1<?xml version="1.0" encoding="ISO-8859-1"?> 2<!DOCTYPE html 3 PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" 4 "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> 5 6<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en"> 7<head> 8 <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1" /> 9 <meta name="AUTHOR" content="pme@gcc.gnu.org (Phil Edwards)" /> 10 <meta name="KEYWORDS" content="HOWTO, libstdc++, GCC, g++, libg++, STL" /> 11 <meta name="DESCRIPTION" content="HOWTO for the libstdc++ chapter 27." /> 12 <meta name="GENERATOR" content="vi and eight fingers" /> 13 <title>libstdc++-v3 HOWTO: Chapter 27: Input/Output</title> 14<link rel="StyleSheet" href="../lib3styles.css" type="text/css" /> 15<link rel="Start" href="../documentation.html" type="text/html" 16 title="GNU C++ Standard Library" /> 17<link rel="Prev" href="../26_numerics/howto.html" type="text/html" 18 title="Numerics" /> 19<link rel="Next" href="../ext/howto.html" type="text/html" 20 title="Extensions" /> 21<link rel="Copyright" href="../17_intro/license.html" type="text/html" /> 22<link rel="Help" href="../faq/index.html" type="text/html" title="F.A.Q." /> 23</head> 24<body> 25 26<h1 class="centered"><a name="top">Chapter 27: Input/Output</a></h1> 27 28<p>Chapter 27 deals with iostreams and all their subcomponents 29 and extensions. All <em>kinds</em> of fun stuff. 30</p> 31 32 33<!-- ####################################################### --> 34<hr /> 35<h1>Contents</h1> 36<ul> 37 <li><a href="#1">Copying a file</a></li> 38 <li><a href="#2">The buffering is screwing up my program!</a></li> 39 <li><a href="#3">Binary I/O</a></li> 40 <li><a href="#5">What is this <sstream>/stringstreams thing?</a></li> 41 <li><a href="#6">Deriving a stream buffer</a></li> 42 <li><a href="#7">More on binary I/O</a></li> 43 <li><a href="#8">Pathetic performance? Ditch C.</a></li> 44 <li><a href="#9">Threads and I/O</a></li> 45 <li><a href="#10">Which header?</a></li> 46 <li><a href="#11">Using FILE*s and file descriptors with IOStreams</a></li> 47</ul> 48 49<hr /> 50 51<!-- ####################################################### --> 52 53<h2><a name="1">Copying a file</a></h2> 54 <p>So you want to copy a file quickly and easily, and most important, 55 completely portably. And since this is C++, you have an open 56 ifstream (call it IN) and an open ofstream (call it OUT): 57 </p> 58 <pre> 59 #include <fstream> 60 61 std::ifstream IN ("input_file"); 62 std::ofstream OUT ("output_file"); </pre> 63 <p>Here's the easiest way to get it completely wrong: 64 </p> 65 <pre> 66 OUT << IN;</pre> 67 <p>For those of you who don't already know why this doesn't work 68 (probably from having done it before), I invite you to quickly 69 create a simple text file called "input_file" containing 70 the sentence 71 </p> 72 <pre> 73 The quick brown fox jumped over the lazy dog.</pre> 74 <p>surrounded by blank lines. Code it up and try it. The contents 75 of "output_file" may surprise you. 76 </p> 77 <p>Seriously, go do it. Get surprised, then come back. It's worth it. 78 </p> 79 <hr width="60%" /> 80 <p>The thing to remember is that the <code>basic_[io]stream</code> classes 81 handle formatting, nothing else. In particular, they break up on 82 whitespace. The actual reading, writing, and storing of data is 83 handled by the <code>basic_streambuf</code> family. Fortunately, the 84 <code>operator<<</code> is overloaded to take an ostream and 85 a pointer-to-streambuf, in order to help with just this kind of 86 "dump the data verbatim" situation. 87 </p> 88 <p>Why a <em>pointer</em> to streambuf and not just a streambuf? Well, 89 the [io]streams hold pointers (or references, depending on the 90 implementation) to their buffers, not the actual 91 buffers. This allows polymorphic behavior on the part of the buffers 92 as well as the streams themselves. The pointer is easily retrieved 93 using the <code>rdbuf()</code> member function. Therefore, the easiest 94 way to copy the file is: 95 </p> 96 <pre> 97 OUT << IN.rdbuf();</pre> 98 <p>So what <em>was</em> happening with OUT<<IN? Undefined 99 behavior, since that particular << isn't defined by the Standard. 100 I have seen instances where it is implemented, but the character 101 extraction process removes all the whitespace, leaving you with no 102 blank lines and only "Thequickbrownfox...". With 103 libraries that do not define that operator, IN (or one of IN's 104 member pointers) sometimes gets converted to a void*, and the output 105 file then contains a perfect text representation of a hexadecimal 106 address (quite a big surprise). Others don't compile at all. 107 </p> 108 <p>Also note that none of this is specific to o<b>*f*</b>streams. 109 The operators shown above are all defined in the parent 110 basic_ostream class and are therefore available with all possible 111 descendants. 112 </p> 113 <p>Return <a href="#top">to top of page</a> or 114 <a href="../faq/index.html">to the FAQ</a>. 115 </p> 116 117<hr /> 118<h2><a name="2">The buffering is screwing up my program!</a></h2> 119<!-- 120 This is not written very well. I need to redo this section. 121--> 122 <p>First, are you sure that you understand buffering? Particularly 123 the fact that C++ may not, in fact, have anything to do with it? 124 </p> 125 <p>The rules for buffering can be a little odd, but they aren't any 126 different from those of C. (Maybe that's why they can be a bit 127 odd.) Many people think that writing a newline to an output 128 stream automatically flushes the output buffer. This is true only 129 when the output stream is, in fact, a terminal and not a file 130 or some other device -- and <em>that</em> may not even be true 131 since C++ says nothing about files nor terminals. All of that is 132 system-dependent. (The "newline-buffer-flushing only occurring 133 on terminals" thing is mostly true on Unix systems, though.) 134 </p> 135 <p>Some people also believe that sending <code>endl</code> down an 136 output stream only writes a newline. This is incorrect; after a 137 newline is written, the buffer is also flushed. Perhaps this 138 is the effect you want when writing to a screen -- get the text 139 out as soon as possible, etc -- but the buffering is largely 140 wasted when doing this to a file: 141 </p> 142 <pre> 143 output << "a line of text" << endl; 144 output << some_data_variable << endl; 145 output << "another line of text" << endl; </pre> 146 <p>The proper thing to do in this case to just write the data out 147 and let the libraries and the system worry about the buffering. 148 If you need a newline, just write a newline: 149 </p> 150 <pre> 151 output << "a line of text\n" 152 << some_data_variable << '\n' 153 << "another line of text\n"; </pre> 154 <p>I have also joined the output statements into a single statement. 155 You could make the code prettier by moving the single newline to 156 the start of the quoted text on the last line, for example. 157 </p> 158 <p>If you do need to flush the buffer above, you can send an 159 <code>endl</code> if you also need a newline, or just flush the buffer 160 yourself: 161 </p> 162 <pre> 163 output << ...... << flush; // can use std::flush manipulator 164 output.flush(); // or call a member fn </pre> 165 <p>On the other hand, there are times when writing to a file should 166 be like writing to standard error; no buffering should be done 167 because the data needs to appear quickly (a prime example is a 168 log file for security-related information). The way to do this is 169 just to turn off the buffering <em>before any I/O operations at 170 all</em> have been done (note that opening counts as an I/O operation): 171 </p> 172 <pre> 173 std::ofstream os; 174 std::ifstream is; 175 int i; 176 177 os.rdbuf()->pubsetbuf(0,0); 178 is.rdbuf()->pubsetbuf(0,0); 179 180 os.open("/foo/bar/baz"); 181 is.open("/qux/quux/quuux"); 182 ... 183 os << "this data is written immediately\n"; 184 is >> i; // and this will probably cause a disk read </pre> 185 <p>Since all aspects of buffering are handled by a streambuf-derived 186 member, it is necessary to get at that member with <code>rdbuf()</code>. 187 Then the public version of <code>setbuf</code> can be called. The 188 arguments are the same as those for the Standard C I/O Library 189 function (a buffer area followed by its size). 190 </p> 191 <p>A great deal of this is implementation-dependent. For example, 192 <code>streambuf</code> does not specify any actions for its own 193 <code>setbuf()</code>-ish functions; the classes derived from 194 <code>streambuf</code> each define behavior that "makes 195 sense" for that class: an argument of (0,0) turns off buffering 196 for <code>filebuf</code> but does nothing at all for its siblings 197 <code>stringbuf</code> and <code>strstreambuf</code>, and specifying 198 anything other than (0,0) has varying effects. 199 User-defined classes derived from <code>streambuf</code> can 200 do whatever they want. (For <code>filebuf</code> and arguments for 201 <code>(p,s)</code> other than zeros, libstdc++ does what you'd expect: 202 the first <code>s</code> bytes of <code>p</code> are used as a buffer, 203 which you must allocate and deallocate.) 204 </p> 205 <p>A last reminder: there are usually more buffers involved than 206 just those at the language/library level. Kernel buffers, disk 207 buffers, and the like will also have an effect. Inspecting and 208 changing those are system-dependent. 209 </p> 210 <p>Return <a href="#top">to top of page</a> or 211 <a href="../faq/index.html">to the FAQ</a>. 212 </p> 213 214<hr /> 215<h2><a name="3">Binary I/O</a></h2> 216 <p>The first and most important thing to remember about binary I/O is 217 that opening a file with <code>ios::binary</code> is not, repeat 218 <em>not</em>, the only thing you have to do. It is not a silver 219 bullet, and will not allow you to use the <code><</>></code> 220 operators of the normal fstreams to do binary I/O. 221 </p> 222 <p>Sorry. Them's the breaks. 223 </p> 224 <p>This isn't going to try and be a complete tutorial on reading and 225 writing binary files (because "binary" 226 <a href="#7">covers a lot of ground)</a>, but we will try and clear 227 up a couple of misconceptions and common errors. 228 </p> 229 <p>First, <code>ios::binary</code> has exactly one defined effect, no more 230 and no less. Normal text mode has to be concerned with the newline 231 characters, and the runtime system will translate between (for 232 example) '\n' and the appropriate end-of-line sequence (LF on Unix, 233 CRLF on DOS, CR on Macintosh, etc). (There are other things that 234 normal mode does, but that's the most obvious.) Opening a file in 235 binary mode disables this conversion, so reading a CRLF sequence 236 under Windows won't accidentally get mapped to a '\n' character, etc. 237 Binary mode is not supposed to suddenly give you a bitstream, and 238 if it is doing so in your program then you've discovered a bug in 239 your vendor's compiler (or some other part of the C++ implementation, 240 possibly the runtime system). 241 </p> 242 <p>Second, using <code><<</code> to write and <code>>></code> to 243 read isn't going to work with the standard file stream classes, even 244 if you use <code>skipws</code> during reading. Why not? Because 245 ifstream and ofstream exist for the purpose of <em>formatting</em>, 246 not reading and writing. Their job is to interpret the data into 247 text characters, and that's exactly what you don't want to happen 248 during binary I/O. 249 </p> 250 <p>Third, using the <code>get()</code> and <code>put()/write()</code> member 251 functions still aren't guaranteed to help you. These are 252 "unformatted" I/O functions, but still character-based. 253 (This may or may not be what you want, see below.) 254 </p> 255 <p>Notice how all the problems here are due to the inappropriate use 256 of <em>formatting</em> functions and classes to perform something 257 which <em>requires</em> that formatting not be done? There are a 258 seemingly infinite number of solutions, and a few are listed here: 259 </p> 260 <ul> 261 <li>"Derive your own fstream-type classes and write your own 262 <</>> operators to do binary I/O on whatever data 263 types you're using." This is a Bad Thing, because while 264 the compiler would probably be just fine with it, other humans 265 are going to be confused. The overloaded bitshift operators 266 have a well-defined meaning (formatting), and this breaks it. 267 </li> 268 <li>"Build the file structure in memory, then <code>mmap()</code> 269 the file and copy the structure." Well, this is easy to 270 make work, and easy to break, and is pretty equivalent to 271 using <code>::read()</code> and <code>::write()</code> directly, and 272 makes no use of the iostream library at all... 273 </li> 274 <li>"Use streambufs, that's what they're there for." 275 While not trivial for the beginner, this is the best of all 276 solutions. The streambuf/filebuf layer is the layer that is 277 responsible for actual I/O. If you want to use the C++ 278 library for binary I/O, this is where you start. 279 </li> 280 </ul> 281 <p>How to go about using streambufs is a bit beyond the scope of this 282 document (at least for now), but while streambufs go a long way, 283 they still leave a couple of things up to you, the programmer. 284 As an example, byte ordering is completely between you and the 285 operating system, and you have to handle it yourself. 286 </p> 287 <p>Deriving a streambuf or filebuf 288 class from the standard ones, one that is specific to your data 289 types (or an abstraction thereof) is probably a good idea, and 290 lots of examples exist in journals and on Usenet. Using the 291 standard filebufs directly (either by declaring your own or by 292 using the pointer returned from an fstream's <code>rdbuf()</code>) 293 is certainly feasible as well. 294 </p> 295 <p>One area that causes problems is trying to do bit-by-bit operations 296 with filebufs. C++ is no different from C in this respect: I/O 297 must be done at the byte level. If you're trying to read or write 298 a few bits at a time, you're going about it the wrong way. You 299 must read/write an integral number of bytes and then process the 300 bytes. (For example, the streambuf functions take and return 301 variables of type <code>int_type</code>.) 302 </p> 303 <p>Another area of problems is opening text files in binary mode. 304 Generally, binary mode is intended for binary files, and opening 305 text files in binary mode means that you now have to deal with all of 306 those end-of-line and end-of-file problems that we mentioned before. 307 An instructive thread from comp.lang.c++.moderated delved off into 308 this topic starting more or less at 309 <a href="http://groups.google.com/groups?oi=djq&selm=an_436187505">this</a> 310 article and continuing to the end of the thread. (You'll have to 311 sort through some flames every couple of paragraphs, but the points 312 made are good ones.) 313 </p> 314 315<hr /> 316<h2><a name="5">What is this <sstream>/stringstreams thing?</a></h2> 317 <p>Stringstreams (defined in the header <code><sstream></code>) 318 are in this author's opinion one of the coolest things since 319 sliced time. An example of their use is in the Received Wisdom 320 section for Chapter 21 (Strings), 321 <a href="../21_strings/howto.html#1.1internal"> describing how to 322 format strings</a>. 323 </p> 324 <p>The quick definition is: they are siblings of ifstream and ofstream, 325 and they do for <code>std::string</code> what their siblings do for 326 files. All that work you put into writing <code><<</code> and 327 <code>>></code> functions for your classes now pays off 328 <em>again!</em> Need to format a string before passing the string 329 to a function? Send your stuff via <code><<</code> to an 330 ostringstream. You've read a string as input and need to parse it? 331 Initialize an istringstream with that string, and then pull pieces 332 out of it with <code>>></code>. Have a stringstream and need to 333 get a copy of the string inside? Just call the <code>str()</code> 334 member function. 335 </p> 336 <p>This only works if you've written your 337 <code><<</code>/<code>>></code> functions correctly, though, 338 and correctly means that they take istreams and ostreams as 339 parameters, not i<b>f</b>streams and o<b>f</b>streams. If they 340 take the latter, then your I/O operators will work fine with 341 file streams, but with nothing else -- including stringstreams. 342 </p> 343 <p>If you are a user of the strstream classes, you need to update 344 your code. You don't have to explicitly append <code>ends</code> to 345 terminate the C-style character array, you don't have to mess with 346 "freezing" functions, and you don't have to manage the 347 memory yourself. The strstreams have been officially deprecated, 348 which means that 1) future revisions of the C++ Standard won't 349 support them, and 2) if you use them, people will laugh at you. 350 </p> 351 352<hr /> 353<h2><a name="6">Deriving a stream buffer</a></h2> 354 <p>Creating your own stream buffers for I/O can be remarkably easy. 355 If you are interested in doing so, we highly recommend two very 356 excellent books: 357 <a href="http://www.langer.camelot.de/iostreams.html">Standard C++ 358 IOStreams and Locales</a> by Langer and Kreft, ISBN 0-201-18395-1, and 359 <a href="http://www.josuttis.com/libbook/">The C++ Standard Library</a> 360 by Nicolai Josuttis, ISBN 0-201-37926-0. Both are published by 361 Addison-Wesley, who isn't paying us a cent for saying that, honest. 362 </p> 363 <p>Here is a simple example, io/outbuf1, from the Josuttis text. It 364 transforms everything sent through it to uppercase. This version 365 assumes many things about the nature of the character type being 366 used (for more information, read the books or the newsgroups): 367 </p> 368 <pre> 369 #include <iostream> 370 #include <streambuf> 371 #include <locale> 372 #include <cstdio> 373 374 class outbuf : public std::streambuf 375 { 376 protected: 377 /* central output function 378 * - print characters in uppercase mode 379 */ 380 virtual int_type overflow (int_type c) { 381 if (c != EOF) { 382 // convert lowercase to uppercase 383 c = std::toupper(static_cast<char>(c),getloc()); 384 385 // and write the character to the standard output 386 if (putchar(c) == EOF) { 387 return EOF; 388 } 389 } 390 return c; 391 } 392 }; 393 394 int main() 395 { 396 // create special output buffer 397 outbuf ob; 398 // initialize output stream with that output buffer 399 std::ostream out(&ob); 400 401 out << "31 hexadecimal: " 402 << std::hex << 31 << std::endl; 403 return 0; 404 } 405 </pre> 406 <p>Try it yourself! More examples can be found in 3.1.x code, in 407 <code>include/ext/*_filebuf.h</code>, and on 408 <a href="http://www.informatik.uni-konstanz.de/~kuehl/c++/iostream/">Dietmar 409 Kühl's IOStreams page</a>. 410 </p> 411 412<hr /> 413<h2><a name="7">More on binary I/O</a></h2> 414 <p>Towards the beginning of February 2001, the subject of 415 "binary" I/O was brought up in a couple of places at the 416 same time. One notable place was Usenet, where James Kanze and 417 Dietmar Kühl separately posted articles on why attempting 418 generic binary I/O was not a good idea. (Here are copies of 419 <a href="binary_iostreams_kanze.txt">Kanze's article</a> and 420 <a href="binary_iostreams_kuehl.txt">Kühl's article</a>.) 421 </p> 422 <p>Briefly, the problems of byte ordering and type sizes mean that 423 the unformatted functions like <code>ostream::put()</code> and 424 <code>istream::get()</code> cannot safely be used to communicate 425 between arbitrary programs, or across a network, or from one 426 invocation of a program to another invocation of the same program 427 on a different platform, etc. 428 </p> 429 <p>The entire Usenet thread is instructive, and took place under the 430 subject heading "binary iostreams" on both comp.std.c++ 431 and comp.lang.c++.moderated in parallel. Also in that thread, 432 Dietmar Kühl mentioned that he had written a pair of stream 433 classes that would read and write XDR, which is a good step towards 434 a portable binary format. 435 </p> 436 437<hr /> 438<h2><a name="8">Pathetic performance? Ditch C.</a></h2> 439 <p>It sounds like a flame on C, but it isn't. Really. Calm down. 440 I'm just saying it to get your attention. 441 </p> 442 <p>Because the C++ library includes the C library, both C-style and 443 C++-style I/O have to work at the same time. For example: 444 </p> 445 <pre> 446 #include <iostream> 447 #include <cstdio> 448 449 std::cout << "Hel"; 450 std::printf ("lo, worl"); 451 std::cout << "d!\n"; 452 </pre> 453 <p>This must do what you think it does. 454 </p> 455 <p>Alert members of the audience will immediately notice that buffering 456 is going to make a hash of the output unless special steps are taken. 457 </p> 458 <p>The special steps taken by libstdc++, at least for version 3.0, 459 involve doing very little buffering for the standard streams, leaving 460 most of the buffering to the underlying C library. (This kind of 461 thing is <a href="../explanations.html#cstdio">tricky to get right</a>.) 462 The upside is that correctness is ensured. The downside is that 463 writing through <code>cout</code> can quite easily lead to awful 464 performance when the C++ I/O library is layered on top of the C I/O 465 library (as it is for 3.0 by default). Some patches have been applied 466 which improve the situation for 3.1. 467 </p> 468 <p>However, the C and C++ standard streams only need to be kept in sync 469 when both libraries' facilities are in use. If your program only uses 470 C++ I/O, then there's no need to sync with the C streams. The right 471 thing to do in this case is to call 472 </p> 473 <pre> 474 #include <em>any of the I/O headers such as ios, iostream, etc</em> 475 476 std::ios::sync_with_stdio(false); 477 </pre> 478 <p>You must do this before performing any I/O via the C++ stream objects. 479 Once you call this, the C++ streams will operate independently of the 480 (unused) C streams. For GCC 3.x, this means that <code>cout</code> and 481 company will become fully buffered on their own. 482 </p> 483 <p>Note, by the way, that the synchronization requirement only applies to 484 the standard streams (<code>cin</code>, <code>cout</code>, 485 <code>cerr</code>, 486 <code>clog</code>, and their wide-character counterparts). File stream 487 objects that you declare yourself have no such requirement and are fully 488 buffered. 489 </p> 490 491<hr /> 492<h2><a name="9">Threads and I/O</a></h2> 493 <p>I'll assume that you have already read the 494 <a href="../17_intro/howto.html#3">general notes on library threads</a>, 495 and the 496 <a href="../23_containers/howto.html#3">notes on threaded container 497 access</a> (you might not think of an I/O stream as a container, but 498 the points made there also hold here). If you have not read them, 499 please do so first. 500 </p> 501 <p>This gets a bit tricky. Please read carefully, and bear with me. 502 </p> 503 <h3>Structure</h3> 504 <p>As described <a href="../explanations.html#cstdio">here</a>, a wrapper 505 type called <code>__basic_file</code> provides our abstraction layer 506 for the <code>std::filebuf</code> classes. Nearly all decisions dealing 507 with actual input and output must be made in <code>__basic_file</code>. 508 </p> 509 <p>A generic locking mechanism is somewhat in place at the filebuf layer, 510 but is not used in the current code. Providing locking at any higher 511 level is akin to providing locking within containers, and is not done 512 for the same reasons (see the links above). 513 </p> 514 <h3>The defaults for 3.0.x</h3> 515 <p>The __basic_file type is simply a collection of small wrappers around 516 the C stdio layer (again, see the link under Structure). We do no 517 locking ourselves, but simply pass through to calls to <code>fopen</code>, 518 <code>fwrite</code>, and so forth. 519 </p> 520 <p>So, for 3.0, the question of "is multithreading safe for I/O" 521 must be answered with, "is your platform's C library threadsafe 522 for I/O?" Some are by default, some are not; many offer multiple 523 implementations of the C library with varying tradeoffs of threadsafety 524 and efficiency. You, the programmer, are always required to take care 525 with multiple threads. 526 </p> 527 <p>(As an example, the POSIX standard requires that C stdio FILE* 528 operations are atomic. POSIX-conforming C libraries (e.g, on Solaris 529 and GNU/Linux) have an internal mutex to serialize operations on 530 FILE*s. However, you still need to not do stupid things like calling 531 <code>fclose(fs)</code> in one thread followed by an access of 532 <code>fs</code> in another.) 533 </p> 534 <p>So, if your platform's C library is threadsafe, then your 535 <code>fstream</code> I/O operations will be threadsafe at the lowest 536 level. For higher-level operations, such as manipulating the data 537 contained in the stream formatting classes (e.g., setting up callbacks 538 inside an <code>std::ofstream</code>), you need to guard such accesses 539 like any other critical shared resource. 540 </p> 541 <h3>The future</h3> 542 <p>As already mentioned <a href="../explanations.html#cstdio">here</a>, a 543 second choice is available for I/O implementations: libio. This is 544 disabled by default, and in fact will not currently work due to other 545 issues. It will be revisited, however. 546 </p> 547 <p>The libio code is a subset of the guts of the GNU libc (glibc) I/O 548 implementation. When libio is in use, the <code>__basic_file</code> 549 type is basically derived from FILE. (The real situation is more 550 complex than that... it's derived from an internal type used to 551 implement FILE. See libio/libioP.h to see scary things done with 552 vtbls.) The result is that there is no "layer" of C stdio 553 to go through; the filebuf makes calls directly into the same 554 functions used to implement <code>fread</code>, <code>fwrite</code>, 555 and so forth, using internal data structures. (And when I say 556 "makes calls directly," I mean the function is literally 557 replaced by a jump into an internal function. Fast but frightening. 558 *grin*) 559 </p> 560 <p>Also, the libio internal locks are used. This requires pulling in 561 large chunks of glibc, such as a pthreads implementation, and is one 562 of the issues preventing widespread use of libio as the libstdc++ 563 cstdio implementation. 564 </p> 565 <p>But we plan to make this work, at least as an option if not a future 566 default. Platforms running a copy of glibc with a recent-enough 567 version will see calls from libstdc++ directly into the glibc already 568 installed. For other platforms, a copy of the libio subsection will 569 be built and included in libstdc++. 570 </p> 571 <h3>Alternatives</h3> 572 <p>Don't forget that other cstdio implementations are possible. You could 573 easily write one to perform your own forms of locking, to solve your 574 "interesting" problems. 575 </p> 576 577<hr /> 578<h2><a name="10">Which header?</a></h2> 579 <p>To minimize the time you have to wait on the compiler, it's good to 580 only include the headers you really need. Many people simply include 581 <iostream> when they don't need to -- and that can <em>penalize 582 your runtime as well.</em> Here are some tips on which header to use 583 for which situations, starting with the simplest. 584 </p> 585 <p><strong><iosfwd></strong> should be included whenever you simply 586 need the <em>name</em> of an I/O-related class, such as 587 "ofstream" or "basic_streambuf". Like the name 588 implies, these are forward declarations. (A word to all you fellow 589 old school programmers: trying to forward declare classes like 590 "class istream;" won't work. Look in the iosfwd header if 591 you'd like to know why.) For example, 592 </p> 593 <pre> 594 #include <iosfwd> 595 596 class MyClass 597 { 598 .... 599 std::ifstream& input_file; 600 }; 601 602 extern std::ostream& operator<< (std::ostream&, MyClass&); 603 </pre> 604 <p><strong><ios></strong> declares the base classes for the entire 605 I/O stream hierarchy, std::ios_base and std::basic_ios<charT>, the 606 counting types std::streamoff and std::streamsize, the file 607 positioning type std::fpos, and the various manipulators like 608 std::hex, std::fixed, std::noshowbase, and so forth. 609 </p> 610 <p>The ios_base class is what holds the format flags, the state flags, 611 and the functions which change them (setf(), width(), precision(), 612 etc). You can also store extra data and register callback functions 613 through ios_base, but that has been historically underused. Anything 614 which doesn't depend on the type of characters stored is consolidated 615 here. 616 </p> 617 <p>The template class basic_ios is the highest template class in the 618 hierarchy; it is the first one depending on the character type, and 619 holds all general state associated with that type: the pointer to the 620 polymorphic stream buffer, the facet information, etc. 621 </p> 622 <p><strong><streambuf></strong> declares the template class 623 basic_streambuf, and two standard instantiations, streambuf and 624 wstreambuf. If you need to work with the vastly useful and capable 625 stream buffer classes, e.g., to create a new form of storage 626 transport, this header is the one to include. 627 </p> 628 <p><strong><istream></strong>/<strong><ostream></strong> are 629 the headers to include when you are using the >>/<< 630 interface, or any of the other abstract stream formatting functions. 631 For example, 632 </p> 633 <pre> 634 #include <istream> 635 636 std::ostream& operator<< (std::ostream& os, MyClass& c) 637 { 638 return os << c.data1() << c.data2(); 639 } 640 </pre> 641 <p>The std::istream and std::ostream classes are the abstract parents of 642 the various concrete implementations. If you are only using the 643 interfaces, then you only need to use the appropriate interface header. 644 </p> 645 <p><strong><iomanip></strong> provides "extractors and inserters 646 that alter information maintained by class ios_base and its derived 647 classes," such as std::setprecision and std::setw. If you need 648 to write expressions like <code>os << setw(3);</code> or 649 <code>is >> setbase(8);</code>, you must include <iomanip>. 650 </p> 651 <p><strong><sstream></strong>/<strong><fstream></strong> 652 declare the six stringstream and fstream classes. As they are the 653 standard concrete descendants of istream and ostream, you will already 654 know about them. 655 </p> 656 <p>Finally, <strong><iostream></strong> provides the eight standard 657 global objects (cin, cout, etc). To do this correctly, this header 658 also provides the contents of the <istream> and <ostream> 659 headers, but nothing else. The contents of this header look like 660 </p> 661 <pre> 662 #include <ostream> 663 #include <istream> 664 665 namespace std 666 { 667 extern istream cin; 668 extern ostream cout; 669 .... 670 671 // this is explained below 672 <strong>static ios_base::Init __foo;</strong> // not its real name 673 } 674 </pre> 675 <p>Now, the runtime penalty mentioned previously: the global objects 676 must be initialized before any of your own code uses them; this is 677 guaranteed by the standard. Like any other global object, they must 678 be initialized once and only once. This is typically done with a 679 construct like the one above, and the nested class ios_base::Init is 680 specified in the standard for just this reason. 681 </p> 682 <p>How does it work? Because the header is included before any of your 683 code, the <strong>__foo</strong> object is constructed before any of 684 your objects. (Global objects are built in the order in which they 685 are declared, and destroyed in reverse order.) The first time the 686 constructor runs, the eight stream objects are set up. 687 </p> 688 <p>The <code>static</code> keyword means that each object file compiled 689 from a source file containing <iostream> will have its own 690 private copy of <strong>__foo</strong>. There is no specified order 691 of construction across object files (it's one of those pesky NP 692 problems that make life so interesting), so one copy in each object 693 file means that the stream objects are guaranteed to be set up before 694 any of your code which uses them could run, thereby meeting the 695 requirements of the standard. 696 </p> 697 <p>The penalty, of course, is that after the first copy of 698 <strong>__foo</strong> is constructed, all the others are just wasted 699 processor time. The time spent is merely for an increment-and-test 700 inside a function call, but over several dozen or hundreds of object 701 files, that time can add up. (It's not in a tight loop, either.) 702 </p> 703 <p>The lesson? Only include <iostream> when you need to use one of 704 the standard objects in that source file; you'll pay less startup 705 time. Only include the header files you need to in general; your 706 compile times will go down when there's less parsing work to do. 707 </p> 708 709 710<hr /> 711<h2><a name="11">Using FILE*s and file descriptors with IOStreams</a></h2> 712 <!-- referenced by ext/howto.html#2, update link if numbering changes --> 713 <p>The v2 library included non-standard extensions to construct 714 <code>std::filebuf</code>s from C stdio types such as 715 <code>FILE*</code>s and POSIX file descriptors. 716 Today the recommended way to use stdio types with libstdc++-v3 717 IOStreams is via the <code>stdio_filebuf</code> class (see below), 718 but earlier releases provided slightly different mechanisms. 719 </p> 720 <ul> 721 <li>3.0.x <code>filebuf</code>s have another ctor with this signature: 722 <br /> 723 <code>basic_filebuf(__c_file_type*, ios_base::openmode, int_type);</code> 724 <br />This comes in very handy in a number of places, such as 725 attaching Unix sockets, pipes, and anything else which uses file 726 descriptors, into the IOStream buffering classes. The three 727 arguments are as follows: 728 <ul> 729 <li><code>__c_file_type* F </code> 730 // the __c_file_type typedef usually boils down to stdio's FILE 731 </li> 732 <li><code>ios_base::openmode M </code> 733 // same as all the other uses of openmode 734 </li> 735 <li><code>int_type B </code> 736 // buffer size, defaults to BUFSIZ if not specified 737 </li> 738 </ul> 739 For those wanting to use file descriptors instead of FILE*'s, I 740 invite you to contemplate the mysteries of C's <code>fdopen()</code>. 741 </li> 742 <li>In library snapshot 3.0.95 and later, <code>filebuf</code>s bring 743 back an old extension: the <code>fd()</code> member function. The 744 integer returned from this function can be used for whatever file 745 descriptors can be used for on your platform. Naturally, the 746 library cannot track what you do on your own with a file descriptor, 747 so if you perform any I/O directly, don't expect the library to be 748 aware of it. 749 </li> 750 <li>Beginning with 3.1, the extra <code>filebuf</code> constructor and 751 the <code>fd()</code> function were removed from the standard 752 filebuf. Instead, <code><ext/stdio_filebuf.h></code> contains 753 a derived class called 754 <a href="http://gcc.gnu.org/onlinedocs/libstdc++/latest-doxygen/class____gnu__cxx_1_1stdio__filebuf.html"><code>__gnu_cxx::stdio_filebuf</code></a>. 755 This class can be constructed from a C <code>FILE*</code> or a file 756 descriptor, and provides the <code>fd()</code> function. 757 </li> 758 </ul> 759 <p>If you want to access a <code>filebuf</code>s file descriptor to 760 implement file locking (e.g. using the <code>fcntl()</code> system 761 call) then you might be interested in Henry Suter's 762 <a href="http://suter.home.cern.ch/suter/RWLock.html">RWLock</a> 763 class. 764 </p> 765 766<!-- ####################################################### --> 767 768<hr /> 769<p class="fineprint"><em> 770See <a href="../17_intro/license.html">license.html</a> for copying conditions. 771Comments and suggestions are welcome, and may be sent to 772<a href="mailto:libstdc++@gcc.gnu.org">the libstdc++ mailing list</a>. 773</em></p> 774 775 776</body> 777</html> 778 779 780