1<chapter xmlns="http://docbook.org/ns/docbook" version="5.0"
2	 xml:id="std.strings" xreflabel="Strings">
3<?dbhtml filename="strings.html"?>
4
5<info><title>
6  Strings
7  <indexterm><primary>Strings</primary></indexterm>
8</title>
9  <keywordset>
10    <keyword>ISO C++</keyword>
11    <keyword>library</keyword>
12  </keywordset>
13</info>
14
15<!-- Sect1 01 : Character Traits -->
16
17<!-- Sect1 02 : String Classes -->
18<section xml:id="std.strings.string" xreflabel="string"><info><title>String Classes</title></info>
19
20
21  <section xml:id="strings.string.simple" xreflabel="Simple Transformations"><info><title>Simple Transformations</title></info>
22
23    <para>
24      Here are Standard, simple, and portable ways to perform common
25      transformations on a <code>string</code> instance, such as
26      "convert to all upper case." The word transformations
27      is especially apt, because the standard template function
28      <code>transform&lt;&gt;</code> is used.
29   </para>
30   <para>
31     This code will go through some iterations.  Here's a simple
32     version:
33   </para>
34   <programlisting>
35   #include &lt;string&gt;
36   #include &lt;algorithm&gt;
37   #include &lt;cctype&gt;      // old &lt;ctype.h&gt;
38
39   struct ToLower
40   {
41     char operator() (char c) const  { return std::tolower(c); }
42   };
43
44   struct ToUpper
45   {
46     char operator() (char c) const  { return std::toupper(c); }
47   };
48
49   int main()
50   {
51     std::string  s ("Some Kind Of Initial Input Goes Here");
52
53     // Change everything into upper case
54     std::transform (s.begin(), s.end(), s.begin(), ToUpper());
55
56     // Change everything into lower case
57     std::transform (s.begin(), s.end(), s.begin(), ToLower());
58
59     // Change everything back into upper case, but store the
60     // result in a different string
61     std::string  capital_s;
62     capital_s.resize(s.size());
63     std::transform (s.begin(), s.end(), capital_s.begin(), ToUpper());
64   }
65   </programlisting>
66   <para>
67     <emphasis>Note</emphasis> that these calls all
68      involve the global C locale through the use of the C functions
69      <code>toupper/tolower</code>.  This is absolutely guaranteed to work --
70      but <emphasis>only</emphasis> if the string contains <emphasis>only</emphasis> characters
71      from the basic source character set, and there are <emphasis>only</emphasis>
72      96 of those.  Which means that not even all English text can be
73      represented (certain British spellings, proper names, and so forth).
74      So, if all your input forevermore consists of only those 96
75      characters (hahahahahaha), then you're done.
76   </para>
77   <para><emphasis>Note</emphasis> that the
78      <code>ToUpper</code> and <code>ToLower</code> function objects
79      are needed because <code>toupper</code> and <code>tolower</code>
80      are overloaded names (declared in <code>&lt;cctype&gt;</code> and
81      <code>&lt;locale&gt;</code>) so the template-arguments for
82      <code>transform&lt;&gt;</code> cannot be deduced, as explained in
83      <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://gcc.gnu.org/ml/libstdc++/2002-11/msg00180.html">this
84      message</link>.
85      <!-- section 14.8.2.4 clause 16 in ISO 14882:1998  -->
86      At minimum, you can write short wrappers like
87   </para>
88   <programlisting>
89   char toLower (char c)
90   {
91      // std::tolower(c) is undefined if c &lt; 0 so cast to unsigned char.
92      return std::tolower((unsigned char)c);
93   } </programlisting>
94   <para>(Thanks to James Kanze for assistance and suggestions on all of this.)
95   </para>
96   <para>Another common operation is trimming off excess whitespace.  Much
97      like transformations, this task is trivial with the use of string's
98      <code>find</code> family.  These examples are broken into multiple
99      statements for readability:
100   </para>
101   <programlisting>
102   std::string  str (" \t blah blah blah    \n ");
103
104   // trim leading whitespace
105   string::size_type  notwhite = str.find_first_not_of(" \t\n");
106   str.erase(0,notwhite);
107
108   // trim trailing whitespace
109   notwhite = str.find_last_not_of(" \t\n");
110   str.erase(notwhite+1); </programlisting>
111   <para>Obviously, the calls to <code>find</code> could be inserted directly
112      into the calls to <code>erase</code>, in case your compiler does not
113      optimize named temporaries out of existence.
114   </para>
115
116  </section>
117  <section xml:id="strings.string.case" xreflabel="Case Sensitivity"><info><title>Case Sensitivity</title></info>
118
119    <para>
120    </para>
121
122   <para>The well-known-and-if-it-isn't-well-known-it-ought-to-be
123      <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://www.gotw.ca/gotw/">Guru of the Week</link>
124      discussions held on Usenet covered this topic in January of 1998.
125      Briefly, the challenge was, <quote>write a 'ci_string' class which
126      is identical to the standard 'string' class, but is
127      case-insensitive in the same way as the (common but nonstandard)
128      C function stricmp()</quote>.
129   </para>
130   <programlisting>
131   ci_string s( "AbCdE" );
132
133   // case insensitive
134   assert( s == "abcde" );
135   assert( s == "ABCDE" );
136
137   // still case-preserving, of course
138   assert( strcmp( s.c_str(), "AbCdE" ) == 0 );
139   assert( strcmp( s.c_str(), "abcde" ) != 0 ); </programlisting>
140
141   <para>The solution is surprisingly easy.  The original answer was
142   posted on Usenet, and a revised version appears in Herb Sutter's
143   book <emphasis>Exceptional C++</emphasis> and on his website as <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://www.gotw.ca/gotw/029.htm">GotW 29</link>.
144   </para>
145   <para>See?  Told you it was easy!</para>
146   <para>
147     <emphasis>Added June 2000:</emphasis> The May 2000 issue of C++
148     Report contains a fascinating <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://lafstern.org/matt/col2_new.pdf"> article</link> by
149     Matt Austern (yes, <emphasis>the</emphasis> Matt Austern) on why
150     case-insensitive comparisons are not as easy as they seem, and
151     why creating a class is the <emphasis>wrong</emphasis> way to go
152     about it in production code.  (The GotW answer mentions one of
153     the principle difficulties; his article mentions more.)
154   </para>
155   <para>Basically, this is "easy" only if you ignore some things,
156      things which may be too important to your program to ignore.  (I chose
157      to ignore them when originally writing this entry, and am surprised
158      that nobody ever called me on it...)  The GotW question and answer
159      remain useful instructional tools, however.
160   </para>
161   <para><emphasis>Added September 2000:</emphasis>  James Kanze provided a link to a
162      <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://www.unicode.org/reports/tr21/tr21-5.html">Unicode
163      Technical Report discussing case handling</link>, which provides some
164      very good information.
165   </para>
166
167  </section>
168  <section xml:id="strings.string.character_types" xreflabel="Arbitrary Characters"><info><title>Arbitrary Character Types</title></info>
169
170    <para>
171    </para>
172
173   <para>The <code>std::basic_string</code> is tantalizingly general, in that
174      it is parameterized on the type of the characters which it holds.
175      In theory, you could whip up a Unicode character class and instantiate
176      <code>std::basic_string&lt;my_unicode_char&gt;</code>, or assuming
177      that integers are wider than characters on your platform, maybe just
178      declare variables of type <code>std::basic_string&lt;int&gt;</code>.
179   </para>
180   <para>That's the theory.  Remember however that basic_string has additional
181      type parameters, which take default arguments based on the character
182      type (called <code>CharT</code> here):
183   </para>
184   <programlisting>
185      template &lt;typename CharT,
186		typename Traits = char_traits&lt;CharT&gt;,
187		typename Alloc = allocator&lt;CharT&gt; &gt;
188      class basic_string { .... };</programlisting>
189   <para>Now, <code>allocator&lt;CharT&gt;</code> will probably Do The Right
190      Thing by default, unless you need to implement your own allocator
191      for your characters.
192   </para>
193   <para>But <code>char_traits</code> takes more work.  The char_traits
194      template is <emphasis>declared</emphasis> but not <emphasis>defined</emphasis>.
195      That means there is only
196   </para>
197   <programlisting>
198      template &lt;typename CharT&gt;
199	struct char_traits
200	{
201	    static void foo (type1 x, type2 y);
202	    ...
203	};</programlisting>
204   <para>and functions such as char_traits&lt;CharT&gt;::foo() are not
205      actually defined anywhere for the general case.  The C++ standard
206      permits this, because writing such a definition to fit all possible
207      CharT's cannot be done.
208   </para>
209   <para>The C++ standard also requires that char_traits be specialized for
210      instantiations of <code>char</code> and <code>wchar_t</code>, and it
211      is these template specializations that permit entities like
212      <code>basic_string&lt;char,char_traits&lt;char&gt;&gt;</code> to work.
213   </para>
214   <para>If you want to use character types other than char and wchar_t,
215      such as <code>unsigned char</code> and <code>int</code>, you will
216      need suitable specializations for them.  For a time, in earlier
217      versions of GCC, there was a mostly-correct implementation that
218      let programmers be lazy but it broke under many situations, so it
219      was removed.  GCC 3.4 introduced a new implementation that mostly
220      works and can be specialized even for <code>int</code> and other
221      built-in types.
222   </para>
223   <para>If you want to use your own special character class, then you have
224      <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://gcc.gnu.org/ml/libstdc++/2002-08/msg00163.html">a lot
225      of work to do</link>, especially if you with to use i18n features
226      (facets require traits information but don't have a traits argument).
227   </para>
228   <para>Another example of how to specialize char_traits was given <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://gcc.gnu.org/ml/libstdc++/2002-08/msg00260.html">on the
229      mailing list</link> and at a later date was put into the file <code>
230      include/ext/pod_char_traits.h</code>.  We agree
231      that the way it's used with basic_string (scroll down to main())
232      doesn't look nice, but that's because <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://gcc.gnu.org/ml/libstdc++/2002-08/msg00236.html">the
233      nice-looking first attempt</link> turned out to <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://gcc.gnu.org/ml/libstdc++/2002-08/msg00242.html">not
234      be conforming C++</link>, due to the rule that CharT must be a POD.
235      (See how tricky this is?)
236   </para>
237
238  </section>
239
240  <section xml:id="strings.string.token" xreflabel="Tokenizing"><info><title>Tokenizing</title></info>
241
242    <para>
243    </para>
244   <para>The Standard C (and C++) function <code>strtok()</code> leaves a lot to
245      be desired in terms of user-friendliness.  It's unintuitive, it
246      destroys the character string on which it operates, and it requires
247      you to handle all the memory problems.  But it does let the client
248      code decide what to use to break the string into pieces; it allows
249      you to choose the "whitespace," so to speak.
250   </para>
251   <para>A C++ implementation lets us keep the good things and fix those
252      annoyances.  The implementation here is more intuitive (you only
253      call it once, not in a loop with varying argument), it does not
254      affect the original string at all, and all the memory allocation
255      is handled for you.
256   </para>
257   <para>It's called stringtok, and it's a template function. Sources are
258   as below, in a less-portable form than it could be, to keep this
259   example simple (for example, see the comments on what kind of
260   string it will accept).
261   </para>
262
263<programlisting>
264#include &lt;string&gt;
265template &lt;typename Container&gt;
266void
267stringtok(Container &amp;container, string const &amp;in,
268	  const char * const delimiters = " \t\n")
269{
270    const string::size_type len = in.length();
271	  string::size_type i = 0;
272
273    while (i &lt; len)
274    {
275	// Eat leading whitespace
276	i = in.find_first_not_of(delimiters, i);
277	if (i == string::npos)
278	  return;   // Nothing left but white space
279
280	// Find the end of the token
281	string::size_type j = in.find_first_of(delimiters, i);
282
283	// Push token
284	if (j == string::npos)
285	{
286	  container.push_back(in.substr(i));
287	  return;
288	}
289	else
290	  container.push_back(in.substr(i, j-i));
291
292	// Set up for next loop
293	i = j + 1;
294    }
295}
296</programlisting>
297
298
299   <para>
300     The author uses a more general (but less readable) form of it for
301     parsing command strings and the like.  If you compiled and ran this
302     code using it:
303   </para>
304
305
306   <programlisting>
307   std::list&lt;string&gt;  ls;
308   stringtok (ls, " this  \t is\t\n  a test  ");
309   for (std::list&lt;string&gt;const_iterator i = ls.begin();
310	i != ls.end(); ++i)
311   {
312       std::cerr &lt;&lt; ':' &lt;&lt; (*i) &lt;&lt; ":\n";
313   } </programlisting>
314   <para>You would see this as output:
315   </para>
316   <programlisting>
317   :this:
318   :is:
319   :a:
320   :test: </programlisting>
321   <para>with all the whitespace removed.  The original <code>s</code> is still
322      available for use, <code>ls</code> will clean up after itself, and
323      <code>ls.size()</code> will return how many tokens there were.
324   </para>
325   <para>As always, there is a price paid here, in that stringtok is not
326      as fast as strtok.  The other benefits usually outweigh that, however.
327   </para>
328
329   <para><emphasis>Added February 2001:</emphasis>  Mark Wilden pointed out that the
330      standard <code>std::getline()</code> function can be used with standard
331      <code>istringstreams</code> to perform
332      tokenizing as well.  Build an istringstream from the input text,
333      and then use std::getline with varying delimiters (the three-argument
334      signature) to extract tokens into a string.
335   </para>
336
337
338  </section>
339  <section xml:id="strings.string.shrink" xreflabel="Shrink to Fit"><info><title>Shrink to Fit</title></info>
340
341    <para>
342    </para>
343   <para>From GCC 3.4 calling <code>s.reserve(res)</code> on a
344      <code>string s</code> with <code>res &lt; s.capacity()</code> will
345      reduce the string's capacity to <code>std::max(s.size(), res)</code>.
346   </para>
347   <para>This behaviour is suggested, but not required by the standard. Prior
348      to GCC 3.4 the following alternative can be used instead
349   </para>
350   <programlisting>
351      std::string(str.data(), str.size()).swap(str);
352   </programlisting>
353   <para>This is similar to the idiom for reducing
354      a <code>vector</code>'s memory usage
355      (see <link linkend="faq.size_equals_capacity">this FAQ
356      entry</link>) but the regular copy constructor cannot be used
357      because libstdc++'s <code>string</code> is Copy-On-Write in GCC 3.
358   </para>
359   <para>In <link linkend="status.iso.2011">C++11</link> mode you can call
360      <code>s.shrink_to_fit()</code> to achieve the same effect as
361      <code>s.reserve(s.size())</code>.
362   </para>
363
364
365  </section>
366
367  <section xml:id="strings.string.Cstring" xreflabel="CString (MFC)"><info><title>CString (MFC)</title></info>
368
369    <para>
370    </para>
371
372   <para>A common lament seen in various newsgroups deals with the Standard
373      string class as opposed to the Microsoft Foundation Class called
374      CString.  Often programmers realize that a standard portable
375      answer is better than a proprietary nonportable one, but in porting
376      their application from a Win32 platform, they discover that they
377      are relying on special functions offered by the CString class.
378   </para>
379   <para>Things are not as bad as they seem.  In
380      <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://gcc.gnu.org/ml/gcc/1999-04n/msg00236.html">this
381      message</link>, Joe Buck points out a few very important things:
382   </para>
383      <itemizedlist>
384	 <listitem><para>The Standard <code>string</code> supports all the operations
385	     that CString does, with three exceptions.
386	 </para></listitem>
387	 <listitem><para>Two of those exceptions (whitespace trimming and case
388	     conversion) are trivial to implement.  In fact, we do so
389	     on this page.
390	 </para></listitem>
391	 <listitem><para>The third is <code>CString::Format</code>, which allows formatting
392	     in the style of <code>sprintf</code>.  This deserves some mention:
393	 </para></listitem>
394      </itemizedlist>
395   <para>
396      The old libg++ library had a function called form(), which did much
397      the same thing.  But for a Standard solution, you should use the
398      stringstream classes.  These are the bridge between the iostream
399      hierarchy and the string class, and they operate with regular
400      streams seamlessly because they inherit from the iostream
401      hierarchy.  An quick example:
402   </para>
403   <programlisting>
404   #include &lt;iostream&gt;
405   #include &lt;string&gt;
406   #include &lt;sstream&gt;
407
408   string f (string&amp; incoming)     // incoming is "foo  N"
409   {
410       istringstream   incoming_stream(incoming);
411       string          the_word;
412       int             the_number;
413
414       incoming_stream &gt;&gt; the_word        // extract "foo"
415		       &gt;&gt; the_number;     // extract N
416
417       ostringstream   output_stream;
418       output_stream &lt;&lt; "The word was " &lt;&lt; the_word
419		     &lt;&lt; " and 3*N was " &lt;&lt; (3*the_number);
420
421       return output_stream.str();
422   } </programlisting>
423   <para>A serious problem with CString is a design bug in its memory
424      allocation.  Specifically, quoting from that same message:
425   </para>
426   <programlisting>
427   CString suffers from a common programming error that results in
428   poor performance.  Consider the following code:
429
430   CString n_copies_of (const CString&amp; foo, unsigned n)
431   {
432	   CString tmp;
433	   for (unsigned i = 0; i &lt; n; i++)
434		   tmp += foo;
435	   return tmp;
436   }
437
438   This function is O(n^2), not O(n).  The reason is that each +=
439   causes a reallocation and copy of the existing string.  Microsoft
440   applications are full of this kind of thing (quadratic performance
441   on tasks that can be done in linear time) -- on the other hand,
442   we should be thankful, as it's created such a big market for high-end
443   ix86 hardware. :-)
444
445   If you replace CString with string in the above function, the
446   performance is O(n).
447   </programlisting>
448   <para>Joe Buck also pointed out some other things to keep in mind when
449      comparing CString and the Standard string class:
450   </para>
451      <itemizedlist>
452	 <listitem><para>CString permits access to its internal representation; coders
453	     who exploited that may have problems moving to <code>string</code>.
454	 </para></listitem>
455	 <listitem><para>Microsoft ships the source to CString (in the files
456	     MFC\SRC\Str{core,ex}.cpp), so you could fix the allocation
457	     bug and rebuild your MFC libraries.
458	     <emphasis><emphasis>Note:</emphasis> It looks like the CString shipped
459	     with VC++6.0 has fixed this, although it may in fact have been
460	     one of the VC++ SPs that did it.</emphasis>
461	 </para></listitem>
462	 <listitem><para><code>string</code> operations like this have O(n) complexity
463	     <emphasis>if the implementors do it correctly</emphasis>.  The libstdc++
464	     implementors did it correctly.  Other vendors might not.
465	 </para></listitem>
466	 <listitem><para>While parts of the SGI STL are used in libstdc++, their
467	     string class is not.  The SGI <code>string</code> is essentially
468	     <code>vector&lt;char&gt;</code> and does not do any reference
469	     counting like libstdc++'s does.  (It is O(n), though.)
470	     So if you're thinking about SGI's string or rope classes,
471	     you're now looking at four possibilities:  CString, the
472	     libstdc++ string, the SGI string, and the SGI rope, and this
473	     is all before any allocator or traits customizations!  (More
474	     choices than you can shake a stick at -- want fries with that?)
475	 </para></listitem>
476      </itemizedlist>
477
478  </section>
479</section>
480
481<!-- Sect1 03 : Interacting with C -->
482
483</chapter>
484