1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
6 <link rel="stylesheet" href="llvm.css" type="text/css">
7 <title>LLVM Coding Standards</title>
16 <li><a href="#introduction">Introduction</a></li>
17 <li><a href="#mechanicalissues">Mechanical Source Issues</a>
19 <li><a href="#sourceformating">Source Code Formatting</a>
21 <li><a href="#scf_commenting">Commenting</a></li>
22 <li><a href="#scf_commentformat">Comment Formatting</a></li>
23 <li><a href="#scf_includes"><tt>#include</tt> Style</a></li>
24 <li><a href="#scf_codewidth">Source Code Width</a></li>
25 <li><a href="#scf_spacestabs">Use Spaces Instead of Tabs</a></li>
26 <li><a href="#scf_indentation">Indent Code Consistently</a></li>
28 <li><a href="#compilerissues">Compiler Issues</a>
30 <li><a href="#ci_warningerrors">Treat Compiler Warnings Like
32 <li><a href="#ci_portable_code">Write Portable Code</a></li>
33 <li><a href="#ci_rtti_exceptions">Do not use RTTI or Exceptions</a></li>
34 <li><a href="#ci_static_ctors">Do not use Static Constructors</a></li>
35 <li><a href="#ci_class_struct">Use of <tt>class</tt>/<tt>struct</tt> Keywords</a></li>
38 <li><a href="#styleissues">Style Issues</a>
40 <li><a href="#macro">The High-Level Issues</a>
42 <li><a href="#hl_module">A Public Header File <b>is</b> a
44 <li><a href="#hl_dontinclude"><tt>#include</tt> as Little as Possible</a></li>
45 <li><a href="#hl_privateheaders">Keep "internal" Headers
47 <li><a href="#hl_earlyexit">Use Early Exits and <tt>continue</tt> to Simplify
49 <li><a href="#hl_else_after_return">Don't use <tt>else</tt> after a
50 <tt>return</tt></a></li>
51 <li><a href="#hl_predicateloops">Turn Predicate Loops into Predicate
54 <li><a href="#micro">The Low-Level Issues</a>
56 <li><a href="#ll_naming">Name Types, Functions, Variables, and Enumerators Properly</a></li>
57 <li><a href="#ll_assert">Assert Liberally</a></li>
58 <li><a href="#ll_ns_std">Do not use '<tt>using namespace std</tt>'</a></li>
59 <li><a href="#ll_virtual_anch">Provide a virtual method anchor for
60 classes in headers</a></li>
61 <li><a href="#ll_end">Don't evaluate <tt>end()</tt> every time through a
63 <li><a href="#ll_iostream"><tt>#include <iostream></tt> is
64 <em>forbidden</em></a></li>
65 <li><a href="#ll_raw_ostream">Use <tt>raw_ostream</tt></a></li>
66 <li><a href="#ll_avoidendl">Avoid <tt>std::endl</tt></a></li>
69 <li><a href="#nano">Microscopic Details</a>
71 <li><a href="#micro_spaceparen">Spaces Before Parentheses</a></li>
72 <li><a href="#micro_preincrement">Prefer Preincrement</a></li>
73 <li><a href="#micro_namespaceindent">Namespace Indentation</a></li>
74 <li><a href="#micro_anonns">Anonymous Namespaces</a></li>
79 <li><a href="#seealso">See Also</a></li>
82 <div class="doc_author">
83 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
87 <!-- *********************************************************************** -->
89 <a name="introduction">Introduction</a>
91 <!-- *********************************************************************** -->
95 <p>This document attempts to describe a few coding standards that are being used
96 in the LLVM source tree. Although no coding standards should be regarded as
97 absolute requirements to be followed in all instances, coding standards can be
100 <p>This document intentionally does not prescribe fixed standards for religious
101 issues such as brace placement and space usage. For issues like this, follow
106 <p><b><a name="goldenrule">If you are adding a significant body of source to a
107 project, feel free to use whatever style you are most comfortable with. If you
108 are extending, enhancing, or bug fixing already implemented code, use the style
109 that is already being used so that the source is uniform and easy to
114 <p>The ultimate goal of these guidelines is the increase readability and
115 maintainability of our common source base. If you have suggestions for topics to
116 be included, please mail them to <a
117 href="mailto:sabre@nondot.org">Chris</a>.</p>
121 <!-- *********************************************************************** -->
123 <a name="mechanicalissues">Mechanical Source Issues</a>
125 <!-- *********************************************************************** -->
129 <!-- ======================================================================= -->
131 <a name="sourceformating">Source Code Formatting</a>
136 <!-- _______________________________________________________________________ -->
138 <a name="scf_commenting">Commenting</a>
143 <p>Comments are one critical part of readability and maintainability. Everyone
144 knows they should comment, so should you. When writing comments, write them as
145 English prose, which means they should use proper capitalization, punctuation,
146 etc. Although we all should probably
147 comment our code more than we do, there are a few very critical places that
148 documentation is very useful:</p>
150 <h5>File Headers</h5>
154 <p>Every source file should have a header on it that describes the basic
155 purpose of the file. If a file does not have a header, it should not be
156 checked into Subversion. Most source trees will probably have a standard
157 file header format. The standard format for the LLVM source tree looks like
160 <div class="doc_code">
162 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
164 // The LLVM Compiler Infrastructure
166 // This file is distributed under the University of Illinois Open Source
167 // License. See LICENSE.TXT for details.
169 //===----------------------------------------------------------------------===//
171 // This file contains the declaration of the Instruction class, which is the
172 // base class for all of the VM instructions.
174 //===----------------------------------------------------------------------===//
178 <p>A few things to note about this particular format: The "<tt>-*- C++
179 -*-</tt>" string on the first line is there to tell Emacs that the source file
180 is a C++ file, not a C file (Emacs assumes <tt>.h</tt> files are C files by default).
181 Note that this tag is not necessary in <tt>.cpp</tt> files. The name of the file is also
182 on the first line, along with a very short description of the purpose of the
183 file. This is important when printing out code and flipping though lots of
186 <p>The next section in the file is a concise note that defines the license
187 that the file is released under. This makes it perfectly clear what terms the
188 source code can be distributed under and should not be modified in any way.</p>
190 <p>The main body of the description does not have to be very long in most cases.
191 Here it's only two lines. If an algorithm is being implemented or something
192 tricky is going on, a reference to the paper where it is published should be
193 included, as well as any notes or "gotchas" in the code to watch out for.</p>
197 <h5>Class overviews</h5>
199 <p>Classes are one fundamental part of a good object oriented design. As such,
200 a class definition should have a comment block that explains what the class is
201 used for... if it's not obvious. If it's so completely obvious your grandma
202 could figure it out, it's probably safe to leave it out. Naming classes
203 something sane goes a long ways towards avoiding writing documentation.</p>
206 <h5>Method information</h5>
210 <p>Methods defined in a class (as well as any global functions) should also be
211 documented properly. A quick note about what it does and a description of the
212 borderline behaviour is all that is necessary here (unless something
213 particularly tricky or insidious is going on). The hope is that people can
214 figure out how to use your interfaces without reading the code itself... that is
217 <p>Good things to talk about here are what happens when something unexpected
218 happens: does the method return null? Abort? Format your hard disk?</p>
224 <!-- _______________________________________________________________________ -->
226 <a name="scf_commentformat">Comment Formatting</a>
231 <p>In general, prefer C++ style (<tt>//</tt>) comments. They take less space,
232 require less typing, don't have nesting problems, etc. There are a few cases
233 when it is useful to use C style (<tt>/* */</tt>) comments however:</p>
236 <li>When writing C code: Obviously if you are writing C code, use C style
238 <li>When writing a header file that may be <tt>#include</tt>d by a C source
240 <li>When writing a source file that is used by a tool that only accepts C
244 <p>To comment out a large block of code, use <tt>#if 0</tt> and <tt>#endif</tt>.
245 These nest properly and are better behaved in general than C style comments.</p>
249 <!-- _______________________________________________________________________ -->
251 <a name="scf_includes"><tt>#include</tt> Style</a>
256 <p>Immediately after the <a href="#scf_commenting">header file comment</a> (and
257 include guards if working on a header file), the <a
258 href="#hl_dontinclude">minimal</a> list of <tt>#include</tt>s required by the
259 file should be listed. We prefer these <tt>#include</tt>s to be listed in this
263 <li><a href="#mmheader">Main Module Header</a></li>
264 <li><a href="#hl_privateheaders">Local/Private Headers</a></li>
265 <li><tt>llvm/*</tt></li>
266 <li><tt>llvm/Analysis/*</tt></li>
267 <li><tt>llvm/Assembly/*</tt></li>
268 <li><tt>llvm/Bitcode/*</tt></li>
269 <li><tt>llvm/CodeGen/*</tt></li>
271 <li><tt>Support/*</tt></li>
272 <li><tt>Config/*</tt></li>
273 <li>System <tt>#includes</tt></li>
276 <p>and each category should be sorted by name.</p>
278 <p><a name="mmheader">The "Main Module Header"</a> file applies to <tt>.cpp</tt> files
279 which implement an interface defined by a <tt>.h</tt> file. This <tt>#include</tt>
280 should always be included <b>first</b> regardless of where it lives on the file
281 system. By including a header file first in the <tt>.cpp</tt> files that implement the
282 interfaces, we ensure that the header does not have any hidden dependencies
283 which are not explicitly #included in the header, but should be. It is also a
284 form of documentation in the <tt>.cpp</tt> file to indicate where the interfaces it
285 implements are defined.</p>
289 <!-- _______________________________________________________________________ -->
291 <a name="scf_codewidth">Source Code Width</a>
296 <p>Write your code to fit within 80 columns of text. This helps those of us who
297 like to print out code and look at your code in an xterm without resizing
300 <p>The longer answer is that there must be some limit to the width of the code
301 in order to reasonably allow developers to have multiple files side-by-side in
302 windows on a modest display. If you are going to pick a width limit, it is
303 somewhat arbitrary but you might as well pick something standard. Going with
304 90 columns (for example) instead of 80 columns wouldn't add any significant
305 value and would be detrimental to printing out code. Also many other projects
306 have standardized on 80 columns, so some people have already configured their
307 editors for it (vs something else, like 90 columns).</p>
309 <p>This is one of many contentious issues in coding standards, but it is not up
314 <!-- _______________________________________________________________________ -->
316 <a name="scf_spacestabs">Use Spaces Instead of Tabs</a>
321 <p>In all cases, prefer spaces to tabs in source files. People have different
322 preferred indentation levels, and different styles of indentation that they
323 like; this is fine. What isn't fine is that different editors/viewers expand
324 tabs out to different tab stops. This can cause your code to look completely
325 unreadable, and it is not worth dealing with.</p>
327 <p>As always, follow the <a href="#goldenrule">Golden Rule</a> above: follow the
328 style of existing code if you are modifying and extending it. If you like four
329 spaces of indentation, <b>DO NOT</b> do that in the middle of a chunk of code
330 with two spaces of indentation. Also, do not reindent a whole source file: it
331 makes for incredible diffs that are absolutely worthless.</p>
335 <!-- _______________________________________________________________________ -->
337 <a name="scf_indentation">Indent Code Consistently</a>
342 <p>Okay, in your first year of programming you were told that indentation is
343 important. If you didn't believe and internalize this then, now is the time.
350 <!-- ======================================================================= -->
352 <a name="compilerissues">Compiler Issues</a>
357 <!-- _______________________________________________________________________ -->
359 <a name="ci_warningerrors">Treat Compiler Warnings Like Errors</a>
364 <p>If your code has compiler warnings in it, something is wrong — you
365 aren't casting values correctly, your have "questionable" constructs in your
366 code, or you are doing something legitimately wrong. Compiler warnings can
367 cover up legitimate errors in output and make dealing with a translation unit
370 <p>It is not possible to prevent all warnings from all compilers, nor is it
371 desirable. Instead, pick a standard compiler (like <tt>gcc</tt>) that provides
372 a good thorough set of warnings, and stick to it. At least in the case of
373 <tt>gcc</tt>, it is possible to work around any spurious errors by changing the
374 syntax of the code slightly. For example, a warning that annoys me occurs when
375 I write code like this:</p>
377 <div class="doc_code">
379 if (V = getValue()) {
385 <p><tt>gcc</tt> will warn me that I probably want to use the <tt>==</tt>
386 operator, and that I probably mistyped it. In most cases, I haven't, and I
387 really don't want the spurious errors. To fix this particular problem, I
388 rewrite the code like this:</p>
390 <div class="doc_code">
392 if ((V = getValue())) {
398 <p>which shuts <tt>gcc</tt> up. Any <tt>gcc</tt> warning that annoys you can
399 be fixed by massaging the code appropriately.</p>
401 <p>These are the <tt>gcc</tt> warnings that I prefer to enable:</p>
403 <div class="doc_code">
405 -Wall -Winline -W -Wwrite-strings -Wno-unused
411 <!-- _______________________________________________________________________ -->
413 <a name="ci_portable_code">Write Portable Code</a>
418 <p>In almost all cases, it is possible and within reason to write completely
419 portable code. If there are cases where it isn't possible to write portable
420 code, isolate it behind a well defined (and well documented) interface.</p>
422 <p>In practice, this means that you shouldn't assume much about the host
423 compiler, and Visual Studio tends to be the lowest common denominator.
424 If advanced features are used, they should only be an implementation detail of
425 a library which has a simple exposed API, and preferably be buried in
430 <!-- _______________________________________________________________________ -->
432 <a name="ci_rtti_exceptions">Do not use RTTI or Exceptions</a>
436 <p>In an effort to reduce code and executable size, LLVM does not use RTTI
437 (e.g. <tt>dynamic_cast<></tt>) or exceptions. These two language features
438 violate the general C++ principle of <i>"you only pay for what you use"</i>,
439 causing executable bloat even if exceptions are never used in the code base, or
440 if RTTI is never used for a class. Because of this, we turn them off globally
443 <p>That said, LLVM does make extensive use of a hand-rolled form of RTTI that
444 use templates like <a href="ProgrammersManual.html#isa"><tt>isa<></tt>,
445 <tt>cast<></tt>, and <tt>dyn_cast<></tt></a>. This form of RTTI is
446 opt-in and can be added to any class. It is also substantially more efficient
447 than <tt>dynamic_cast<></tt>.</p>
451 <!-- _______________________________________________________________________ -->
453 <a name="ci_static_ctors">Do not use Static Constructors</a>
457 <p>Static constructors and destructors (e.g. global variables whose types have
458 a constructor or destructor) should not be added to the code base, and should be
459 removed wherever possible. Besides <a
460 href="http://yosefk.com/c++fqa/ctors.html#fqa-10.12">well known problems</a>
461 where the order of initialization is undefined between globals in different
462 source files, the entire concept of static constructors is at odds with the
463 common use case of LLVM as a library linked into a larger application.</p>
465 <p>Consider the use of LLVM as a JIT linked into another application (perhaps
466 for <a href="http://llvm.org/Users.html">OpenGL, custom languages</a>,
467 <a href="http://llvm.org/devmtg/2010-11/Gritz-OpenShadingLang.pdf">shaders in
468 movies</a>, etc</a>). Due to the design of static constructors, they must be
469 executed at startup time of the entire application, regardless of whether or
470 how LLVM is used in that larger application. There are two problems with
474 <li>The time to run the static constructors impacts startup time of
475 applications — a critical time for GUI apps, among others.</li>
477 <li>The static constructors cause the app to pull many extra pages of memory
478 off the disk: both the code for the constructor in each <tt>.o</tt> file and
479 the small amount of data that gets touched. In addition, touched/dirty pages
480 put more pressure on the VM system on low-memory machines.</li>
483 <p>We would really like for there to be zero cost for linking in an additional
484 LLVM target or other library into an application, but static constructors
485 violate this goal.</p>
487 <p>That said, LLVM unfortunately does contain static constructors. It would be
488 a <a href="http://llvm.org/PR11944">great project</a> for someone to purge all
489 static constructors from LLVM, and then enable the
490 <tt>-Wglobal-constructors</tt> warning flag (when building with Clang) to ensure
491 we do not regress in the future.
496 <!-- _______________________________________________________________________ -->
498 <a name="ci_class_struct">Use of <tt>class</tt> and <tt>struct</tt> Keywords</a>
502 <p>In C++, the <tt>class</tt> and <tt>struct</tt> keywords can be used almost
503 interchangeably. The only difference is when they are used to declare a class:
504 <tt>class</tt> makes all members private by default while <tt>struct</tt> makes
505 all members public by default.</p>
507 <p>Unfortunately, not all compilers follow the rules and some will generate
508 different symbols based on whether <tt>class</tt> or <tt>struct</tt> was used to
509 declare the symbol. This can lead to problems at link time.</p>
511 <p>So, the rule for LLVM is to always use the <tt>class</tt> keyword, unless
512 <b>all</b> members are public and the type is a C++
513 <a href="http://en.wikipedia.org/wiki/Plain_old_data_structure">POD</a> type, in
514 which case <tt>struct</tt> is allowed.</p>
522 <!-- *********************************************************************** -->
524 <a name="styleissues">Style Issues</a>
526 <!-- *********************************************************************** -->
530 <!-- ======================================================================= -->
532 <a name="macro">The High-Level Issues</a>
534 <!-- ======================================================================= -->
538 <!-- _______________________________________________________________________ -->
540 <a name="hl_module">A Public Header File <b>is</b> a Module</a>
545 <p>C++ doesn't do too well in the modularity department. There is no real
546 encapsulation or data hiding (unless you use expensive protocol classes), but it
547 is what we have to work with. When you write a public header file (in the LLVM
548 source tree, they live in the top level "<tt>include</tt>" directory), you are
549 defining a module of functionality.</p>
551 <p>Ideally, modules should be completely independent of each other, and their
552 header files should only <tt>#include</tt> the absolute minimum number of
553 headers possible. A module is not just a class, a function, or a
554 namespace: <a href="http://www.cuj.com/articles/2000/0002/0002c/0002c.htm">it's
555 a collection of these</a> that defines an interface. This interface may be
556 several functions, classes, or data structures, but the important issue is how
557 they work together.</p>
559 <p>In general, a module should be implemented by one or more <tt>.cpp</tt>
560 files. Each of these <tt>.cpp</tt> files should include the header that defines
561 their interface first. This ensures that all of the dependences of the module
562 header have been properly added to the module header itself, and are not
563 implicit. System headers should be included after user headers for a
564 translation unit.</p>
568 <!-- _______________________________________________________________________ -->
570 <a name="hl_dontinclude"><tt>#include</tt> as Little as Possible</a>
575 <p><tt>#include</tt> hurts compile time performance. Don't do it unless you
576 have to, especially in header files.</p>
578 <p>But wait! Sometimes you need to have the definition of a class to use it, or
579 to inherit from it. In these cases go ahead and <tt>#include</tt> that header
580 file. Be aware however that there are many cases where you don't need to have
581 the full definition of a class. If you are using a pointer or reference to a
582 class, you don't need the header file. If you are simply returning a class
583 instance from a prototyped function or method, you don't need it. In fact, for
584 most cases, you simply don't need the definition of a class. And not
585 <tt>#include</tt>'ing speeds up compilation.</p>
587 <p>It is easy to try to go too overboard on this recommendation, however. You
588 <b>must</b> include all of the header files that you are using — you can
589 include them either directly or indirectly (through another header file). To
590 make sure that you don't accidentally forget to include a header file in your
591 module header, make sure to include your module header <b>first</b> in the
592 implementation file (as mentioned above). This way there won't be any hidden
593 dependencies that you'll find out about later.</p>
597 <!-- _______________________________________________________________________ -->
599 <a name="hl_privateheaders">Keep "Internal" Headers Private</a>
604 <p>Many modules have a complex implementation that causes them to use more than
605 one implementation (<tt>.cpp</tt>) file. It is often tempting to put the
606 internal communication interface (helper classes, extra functions, etc) in the
607 public module header file. Don't do this!</p>
609 <p>If you really need to do something like this, put a private header file in
610 the same directory as the source files, and include it locally. This ensures
611 that your private interface remains private and undisturbed by outsiders.</p>
613 <p>Note however, that it's okay to put extra implementation methods in a public
614 class itself. Just make them private (or protected) and all is well.</p>
618 <!-- _______________________________________________________________________ -->
620 <a name="hl_earlyexit">Use Early Exits and <tt>continue</tt> to Simplify Code</a>
625 <p>When reading code, keep in mind how much state and how many previous
626 decisions have to be remembered by the reader to understand a block of code.
627 Aim to reduce indentation where possible when it doesn't make it more difficult
628 to understand the code. One great way to do this is by making use of early
629 exits and the <tt>continue</tt> keyword in long loops. As an example of using
630 an early exit from a function, consider this "bad" code:</p>
632 <div class="doc_code">
634 Value *DoSomething(Instruction *I) {
635 if (!isa<TerminatorInst>(I) &&
636 I->hasOneUse() && SomeOtherThing(I)) {
637 ... some long code ....
645 <p>This code has several problems if the body of the '<tt>if</tt>' is large.
646 When you're looking at the top of the function, it isn't immediately clear that
647 this <em>only</em> does interesting things with non-terminator instructions, and
648 only applies to things with the other predicates. Second, it is relatively
649 difficult to describe (in comments) why these predicates are important because
650 the <tt>if</tt> statement makes it difficult to lay out the comments. Third,
651 when you're deep within the body of the code, it is indented an extra level.
652 Finally, when reading the top of the function, it isn't clear what the result is
653 if the predicate isn't true; you have to read to the end of the function to know
654 that it returns null.</p>
656 <p>It is much preferred to format the code like this:</p>
658 <div class="doc_code">
660 Value *DoSomething(Instruction *I) {
661 // Terminators never need 'something' done to them because ...
662 if (isa<TerminatorInst>(I))
665 // We conservatively avoid transforming instructions with multiple uses
666 // because goats like cheese.
667 if (!I->hasOneUse())
670 // This is really just here for example.
671 if (!SomeOtherThing(I))
674 ... some long code ....
679 <p>This fixes these problems. A similar problem frequently happens in <tt>for</tt>
680 loops. A silly example is something like this:</p>
682 <div class="doc_code">
684 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
685 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
686 Value *LHS = BO->getOperand(0);
687 Value *RHS = BO->getOperand(1);
696 <p>When you have very, very small loops, this sort of structure is fine. But if
697 it exceeds more than 10-15 lines, it becomes difficult for people to read and
698 understand at a glance. The problem with this sort of code is that it gets very
699 nested very quickly. Meaning that the reader of the code has to keep a lot of
700 context in their brain to remember what is going immediately on in the loop,
701 because they don't know if/when the <tt>if</tt> conditions will have elses etc.
702 It is strongly preferred to structure the loop like this:</p>
704 <div class="doc_code">
706 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
707 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
710 Value *LHS = BO->getOperand(0);
711 Value *RHS = BO->getOperand(1);
712 if (LHS == RHS) continue;
719 <p>This has all the benefits of using early exits for functions: it reduces
720 nesting of the loop, it makes it easier to describe why the conditions are true,
721 and it makes it obvious to the reader that there is no <tt>else</tt> coming up
722 that they have to push context into their brain for. If a loop is large, this
723 can be a big understandability win.</p>
727 <!-- _______________________________________________________________________ -->
729 <a name="hl_else_after_return">Don't use <tt>else</tt> after a <tt>return</tt></a>
734 <p>For similar reasons above (reduction of indentation and easier reading),
735 please do not use '<tt>else</tt>' or '<tt>else if</tt>' after something that
736 interrupts control flow — like <tt>return</tt>, <tt>break</tt>,
737 <tt>continue</tt>, <tt>goto</tt>, etc. For example, this is <em>bad</em>:</p>
739 <div class="doc_code">
743 Type = Context.getsigjmp_bufType();
745 Error = ASTContext::GE_Missing_sigjmp_buf;
751 Type = Context.getjmp_bufType();
753 Error = ASTContext::GE_Missing_jmp_buf;
764 <p>It is better to write it like this:</p>
766 <div class="doc_code">
770 Type = Context.getsigjmp_bufType();
772 Error = ASTContext::GE_Missing_sigjmp_buf;
776 Type = Context.getjmp_bufType();
778 Error = ASTContext::GE_Missing_jmp_buf;
786 <p>Or better yet (in this case) as:</p>
788 <div class="doc_code">
792 Type = Context.getsigjmp_bufType();
794 Type = Context.getjmp_bufType();
797 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
798 ASTContext::GE_Missing_jmp_buf;
805 <p>The idea is to reduce indentation and the amount of code you have to keep
806 track of when reading the code.</p>
810 <!-- _______________________________________________________________________ -->
812 <a name="hl_predicateloops">Turn Predicate Loops into Predicate Functions</a>
817 <p>It is very common to write small loops that just compute a boolean value.
818 There are a number of ways that people commonly write these, but an example of
819 this sort of thing is:</p>
821 <div class="doc_code">
823 <b>bool FoundFoo = false;</b>
824 for (unsigned i = 0, e = BarList.size(); i != e; ++i)
825 if (BarList[i]->isFoo()) {
826 <b>FoundFoo = true;</b>
830 <b>if (FoundFoo) {</b>
836 <p>This sort of code is awkward to write, and is almost always a bad sign.
837 Instead of this sort of loop, we strongly prefer to use a predicate function
838 (which may be <a href="#micro_anonns">static</a>) that uses
839 <a href="#hl_earlyexit">early exits</a> to compute the predicate. We prefer
840 the code to be structured like this:</p>
842 <div class="doc_code">
844 /// ListContainsFoo - Return true if the specified list has an element that is
846 static bool ListContainsFoo(const std::vector<Bar*> &List) {
847 for (unsigned i = 0, e = List.size(); i != e; ++i)
848 if (List[i]->isFoo())
854 <b>if (ListContainsFoo(BarList)) {</b>
860 <p>There are many reasons for doing this: it reduces indentation and factors out
861 code which can often be shared by other code that checks for the same predicate.
862 More importantly, it <em>forces you to pick a name</em> for the function, and
863 forces you to write a comment for it. In this silly example, this doesn't add
864 much value. However, if the condition is complex, this can make it a lot easier
865 for the reader to understand the code that queries for this predicate. Instead
866 of being faced with the in-line details of how we check to see if the BarList
867 contains a foo, we can trust the function name and continue reading with better
874 <!-- ======================================================================= -->
876 <a name="micro">The Low-Level Issues</a>
878 <!-- ======================================================================= -->
882 <!-- _______________________________________________________________________ -->
885 Name Types, Functions, Variables, and Enumerators Properly
891 <p>Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
892 enough how important it is to use <em>descriptive</em> names. Pick names that
893 match the semantics and role of the underlying entities, within reason. Avoid
894 abbreviations unless they are well known. After picking a good name, make sure
895 to use consistent capitalization for the name, as inconsistency requires clients
896 to either memorize the APIs or to look it up to find the exact spelling.</p>
898 <p>In general, names should be in camel case (e.g. <tt>TextFileReader</tt>
899 and <tt>isLValue()</tt>). Different kinds of declarations have different
903 <li><p><b>Type names</b> (including classes, structs, enums, typedefs, etc)
904 should be nouns and start with an upper-case letter (e.g.
905 <tt>TextFileReader</tt>).</p></li>
907 <li><p><b>Variable names</b> should be nouns (as they represent state). The
908 name should be camel case, and start with an upper case letter (e.g.
909 <tt>Leader</tt> or <tt>Boats</tt>).</p></li>
911 <li><p><b>Function names</b> should be verb phrases (as they represent
912 actions), and command-like function should be imperative. The name should
913 be camel case, and start with a lower case letter (e.g. <tt>openFile()</tt>
914 or <tt>isFoo()</tt>).</p></li>
916 <li><p><b>Enum declarations</b> (e.g. <tt>enum Foo {...}</tt>) are types, so
917 they should follow the naming conventions for types. A common use for enums
918 is as a discriminator for a union, or an indicator of a subclass. When an
919 enum is used for something like this, it should have a <tt>Kind</tt> suffix
920 (e.g. <tt>ValueKind</tt>).</p></li>
922 <li><p><b>Enumerators</b> (e.g. <tt>enum { Foo, Bar }</tt>) and <b>public member
923 variables</b> should start with an upper-case letter, just like types.
924 Unless the enumerators are defined in their own small namespace or inside a
925 class, enumerators should have a prefix corresponding to the enum
926 declaration name. For example, <tt>enum ValueKind { ... };</tt> may contain
927 enumerators like <tt>VK_Argument</tt>, <tt>VK_BasicBlock</tt>, etc.
928 Enumerators that are just convenience constants are exempt from the
929 requirement for a prefix. For instance:</p>
931 <div class="doc_code">
943 <p>As an exception, classes that mimic STL classes can have member names in
944 STL's style of lower-case words separated by underscores (e.g. <tt>begin()</tt>,
945 <tt>push_back()</tt>, and <tt>empty()</tt>).</p>
947 <p>Here are some examples of good and bad names:</p>
949 <div class="doc_code">
953 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
954 Factory<Tire> Factory; // Better.
955 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
956 // kind of factories.
959 Vehicle MakeVehicle(VehicleType Type) {
960 VehicleMaker M; // Might be OK if having a short life-span.
961 Tire tmp1 = M.makeTire(); // Bad -- 'tmp1' provides no information.
962 Light headlight = M.makeLight("head"); // Good -- descriptive.
971 <!-- _______________________________________________________________________ -->
973 <a name="ll_assert">Assert Liberally</a>
978 <p>Use the "<tt>assert</tt>" macro to its fullest. Check all of your
979 preconditions and assumptions, you never know when a bug (not necessarily even
980 yours) might be caught early by an assertion, which reduces debugging time
981 dramatically. The "<tt><cassert></tt>" header file is probably already
982 included by the header files you are using, so it doesn't cost anything to use
985 <p>To further assist with debugging, make sure to put some kind of error message
986 in the assertion statement, which is printed if the assertion is tripped. This
987 helps the poor debugger make sense of why an assertion is being made and
988 enforced, and hopefully what to do about it. Here is one complete example:</p>
990 <div class="doc_code">
992 inline Value *getOperand(unsigned i) {
993 assert(i < Operands.size() && "getOperand() out of range!");
999 <p>Here are more examples:</p>
1001 <div class="doc_code">
1003 assert(Ty->isPointerType() && "Can't allocate a non pointer type!");
1005 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
1007 assert(idx < getNumSuccessors() && "Successor # out of range!");
1009 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
1011 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
1015 <p>You get the idea.</p>
1017 <p>Please be aware that, when adding assert statements, not all compilers are aware of
1018 the semantics of the assert. In some places, asserts are used to indicate a piece of
1019 code that should not be reached. These are typically of the form:</p>
1021 <div class="doc_code">
1023 assert(0 && "Some helpful error message");
1027 <p>When used in a function that returns a value, they should be followed with a return
1028 statement and a comment indicating that this line is never reached. This will prevent
1029 a compiler which is unable to deduce that the assert statement never returns from
1030 generating a warning.</p>
1032 <div class="doc_code">
1034 assert(0 && "Some helpful error message");
1040 <p>Another issue is that values used only by assertions will produce an "unused
1041 value" warning when assertions are disabled. For example, this code will
1044 <div class="doc_code">
1046 unsigned Size = V.size();
1047 assert(Size > 42 && "Vector smaller than it should be");
1049 bool NewToSet = Myset.insert(Value);
1050 assert(NewToSet && "The value shouldn't be in the set yet");
1054 <p>These are two interesting different cases. In the first case, the call to
1055 V.size() is only useful for the assert, and we don't want it executed when
1056 assertions are disabled. Code like this should move the call into the assert
1057 itself. In the second case, the side effects of the call must happen whether
1058 the assert is enabled or not. In this case, the value should be cast to void to
1059 disable the warning. To be specific, it is preferred to write the code like
1062 <div class="doc_code">
1064 assert(V.size() > 42 && "Vector smaller than it should be");
1066 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1067 assert(NewToSet && "The value shouldn't be in the set yet");
1074 <!-- _______________________________________________________________________ -->
1076 <a name="ll_ns_std">Do Not Use '<tt>using namespace std</tt>'</a>
1081 <p>In LLVM, we prefer to explicitly prefix all identifiers from the standard
1082 namespace with an "<tt>std::</tt>" prefix, rather than rely on
1083 "<tt>using namespace std;</tt>".</p>
1085 <p> In header files, adding a '<tt>using namespace XXX</tt>' directive pollutes
1086 the namespace of any source file that <tt>#include</tt>s the header. This is
1087 clearly a bad thing.</p>
1089 <p>In implementation files (e.g. <tt>.cpp</tt> files), the rule is more of a stylistic
1090 rule, but is still important. Basically, using explicit namespace prefixes
1091 makes the code <b>clearer</b>, because it is immediately obvious what facilities
1092 are being used and where they are coming from. And <b>more portable</b>, because
1093 namespace clashes cannot occur between LLVM code and other namespaces. The
1094 portability rule is important because different standard library implementations
1095 expose different symbols (potentially ones they shouldn't), and future revisions
1096 to the C++ standard will add more symbols to the <tt>std</tt> namespace. As
1097 such, we never use '<tt>using namespace std;</tt>' in LLVM.</p>
1099 <p>The exception to the general rule (i.e. it's not an exception for
1100 the <tt>std</tt> namespace) is for implementation files. For example, all of
1101 the code in the LLVM project implements code that lives in the 'llvm' namespace.
1102 As such, it is ok, and actually clearer, for the <tt>.cpp</tt> files to have a
1103 '<tt>using namespace llvm;</tt>' directive at the top, after the
1104 <tt>#include</tt>s. This reduces indentation in the body of the file for source
1105 editors that indent based on braces, and keeps the conceptual context cleaner.
1106 The general form of this rule is that any <tt>.cpp</tt> file that implements
1107 code in any namespace may use that namespace (and its parents'), but should not
1112 <!-- _______________________________________________________________________ -->
1114 <a name="ll_virtual_anch">
1115 Provide a Virtual Method Anchor for Classes in Headers
1121 <p>If a class is defined in a header file and has a v-table (either it has
1122 virtual methods or it derives from classes with virtual methods), it must
1123 always have at least one out-of-line virtual method in the class. Without
1124 this, the compiler will copy the vtable and RTTI into every <tt>.o</tt> file
1125 that <tt>#include</tt>s the header, bloating <tt>.o</tt> file sizes and
1126 increasing link times.</p>
1130 <!-- _______________________________________________________________________ -->
1132 <a name="ll_end">Don't evaluate <tt>end()</tt> every time through a loop</a>
1137 <p>Because C++ doesn't have a standard "<tt>foreach</tt>" loop (though it can be
1138 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1139 loops that manually iterate from begin to end on a variety of containers or
1140 through other data structures. One common mistake is to write a loop in this
1143 <div class="doc_code">
1145 BasicBlock *BB = ...
1146 for (BasicBlock::iterator I = BB->begin(); I != <b>BB->end()</b>; ++I)
1151 <p>The problem with this construct is that it evaluates "<tt>BB->end()</tt>"
1152 every time through the loop. Instead of writing the loop like this, we strongly
1153 prefer loops to be written so that they evaluate it once before the loop starts.
1154 A convenient way to do this is like so:</p>
1156 <div class="doc_code">
1158 BasicBlock *BB = ...
1159 for (BasicBlock::iterator I = BB->begin(), E = <b>BB->end()</b>; I != E; ++I)
1164 <p>The observant may quickly point out that these two loops may have different
1165 semantics: if the container (a basic block in this case) is being mutated, then
1166 "<tt>BB->end()</tt>" may change its value every time through the loop and the
1167 second loop may not in fact be correct. If you actually do depend on this
1168 behavior, please write the loop in the first form and add a comment indicating
1169 that you did it intentionally.</p>
1171 <p>Why do we prefer the second form (when correct)? Writing the loop in the
1172 first form has two problems. First it may be less efficient than evaluating it
1173 at the start of the loop. In this case, the cost is probably minor — a
1174 few extra loads every time through the loop. However, if the base expression is
1175 more complex, then the cost can rise quickly. I've seen loops where the end
1176 expression was actually something like: "<tt>SomeMap[x]->end()</tt>" and map
1177 lookups really aren't cheap. By writing it in the second form consistently, you
1178 eliminate the issue entirely and don't even have to think about it.</p>
1180 <p>The second (even bigger) issue is that writing the loop in the first form
1181 hints to the reader that the loop is mutating the container (a fact that a
1182 comment would handily confirm!). If you write the loop in the second form, it
1183 is immediately obvious without even looking at the body of the loop that the
1184 container isn't being modified, which makes it easier to read the code and
1185 understand what it does.</p>
1187 <p>While the second form of the loop is a few extra keystrokes, we do strongly
1192 <!-- _______________________________________________________________________ -->
1194 <a name="ll_iostream"><tt>#include <iostream></tt> is Forbidden</a>
1199 <p>The use of <tt>#include <iostream></tt> in library files is
1200 hereby <b><em>forbidden</em></b>, because many common implementations
1201 transparently inject a <a href="#ci_static_ctors">static constructor</a> into
1202 every translation unit that includes it.</p>
1204 <p>Note that using the other stream headers (<tt><sstream></tt> for
1205 example) is not problematic in this regard —
1206 just <tt><iostream></tt>. However, <tt>raw_ostream</tt> provides various
1207 APIs that are better performing for almost every use than <tt>std::ostream</tt>
1208 style APIs. <b>Therefore new code should always
1209 use <a href="#ll_raw_ostream"><tt>raw_ostream</tt></a> for writing, or
1210 the <tt>llvm::MemoryBuffer</tt> API for reading files.</b></p>
1215 <!-- _______________________________________________________________________ -->
1217 <a name="ll_raw_ostream">Use <tt>raw_ostream</tt></a>
1222 <p>LLVM includes a lightweight, simple, and efficient stream implementation
1223 in <tt>llvm/Support/raw_ostream.h</tt>, which provides all of the common
1224 features of <tt>std::ostream</tt>. All new code should use <tt>raw_ostream</tt>
1225 instead of <tt>ostream</tt>.</p>
1227 <p>Unlike <tt>std::ostream</tt>, <tt>raw_ostream</tt> is not a template and can
1228 be forward declared as <tt>class raw_ostream</tt>. Public headers should
1229 generally not include the <tt>raw_ostream</tt> header, but use forward
1230 declarations and constant references to <tt>raw_ostream</tt> instances.</p>
1235 <!-- _______________________________________________________________________ -->
1237 <a name="ll_avoidendl">Avoid <tt>std::endl</tt></a>
1242 <p>The <tt>std::endl</tt> modifier, when used with <tt>iostreams</tt> outputs a
1243 newline to the output stream specified. In addition to doing this, however, it
1244 also flushes the output stream. In other words, these are equivalent:</p>
1246 <div class="doc_code">
1248 std::cout << std::endl;
1249 std::cout << '\n' << std::flush;
1253 <p>Most of the time, you probably have no reason to flush the output stream, so
1254 it's better to use a literal <tt>'\n'</tt>.</p>
1260 <!-- ======================================================================= -->
1262 <a name="nano">Microscopic Details</a>
1264 <!-- ======================================================================= -->
1268 <p>This section describes preferred low-level formatting guidelines along with
1269 reasoning on why we prefer them.</p>
1271 <!-- _______________________________________________________________________ -->
1273 <a name="micro_spaceparen">Spaces Before Parentheses</a>
1278 <p>We prefer to put a space before an open parenthesis only in control flow
1279 statements, but not in normal function call expressions and function-like
1280 macros. For example, this is good:</p>
1282 <div class="doc_code">
1285 <b>for (</b>i = 0; i != 100; ++i) ...
1286 <b>while (</b>llvm_rocks) ...
1288 <b>somefunc(</b>42);
1289 <b><a href="#ll_assert">assert</a>(</b>3 != 4 && "laws of math are failing me");
1291 a = <b>foo(</b>42, 92) + <b>bar(</b>x);
1295 <p>and this is bad:</p>
1297 <div class="doc_code">
1300 <b>for(</b>i = 0; i != 100; ++i) ...
1301 <b>while(</b>llvm_rocks) ...
1303 <b>somefunc (</b>42);
1304 <b><a href="#ll_assert">assert</a> (</b>3 != 4 && "laws of math are failing me");
1306 a = <b>foo (</b>42, 92) + <b>bar (</b>x);
1310 <p>The reason for doing this is not completely arbitrary. This style makes
1311 control flow operators stand out more, and makes expressions flow better. The
1312 function call operator binds very tightly as a postfix operator. Putting a
1313 space after a function name (as in the last example) makes it appear that the
1314 code might bind the arguments of the left-hand-side of a binary operator with
1315 the argument list of a function and the name of the right side. More
1316 specifically, it is easy to misread the "a" example as:</p>
1318 <div class="doc_code">
1320 a = foo <b>(</b>(42, 92) + bar<b>)</b> (x);
1324 <p>when skimming through the code. By avoiding a space in a function, we avoid
1325 this misinterpretation.</p>
1329 <!-- _______________________________________________________________________ -->
1331 <a name="micro_preincrement">Prefer Preincrement</a>
1336 <p>Hard fast rule: Preincrement (<tt>++X</tt>) may be no slower than
1337 postincrement (<tt>X++</tt>) and could very well be a lot faster than it. Use
1338 preincrementation whenever possible.</p>
1340 <p>The semantics of postincrement include making a copy of the value being
1341 incremented, returning it, and then preincrementing the "work value". For
1342 primitive types, this isn't a big deal... but for iterators, it can be a huge
1343 issue (for example, some iterators contains stack and set objects in them...
1344 copying an iterator could invoke the copy ctor's of these as well). In general,
1345 get in the habit of always using preincrement, and you won't have a problem.</p>
1349 <!-- _______________________________________________________________________ -->
1351 <a name="micro_namespaceindent">Namespace Indentation</a>
1357 In general, we strive to reduce indentation wherever possible. This is useful
1358 because we want code to <a href="#scf_codewidth">fit into 80 columns</a> without
1359 wrapping horribly, but also because it makes it easier to understand the code.
1360 Namespaces are a funny thing: they are often large, and we often desire to put
1361 lots of stuff into them (so they can be large). Other times they are tiny,
1362 because they just hold an enum or something similar. In order to balance this,
1363 we use different approaches for small versus large namespaces.
1367 If a namespace definition is small and <em>easily</em> fits on a screen (say,
1368 less than 35 lines of code), then you should indent its body. Here's an
1372 <div class="doc_code">
1376 /// RelocationType - An enum for the x86 relocation codes. Note that
1377 /// the terminology here doesn't follow x86 convention - word means
1378 /// 32-bit and dword means 64-bit.
1379 enum RelocationType {
1380 /// reloc_pcrel_word - PC relative relocation, add the relocated value to
1381 /// the value already in memory, after we adjust it for where the PC is.
1382 reloc_pcrel_word = 0,
1384 /// reloc_picrel_word - PIC base relative relocation, add the relocated
1385 /// value to the value already in memory, after we adjust it for where the
1387 reloc_picrel_word = 1,
1389 /// reloc_absolute_word, reloc_absolute_dword - Absolute relocation, just
1390 /// add the relocated value to the value already in memory.
1391 reloc_absolute_word = 2,
1392 reloc_absolute_dword = 3
1399 <p>Since the body is small, indenting adds value because it makes it very clear
1400 where the namespace starts and ends, and it is easy to take the whole thing in
1401 in one "gulp" when reading the code. If the blob of code in the namespace is
1402 larger (as it typically is in a header in the <tt>llvm</tt> or <tt>clang</tt> namespaces), do not
1403 indent the code, and add a comment indicating what namespace is being closed.
1406 <div class="doc_code">
1409 namespace knowledge {
1411 /// Grokable - This class represents things that Smith can have an intimate
1412 /// understanding of and contains the data associated with it.
1416 explicit Grokable() { ... }
1417 virtual ~Grokable() = 0;
1423 } // end namespace knowledge
1424 } // end namespace llvm
1428 <p>Because the class is large, we don't expect that the reader can easily
1429 understand the entire concept in a glance, and the end of the file (where the
1430 namespaces end) may be a long ways away from the place they open. As such,
1431 indenting the contents of the namespace doesn't add any value, and detracts from
1432 the readability of the class. In these cases it is best to <em>not</em> indent
1433 the contents of the namespace.</p>
1437 <!-- _______________________________________________________________________ -->
1439 <a name="micro_anonns">Anonymous Namespaces</a>
1444 <p>After talking about namespaces in general, you may be wondering about
1445 anonymous namespaces in particular.
1446 Anonymous namespaces are a great language feature that tells the C++ compiler
1447 that the contents of the namespace are only visible within the current
1448 translation unit, allowing more aggressive optimization and eliminating the
1449 possibility of symbol name collisions. Anonymous namespaces are to C++ as
1450 "static" is to C functions and global variables. While "static" is available
1451 in C++, anonymous namespaces are more general: they can make entire classes
1452 private to a file.</p>
1454 <p>The problem with anonymous namespaces is that they naturally want to
1455 encourage indentation of their body, and they reduce locality of reference: if
1456 you see a random function definition in a C++ file, it is easy to see if it is
1457 marked static, but seeing if it is in an anonymous namespace requires scanning
1458 a big chunk of the file.</p>
1460 <p>Because of this, we have a simple guideline: make anonymous namespaces as
1461 small as possible, and only use them for class declarations. For example, this
1464 <div class="doc_code">
1471 bool operator<(const char *RHS) const;
1473 <b>} // end anonymous namespace</b>
1475 static void Helper() {
1479 bool StringSort::operator<(const char *RHS) const {
1489 <div class="doc_code">
1496 bool operator<(const char *RHS) const;
1503 bool StringSort::operator<(const char *RHS) const {
1507 <b>} // end anonymous namespace</b>
1513 <p>This is bad specifically because if you're looking at "Helper" in the middle
1514 of a large C++ file, that you have no immediate way to tell if it is local to
1515 the file. When it is marked static explicitly, this is immediately obvious.
1516 Also, there is no reason to enclose the definition of "operator<" in the
1517 namespace just because it was declared there.
1526 <!-- *********************************************************************** -->
1528 <a name="seealso">See Also</a>
1530 <!-- *********************************************************************** -->
1534 <p>A lot of these comments and recommendations have been culled for other
1535 sources. Two particularly important books for our work are:</p>
1539 <li><a href="http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876">Effective
1540 C++</a> by Scott Meyers. Also
1541 interesting and useful are "More Effective C++" and "Effective STL" by the same
1544 <li>Large-Scale C++ Software Design by John Lakos</li>
1548 <p>If you get some free time, and you haven't read them: do so, you might learn
1553 <!-- *********************************************************************** -->
1557 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1558 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1559 <a href="http://validator.w3.org/check/referer"><img
1560 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1562 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1563 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1564 Last modified: $Date$