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>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. 3 Strings 4 5</th></tr><tr><td width="20%" align="left"><a accesskey="p" href="traits.html">Prev</a> </td><th width="60%" align="center">Part II. 6 Standard Contents 7 </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. 8 Strings 9 <a id="id-1.3.4.5.1.1.1" class="indexterm"></a> 10</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> 11 Here are Standard, simple, and portable ways to perform common 12 transformations on a <code class="code">string</code> instance, such as 13 "convert to all upper case." The word transformations 14 is especially apt, because the standard template function 15 <code class="code">transform<></code> is used. 16 </p><p> 17 This code will go through some iterations. Here's a simple 18 version: 19 </p><pre class="programlisting"> 20 #include <string> 21 #include <algorithm> 22 #include <cctype> // old <ctype.h> 23 24 struct ToLower 25 { 26 char operator() (char c) const { return std::tolower(c); } 27 }; 28 29 struct ToUpper 30 { 31 char operator() (char c) const { return std::toupper(c); } 32 }; 33 34 int main() 35 { 36 std::string s ("Some Kind Of Initial Input Goes Here"); 37 38 // Change everything into upper case 39 std::transform (s.begin(), s.end(), s.begin(), ToUpper()); 40 41 // Change everything into lower case 42 std::transform (s.begin(), s.end(), s.begin(), ToLower()); 43 44 // Change everything back into upper case, but store the 45 // result in a different string 46 std::string capital_s; 47 capital_s.resize(s.size()); 48 std::transform (s.begin(), s.end(), capital_s.begin(), ToUpper()); 49 } 50 </pre><p> 51 <span class="emphasis"><em>Note</em></span> that these calls all 52 involve the global C locale through the use of the C functions 53 <code class="code">toupper/tolower</code>. This is absolutely guaranteed to work -- 54 but <span class="emphasis"><em>only</em></span> if the string contains <span class="emphasis"><em>only</em></span> characters 55 from the basic source character set, and there are <span class="emphasis"><em>only</em></span> 56 96 of those. Which means that not even all English text can be 57 represented (certain British spellings, proper names, and so forth). 58 So, if all your input forevermore consists of only those 96 59 characters (hahahahahaha), then you're done. 60 </p><p><span class="emphasis"><em>Note</em></span> that the 61 <code class="code">ToUpper</code> and <code class="code">ToLower</code> function objects 62 are needed because <code class="code">toupper</code> and <code class="code">tolower</code> 63 are overloaded names (declared in <code class="code"><cctype></code> and 64 <code class="code"><locale></code>) so the template-arguments for 65 <code class="code">transform<></code> cannot be deduced, as explained in 66 <a class="link" href="http://gcc.gnu.org/ml/libstdc++/2002-11/msg00180.html" target="_top">this 67 message</a>. 68 69 At minimum, you can write short wrappers like 70 </p><pre class="programlisting"> 71 char toLower (char c) 72 { 73 return std::tolower(c); 74 } </pre><p>(Thanks to James Kanze for assistance and suggestions on all of this.) 75 </p><p>Another common operation is trimming off excess whitespace. Much 76 like transformations, this task is trivial with the use of string's 77 <code class="code">find</code> family. These examples are broken into multiple 78 statements for readability: 79 </p><pre class="programlisting"> 80 std::string str (" \t blah blah blah \n "); 81 82 // trim leading whitespace 83 string::size_type notwhite = str.find_first_not_of(" \t\n"); 84 str.erase(0,notwhite); 85 86 // trim trailing whitespace 87 notwhite = str.find_last_not_of(" \t\n"); 88 str.erase(notwhite+1); </pre><p>Obviously, the calls to <code class="code">find</code> could be inserted directly 89 into the calls to <code class="code">erase</code>, in case your compiler does not 90 optimize named temporaries out of existence. 91 </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> 92 </p><p>The well-known-and-if-it-isn't-well-known-it-ought-to-be 93 <a class="link" href="http://www.gotw.ca/gotw/" target="_top">Guru of the Week</a> 94 discussions held on Usenet covered this topic in January of 1998. 95 Briefly, the challenge was, <span class="quote">“<span class="quote">write a 'ci_string' class which 96 is identical to the standard 'string' class, but is 97 case-insensitive in the same way as the (common but nonstandard) 98 C function stricmp()</span>”</span>. 99 </p><pre class="programlisting"> 100 ci_string s( "AbCdE" ); 101 102 // case insensitive 103 assert( s == "abcde" ); 104 assert( s == "ABCDE" ); 105 106 // still case-preserving, of course 107 assert( strcmp( s.c_str(), "AbCdE" ) == 0 ); 108 assert( strcmp( s.c_str(), "abcde" ) != 0 ); </pre><p>The solution is surprisingly easy. The original answer was 109 posted on Usenet, and a revised version appears in Herb Sutter's 110 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>. 111 </p><p>See? Told you it was easy!</p><p> 112 <span class="emphasis"><em>Added June 2000:</em></span> The May 2000 issue of C++ 113 Report contains a fascinating <a class="link" href="http://lafstern.org/matt/col2_new.pdf" target="_top"> article</a> by 114 Matt Austern (yes, <span class="emphasis"><em>the</em></span> Matt Austern) on why 115 case-insensitive comparisons are not as easy as they seem, and 116 why creating a class is the <span class="emphasis"><em>wrong</em></span> way to go 117 about it in production code. (The GotW answer mentions one of 118 the principle difficulties; his article mentions more.) 119 </p><p>Basically, this is "easy" only if you ignore some things, 120 things which may be too important to your program to ignore. (I chose 121 to ignore them when originally writing this entry, and am surprised 122 that nobody ever called me on it...) The GotW question and answer 123 remain useful instructional tools, however. 124 </p><p><span class="emphasis"><em>Added September 2000:</em></span> James Kanze provided a link to a 125 <a class="link" href="http://www.unicode.org/reports/tr21/tr21-5.html" target="_top">Unicode 126 Technical Report discussing case handling</a>, which provides some 127 very good information. 128 </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> 129 </p><p>The <code class="code">std::basic_string</code> is tantalizingly general, in that 130 it is parameterized on the type of the characters which it holds. 131 In theory, you could whip up a Unicode character class and instantiate 132 <code class="code">std::basic_string<my_unicode_char></code>, or assuming 133 that integers are wider than characters on your platform, maybe just 134 declare variables of type <code class="code">std::basic_string<int></code>. 135 </p><p>That's the theory. Remember however that basic_string has additional 136 type parameters, which take default arguments based on the character 137 type (called <code class="code">CharT</code> here): 138 </p><pre class="programlisting"> 139 template <typename CharT, 140 typename Traits = char_traits<CharT>, 141 typename Alloc = allocator<CharT> > 142 class basic_string { .... };</pre><p>Now, <code class="code">allocator<CharT></code> will probably Do The Right 143 Thing by default, unless you need to implement your own allocator 144 for your characters. 145 </p><p>But <code class="code">char_traits</code> takes more work. The char_traits 146 template is <span class="emphasis"><em>declared</em></span> but not <span class="emphasis"><em>defined</em></span>. 147 That means there is only 148 </p><pre class="programlisting"> 149 template <typename CharT> 150 struct char_traits 151 { 152 static void foo (type1 x, type2 y); 153 ... 154 };</pre><p>and functions such as char_traits<CharT>::foo() are not 155 actually defined anywhere for the general case. The C++ standard 156 permits this, because writing such a definition to fit all possible 157 CharT's cannot be done. 158 </p><p>The C++ standard also requires that char_traits be specialized for 159 instantiations of <code class="code">char</code> and <code class="code">wchar_t</code>, and it 160 is these template specializations that permit entities like 161 <code class="code">basic_string<char,char_traits<char>></code> to work. 162 </p><p>If you want to use character types other than char and wchar_t, 163 such as <code class="code">unsigned char</code> and <code class="code">int</code>, you will 164 need suitable specializations for them. For a time, in earlier 165 versions of GCC, there was a mostly-correct implementation that 166 let programmers be lazy but it broke under many situations, so it 167 was removed. GCC 3.4 introduced a new implementation that mostly 168 works and can be specialized even for <code class="code">int</code> and other 169 built-in types. 170 </p><p>If you want to use your own special character class, then you have 171 <a class="link" href="http://gcc.gnu.org/ml/libstdc++/2002-08/msg00163.html" target="_top">a lot 172 of work to do</a>, especially if you with to use i18n features 173 (facets require traits information but don't have a traits argument). 174 </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 175 mailing list</a> and at a later date was put into the file <code class="code"> 176 include/ext/pod_char_traits.h</code>. We agree 177 that the way it's used with basic_string (scroll down to main()) 178 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 179 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 180 be conforming C++</a>, due to the rule that CharT must be a POD. 181 (See how tricky this is?) 182 </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> 183 </p><p>The Standard C (and C++) function <code class="code">strtok()</code> leaves a lot to 184 be desired in terms of user-friendliness. It's unintuitive, it 185 destroys the character string on which it operates, and it requires 186 you to handle all the memory problems. But it does let the client 187 code decide what to use to break the string into pieces; it allows 188 you to choose the "whitespace," so to speak. 189 </p><p>A C++ implementation lets us keep the good things and fix those 190 annoyances. The implementation here is more intuitive (you only 191 call it once, not in a loop with varying argument), it does not 192 affect the original string at all, and all the memory allocation 193 is handled for you. 194 </p><p>It's called stringtok, and it's a template function. Sources are 195 as below, in a less-portable form than it could be, to keep this 196 example simple (for example, see the comments on what kind of 197 string it will accept). 198 </p><pre class="programlisting"> 199#include <string> 200template <typename Container> 201void 202stringtok(Container &container, string const &in, 203 const char * const delimiters = " \t\n") 204{ 205 const string::size_type len = in.length(); 206 string::size_type i = 0; 207 208 while (i < len) 209 { 210 // Eat leading whitespace 211 i = in.find_first_not_of(delimiters, i); 212 if (i == string::npos) 213 return; // Nothing left but white space 214 215 // Find the end of the token 216 string::size_type j = in.find_first_of(delimiters, i); 217 218 // Push token 219 if (j == string::npos) 220 { 221 container.push_back(in.substr(i)); 222 return; 223 } 224 else 225 container.push_back(in.substr(i, j-i)); 226 227 // Set up for next loop 228 i = j + 1; 229 } 230} 231</pre><p> 232 The author uses a more general (but less readable) form of it for 233 parsing command strings and the like. If you compiled and ran this 234 code using it: 235 </p><pre class="programlisting"> 236 std::list<string> ls; 237 stringtok (ls, " this \t is\t\n a test "); 238 for (std::list<string>const_iterator i = ls.begin(); 239 i != ls.end(); ++i) 240 { 241 std::cerr << ':' << (*i) << ":\n"; 242 } </pre><p>You would see this as output: 243 </p><pre class="programlisting"> 244 :this: 245 :is: 246 :a: 247 :test: </pre><p>with all the whitespace removed. The original <code class="code">s</code> is still 248 available for use, <code class="code">ls</code> will clean up after itself, and 249 <code class="code">ls.size()</code> will return how many tokens there were. 250 </p><p>As always, there is a price paid here, in that stringtok is not 251 as fast as strtok. The other benefits usually outweigh that, however. 252 </p><p><span class="emphasis"><em>Added February 2001:</em></span> Mark Wilden pointed out that the 253 standard <code class="code">std::getline()</code> function can be used with standard 254 <code class="code">istringstreams</code> to perform 255 tokenizing as well. Build an istringstream from the input text, 256 and then use std::getline with varying delimiters (the three-argument 257 signature) to extract tokens into a string. 258 </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> 259 </p><p>From GCC 3.4 calling <code class="code">s.reserve(res)</code> on a 260 <code class="code">string s</code> with <code class="code">res < s.capacity()</code> will 261 reduce the string's capacity to <code class="code">std::max(s.size(), res)</code>. 262 </p><p>This behaviour is suggested, but not required by the standard. Prior 263 to GCC 3.4 the following alternative can be used instead 264 </p><pre class="programlisting"> 265 std::string(str.data(), str.size()).swap(str); 266 </pre><p>This is similar to the idiom for reducing 267 a <code class="code">vector</code>'s memory usage 268 (see <a class="link" href="../faq.html#faq.size_equals_capacity" title="7.8.">this FAQ 269 entry</a>) but the regular copy constructor cannot be used 270 because libstdc++'s <code class="code">string</code> is Copy-On-Write in GCC 3. 271 </p><p>In <a class="link" href="status.html#status.iso.2011" title="C++ 2011">C++11</a> mode you can call 272 <code class="code">s.shrink_to_fit()</code> to achieve the same effect as 273 <code class="code">s.reserve(s.size())</code>. 274 </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> 275 </p><p>A common lament seen in various newsgroups deals with the Standard 276 string class as opposed to the Microsoft Foundation Class called 277 CString. Often programmers realize that a standard portable 278 answer is better than a proprietary nonportable one, but in porting 279 their application from a Win32 platform, they discover that they 280 are relying on special functions offered by the CString class. 281 </p><p>Things are not as bad as they seem. In 282 <a class="link" href="http://gcc.gnu.org/ml/gcc/1999-04n/msg00236.html" target="_top">this 283 message</a>, Joe Buck points out a few very important things: 284 </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 285 that CString does, with three exceptions. 286 </p></li><li class="listitem"><p>Two of those exceptions (whitespace trimming and case 287 conversion) are trivial to implement. In fact, we do so 288 on this page. 289 </p></li><li class="listitem"><p>The third is <code class="code">CString::Format</code>, which allows formatting 290 in the style of <code class="code">sprintf</code>. This deserves some mention: 291 </p></li></ul></div><p> 292 The old libg++ library had a function called form(), which did much 293 the same thing. But for a Standard solution, you should use the 294 stringstream classes. These are the bridge between the iostream 295 hierarchy and the string class, and they operate with regular 296 streams seamlessly because they inherit from the iostream 297 hierarchy. An quick example: 298 </p><pre class="programlisting"> 299 #include <iostream> 300 #include <string> 301 #include <sstream> 302 303 string f (string& incoming) // incoming is "foo N" 304 { 305 istringstream incoming_stream(incoming); 306 string the_word; 307 int the_number; 308 309 incoming_stream >> the_word // extract "foo" 310 >> the_number; // extract N 311 312 ostringstream output_stream; 313 output_stream << "The word was " << the_word 314 << " and 3*N was " << (3*the_number); 315 316 return output_stream.str(); 317 } </pre><p>A serious problem with CString is a design bug in its memory 318 allocation. Specifically, quoting from that same message: 319 </p><pre class="programlisting"> 320 CString suffers from a common programming error that results in 321 poor performance. Consider the following code: 322 323 CString n_copies_of (const CString& foo, unsigned n) 324 { 325 CString tmp; 326 for (unsigned i = 0; i < n; i++) 327 tmp += foo; 328 return tmp; 329 } 330 331 This function is O(n^2), not O(n). The reason is that each += 332 causes a reallocation and copy of the existing string. Microsoft 333 applications are full of this kind of thing (quadratic performance 334 on tasks that can be done in linear time) -- on the other hand, 335 we should be thankful, as it's created such a big market for high-end 336 ix86 hardware. :-) 337 338 If you replace CString with string in the above function, the 339 performance is O(n). 340 </pre><p>Joe Buck also pointed out some other things to keep in mind when 341 comparing CString and the Standard string class: 342 </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>CString permits access to its internal representation; coders 343 who exploited that may have problems moving to <code class="code">string</code>. 344 </p></li><li class="listitem"><p>Microsoft ships the source to CString (in the files 345 MFC\SRC\Str{core,ex}.cpp), so you could fix the allocation 346 bug and rebuild your MFC libraries. 347 <span class="emphasis"><em><span class="emphasis"><em>Note:</em></span> It looks like the CString shipped 348 with VC++6.0 has fixed this, although it may in fact have been 349 one of the VC++ SPs that did it.</em></span> 350 </p></li><li class="listitem"><p><code class="code">string</code> operations like this have O(n) complexity 351 <span class="emphasis"><em>if the implementors do it correctly</em></span>. The libstdc++ 352 implementors did it correctly. Other vendors might not. 353 </p></li><li class="listitem"><p>While parts of the SGI STL are used in libstdc++, their 354 string class is not. The SGI <code class="code">string</code> is essentially 355 <code class="code">vector<char></code> and does not do any reference 356 counting like libstdc++'s does. (It is O(n), though.) 357 So if you're thinking about SGI's string or rope classes, 358 you're now looking at four possibilities: CString, the 359 libstdc++ string, the SGI string, and the SGI rope, and this 360 is all before any allocator or traits customizations! (More 361 choices than you can shake a stick at -- want fries with that?) 362 </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. 363 Localization 364 365</td></tr></table></div></body></html>