end
 345    ix86 hardware. :-)
 346
 347    If you replace CString with string in the above function, the
 348    performance is O(n).
 349    </pre><p>Joe Buck also pointed out some other things to keep in mind when
 350       comparing CString and the Standard string class:
 351    </p><div class="itemizedlist"><ul class="itemizedlist" type="disc"><li class="listitem"><p>CString permits access to its internal representation; coders
 352              who exploited that may have problems moving to <code class="code">string</code>.
 353          </p></li><li class="listitem"><p>Microsoft ships the source to CString (in the files
 354              MFC\SRC\Str{core,ex}.cpp), so you could fix the allocation
 355              bug and rebuild your MFC libraries.
 356              <span class="emphasis"><em><span class="emphasis"><em>Note:</em></span> It looks like the CString shipped
 357              with VC++6.0 has fixed this, although it may in fact have been
 358              one of the VC++ SPs that did it.</em></span>
 359          </p></li><li class="listitem"><p><code class="code">string</code> operations like this have O(n) complexity
 360              <span class="emphasis"><em>if the implementors do it correctly</em></span>.  The libstdc++
 361              implementors did it correctly.  Other vendors might not.
 362          </p></li><li class="listitem"><p>While chapters of the SGI STL are used in libstdc++, their
 363              string class is not.  The SGI <code class="code">string</code> is essentially
 364              <code class="code">vector&lt;char&gt;</code> and does not do any reference
 365              counting like libstdc++'s does.  (It is O(n), though.)
 366              So if you're thinking about SGI's string or rope classes,
 367              you're now looking at four possibilities:  CString, the
 368              libstdc++ string, the SGI string, and the SGI rope, and this
 369              is all before any allocator or traits customizations!  (More
 370              choices than you can shake a stick at -- want fries with that?)
 371          </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="bk01pt02.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.
 372   Localization
 373
 374 </td></tr></table></div></body></html>