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