• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2                      "http://www.w3.org/TR/html4/strict.dtd">
3<html>
4<head>
5  <link rel="stylesheet" href="llvm.css" type="text/css">
6  <title>LLVM Coding Standards</title>
7</head>
8<body>
9
10<h1>
11  LLVM Coding Standards
12</h1>
13
14<ol>
15  <li><a href="#introduction">Introduction</a></li>
16  <li><a href="#mechanicalissues">Mechanical Source Issues</a>
17    <ol>
18      <li><a href="#sourceformating">Source Code Formatting</a>
19        <ol>
20          <li><a href="#scf_commenting">Commenting</a></li>
21          <li><a href="#scf_commentformat">Comment Formatting</a></li>
22          <li><a href="#scf_includes"><tt>#include</tt> Style</a></li>
23          <li><a href="#scf_codewidth">Source Code Width</a></li>
24          <li><a href="#scf_spacestabs">Use Spaces Instead of Tabs</a></li>
25          <li><a href="#scf_indentation">Indent Code Consistently</a></li>
26        </ol></li>
27      <li><a href="#compilerissues">Compiler Issues</a>
28        <ol>
29          <li><a href="#ci_warningerrors">Treat Compiler Warnings Like
30              Errors</a></li>
31          <li><a href="#ci_portable_code">Write Portable Code</a></li>
32          <li><a href="#ci_rtti_exceptions">Do not use RTTI or Exceptions</a></li>
33          <li><a href="#ci_class_struct">Use of <tt>class</tt>/<tt>struct</tt> Keywords</a></li>
34        </ol></li>
35    </ol></li>
36  <li><a href="#styleissues">Style Issues</a>
37    <ol>
38      <li><a href="#macro">The High-Level Issues</a>
39        <ol>
40          <li><a href="#hl_module">A Public Header File <b>is</b> a
41              Module</a></li>
42          <li><a href="#hl_dontinclude"><tt>#include</tt> as Little as Possible</a></li>
43          <li><a href="#hl_privateheaders">Keep "internal" Headers
44              Private</a></li>
45          <li><a href="#hl_earlyexit">Use Early Exits and <tt>continue</tt> to Simplify
46              Code</a></li>
47          <li><a href="#hl_else_after_return">Don't use <tt>else</tt> after a
48              <tt>return</tt></a></li>
49          <li><a href="#hl_predicateloops">Turn Predicate Loops into Predicate
50              Functions</a></li>
51        </ol></li>
52      <li><a href="#micro">The Low-Level Issues</a>
53        <ol>
54          <li><a href="#ll_naming">Name Types, Functions, Variables, and Enumerators Properly</a></li>
55          <li><a href="#ll_assert">Assert Liberally</a></li>
56          <li><a href="#ll_ns_std">Do not use '<tt>using namespace std</tt>'</a></li>
57          <li><a href="#ll_virtual_anch">Provide a virtual method anchor for
58              classes in headers</a></li>
59          <li><a href="#ll_end">Don't evaluate <tt>end()</tt> every time through a
60              loop</a></li>
61          <li><a href="#ll_iostream"><tt>#include &lt;iostream&gt;</tt> is
62              <em>forbidden</em></a></li>
63          <li><a href="#ll_raw_ostream">Use <tt>raw_ostream</tt></a></li>
64          <li><a href="#ll_avoidendl">Avoid <tt>std::endl</tt></a></li>
65        </ol></li>
66
67      <li><a href="#nano">Microscopic Details</a>
68        <ol>
69          <li><a href="#micro_spaceparen">Spaces Before Parentheses</a></li>
70          <li><a href="#micro_preincrement">Prefer Preincrement</a></li>
71          <li><a href="#micro_namespaceindent">Namespace Indentation</a></li>
72          <li><a href="#micro_anonns">Anonymous Namespaces</a></li>
73        </ol></li>
74
75
76    </ol></li>
77  <li><a href="#seealso">See Also</a></li>
78</ol>
79
80<div class="doc_author">
81  <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
82</div>
83
84
85<!-- *********************************************************************** -->
86<h2>
87  <a name="introduction">Introduction</a>
88</h2>
89<!-- *********************************************************************** -->
90
91<div>
92
93<p>This document attempts to describe a few coding standards that are being used
94in the LLVM source tree.  Although no coding standards should be regarded as
95absolute requirements to be followed in all instances, coding standards can be
96useful.</p>
97
98<p>This document intentionally does not prescribe fixed standards for religious
99issues such as brace placement and space usage.  For issues like this, follow
100the golden rule:</p>
101
102<blockquote>
103
104<p><b><a name="goldenrule">If you are adding a significant body of source to a
105project, feel free to use whatever style you are most comfortable with.  If you
106are extending, enhancing, or bug fixing already implemented code, use the style
107that is already being used so that the source is uniform and easy to
108follow.</a></b></p>
109
110</blockquote>
111
112<p>The ultimate goal of these guidelines is the increase readability and
113maintainability of our common source base. If you have suggestions for topics to
114be included, please mail them to <a
115href="mailto:sabre@nondot.org">Chris</a>.</p>
116
117</div>
118
119<!-- *********************************************************************** -->
120<h2>
121  <a name="mechanicalissues">Mechanical Source Issues</a>
122</h2>
123<!-- *********************************************************************** -->
124
125<div>
126
127<!-- ======================================================================= -->
128<h3>
129  <a name="sourceformating">Source Code Formatting</a>
130</h3>
131
132<div>
133
134<!-- _______________________________________________________________________ -->
135<h4>
136  <a name="scf_commenting">Commenting</a>
137</h4>
138
139<div>
140
141<p>Comments are one critical part of readability and maintainability.  Everyone
142knows they should comment, so should you.  When writing comments, write them as
143English prose, which means they should use proper capitalization, punctuation,
144etc.  Although we all should probably
145comment our code more than we do, there are a few very critical places that
146documentation is very useful:</p>
147
148<h5>File Headers</h5>
149
150<div>
151
152<p>Every source file should have a header on it that describes the basic
153purpose of the file.  If a file does not have a header, it should not be
154checked into Subversion.  Most source trees will probably have a standard
155file header format.  The standard format for the LLVM source tree looks like
156this:</p>
157
158<div class="doc_code">
159<pre>
160//===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
161//
162//                     The LLVM Compiler Infrastructure
163//
164// This file is distributed under the University of Illinois Open Source
165// License. See LICENSE.TXT for details.
166//
167//===----------------------------------------------------------------------===//
168//
169// This file contains the declaration of the Instruction class, which is the
170// base class for all of the VM instructions.
171//
172//===----------------------------------------------------------------------===//
173</pre>
174</div>
175
176<p>A few things to note about this particular format:  The "<tt>-*- C++
177-*-</tt>" string on the first line is there to tell Emacs that the source file
178is a C++ file, not a C file (Emacs assumes <tt>.h</tt> files are C files by default).
179Note that this tag is not necessary in <tt>.cpp</tt> files.  The name of the file is also
180on the first line, along with a very short description of the purpose of the
181file.  This is important when printing out code and flipping though lots of
182pages.</p>
183
184<p>The next section in the file is a concise note that defines the license
185that the file is released under.  This makes it perfectly clear what terms the
186source code can be distributed under and should not be modified in any way.</p>
187
188<p>The main body of the description does not have to be very long in most cases.
189Here it's only two lines.  If an algorithm is being implemented or something
190tricky is going on, a reference to the paper where it is published should be
191included, as well as any notes or "gotchas" in the code to watch out for.</p>
192
193</div>
194
195<h5>Class overviews</h5>
196
197<p>Classes are one fundamental part of a good object oriented design.  As such,
198a class definition should have a comment block that explains what the class is
199used for... if it's not obvious.  If it's so completely obvious your grandma
200could figure it out, it's probably safe to leave it out.  Naming classes
201something sane goes a long ways towards avoiding writing documentation.</p>
202
203
204<h5>Method information</h5>
205
206<div>
207
208<p>Methods defined in a class (as well as any global functions) should also be
209documented properly.  A quick note about what it does and a description of the
210borderline behaviour is all that is necessary here (unless something
211particularly tricky or insidious is going on).  The hope is that people can
212figure out how to use your interfaces without reading the code itself... that is
213the goal metric.</p>
214
215<p>Good things to talk about here are what happens when something unexpected
216happens: does the method return null?  Abort?  Format your hard disk?</p>
217
218</div>
219
220</div>
221
222<!-- _______________________________________________________________________ -->
223<h4>
224  <a name="scf_commentformat">Comment Formatting</a>
225</h4>
226
227<div>
228
229<p>In general, prefer C++ style (<tt>//</tt>) comments.  They take less space,
230require less typing, don't have nesting problems, etc.  There are a few cases
231when it is useful to use C style (<tt>/* */</tt>) comments however:</p>
232
233<ol>
234  <li>When writing C code: Obviously if you are writing C code, use C style
235      comments.</li>
236  <li>When writing a header file that may be <tt>#include</tt>d by a C source
237      file.</li>
238  <li>When writing a source file that is used by a tool that only accepts C
239      style comments.</li>
240</ol>
241
242<p>To comment out a large block of code, use <tt>#if 0</tt> and <tt>#endif</tt>.
243These nest properly and are better behaved in general than C style comments.</p>
244
245</div>
246
247<!-- _______________________________________________________________________ -->
248<h4>
249  <a name="scf_includes"><tt>#include</tt> Style</a>
250</h4>
251
252<div>
253
254<p>Immediately after the <a href="#scf_commenting">header file comment</a> (and
255include guards if working on a header file), the <a
256href="#hl_dontinclude">minimal</a> list of <tt>#include</tt>s required by the
257file should be listed.  We prefer these <tt>#include</tt>s to be listed in this
258order:</p>
259
260<ol>
261  <li><a href="#mmheader">Main Module Header</a></li>
262  <li><a href="#hl_privateheaders">Local/Private Headers</a></li>
263  <li><tt>llvm/*</tt></li>
264  <li><tt>llvm/Analysis/*</tt></li>
265  <li><tt>llvm/Assembly/*</tt></li>
266  <li><tt>llvm/Bitcode/*</tt></li>
267  <li><tt>llvm/CodeGen/*</tt></li>
268  <li>...</li>
269  <li><tt>Support/*</tt></li>
270  <li><tt>Config/*</tt></li>
271  <li>System <tt>#includes</tt></li>
272</ol>
273
274<p>and each category should be sorted by name.</p>
275
276<p><a name="mmheader">The "Main Module Header"</a> file applies to <tt>.cpp</tt> files
277which implement an interface defined by a <tt>.h</tt> file.  This <tt>#include</tt>
278should always be included <b>first</b> regardless of where it lives on the file
279system.  By including a header file first in the <tt>.cpp</tt> files that implement the
280interfaces, we ensure that the header does not have any hidden dependencies
281which are not explicitly #included in the header, but should be.  It is also a
282form of documentation in the <tt>.cpp</tt> file to indicate where the interfaces it
283implements are defined.</p>
284
285</div>
286
287<!-- _______________________________________________________________________ -->
288<h4>
289  <a name="scf_codewidth">Source Code Width</a>
290</h4>
291
292<div>
293
294<p>Write your code to fit within 80 columns of text.  This helps those of us who
295like to print out code and look at your code in an xterm without resizing
296it.</p>
297
298<p>The longer answer is that there must be some limit to the width of the code
299in order to reasonably allow developers to have multiple files side-by-side in
300windows on a modest display.  If you are going to pick a width limit, it is
301somewhat arbitrary but you might as well pick something standard.  Going with
30290 columns (for example) instead of 80 columns wouldn't add any significant
303value and would be detrimental to printing out code.  Also many other projects
304have standardized on 80 columns, so some people have already configured their
305editors for it (vs something else, like 90 columns).</p>
306
307<p>This is one of many contentious issues in coding standards, but it is not up
308for debate.</p>
309
310</div>
311
312<!-- _______________________________________________________________________ -->
313<h4>
314  <a name="scf_spacestabs">Use Spaces Instead of Tabs</a>
315</h4>
316
317<div>
318
319<p>In all cases, prefer spaces to tabs in source files.  People have different
320preferred indentation levels, and different styles of indentation that they
321like; this is fine.  What isn't fine is that different editors/viewers expand
322tabs out to different tab stops.  This can cause your code to look completely
323unreadable, and it is not worth dealing with.</p>
324
325<p>As always, follow the <a href="#goldenrule">Golden Rule</a> above: follow the
326style of existing code if you are modifying and extending it.  If you like four
327spaces of indentation, <b>DO NOT</b> do that in the middle of a chunk of code
328with two spaces of indentation.  Also, do not reindent a whole source file: it
329makes for incredible diffs that are absolutely worthless.</p>
330
331</div>
332
333<!-- _______________________________________________________________________ -->
334<h4>
335  <a name="scf_indentation">Indent Code Consistently</a>
336</h4>
337
338<div>
339
340<p>Okay, in your first year of programming you were told that indentation is
341important.  If you didn't believe and internalize this then, now is the time.
342Just do it.</p>
343
344</div>
345
346</div>
347
348<!-- ======================================================================= -->
349<h3>
350  <a name="compilerissues">Compiler Issues</a>
351</h3>
352
353<div>
354
355<!-- _______________________________________________________________________ -->
356<h4>
357  <a name="ci_warningerrors">Treat Compiler Warnings Like Errors</a>
358</h4>
359
360<div>
361
362<p>If your code has compiler warnings in it, something is wrong &mdash; you
363aren't casting values correctly, your have "questionable" constructs in your
364code, or you are doing something legitimately wrong.  Compiler warnings can
365cover up legitimate errors in output and make dealing with a translation unit
366difficult.</p>
367
368<p>It is not possible to prevent all warnings from all compilers, nor is it
369desirable.  Instead, pick a standard compiler (like <tt>gcc</tt>) that provides
370a good thorough set of warnings, and stick to it.  At least in the case of
371<tt>gcc</tt>, it is possible to work around any spurious errors by changing the
372syntax of the code slightly.  For example, a warning that annoys me occurs when
373I write code like this:</p>
374
375<div class="doc_code">
376<pre>
377if (V = getValue()) {
378  ...
379}
380</pre>
381</div>
382
383<p><tt>gcc</tt> will warn me that I probably want to use the <tt>==</tt>
384operator, and that I probably mistyped it.  In most cases, I haven't, and I
385really don't want the spurious errors.  To fix this particular problem, I
386rewrite the code like this:</p>
387
388<div class="doc_code">
389<pre>
390if ((V = getValue())) {
391  ...
392}
393</pre>
394</div>
395
396<p>which shuts <tt>gcc</tt> up.  Any <tt>gcc</tt> warning that annoys you can
397be fixed by massaging the code appropriately.</p>
398
399<p>These are the <tt>gcc</tt> warnings that I prefer to enable:</p>
400
401<div class="doc_code">
402<pre>
403-Wall -Winline -W -Wwrite-strings -Wno-unused
404</pre>
405</div>
406
407</div>
408
409<!-- _______________________________________________________________________ -->
410<h4>
411  <a name="ci_portable_code">Write Portable Code</a>
412</h4>
413
414<div>
415
416<p>In almost all cases, it is possible and within reason to write completely
417portable code.  If there are cases where it isn't possible to write portable
418code, isolate it behind a well defined (and well documented) interface.</p>
419
420<p>In practice, this means that you shouldn't assume much about the host
421compiler, and Visual Studio tends to be the lowest common denominator.
422If advanced features are used, they should only be an implementation detail of
423a library which has a simple exposed API, and preferably be buried in
424libSystem.</p>
425
426</div>
427
428<!-- _______________________________________________________________________ -->
429<h4>
430<a name="ci_rtti_exceptions">Do not use RTTI or Exceptions</a>
431</h4>
432<div>
433
434<p>In an effort to reduce code and executable size, LLVM does not use RTTI
435(e.g. <tt>dynamic_cast&lt;&gt;</tt>) or exceptions.  These two language features
436violate the general C++ principle of <i>"you only pay for what you use"</i>,
437causing executable bloat even if exceptions are never used in the code base, or
438if RTTI is never used for a class.  Because of this, we turn them off globally
439in the code.</p>
440
441<p>That said, LLVM does make extensive use of a hand-rolled form of RTTI that
442use templates like <a href="ProgrammersManual.html#isa"><tt>isa&lt;&gt;</tt>,
443<tt>cast&lt;&gt;</tt>, and <tt>dyn_cast&lt;&gt;</tt></a>.  This form of RTTI is
444opt-in and can be added to any class.  It is also substantially more efficient
445than <tt>dynamic_cast&lt;&gt;</tt>.</p>
446
447</div>
448
449<!-- _______________________________________________________________________ -->
450<h4>
451<a name="ci_class_struct">Use of <tt>class</tt> and <tt>struct</tt> Keywords</a>
452</h4>
453<div>
454
455<p>In C++, the <tt>class</tt> and <tt>struct</tt> keywords can be used almost
456interchangeably. The only difference is when they are used to declare a class:
457<tt>class</tt> makes all members private by default while <tt>struct</tt> makes
458all members public by default.</p>
459
460<p>Unfortunately, not all compilers follow the rules and some will generate
461different symbols based on whether <tt>class</tt> or <tt>struct</tt> was used to
462declare the symbol.  This can lead to problems at link time.</p>
463
464<p>So, the rule for LLVM is to always use the <tt>class</tt> keyword, unless
465<b>all</b> members are public and the type is a C++
466<a href="http://en.wikipedia.org/wiki/Plain_old_data_structure">POD</a> type, in
467which case <tt>struct</tt> is allowed.</p>
468
469</div>
470
471</div>
472
473</div>
474
475<!-- *********************************************************************** -->
476<h2>
477  <a name="styleissues">Style Issues</a>
478</h2>
479<!-- *********************************************************************** -->
480
481<div>
482
483<!-- ======================================================================= -->
484<h3>
485  <a name="macro">The High-Level Issues</a>
486</h3>
487<!-- ======================================================================= -->
488
489<div>
490
491<!-- _______________________________________________________________________ -->
492<h4>
493  <a name="hl_module">A Public Header File <b>is</b> a Module</a>
494</h4>
495
496<div>
497
498<p>C++ doesn't do too well in the modularity department.  There is no real
499encapsulation or data hiding (unless you use expensive protocol classes), but it
500is what we have to work with.  When you write a public header file (in the LLVM
501source tree, they live in the top level "<tt>include</tt>" directory), you are
502defining a module of functionality.</p>
503
504<p>Ideally, modules should be completely independent of each other, and their
505header files should only <tt>#include</tt> the absolute minimum number of
506headers possible. A module is not just a class, a function, or a
507namespace: <a href="http://www.cuj.com/articles/2000/0002/0002c/0002c.htm">it's
508a collection of these</a> that defines an interface.  This interface may be
509several functions, classes, or data structures, but the important issue is how
510they work together.</p>
511
512<p>In general, a module should be implemented by one or more <tt>.cpp</tt>
513files.  Each of these <tt>.cpp</tt> files should include the header that defines
514their interface first.  This ensures that all of the dependences of the module
515header have been properly added to the module header itself, and are not
516implicit.  System headers should be included after user headers for a
517translation unit.</p>
518
519</div>
520
521<!-- _______________________________________________________________________ -->
522<h4>
523  <a name="hl_dontinclude"><tt>#include</tt> as Little as Possible</a>
524</h4>
525
526<div>
527
528<p><tt>#include</tt> hurts compile time performance.  Don't do it unless you
529have to, especially in header files.</p>
530
531<p>But wait! Sometimes you need to have the definition of a class to use it, or
532to inherit from it.  In these cases go ahead and <tt>#include</tt> that header
533file.  Be aware however that there are many cases where you don't need to have
534the full definition of a class.  If you are using a pointer or reference to a
535class, you don't need the header file.  If you are simply returning a class
536instance from a prototyped function or method, you don't need it.  In fact, for
537most cases, you simply don't need the definition of a class. And not
538<tt>#include</tt>'ing speeds up compilation.</p>
539
540<p>It is easy to try to go too overboard on this recommendation, however.  You
541<b>must</b> include all of the header files that you are using &mdash; you can
542include them either directly or indirectly (through another header file).  To
543make sure that you don't accidentally forget to include a header file in your
544module header, make sure to include your module header <b>first</b> in the
545implementation file (as mentioned above).  This way there won't be any hidden
546dependencies that you'll find out about later.</p>
547
548</div>
549
550<!-- _______________________________________________________________________ -->
551<h4>
552  <a name="hl_privateheaders">Keep "Internal" Headers Private</a>
553</h4>
554
555<div>
556
557<p>Many modules have a complex implementation that causes them to use more than
558one implementation (<tt>.cpp</tt>) file.  It is often tempting to put the
559internal communication interface (helper classes, extra functions, etc) in the
560public module header file.  Don't do this!</p>
561
562<p>If you really need to do something like this, put a private header file in
563the same directory as the source files, and include it locally.  This ensures
564that your private interface remains private and undisturbed by outsiders.</p>
565
566<p>Note however, that it's okay to put extra implementation methods in a public
567class itself. Just make them private (or protected) and all is well.</p>
568
569</div>
570
571<!-- _______________________________________________________________________ -->
572<h4>
573  <a name="hl_earlyexit">Use Early Exits and <tt>continue</tt> to Simplify Code</a>
574</h4>
575
576<div>
577
578<p>When reading code, keep in mind how much state and how many previous
579decisions have to be remembered by the reader to understand a block of code.
580Aim to reduce indentation where possible when it doesn't make it more difficult
581to understand the code.  One great way to do this is by making use of early
582exits and the <tt>continue</tt> keyword in long loops.  As an example of using
583an early exit from a function, consider this "bad" code:</p>
584
585<div class="doc_code">
586<pre>
587Value *DoSomething(Instruction *I) {
588  if (!isa&lt;TerminatorInst&gt;(I) &amp;&amp;
589      I-&gt;hasOneUse() &amp;&amp; SomeOtherThing(I)) {
590    ... some long code ....
591  }
592
593  return 0;
594}
595</pre>
596</div>
597
598<p>This code has several problems if the body of the '<tt>if</tt>' is large.
599When you're looking at the top of the function, it isn't immediately clear that
600this <em>only</em> does interesting things with non-terminator instructions, and
601only applies to things with the other predicates.  Second, it is relatively
602difficult to describe (in comments) why these predicates are important because
603the <tt>if</tt> statement makes it difficult to lay out the comments.  Third,
604when you're deep within the body of the code, it is indented an extra level.
605Finally, when reading the top of the function, it isn't clear what the result is
606if the predicate isn't true; you have to read to the end of the function to know
607that it returns null.</p>
608
609<p>It is much preferred to format the code like this:</p>
610
611<div class="doc_code">
612<pre>
613Value *DoSomething(Instruction *I) {
614  // Terminators never need 'something' done to them because ...
615  if (isa&lt;TerminatorInst&gt;(I))
616    return 0;
617
618  // We conservatively avoid transforming instructions with multiple uses
619  // because goats like cheese.
620  if (!I-&gt;hasOneUse())
621    return 0;
622
623  // This is really just here for example.
624  if (!SomeOtherThing(I))
625    return 0;
626
627  ... some long code ....
628}
629</pre>
630</div>
631
632<p>This fixes these problems.  A similar problem frequently happens in <tt>for</tt>
633loops.  A silly example is something like this:</p>
634
635<div class="doc_code">
636<pre>
637  for (BasicBlock::iterator II = BB-&gt;begin(), E = BB-&gt;end(); II != E; ++II) {
638    if (BinaryOperator *BO = dyn_cast&lt;BinaryOperator&gt;(II)) {
639      Value *LHS = BO-&gt;getOperand(0);
640      Value *RHS = BO-&gt;getOperand(1);
641      if (LHS != RHS) {
642        ...
643      }
644    }
645  }
646</pre>
647</div>
648
649<p>When you have very, very small loops, this sort of structure is fine. But if
650it exceeds more than 10-15 lines, it becomes difficult for people to read and
651understand at a glance. The problem with this sort of code is that it gets very
652nested very quickly. Meaning that the reader of the code has to keep a lot of
653context in their brain to remember what is going immediately on in the loop,
654because they don't know if/when the <tt>if</tt> conditions will have elses etc.
655It is strongly preferred to structure the loop like this:</p>
656
657<div class="doc_code">
658<pre>
659  for (BasicBlock::iterator II = BB-&gt;begin(), E = BB-&gt;end(); II != E; ++II) {
660    BinaryOperator *BO = dyn_cast&lt;BinaryOperator&gt;(II);
661    if (!BO) continue;
662
663    Value *LHS = BO-&gt;getOperand(0);
664    Value *RHS = BO-&gt;getOperand(1);
665    if (LHS == RHS) continue;
666
667    ...
668  }
669</pre>
670</div>
671
672<p>This has all the benefits of using early exits for functions: it reduces
673nesting of the loop, it makes it easier to describe why the conditions are true,
674and it makes it obvious to the reader that there is no <tt>else</tt> coming up
675that they have to push context into their brain for.  If a loop is large, this
676can be a big understandability win.</p>
677
678</div>
679
680<!-- _______________________________________________________________________ -->
681<h4>
682  <a name="hl_else_after_return">Don't use <tt>else</tt> after a <tt>return</tt></a>
683</h4>
684
685<div>
686
687<p>For similar reasons above (reduction of indentation and easier reading),
688please do not use '<tt>else</tt>' or '<tt>else if</tt>' after something that
689interrupts control flow &mdash; like <tt>return</tt>, <tt>break</tt>,
690<tt>continue</tt>, <tt>goto</tt>, etc. For example, this is <em>bad</em>:</p>
691
692<div class="doc_code">
693<pre>
694  case 'J': {
695    if (Signed) {
696      Type = Context.getsigjmp_bufType();
697      if (Type.isNull()) {
698        Error = ASTContext::GE_Missing_sigjmp_buf;
699        return QualType();
700      <b>} else {
701        break;
702      }</b>
703    } else {
704      Type = Context.getjmp_bufType();
705      if (Type.isNull()) {
706        Error = ASTContext::GE_Missing_jmp_buf;
707        return QualType();
708      <b>} else {
709        break;
710      }</b>
711    }
712  }
713  }
714</pre>
715</div>
716
717<p>It is better to write it like this:</p>
718
719<div class="doc_code">
720<pre>
721  case 'J':
722    if (Signed) {
723      Type = Context.getsigjmp_bufType();
724      if (Type.isNull()) {
725        Error = ASTContext::GE_Missing_sigjmp_buf;
726        return QualType();
727      }
728    } else {
729      Type = Context.getjmp_bufType();
730      if (Type.isNull()) {
731        Error = ASTContext::GE_Missing_jmp_buf;
732        return QualType();
733      }
734    }
735    <b>break;</b>
736</pre>
737</div>
738
739<p>Or better yet (in this case) as:</p>
740
741<div class="doc_code">
742<pre>
743  case 'J':
744    if (Signed)
745      Type = Context.getsigjmp_bufType();
746    else
747      Type = Context.getjmp_bufType();
748
749    if (Type.isNull()) {
750      Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
751                       ASTContext::GE_Missing_jmp_buf;
752      return QualType();
753    }
754    <b>break;</b>
755</pre>
756</div>
757
758<p>The idea is to reduce indentation and the amount of code you have to keep
759track of when reading the code.</p>
760
761</div>
762
763<!-- _______________________________________________________________________ -->
764<h4>
765  <a name="hl_predicateloops">Turn Predicate Loops into Predicate Functions</a>
766</h4>
767
768<div>
769
770<p>It is very common to write small loops that just compute a boolean value.
771There are a number of ways that people commonly write these, but an example of
772this sort of thing is:</p>
773
774<div class="doc_code">
775<pre>
776  <b>bool FoundFoo = false;</b>
777  for (unsigned i = 0, e = BarList.size(); i != e; ++i)
778    if (BarList[i]-&gt;isFoo()) {
779      <b>FoundFoo = true;</b>
780      break;
781    }
782
783  <b>if (FoundFoo) {</b>
784    ...
785  }
786</pre>
787</div>
788
789<p>This sort of code is awkward to write, and is almost always a bad sign.
790Instead of this sort of loop, we strongly prefer to use a predicate function
791(which may be <a href="#micro_anonns">static</a>) that uses
792<a href="#hl_earlyexit">early exits</a> to compute the predicate.  We prefer
793the code to be structured like this:</p>
794
795<div class="doc_code">
796<pre>
797/// ListContainsFoo - Return true if the specified list has an element that is
798/// a foo.
799static bool ListContainsFoo(const std::vector&lt;Bar*&gt; &amp;List) {
800  for (unsigned i = 0, e = List.size(); i != e; ++i)
801    if (List[i]-&gt;isFoo())
802      return true;
803  return false;
804}
805...
806
807  <b>if (ListContainsFoo(BarList)) {</b>
808    ...
809  }
810</pre>
811</div>
812
813<p>There are many reasons for doing this: it reduces indentation and factors out
814code which can often be shared by other code that checks for the same predicate.
815More importantly, it <em>forces you to pick a name</em> for the function, and
816forces you to write a comment for it.  In this silly example, this doesn't add
817much value.  However, if the condition is complex, this can make it a lot easier
818for the reader to understand the code that queries for this predicate.  Instead
819of being faced with the in-line details of how we check to see if the BarList
820contains a foo, we can trust the function name and continue reading with better
821locality.</p>
822
823</div>
824
825</div>
826
827<!-- ======================================================================= -->
828<h3>
829  <a name="micro">The Low-Level Issues</a>
830</h3>
831<!-- ======================================================================= -->
832
833<div>
834
835<!-- _______________________________________________________________________ -->
836<h4>
837  <a name="ll_naming">
838    Name Types, Functions, Variables, and Enumerators Properly
839  </a>
840</h4>
841
842<div>
843
844<p>Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
845enough how important it is to use <em>descriptive</em> names.  Pick names that
846match the semantics and role of the underlying entities, within reason.  Avoid
847abbreviations unless they are well known.  After picking a good name, make sure
848to use consistent capitalization for the name, as inconsistency requires clients
849to either memorize the APIs or to look it up to find the exact spelling.</p>
850
851<p>In general, names should be in camel case (e.g. <tt>TextFileReader</tt>
852and <tt>isLValue()</tt>).  Different kinds of declarations have different
853rules:</p>
854
855<ul>
856<li><p><b>Type names</b> (including classes, structs, enums, typedefs, etc)
857    should be nouns and start with an upper-case letter (e.g.
858    <tt>TextFileReader</tt>).</p></li>
859
860<li><p><b>Variable names</b> should be nouns (as they represent state).  The
861    name should be camel case, and start with an upper case letter (e.g.
862    <tt>Leader</tt> or <tt>Boats</tt>).</p></li>
863
864<li><p><b>Function names</b> should be verb phrases (as they represent
865    actions), and command-like function should be imperative.  The name should
866    be camel case, and start with a lower case letter (e.g. <tt>openFile()</tt>
867    or <tt>isFoo()</tt>).</p></li>
868
869<li><p><b>Enum declarations</b> (e.g. <tt>enum Foo {...}</tt>) are types, so
870    they should follow the naming conventions for types.  A common use for enums
871    is as a discriminator for a union, or an indicator of a subclass.  When an
872    enum is used for something like this, it should have a <tt>Kind</tt> suffix
873    (e.g. <tt>ValueKind</tt>).</p></li>
874
875<li><p><b>Enumerators</b> (e.g. <tt>enum { Foo, Bar }</tt>) and <b>public member
876    variables</b> should start with an upper-case letter, just like types.
877    Unless the enumerators are defined in their own small namespace or inside a
878    class, enumerators should have a prefix corresponding to the enum
879    declaration name.  For example, <tt>enum ValueKind { ... };</tt> may contain
880    enumerators like <tt>VK_Argument</tt>, <tt>VK_BasicBlock</tt>, etc.
881    Enumerators that are just convenience constants are exempt from the
882    requirement for a prefix.  For instance:</p>
883
884<div class="doc_code">
885<pre>
886enum {
887  MaxSize = 42,
888  Density = 12
889};
890</pre>
891</div>
892</li>
893
894</ul>
895
896<p>As an exception, classes that mimic STL classes can have member names in
897STL's style of lower-case words separated by underscores (e.g. <tt>begin()</tt>,
898<tt>push_back()</tt>, and <tt>empty()</tt>).</p>
899
900<p>Here are some examples of good and bad names:</p>
901
902<div class="doc_code">
903<pre>
904class VehicleMaker {
905  ...
906  Factory&lt;Tire&gt; F;            // Bad -- abbreviation and non-descriptive.
907  Factory&lt;Tire&gt; Factory;      // Better.
908  Factory&lt;Tire&gt; TireFactory;  // Even better -- if VehicleMaker has more than one
909                              // kind of factories.
910};
911
912Vehicle MakeVehicle(VehicleType Type) {
913  VehicleMaker M;                         // Might be OK if having a short life-span.
914  Tire tmp1 = M.makeTire();               // Bad -- 'tmp1' provides no information.
915  Light headlight = M.makeLight("head");  // Good -- descriptive.
916  ...
917}
918</pre>
919</div>
920
921</div>
922
923
924<!-- _______________________________________________________________________ -->
925<h4>
926  <a name="ll_assert">Assert Liberally</a>
927</h4>
928
929<div>
930
931<p>Use the "<tt>assert</tt>" macro to its fullest.  Check all of your
932preconditions and assumptions, you never know when a bug (not necessarily even
933yours) might be caught early by an assertion, which reduces debugging time
934dramatically.  The "<tt>&lt;cassert&gt;</tt>" header file is probably already
935included by the header files you are using, so it doesn't cost anything to use
936it.</p>
937
938<p>To further assist with debugging, make sure to put some kind of error message
939in the assertion statement, which is printed if the assertion is tripped. This
940helps the poor debugger make sense of why an assertion is being made and
941enforced, and hopefully what to do about it.  Here is one complete example:</p>
942
943<div class="doc_code">
944<pre>
945inline Value *getOperand(unsigned i) {
946  assert(i &lt; Operands.size() &amp;&amp; "getOperand() out of range!");
947  return Operands[i];
948}
949</pre>
950</div>
951
952<p>Here are more examples:</p>
953
954<div class="doc_code">
955<pre>
956assert(Ty-&gt;isPointerType() &amp;&amp; "Can't allocate a non pointer type!");
957
958assert((Opcode == Shl || Opcode == Shr) &amp;&amp; "ShiftInst Opcode invalid!");
959
960assert(idx &lt; getNumSuccessors() &amp;&amp; "Successor # out of range!");
961
962assert(V1.getType() == V2.getType() &amp;&amp; "Constant types must be identical!");
963
964assert(isa&lt;PHINode&gt;(Succ-&gt;front()) &amp;&amp; "Only works on PHId BBs!");
965</pre>
966</div>
967
968<p>You get the idea.</p>
969
970<p>Please be aware that, when adding assert statements, not all compilers are aware of
971the semantics of the assert.  In some places, asserts are used to indicate a piece of
972code that should not be reached.  These are typically of the form:</p>
973
974<div class="doc_code">
975<pre>
976assert(0 &amp;&amp; "Some helpful error message");
977</pre>
978</div>
979
980<p>When used in a function that returns a value, they should be followed with a return
981statement and a comment indicating that this line is never reached.  This will prevent
982a compiler which is unable to deduce that the assert statement never returns from
983generating a warning.</p>
984
985<div class="doc_code">
986<pre>
987assert(0 &amp;&amp; "Some helpful error message");
988// Not reached
989return 0;
990</pre>
991</div>
992
993<p>Another issue is that values used only by assertions will produce an "unused
994value" warning when assertions are disabled.  For example, this code will
995warn:</p>
996
997<div class="doc_code">
998<pre>
999unsigned Size = V.size();
1000assert(Size &gt; 42 &amp;&amp; "Vector smaller than it should be");
1001
1002bool NewToSet = Myset.insert(Value);
1003assert(NewToSet &amp;&amp; "The value shouldn't be in the set yet");
1004</pre>
1005</div>
1006
1007<p>These are two interesting different cases. In the first case, the call to
1008V.size() is only useful for the assert, and we don't want it executed when
1009assertions are disabled.  Code like this should move the call into the assert
1010itself.  In the second case, the side effects of the call must happen whether
1011the assert is enabled or not.  In this case, the value should be cast to void to
1012disable the warning.  To be specific, it is preferred to write the code like
1013this:</p>
1014
1015<div class="doc_code">
1016<pre>
1017assert(V.size() &gt; 42 &amp;&amp; "Vector smaller than it should be");
1018
1019bool NewToSet = Myset.insert(Value); (void)NewToSet;
1020assert(NewToSet &amp;&amp; "The value shouldn't be in the set yet");
1021</pre>
1022</div>
1023
1024
1025</div>
1026
1027<!-- _______________________________________________________________________ -->
1028<h4>
1029  <a name="ll_ns_std">Do Not Use '<tt>using namespace std</tt>'</a>
1030</h4>
1031
1032<div>
1033
1034<p>In LLVM, we prefer to explicitly prefix all identifiers from the standard
1035namespace with an "<tt>std::</tt>" prefix, rather than rely on
1036"<tt>using namespace std;</tt>".</p>
1037
1038<p> In header files, adding a '<tt>using namespace XXX</tt>' directive pollutes
1039the namespace of any source file that <tt>#include</tt>s the header.  This is
1040clearly a bad thing.</p>
1041
1042<p>In implementation files (e.g. <tt>.cpp</tt> files), the rule is more of a stylistic
1043rule, but is still important.  Basically, using explicit namespace prefixes
1044makes the code <b>clearer</b>, because it is immediately obvious what facilities
1045are being used and where they are coming from. And <b>more portable</b>, because
1046namespace clashes cannot occur between LLVM code and other namespaces.  The
1047portability rule is important because different standard library implementations
1048expose different symbols (potentially ones they shouldn't), and future revisions
1049to the C++ standard will add more symbols to the <tt>std</tt> namespace.  As
1050such, we never use '<tt>using namespace std;</tt>' in LLVM.</p>
1051
1052<p>The exception to the general rule (i.e. it's not an exception for
1053the <tt>std</tt> namespace) is for implementation files.  For example, all of
1054the code in the LLVM project implements code that lives in the 'llvm' namespace.
1055As such, it is ok, and actually clearer, for the <tt>.cpp</tt> files to have a
1056'<tt>using namespace llvm;</tt>' directive at the top, after the
1057<tt>#include</tt>s.  This reduces indentation in the body of the file for source
1058editors that indent based on braces, and keeps the conceptual context cleaner.
1059The general form of this rule is that any <tt>.cpp</tt> file that implements
1060code in any namespace may use that namespace (and its parents'), but should not
1061use any others.</p>
1062
1063</div>
1064
1065<!-- _______________________________________________________________________ -->
1066<h4>
1067  <a name="ll_virtual_anch">
1068    Provide a Virtual Method Anchor for Classes in Headers
1069  </a>
1070</h4>
1071
1072<div>
1073
1074<p>If a class is defined in a header file and has a v-table (either it has
1075virtual methods or it derives from classes with virtual methods), it must
1076always have at least one out-of-line virtual method in the class.  Without
1077this, the compiler will copy the vtable and RTTI into every <tt>.o</tt> file
1078that <tt>#include</tt>s the header, bloating <tt>.o</tt> file sizes and
1079increasing link times.</p>
1080
1081</div>
1082
1083<!-- _______________________________________________________________________ -->
1084<h4>
1085  <a name="ll_end">Don't evaluate <tt>end()</tt> every time through a loop</a>
1086</h4>
1087
1088<div>
1089
1090<p>Because C++ doesn't have a standard "<tt>foreach</tt>" loop (though it can be
1091emulated with macros and may be coming in C++'0x) we end up writing a lot of
1092loops that manually iterate from begin to end on a variety of containers or
1093through other data structures.  One common mistake is to write a loop in this
1094style:</p>
1095
1096<div class="doc_code">
1097<pre>
1098  BasicBlock *BB = ...
1099  for (BasicBlock::iterator I = BB->begin(); I != <b>BB->end()</b>; ++I)
1100     ... use I ...
1101</pre>
1102</div>
1103
1104<p>The problem with this construct is that it evaluates "<tt>BB->end()</tt>"
1105every time through the loop.  Instead of writing the loop like this, we strongly
1106prefer loops to be written so that they evaluate it once before the loop starts.
1107A convenient way to do this is like so:</p>
1108
1109<div class="doc_code">
1110<pre>
1111  BasicBlock *BB = ...
1112  for (BasicBlock::iterator I = BB->begin(), E = <b>BB->end()</b>; I != E; ++I)
1113     ... use I ...
1114</pre>
1115</div>
1116
1117<p>The observant may quickly point out that these two loops may have different
1118semantics: if the container (a basic block in this case) is being mutated, then
1119"<tt>BB->end()</tt>" may change its value every time through the loop and the
1120second loop may not in fact be correct.  If you actually do depend on this
1121behavior, please write the loop in the first form and add a comment indicating
1122that you did it intentionally.</p>
1123
1124<p>Why do we prefer the second form (when correct)?  Writing the loop in the
1125first form has two problems. First it may be less efficient than evaluating it
1126at the start of the loop.  In this case, the cost is probably minor &mdash; a
1127few extra loads every time through the loop.  However, if the base expression is
1128more complex, then the cost can rise quickly.  I've seen loops where the end
1129expression was actually something like: "<tt>SomeMap[x]->end()</tt>" and map
1130lookups really aren't cheap.  By writing it in the second form consistently, you
1131eliminate the issue entirely and don't even have to think about it.</p>
1132
1133<p>The second (even bigger) issue is that writing the loop in the first form
1134hints to the reader that the loop is mutating the container (a fact that a
1135comment would handily confirm!).  If you write the loop in the second form, it
1136is immediately obvious without even looking at the body of the loop that the
1137container isn't being modified, which makes it easier to read the code and
1138understand what it does.</p>
1139
1140<p>While the second form of the loop is a few extra keystrokes, we do strongly
1141prefer it.</p>
1142
1143</div>
1144
1145<!-- _______________________________________________________________________ -->
1146<h4>
1147  <a name="ll_iostream"><tt>#include &lt;iostream&gt;</tt> is Forbidden</a>
1148</h4>
1149
1150<div>
1151
1152<p>The use of <tt>#include &lt;iostream&gt;</tt> in library files is
1153hereby <b><em>forbidden</em></b>. The primary reason for doing this is to
1154support clients using LLVM libraries as part of larger systems. In particular,
1155we statically link LLVM into some dynamic libraries. Even if LLVM isn't used,
1156the static constructors are run whenever an application starts up that uses the
1157dynamic library. There are two problems with this:</p>
1158
1159<ol>
1160  <li>The time to run the static c'tors impacts startup time of applications
1161      &mdash; a critical time for GUI apps.</li>
1162
1163  <li>The static c'tors cause the app to pull many extra pages of memory off the
1164      disk: both the code for the static c'tors in each <tt>.o</tt> file and the
1165      small amount of data that gets touched. In addition, touched/dirty pages
1166      put more pressure on the VM system on low-memory machines.</li>
1167</ol>
1168
1169<p>Note that using the other stream headers (<tt>&lt;sstream&gt;</tt> for
1170example) is not problematic in this regard &mdash;
1171just <tt>&lt;iostream&gt;</tt>. However, <tt>raw_ostream</tt> provides various
1172APIs that are better performing for almost every use than <tt>std::ostream</tt>
1173style APIs. <b>Therefore new code should always
1174use <a href="#ll_raw_ostream"><tt>raw_ostream</tt></a> for writing, or
1175the <tt>llvm::MemoryBuffer</tt> API for reading files.</b></p>
1176
1177</div>
1178
1179
1180<!-- _______________________________________________________________________ -->
1181<h4>
1182  <a name="ll_raw_ostream">Use <tt>raw_ostream</tt></a>
1183</h4>
1184
1185<div>
1186
1187<p>LLVM includes a lightweight, simple, and efficient stream implementation
1188in <tt>llvm/Support/raw_ostream.h</tt>, which provides all of the common
1189features of <tt>std::ostream</tt>.  All new code should use <tt>raw_ostream</tt>
1190instead of <tt>ostream</tt>.</p>
1191
1192<p>Unlike <tt>std::ostream</tt>, <tt>raw_ostream</tt> is not a template and can
1193be forward declared as <tt>class raw_ostream</tt>.  Public headers should
1194generally not include the <tt>raw_ostream</tt> header, but use forward
1195declarations and constant references to <tt>raw_ostream</tt> instances.</p>
1196
1197</div>
1198
1199
1200<!-- _______________________________________________________________________ -->
1201<h4>
1202  <a name="ll_avoidendl">Avoid <tt>std::endl</tt></a>
1203</h4>
1204
1205<div>
1206
1207<p>The <tt>std::endl</tt> modifier, when used with <tt>iostreams</tt> outputs a
1208newline to the output stream specified.  In addition to doing this, however, it
1209also flushes the output stream.  In other words, these are equivalent:</p>
1210
1211<div class="doc_code">
1212<pre>
1213std::cout &lt;&lt; std::endl;
1214std::cout &lt;&lt; '\n' &lt;&lt; std::flush;
1215</pre>
1216</div>
1217
1218<p>Most of the time, you probably have no reason to flush the output stream, so
1219it's better to use a literal <tt>'\n'</tt>.</p>
1220
1221</div>
1222
1223</div>
1224
1225<!-- ======================================================================= -->
1226<h3>
1227  <a name="nano">Microscopic Details</a>
1228</h3>
1229<!-- ======================================================================= -->
1230
1231<div>
1232
1233<p>This section describes preferred low-level formatting guidelines along with
1234reasoning on why we prefer them.</p>
1235
1236<!-- _______________________________________________________________________ -->
1237<h4>
1238  <a name="micro_spaceparen">Spaces Before Parentheses</a>
1239</h4>
1240
1241<div>
1242
1243<p>We prefer to put a space before an open parenthesis only in control flow
1244statements, but not in normal function call expressions and function-like
1245macros.  For example, this is good:</p>
1246
1247<div class="doc_code">
1248<pre>
1249<b>if (</b>x) ...
1250<b>for (</b>i = 0; i != 100; ++i) ...
1251<b>while (</b>llvm_rocks) ...
1252
1253<b>somefunc(</b>42);
1254<b><a href="#ll_assert">assert</a>(</b>3 != 4 &amp;&amp; "laws of math are failing me");
1255
1256a = <b>foo(</b>42, 92) + <b>bar(</b>x);
1257</pre>
1258</div>
1259
1260<p>and this is bad:</p>
1261
1262<div class="doc_code">
1263<pre>
1264<b>if(</b>x) ...
1265<b>for(</b>i = 0; i != 100; ++i) ...
1266<b>while(</b>llvm_rocks) ...
1267
1268<b>somefunc (</b>42);
1269<b><a href="#ll_assert">assert</a> (</b>3 != 4 &amp;&amp; "laws of math are failing me");
1270
1271a = <b>foo (</b>42, 92) + <b>bar (</b>x);
1272</pre>
1273</div>
1274
1275<p>The reason for doing this is not completely arbitrary.  This style makes
1276control flow operators stand out more, and makes expressions flow better. The
1277function call operator binds very tightly as a postfix operator.  Putting a
1278space after a function name (as in the last example) makes it appear that the
1279code might bind the arguments of the left-hand-side of a binary operator with
1280the argument list of a function and the name of the right side.  More
1281specifically, it is easy to misread the "a" example as:</p>
1282
1283<div class="doc_code">
1284<pre>
1285a = foo <b>(</b>(42, 92) + bar<b>)</b> (x);
1286</pre>
1287</div>
1288
1289<p>when skimming through the code.  By avoiding a space in a function, we avoid
1290this misinterpretation.</p>
1291
1292</div>
1293
1294<!-- _______________________________________________________________________ -->
1295<h4>
1296  <a name="micro_preincrement">Prefer Preincrement</a>
1297</h4>
1298
1299<div>
1300
1301<p>Hard fast rule: Preincrement (<tt>++X</tt>) may be no slower than
1302postincrement (<tt>X++</tt>) and could very well be a lot faster than it.  Use
1303preincrementation whenever possible.</p>
1304
1305<p>The semantics of postincrement include making a copy of the value being
1306incremented, returning it, and then preincrementing the "work value".  For
1307primitive types, this isn't a big deal... but for iterators, it can be a huge
1308issue (for example, some iterators contains stack and set objects in them...
1309copying an iterator could invoke the copy ctor's of these as well).  In general,
1310get in the habit of always using preincrement, and you won't have a problem.</p>
1311
1312</div>
1313
1314<!-- _______________________________________________________________________ -->
1315<h4>
1316  <a name="micro_namespaceindent">Namespace Indentation</a>
1317</h4>
1318
1319<div>
1320
1321<p>
1322In general, we strive to reduce indentation wherever possible.  This is useful
1323because we want code to <a href="#scf_codewidth">fit into 80 columns</a> without
1324wrapping horribly, but also because it makes it easier to understand the code.
1325Namespaces are a funny thing: they are often large, and we often desire to put
1326lots of stuff into them (so they can be large).  Other times they are tiny,
1327because they just hold an enum or something similar.  In order to balance this,
1328we use different approaches for small versus large namespaces.
1329</p>
1330
1331<p>
1332If a namespace definition is small and <em>easily</em> fits on a screen (say,
1333less than 35 lines of code), then you should indent its body.  Here's an
1334example:
1335</p>
1336
1337<div class="doc_code">
1338<pre>
1339namespace llvm {
1340  namespace X86 {
1341    /// RelocationType - An enum for the x86 relocation codes. Note that
1342    /// the terminology here doesn't follow x86 convention - word means
1343    /// 32-bit and dword means 64-bit.
1344    enum RelocationType {
1345      /// reloc_pcrel_word - PC relative relocation, add the relocated value to
1346      /// the value already in memory, after we adjust it for where the PC is.
1347      reloc_pcrel_word = 0,
1348
1349      /// reloc_picrel_word - PIC base relative relocation, add the relocated
1350      /// value to the value already in memory, after we adjust it for where the
1351      /// PIC base is.
1352      reloc_picrel_word = 1,
1353
1354      /// reloc_absolute_word, reloc_absolute_dword - Absolute relocation, just
1355      /// add the relocated value to the value already in memory.
1356      reloc_absolute_word = 2,
1357      reloc_absolute_dword = 3
1358    };
1359  }
1360}
1361</pre>
1362</div>
1363
1364<p>Since the body is small, indenting adds value because it makes it very clear
1365where the namespace starts and ends, and it is easy to take the whole thing in
1366in one "gulp" when reading the code.  If the blob of code in the namespace is
1367larger (as it typically is in a header in the <tt>llvm</tt> or <tt>clang</tt> namespaces), do not
1368indent the code, and add a comment indicating what namespace is being closed.
1369For example:</p>
1370
1371<div class="doc_code">
1372<pre>
1373namespace llvm {
1374namespace knowledge {
1375
1376/// Grokable - This class represents things that Smith can have an intimate
1377/// understanding of and contains the data associated with it.
1378class Grokable {
1379...
1380public:
1381  explicit Grokable() { ... }
1382  virtual ~Grokable() = 0;
1383
1384  ...
1385
1386};
1387
1388} // end namespace knowledge
1389} // end namespace llvm
1390</pre>
1391</div>
1392
1393<p>Because the class is large, we don't expect that the reader can easily
1394understand the entire concept in a glance, and the end of the file (where the
1395namespaces end) may be a long ways away from the place they open.  As such,
1396indenting the contents of the namespace doesn't add any value, and detracts from
1397the readability of the class.  In these cases it is best to <em>not</em> indent
1398the contents of the namespace.</p>
1399
1400</div>
1401
1402<!-- _______________________________________________________________________ -->
1403<h4>
1404  <a name="micro_anonns">Anonymous Namespaces</a>
1405</h4>
1406
1407<div>
1408
1409<p>After talking about namespaces in general, you may be wondering about
1410anonymous namespaces in particular.
1411Anonymous namespaces are a great language feature that tells the C++ compiler
1412that the contents of the namespace are only visible within the current
1413translation unit, allowing more aggressive optimization and eliminating the
1414possibility of symbol name collisions.  Anonymous namespaces are to C++ as
1415"static" is to C functions and global variables.  While "static" is available
1416in C++, anonymous namespaces are more general: they can make entire classes
1417private to a file.</p>
1418
1419<p>The problem with anonymous namespaces is that they naturally want to
1420encourage indentation of their body, and they reduce locality of reference: if
1421you see a random function definition in a C++ file, it is easy to see if it is
1422marked static, but seeing if it is in an anonymous namespace requires scanning
1423a big chunk of the file.</p>
1424
1425<p>Because of this, we have a simple guideline: make anonymous namespaces as
1426small as possible, and only use them for class declarations.  For example, this
1427is good:</p>
1428
1429<div class="doc_code">
1430<pre>
1431<b>namespace {</b>
1432  class StringSort {
1433  ...
1434  public:
1435    StringSort(...)
1436    bool operator&lt;(const char *RHS) const;
1437  };
1438<b>} // end anonymous namespace</b>
1439
1440static void Helper() {
1441  ...
1442}
1443
1444bool StringSort::operator&lt;(const char *RHS) const {
1445  ...
1446}
1447
1448</pre>
1449</div>
1450
1451<p>This is bad:</p>
1452
1453
1454<div class="doc_code">
1455<pre>
1456<b>namespace {</b>
1457class StringSort {
1458...
1459public:
1460  StringSort(...)
1461  bool operator&lt;(const char *RHS) const;
1462};
1463
1464void Helper() {
1465  ...
1466}
1467
1468bool StringSort::operator&lt;(const char *RHS) const {
1469  ...
1470}
1471
1472<b>} // end anonymous namespace</b>
1473
1474</pre>
1475</div>
1476
1477
1478<p>This is bad specifically because if you're looking at "Helper" in the middle
1479of a large C++ file, that you have no immediate way to tell if it is local to
1480the file.  When it is marked static explicitly, this is immediately obvious.
1481Also, there is no reason to enclose the definition of "operator&lt;" in the
1482namespace just because it was declared there.
1483</p>
1484
1485</div>
1486
1487</div>
1488
1489</div>
1490
1491<!-- *********************************************************************** -->
1492<h2>
1493  <a name="seealso">See Also</a>
1494</h2>
1495<!-- *********************************************************************** -->
1496
1497<div>
1498
1499<p>A lot of these comments and recommendations have been culled for other
1500sources.  Two particularly important books for our work are:</p>
1501
1502<ol>
1503
1504<li><a href="http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876">Effective
1505C++</a> by Scott Meyers.  Also
1506interesting and useful are "More Effective C++" and "Effective STL" by the same
1507author.</li>
1508
1509<li>Large-Scale C++ Software Design by John Lakos</li>
1510
1511</ol>
1512
1513<p>If you get some free time, and you haven't read them: do so, you might learn
1514something.</p>
1515
1516</div>
1517
1518<!-- *********************************************************************** -->
1519
1520<hr>
1521<address>
1522  <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1523  src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1524  <a href="http://validator.w3.org/check/referer"><img
1525  src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1526
1527  <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1528  <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1529  Last modified: $Date: 2011-08-12 15:49:16 -0400 (Fri, 12 Aug 2011) $
1530</address>
1531
1532</body>
1533</html>
1534