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_class_struct">Use of <tt>class</tt>/<tt>struct</tt> Keywords</a></li>
37 <li><a href="#styleissues">Style Issues</a>
39 <li><a href="#macro">The High-Level Issues</a>
41 <li><a href="#hl_module">A Public Header File <b>is</b> a
43 <li><a href="#hl_dontinclude"><tt>#include</tt> as Little as Possible</a></li>
44 <li><a href="#hl_privateheaders">Keep "internal" Headers
46 <li><a href="#hl_earlyexit">Use Early Exits and <tt>continue</tt> to Simplify
48 <li><a href="#hl_else_after_return">Don't use <tt>else</tt> after a
49 <tt>return</tt></a></li>
50 <li><a href="#hl_predicateloops">Turn Predicate Loops into Predicate
53 <li><a href="#micro">The Low-Level Issues</a>
55 <li><a href="#ll_naming">Name Types, Functions, Variables, and Enumerators Properly</a></li>
56 <li><a href="#ll_assert">Assert Liberally</a></li>
57 <li><a href="#ll_ns_std">Do not use '<tt>using namespace std</tt>'</a></li>
58 <li><a href="#ll_virtual_anch">Provide a virtual method anchor for
59 classes in headers</a></li>
60 <li><a href="#ll_end">Don't evaluate <tt>end()</tt> every time through a
62 <li><a href="#ll_iostream"><tt>#include <iostream></tt> is
63 <em>forbidden</em></a></li>
64 <li><a href="#ll_raw_ostream">Use <tt>raw_ostream</tt></a></li>
65 <li><a href="#ll_avoidendl">Avoid <tt>std::endl</tt></a></li>
68 <li><a href="#nano">Microscopic Details</a>
70 <li><a href="#micro_spaceparen">Spaces Before Parentheses</a></li>
71 <li><a href="#micro_preincrement">Prefer Preincrement</a></li>
72 <li><a href="#micro_namespaceindent">Namespace Indentation</a></li>
73 <li><a href="#micro_anonns">Anonymous Namespaces</a></li>
78 <li><a href="#seealso">See Also</a></li>
81 <div class="doc_author">
82 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
86 <!-- *********************************************************************** -->
88 <a name="introduction">Introduction</a>
90 <!-- *********************************************************************** -->
94 <p>This document attempts to describe a few coding standards that are being used
95 in the LLVM source tree. Although no coding standards should be regarded as
96 absolute requirements to be followed in all instances, coding standards can be
99 <p>This document intentionally does not prescribe fixed standards for religious
100 issues such as brace placement and space usage. For issues like this, follow
105 <p><b><a name="goldenrule">If you are adding a significant body of source to a
106 project, feel free to use whatever style you are most comfortable with. If you
107 are extending, enhancing, or bug fixing already implemented code, use the style
108 that is already being used so that the source is uniform and easy to
113 <p>The ultimate goal of these guidelines is the increase readability and
114 maintainability of our common source base. If you have suggestions for topics to
115 be included, please mail them to <a
116 href="mailto:sabre@nondot.org">Chris</a>.</p>
120 <!-- *********************************************************************** -->
122 <a name="mechanicalissues">Mechanical Source Issues</a>
124 <!-- *********************************************************************** -->
128 <!-- ======================================================================= -->
130 <a name="sourceformating">Source Code Formatting</a>
135 <!-- _______________________________________________________________________ -->
137 <a name="scf_commenting">Commenting</a>
142 <p>Comments are one critical part of readability and maintainability. Everyone
143 knows they should comment, so should you. When writing comments, write them as
144 English prose, which means they should use proper capitalization, punctuation,
145 etc. Although we all should probably
146 comment our code more than we do, there are a few very critical places that
147 documentation is very useful:</p>
149 <h5>File Headers</h5>
153 <p>Every source file should have a header on it that describes the basic
154 purpose of the file. If a file does not have a header, it should not be
155 checked into Subversion. Most source trees will probably have a standard
156 file header format. The standard format for the LLVM source tree looks like
159 <div class="doc_code">
161 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
163 // The LLVM Compiler Infrastructure
165 // This file is distributed under the University of Illinois Open Source
166 // License. See LICENSE.TXT for details.
168 //===----------------------------------------------------------------------===//
170 // This file contains the declaration of the Instruction class, which is the
171 // base class for all of the VM instructions.
173 //===----------------------------------------------------------------------===//
177 <p>A few things to note about this particular format: The "<tt>-*- C++
178 -*-</tt>" string on the first line is there to tell Emacs that the source file
179 is a C++ file, not a C file (Emacs assumes <tt>.h</tt> files are C files by default).
180 Note that this tag is not necessary in <tt>.cpp</tt> files. The name of the file is also
181 on the first line, along with a very short description of the purpose of the
182 file. This is important when printing out code and flipping though lots of
185 <p>The next section in the file is a concise note that defines the license
186 that the file is released under. This makes it perfectly clear what terms the
187 source code can be distributed under and should not be modified in any way.</p>
189 <p>The main body of the description does not have to be very long in most cases.
190 Here it's only two lines. If an algorithm is being implemented or something
191 tricky is going on, a reference to the paper where it is published should be
192 included, as well as any notes or "gotchas" in the code to watch out for.</p>
196 <h5>Class overviews</h5>
198 <p>Classes are one fundamental part of a good object oriented design. As such,
199 a class definition should have a comment block that explains what the class is
200 used for... if it's not obvious. If it's so completely obvious your grandma
201 could figure it out, it's probably safe to leave it out. Naming classes
202 something sane goes a long ways towards avoiding writing documentation.</p>
205 <h5>Method information</h5>
209 <p>Methods defined in a class (as well as any global functions) should also be
210 documented properly. A quick note about what it does and a description of the
211 borderline behaviour is all that is necessary here (unless something
212 particularly tricky or insidious is going on). The hope is that people can
213 figure out how to use your interfaces without reading the code itself... that is
216 <p>Good things to talk about here are what happens when something unexpected
217 happens: does the method return null? Abort? Format your hard disk?</p>
223 <!-- _______________________________________________________________________ -->
225 <a name="scf_commentformat">Comment Formatting</a>
230 <p>In general, prefer C++ style (<tt>//</tt>) comments. They take less space,
231 require less typing, don't have nesting problems, etc. There are a few cases
232 when it is useful to use C style (<tt>/* */</tt>) comments however:</p>
235 <li>When writing C code: Obviously if you are writing C code, use C style
237 <li>When writing a header file that may be <tt>#include</tt>d by a C source
239 <li>When writing a source file that is used by a tool that only accepts C
243 <p>To comment out a large block of code, use <tt>#if 0</tt> and <tt>#endif</tt>.
244 These nest properly and are better behaved in general than C style comments.</p>
248 <!-- _______________________________________________________________________ -->
250 <a name="scf_includes"><tt>#include</tt> Style</a>
255 <p>Immediately after the <a href="#scf_commenting">header file comment</a> (and
256 include guards if working on a header file), the <a
257 href="#hl_dontinclude">minimal</a> list of <tt>#include</tt>s required by the
258 file should be listed. We prefer these <tt>#include</tt>s to be listed in this
262 <li><a href="#mmheader">Main Module Header</a></li>
263 <li><a href="#hl_privateheaders">Local/Private Headers</a></li>
264 <li><tt>llvm/*</tt></li>
265 <li><tt>llvm/Analysis/*</tt></li>
266 <li><tt>llvm/Assembly/*</tt></li>
267 <li><tt>llvm/Bitcode/*</tt></li>
268 <li><tt>llvm/CodeGen/*</tt></li>
270 <li><tt>Support/*</tt></li>
271 <li><tt>Config/*</tt></li>
272 <li>System <tt>#includes</tt></li>
275 <p>and each category should be sorted by name.</p>
277 <p><a name="mmheader">The "Main Module Header"</a> file applies to <tt>.cpp</tt> files
278 which implement an interface defined by a <tt>.h</tt> file. This <tt>#include</tt>
279 should always be included <b>first</b> regardless of where it lives on the file
280 system. By including a header file first in the <tt>.cpp</tt> files that implement the
281 interfaces, we ensure that the header does not have any hidden dependencies
282 which are not explicitly #included in the header, but should be. It is also a
283 form of documentation in the <tt>.cpp</tt> file to indicate where the interfaces it
284 implements are defined.</p>
288 <!-- _______________________________________________________________________ -->
290 <a name="scf_codewidth">Source Code Width</a>
295 <p>Write your code to fit within 80 columns of text. This helps those of us who
296 like to print out code and look at your code in an xterm without resizing
299 <p>The longer answer is that there must be some limit to the width of the code
300 in order to reasonably allow developers to have multiple files side-by-side in
301 windows on a modest display. If you are going to pick a width limit, it is
302 somewhat arbitrary but you might as well pick something standard. Going with
303 90 columns (for example) instead of 80 columns wouldn't add any significant
304 value and would be detrimental to printing out code. Also many other projects
305 have standardized on 80 columns, so some people have already configured their
306 editors for it (vs something else, like 90 columns).</p>
308 <p>This is one of many contentious issues in coding standards, but it is not up
313 <!-- _______________________________________________________________________ -->
315 <a name="scf_spacestabs">Use Spaces Instead of Tabs</a>
320 <p>In all cases, prefer spaces to tabs in source files. People have different
321 preferred indentation levels, and different styles of indentation that they
322 like; this is fine. What isn't fine is that different editors/viewers expand
323 tabs out to different tab stops. This can cause your code to look completely
324 unreadable, and it is not worth dealing with.</p>
326 <p>As always, follow the <a href="#goldenrule">Golden Rule</a> above: follow the
327 style of existing code if you are modifying and extending it. If you like four
328 spaces of indentation, <b>DO NOT</b> do that in the middle of a chunk of code
329 with two spaces of indentation. Also, do not reindent a whole source file: it
330 makes for incredible diffs that are absolutely worthless.</p>
334 <!-- _______________________________________________________________________ -->
336 <a name="scf_indentation">Indent Code Consistently</a>
341 <p>Okay, in your first year of programming you were told that indentation is
342 important. If you didn't believe and internalize this then, now is the time.
349 <!-- ======================================================================= -->
351 <a name="compilerissues">Compiler Issues</a>
356 <!-- _______________________________________________________________________ -->
358 <a name="ci_warningerrors">Treat Compiler Warnings Like Errors</a>
363 <p>If your code has compiler warnings in it, something is wrong — you
364 aren't casting values correctly, your have "questionable" constructs in your
365 code, or you are doing something legitimately wrong. Compiler warnings can
366 cover up legitimate errors in output and make dealing with a translation unit
369 <p>It is not possible to prevent all warnings from all compilers, nor is it
370 desirable. Instead, pick a standard compiler (like <tt>gcc</tt>) that provides
371 a good thorough set of warnings, and stick to it. At least in the case of
372 <tt>gcc</tt>, it is possible to work around any spurious errors by changing the
373 syntax of the code slightly. For example, a warning that annoys me occurs when
374 I write code like this:</p>
376 <div class="doc_code">
378 if (V = getValue()) {
384 <p><tt>gcc</tt> will warn me that I probably want to use the <tt>==</tt>
385 operator, and that I probably mistyped it. In most cases, I haven't, and I
386 really don't want the spurious errors. To fix this particular problem, I
387 rewrite the code like this:</p>
389 <div class="doc_code">
391 if ((V = getValue())) {
397 <p>which shuts <tt>gcc</tt> up. Any <tt>gcc</tt> warning that annoys you can
398 be fixed by massaging the code appropriately.</p>
400 <p>These are the <tt>gcc</tt> warnings that I prefer to enable:</p>
402 <div class="doc_code">
404 -Wall -Winline -W -Wwrite-strings -Wno-unused
410 <!-- _______________________________________________________________________ -->
412 <a name="ci_portable_code">Write Portable Code</a>
417 <p>In almost all cases, it is possible and within reason to write completely
418 portable code. If there are cases where it isn't possible to write portable
419 code, isolate it behind a well defined (and well documented) interface.</p>
421 <p>In practice, this means that you shouldn't assume much about the host
422 compiler, and Visual Studio tends to be the lowest common denominator.
423 If advanced features are used, they should only be an implementation detail of
424 a library which has a simple exposed API, and preferably be buried in
429 <!-- _______________________________________________________________________ -->
431 <a name="ci_rtti_exceptions">Do not use RTTI or Exceptions</a>
435 <p>In an effort to reduce code and executable size, LLVM does not use RTTI
436 (e.g. <tt>dynamic_cast<></tt>) or exceptions. These two language features
437 violate the general C++ principle of <i>"you only pay for what you use"</i>,
438 causing executable bloat even if exceptions are never used in the code base, or
439 if RTTI is never used for a class. Because of this, we turn them off globally
442 <p>That said, LLVM does make extensive use of a hand-rolled form of RTTI that
443 use templates like <a href="ProgrammersManual.html#isa"><tt>isa<></tt>,
444 <tt>cast<></tt>, and <tt>dyn_cast<></tt></a>. This form of RTTI is
445 opt-in and can be added to any class. It is also substantially more efficient
446 than <tt>dynamic_cast<></tt>.</p>
450 <!-- _______________________________________________________________________ -->
452 <a name="ci_class_struct">Use of <tt>class</tt> and <tt>struct</tt> Keywords</a>
456 <p>In C++, the <tt>class</tt> and <tt>struct</tt> keywords can be used almost
457 interchangeably. The only difference is when they are used to declare a class:
458 <tt>class</tt> makes all members private by default while <tt>struct</tt> makes
459 all members public by default.</p>
461 <p>Unfortunately, not all compilers follow the rules and some will generate
462 different symbols based on whether <tt>class</tt> or <tt>struct</tt> was used to
463 declare the symbol. This can lead to problems at link time.</p>
465 <p>So, the rule for LLVM is to always use the <tt>class</tt> keyword, unless
466 <b>all</b> members are public and the type is a C++
467 <a href="http://en.wikipedia.org/wiki/Plain_old_data_structure">POD</a> type, in
468 which case <tt>struct</tt> is allowed.</p>
476 <!-- *********************************************************************** -->
478 <a name="styleissues">Style Issues</a>
480 <!-- *********************************************************************** -->
484 <!-- ======================================================================= -->
486 <a name="macro">The High-Level Issues</a>
488 <!-- ======================================================================= -->
492 <!-- _______________________________________________________________________ -->
494 <a name="hl_module">A Public Header File <b>is</b> a Module</a>
499 <p>C++ doesn't do too well in the modularity department. There is no real
500 encapsulation or data hiding (unless you use expensive protocol classes), but it
501 is what we have to work with. When you write a public header file (in the LLVM
502 source tree, they live in the top level "<tt>include</tt>" directory), you are
503 defining a module of functionality.</p>
505 <p>Ideally, modules should be completely independent of each other, and their
506 header files should only <tt>#include</tt> the absolute minimum number of
507 headers possible. A module is not just a class, a function, or a
508 namespace: <a href="http://www.cuj.com/articles/2000/0002/0002c/0002c.htm">it's
509 a collection of these</a> that defines an interface. This interface may be
510 several functions, classes, or data structures, but the important issue is how
511 they work together.</p>
513 <p>In general, a module should be implemented by one or more <tt>.cpp</tt>
514 files. Each of these <tt>.cpp</tt> files should include the header that defines
515 their interface first. This ensures that all of the dependences of the module
516 header have been properly added to the module header itself, and are not
517 implicit. System headers should be included after user headers for a
518 translation unit.</p>
522 <!-- _______________________________________________________________________ -->
524 <a name="hl_dontinclude"><tt>#include</tt> as Little as Possible</a>
529 <p><tt>#include</tt> hurts compile time performance. Don't do it unless you
530 have to, especially in header files.</p>
532 <p>But wait! Sometimes you need to have the definition of a class to use it, or
533 to inherit from it. In these cases go ahead and <tt>#include</tt> that header
534 file. Be aware however that there are many cases where you don't need to have
535 the full definition of a class. If you are using a pointer or reference to a
536 class, you don't need the header file. If you are simply returning a class
537 instance from a prototyped function or method, you don't need it. In fact, for
538 most cases, you simply don't need the definition of a class. And not
539 <tt>#include</tt>'ing speeds up compilation.</p>
541 <p>It is easy to try to go too overboard on this recommendation, however. You
542 <b>must</b> include all of the header files that you are using — you can
543 include them either directly or indirectly (through another header file). To
544 make sure that you don't accidentally forget to include a header file in your
545 module header, make sure to include your module header <b>first</b> in the
546 implementation file (as mentioned above). This way there won't be any hidden
547 dependencies that you'll find out about later.</p>
551 <!-- _______________________________________________________________________ -->
553 <a name="hl_privateheaders">Keep "Internal" Headers Private</a>
558 <p>Many modules have a complex implementation that causes them to use more than
559 one implementation (<tt>.cpp</tt>) file. It is often tempting to put the
560 internal communication interface (helper classes, extra functions, etc) in the
561 public module header file. Don't do this!</p>
563 <p>If you really need to do something like this, put a private header file in
564 the same directory as the source files, and include it locally. This ensures
565 that your private interface remains private and undisturbed by outsiders.</p>
567 <p>Note however, that it's okay to put extra implementation methods in a public
568 class itself. Just make them private (or protected) and all is well.</p>
572 <!-- _______________________________________________________________________ -->
574 <a name="hl_earlyexit">Use Early Exits and <tt>continue</tt> to Simplify Code</a>
579 <p>When reading code, keep in mind how much state and how many previous
580 decisions have to be remembered by the reader to understand a block of code.
581 Aim to reduce indentation where possible when it doesn't make it more difficult
582 to understand the code. One great way to do this is by making use of early
583 exits and the <tt>continue</tt> keyword in long loops. As an example of using
584 an early exit from a function, consider this "bad" code:</p>
586 <div class="doc_code">
588 Value *DoSomething(Instruction *I) {
589 if (!isa<TerminatorInst>(I) &&
590 I->hasOneUse() && SomeOtherThing(I)) {
591 ... some long code ....
599 <p>This code has several problems if the body of the '<tt>if</tt>' is large.
600 When you're looking at the top of the function, it isn't immediately clear that
601 this <em>only</em> does interesting things with non-terminator instructions, and
602 only applies to things with the other predicates. Second, it is relatively
603 difficult to describe (in comments) why these predicates are important because
604 the <tt>if</tt> statement makes it difficult to lay out the comments. Third,
605 when you're deep within the body of the code, it is indented an extra level.
606 Finally, when reading the top of the function, it isn't clear what the result is
607 if the predicate isn't true; you have to read to the end of the function to know
608 that it returns null.</p>
610 <p>It is much preferred to format the code like this:</p>
612 <div class="doc_code">
614 Value *DoSomething(Instruction *I) {
615 // Terminators never need 'something' done to them because ...
616 if (isa<TerminatorInst>(I))
619 // We conservatively avoid transforming instructions with multiple uses
620 // because goats like cheese.
621 if (!I->hasOneUse())
624 // This is really just here for example.
625 if (!SomeOtherThing(I))
628 ... some long code ....
633 <p>This fixes these problems. A similar problem frequently happens in <tt>for</tt>
634 loops. A silly example is something like this:</p>
636 <div class="doc_code">
638 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
639 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
640 Value *LHS = BO->getOperand(0);
641 Value *RHS = BO->getOperand(1);
650 <p>When you have very, very small loops, this sort of structure is fine. But if
651 it exceeds more than 10-15 lines, it becomes difficult for people to read and
652 understand at a glance. The problem with this sort of code is that it gets very
653 nested very quickly. Meaning that the reader of the code has to keep a lot of
654 context in their brain to remember what is going immediately on in the loop,
655 because they don't know if/when the <tt>if</tt> conditions will have elses etc.
656 It is strongly preferred to structure the loop like this:</p>
658 <div class="doc_code">
660 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
661 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
664 Value *LHS = BO->getOperand(0);
665 Value *RHS = BO->getOperand(1);
666 if (LHS == RHS) continue;
673 <p>This has all the benefits of using early exits for functions: it reduces
674 nesting of the loop, it makes it easier to describe why the conditions are true,
675 and it makes it obvious to the reader that there is no <tt>else</tt> coming up
676 that they have to push context into their brain for. If a loop is large, this
677 can be a big understandability win.</p>
681 <!-- _______________________________________________________________________ -->
683 <a name="hl_else_after_return">Don't use <tt>else</tt> after a <tt>return</tt></a>
688 <p>For similar reasons above (reduction of indentation and easier reading),
689 please do not use '<tt>else</tt>' or '<tt>else if</tt>' after something that
690 interrupts control flow — like <tt>return</tt>, <tt>break</tt>,
691 <tt>continue</tt>, <tt>goto</tt>, etc. For example, this is <em>bad</em>:</p>
693 <div class="doc_code">
697 Type = Context.getsigjmp_bufType();
699 Error = ASTContext::GE_Missing_sigjmp_buf;
705 Type = Context.getjmp_bufType();
707 Error = ASTContext::GE_Missing_jmp_buf;
718 <p>It is better to write it like this:</p>
720 <div class="doc_code">
724 Type = Context.getsigjmp_bufType();
726 Error = ASTContext::GE_Missing_sigjmp_buf;
730 Type = Context.getjmp_bufType();
732 Error = ASTContext::GE_Missing_jmp_buf;
740 <p>Or better yet (in this case) as:</p>
742 <div class="doc_code">
746 Type = Context.getsigjmp_bufType();
748 Type = Context.getjmp_bufType();
751 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
752 ASTContext::GE_Missing_jmp_buf;
759 <p>The idea is to reduce indentation and the amount of code you have to keep
760 track of when reading the code.</p>
764 <!-- _______________________________________________________________________ -->
766 <a name="hl_predicateloops">Turn Predicate Loops into Predicate Functions</a>
771 <p>It is very common to write small loops that just compute a boolean value.
772 There are a number of ways that people commonly write these, but an example of
773 this sort of thing is:</p>
775 <div class="doc_code">
777 <b>bool FoundFoo = false;</b>
778 for (unsigned i = 0, e = BarList.size(); i != e; ++i)
779 if (BarList[i]->isFoo()) {
780 <b>FoundFoo = true;</b>
784 <b>if (FoundFoo) {</b>
790 <p>This sort of code is awkward to write, and is almost always a bad sign.
791 Instead of this sort of loop, we strongly prefer to use a predicate function
792 (which may be <a href="#micro_anonns">static</a>) that uses
793 <a href="#hl_earlyexit">early exits</a> to compute the predicate. We prefer
794 the code to be structured like this:</p>
796 <div class="doc_code">
798 /// ListContainsFoo - Return true if the specified list has an element that is
800 static bool ListContainsFoo(const std::vector<Bar*> &List) {
801 for (unsigned i = 0, e = List.size(); i != e; ++i)
802 if (List[i]->isFoo())
808 <b>if (ListContainsFoo(BarList)) {</b>
814 <p>There are many reasons for doing this: it reduces indentation and factors out
815 code which can often be shared by other code that checks for the same predicate.
816 More importantly, it <em>forces you to pick a name</em> for the function, and
817 forces you to write a comment for it. In this silly example, this doesn't add
818 much value. However, if the condition is complex, this can make it a lot easier
819 for the reader to understand the code that queries for this predicate. Instead
820 of being faced with the in-line details of how we check to see if the BarList
821 contains a foo, we can trust the function name and continue reading with better
828 <!-- ======================================================================= -->
830 <a name="micro">The Low-Level Issues</a>
832 <!-- ======================================================================= -->
836 <!-- _______________________________________________________________________ -->
839 Name Types, Functions, Variables, and Enumerators Properly
845 <p>Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
846 enough how important it is to use <em>descriptive</em> names. Pick names that
847 match the semantics and role of the underlying entities, within reason. Avoid
848 abbreviations unless they are well known. After picking a good name, make sure
849 to use consistent capitalization for the name, as inconsistency requires clients
850 to either memorize the APIs or to look it up to find the exact spelling.</p>
852 <p>In general, names should be in camel case (e.g. <tt>TextFileReader</tt>
853 and <tt>isLValue()</tt>). Different kinds of declarations have different
857 <li><p><b>Type names</b> (including classes, structs, enums, typedefs, etc)
858 should be nouns and start with an upper-case letter (e.g.
859 <tt>TextFileReader</tt>).</p></li>
861 <li><p><b>Variable names</b> should be nouns (as they represent state). The
862 name should be camel case, and start with an upper case letter (e.g.
863 <tt>Leader</tt> or <tt>Boats</tt>).</p></li>
865 <li><p><b>Function names</b> should be verb phrases (as they represent
866 actions), and command-like function should be imperative. The name should
867 be camel case, and start with a lower case letter (e.g. <tt>openFile()</tt>
868 or <tt>isFoo()</tt>).</p></li>
870 <li><p><b>Enum declarations</b> (e.g. <tt>enum Foo {...}</tt>) are types, so
871 they should follow the naming conventions for types. A common use for enums
872 is as a discriminator for a union, or an indicator of a subclass. When an
873 enum is used for something like this, it should have a <tt>Kind</tt> suffix
874 (e.g. <tt>ValueKind</tt>).</p></li>
876 <li><p><b>Enumerators</b> (e.g. <tt>enum { Foo, Bar }</tt>) and <b>public member
877 variables</b> should start with an upper-case letter, just like types.
878 Unless the enumerators are defined in their own small namespace or inside a
879 class, enumerators should have a prefix corresponding to the enum
880 declaration name. For example, <tt>enum ValueKind { ... };</tt> may contain
881 enumerators like <tt>VK_Argument</tt>, <tt>VK_BasicBlock</tt>, etc.
882 Enumerators that are just convenience constants are exempt from the
883 requirement for a prefix. For instance:</p>
885 <div class="doc_code">
897 <p>As an exception, classes that mimic STL classes can have member names in
898 STL's style of lower-case words separated by underscores (e.g. <tt>begin()</tt>,
899 <tt>push_back()</tt>, and <tt>empty()</tt>).</p>
901 <p>Here are some examples of good and bad names:</p>
903 <div class="doc_code">
907 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
908 Factory<Tire> Factory; // Better.
909 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
910 // kind of factories.
913 Vehicle MakeVehicle(VehicleType Type) {
914 VehicleMaker M; // Might be OK if having a short life-span.
915 Tire tmp1 = M.makeTire(); // Bad -- 'tmp1' provides no information.
916 Light headlight = M.makeLight("head"); // Good -- descriptive.
925 <!-- _______________________________________________________________________ -->
927 <a name="ll_assert">Assert Liberally</a>
932 <p>Use the "<tt>assert</tt>" macro to its fullest. Check all of your
933 preconditions and assumptions, you never know when a bug (not necessarily even
934 yours) might be caught early by an assertion, which reduces debugging time
935 dramatically. The "<tt><cassert></tt>" header file is probably already
936 included by the header files you are using, so it doesn't cost anything to use
939 <p>To further assist with debugging, make sure to put some kind of error message
940 in the assertion statement, which is printed if the assertion is tripped. This
941 helps the poor debugger make sense of why an assertion is being made and
942 enforced, and hopefully what to do about it. Here is one complete example:</p>
944 <div class="doc_code">
946 inline Value *getOperand(unsigned i) {
947 assert(i < Operands.size() && "getOperand() out of range!");
953 <p>Here are more examples:</p>
955 <div class="doc_code">
957 assert(Ty->isPointerType() && "Can't allocate a non pointer type!");
959 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
961 assert(idx < getNumSuccessors() && "Successor # out of range!");
963 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
965 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
969 <p>You get the idea.</p>
971 <p>Please be aware that, when adding assert statements, not all compilers are aware of
972 the semantics of the assert. In some places, asserts are used to indicate a piece of
973 code that should not be reached. These are typically of the form:</p>
975 <div class="doc_code">
977 assert(0 && "Some helpful error message");
981 <p>When used in a function that returns a value, they should be followed with a return
982 statement and a comment indicating that this line is never reached. This will prevent
983 a compiler which is unable to deduce that the assert statement never returns from
984 generating a warning.</p>
986 <div class="doc_code">
988 assert(0 && "Some helpful error message");
994 <p>Another issue is that values used only by assertions will produce an "unused
995 value" warning when assertions are disabled. For example, this code will
998 <div class="doc_code">
1000 unsigned Size = V.size();
1001 assert(Size > 42 && "Vector smaller than it should be");
1003 bool NewToSet = Myset.insert(Value);
1004 assert(NewToSet && "The value shouldn't be in the set yet");
1008 <p>These are two interesting different cases. In the first case, the call to
1009 V.size() is only useful for the assert, and we don't want it executed when
1010 assertions are disabled. Code like this should move the call into the assert
1011 itself. In the second case, the side effects of the call must happen whether
1012 the assert is enabled or not. In this case, the value should be cast to void to
1013 disable the warning. To be specific, it is preferred to write the code like
1016 <div class="doc_code">
1018 assert(V.size() > 42 && "Vector smaller than it should be");
1020 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1021 assert(NewToSet && "The value shouldn't be in the set yet");
1028 <!-- _______________________________________________________________________ -->
1030 <a name="ll_ns_std">Do Not Use '<tt>using namespace std</tt>'</a>
1035 <p>In LLVM, we prefer to explicitly prefix all identifiers from the standard
1036 namespace with an "<tt>std::</tt>" prefix, rather than rely on
1037 "<tt>using namespace std;</tt>".</p>
1039 <p> In header files, adding a '<tt>using namespace XXX</tt>' directive pollutes
1040 the namespace of any source file that <tt>#include</tt>s the header. This is
1041 clearly a bad thing.</p>
1043 <p>In implementation files (e.g. <tt>.cpp</tt> files), the rule is more of a stylistic
1044 rule, but is still important. Basically, using explicit namespace prefixes
1045 makes the code <b>clearer</b>, because it is immediately obvious what facilities
1046 are being used and where they are coming from. And <b>more portable</b>, because
1047 namespace clashes cannot occur between LLVM code and other namespaces. The
1048 portability rule is important because different standard library implementations
1049 expose different symbols (potentially ones they shouldn't), and future revisions
1050 to the C++ standard will add more symbols to the <tt>std</tt> namespace. As
1051 such, we never use '<tt>using namespace std;</tt>' in LLVM.</p>
1053 <p>The exception to the general rule (i.e. it's not an exception for
1054 the <tt>std</tt> namespace) is for implementation files. For example, all of
1055 the code in the LLVM project implements code that lives in the 'llvm' namespace.
1056 As such, it is ok, and actually clearer, for the <tt>.cpp</tt> files to have a
1057 '<tt>using namespace llvm;</tt>' directive at the top, after the
1058 <tt>#include</tt>s. This reduces indentation in the body of the file for source
1059 editors that indent based on braces, and keeps the conceptual context cleaner.
1060 The general form of this rule is that any <tt>.cpp</tt> file that implements
1061 code in any namespace may use that namespace (and its parents'), but should not
1066 <!-- _______________________________________________________________________ -->
1068 <a name="ll_virtual_anch">
1069 Provide a Virtual Method Anchor for Classes in Headers
1075 <p>If a class is defined in a header file and has a v-table (either it has
1076 virtual methods or it derives from classes with virtual methods), it must
1077 always have at least one out-of-line virtual method in the class. Without
1078 this, the compiler will copy the vtable and RTTI into every <tt>.o</tt> file
1079 that <tt>#include</tt>s the header, bloating <tt>.o</tt> file sizes and
1080 increasing link times.</p>
1084 <!-- _______________________________________________________________________ -->
1086 <a name="ll_end">Don't evaluate <tt>end()</tt> every time through a loop</a>
1091 <p>Because C++ doesn't have a standard "<tt>foreach</tt>" loop (though it can be
1092 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1093 loops that manually iterate from begin to end on a variety of containers or
1094 through other data structures. One common mistake is to write a loop in this
1097 <div class="doc_code">
1099 BasicBlock *BB = ...
1100 for (BasicBlock::iterator I = BB->begin(); I != <b>BB->end()</b>; ++I)
1105 <p>The problem with this construct is that it evaluates "<tt>BB->end()</tt>"
1106 every time through the loop. Instead of writing the loop like this, we strongly
1107 prefer loops to be written so that they evaluate it once before the loop starts.
1108 A convenient way to do this is like so:</p>
1110 <div class="doc_code">
1112 BasicBlock *BB = ...
1113 for (BasicBlock::iterator I = BB->begin(), E = <b>BB->end()</b>; I != E; ++I)
1118 <p>The observant may quickly point out that these two loops may have different
1119 semantics: if the container (a basic block in this case) is being mutated, then
1120 "<tt>BB->end()</tt>" may change its value every time through the loop and the
1121 second loop may not in fact be correct. If you actually do depend on this
1122 behavior, please write the loop in the first form and add a comment indicating
1123 that you did it intentionally.</p>
1125 <p>Why do we prefer the second form (when correct)? Writing the loop in the
1126 first form has two problems. First it may be less efficient than evaluating it
1127 at the start of the loop. In this case, the cost is probably minor — a
1128 few extra loads every time through the loop. However, if the base expression is
1129 more complex, then the cost can rise quickly. I've seen loops where the end
1130 expression was actually something like: "<tt>SomeMap[x]->end()</tt>" and map
1131 lookups really aren't cheap. By writing it in the second form consistently, you
1132 eliminate the issue entirely and don't even have to think about it.</p>
1134 <p>The second (even bigger) issue is that writing the loop in the first form
1135 hints to the reader that the loop is mutating the container (a fact that a
1136 comment would handily confirm!). If you write the loop in the second form, it
1137 is immediately obvious without even looking at the body of the loop that the
1138 container isn't being modified, which makes it easier to read the code and
1139 understand what it does.</p>
1141 <p>While the second form of the loop is a few extra keystrokes, we do strongly
1146 <!-- _______________________________________________________________________ -->
1148 <a name="ll_iostream"><tt>#include <iostream></tt> is Forbidden</a>
1153 <p>The use of <tt>#include <iostream></tt> in library files is
1154 hereby <b><em>forbidden</em></b>. The primary reason for doing this is to
1155 support clients using LLVM libraries as part of larger systems. In particular,
1156 we statically link LLVM into some dynamic libraries. Even if LLVM isn't used,
1157 the static constructors are run whenever an application starts up that uses the
1158 dynamic library. There are two problems with this:</p>
1161 <li>The time to run the static c'tors impacts startup time of applications
1162 — a critical time for GUI apps.</li>
1164 <li>The static c'tors cause the app to pull many extra pages of memory off the
1165 disk: both the code for the static c'tors in each <tt>.o</tt> file and the
1166 small amount of data that gets touched. In addition, touched/dirty pages
1167 put more pressure on the VM system on low-memory machines.</li>
1170 <p>Note that using the other stream headers (<tt><sstream></tt> for
1171 example) is not problematic in this regard —
1172 just <tt><iostream></tt>. However, <tt>raw_ostream</tt> provides various
1173 APIs that are better performing for almost every use than <tt>std::ostream</tt>
1174 style APIs. <b>Therefore new code should always
1175 use <a href="#ll_raw_ostream"><tt>raw_ostream</tt></a> for writing, or
1176 the <tt>llvm::MemoryBuffer</tt> API for reading files.</b></p>
1181 <!-- _______________________________________________________________________ -->
1183 <a name="ll_raw_ostream">Use <tt>raw_ostream</tt></a>
1188 <p>LLVM includes a lightweight, simple, and efficient stream implementation
1189 in <tt>llvm/Support/raw_ostream.h</tt>, which provides all of the common
1190 features of <tt>std::ostream</tt>. All new code should use <tt>raw_ostream</tt>
1191 instead of <tt>ostream</tt>.</p>
1193 <p>Unlike <tt>std::ostream</tt>, <tt>raw_ostream</tt> is not a template and can
1194 be forward declared as <tt>class raw_ostream</tt>. Public headers should
1195 generally not include the <tt>raw_ostream</tt> header, but use forward
1196 declarations and constant references to <tt>raw_ostream</tt> instances.</p>
1201 <!-- _______________________________________________________________________ -->
1203 <a name="ll_avoidendl">Avoid <tt>std::endl</tt></a>
1208 <p>The <tt>std::endl</tt> modifier, when used with <tt>iostreams</tt> outputs a
1209 newline to the output stream specified. In addition to doing this, however, it
1210 also flushes the output stream. In other words, these are equivalent:</p>
1212 <div class="doc_code">
1214 std::cout << std::endl;
1215 std::cout << '\n' << std::flush;
1219 <p>Most of the time, you probably have no reason to flush the output stream, so
1220 it's better to use a literal <tt>'\n'</tt>.</p>
1226 <!-- ======================================================================= -->
1228 <a name="nano">Microscopic Details</a>
1230 <!-- ======================================================================= -->
1234 <p>This section describes preferred low-level formatting guidelines along with
1235 reasoning on why we prefer them.</p>
1237 <!-- _______________________________________________________________________ -->
1239 <a name="micro_spaceparen">Spaces Before Parentheses</a>
1244 <p>We prefer to put a space before an open parenthesis only in control flow
1245 statements, but not in normal function call expressions and function-like
1246 macros. For example, this is good:</p>
1248 <div class="doc_code">
1251 <b>for (</b>i = 0; i != 100; ++i) ...
1252 <b>while (</b>llvm_rocks) ...
1254 <b>somefunc(</b>42);
1255 <b><a href="#ll_assert">assert</a>(</b>3 != 4 && "laws of math are failing me");
1257 a = <b>foo(</b>42, 92) + <b>bar(</b>x);
1261 <p>and this is bad:</p>
1263 <div class="doc_code">
1266 <b>for(</b>i = 0; i != 100; ++i) ...
1267 <b>while(</b>llvm_rocks) ...
1269 <b>somefunc (</b>42);
1270 <b><a href="#ll_assert">assert</a> (</b>3 != 4 && "laws of math are failing me");
1272 a = <b>foo (</b>42, 92) + <b>bar (</b>x);
1276 <p>The reason for doing this is not completely arbitrary. This style makes
1277 control flow operators stand out more, and makes expressions flow better. The
1278 function call operator binds very tightly as a postfix operator. Putting a
1279 space after a function name (as in the last example) makes it appear that the
1280 code might bind the arguments of the left-hand-side of a binary operator with
1281 the argument list of a function and the name of the right side. More
1282 specifically, it is easy to misread the "a" example as:</p>
1284 <div class="doc_code">
1286 a = foo <b>(</b>(42, 92) + bar<b>)</b> (x);
1290 <p>when skimming through the code. By avoiding a space in a function, we avoid
1291 this misinterpretation.</p>
1295 <!-- _______________________________________________________________________ -->
1297 <a name="micro_preincrement">Prefer Preincrement</a>
1302 <p>Hard fast rule: Preincrement (<tt>++X</tt>) may be no slower than
1303 postincrement (<tt>X++</tt>) and could very well be a lot faster than it. Use
1304 preincrementation whenever possible.</p>
1306 <p>The semantics of postincrement include making a copy of the value being
1307 incremented, returning it, and then preincrementing the "work value". For
1308 primitive types, this isn't a big deal... but for iterators, it can be a huge
1309 issue (for example, some iterators contains stack and set objects in them...
1310 copying an iterator could invoke the copy ctor's of these as well). In general,
1311 get in the habit of always using preincrement, and you won't have a problem.</p>
1315 <!-- _______________________________________________________________________ -->
1317 <a name="micro_namespaceindent">Namespace Indentation</a>
1323 In general, we strive to reduce indentation wherever possible. This is useful
1324 because we want code to <a href="#scf_codewidth">fit into 80 columns</a> without
1325 wrapping horribly, but also because it makes it easier to understand the code.
1326 Namespaces are a funny thing: they are often large, and we often desire to put
1327 lots of stuff into them (so they can be large). Other times they are tiny,
1328 because they just hold an enum or something similar. In order to balance this,
1329 we use different approaches for small versus large namespaces.
1333 If a namespace definition is small and <em>easily</em> fits on a screen (say,
1334 less than 35 lines of code), then you should indent its body. Here's an
1338 <div class="doc_code">
1342 /// RelocationType - An enum for the x86 relocation codes. Note that
1343 /// the terminology here doesn't follow x86 convention - word means
1344 /// 32-bit and dword means 64-bit.
1345 enum RelocationType {
1346 /// reloc_pcrel_word - PC relative relocation, add the relocated value to
1347 /// the value already in memory, after we adjust it for where the PC is.
1348 reloc_pcrel_word = 0,
1350 /// reloc_picrel_word - PIC base relative relocation, add the relocated
1351 /// value to the value already in memory, after we adjust it for where the
1353 reloc_picrel_word = 1,
1355 /// reloc_absolute_word, reloc_absolute_dword - Absolute relocation, just
1356 /// add the relocated value to the value already in memory.
1357 reloc_absolute_word = 2,
1358 reloc_absolute_dword = 3
1365 <p>Since the body is small, indenting adds value because it makes it very clear
1366 where the namespace starts and ends, and it is easy to take the whole thing in
1367 in one "gulp" when reading the code. If the blob of code in the namespace is
1368 larger (as it typically is in a header in the <tt>llvm</tt> or <tt>clang</tt> namespaces), do not
1369 indent the code, and add a comment indicating what namespace is being closed.
1372 <div class="doc_code">
1375 namespace knowledge {
1377 /// Grokable - This class represents things that Smith can have an intimate
1378 /// understanding of and contains the data associated with it.
1382 explicit Grokable() { ... }
1383 virtual ~Grokable() = 0;
1389 } // end namespace knowledge
1390 } // end namespace llvm
1394 <p>Because the class is large, we don't expect that the reader can easily
1395 understand the entire concept in a glance, and the end of the file (where the
1396 namespaces end) may be a long ways away from the place they open. As such,
1397 indenting the contents of the namespace doesn't add any value, and detracts from
1398 the readability of the class. In these cases it is best to <em>not</em> indent
1399 the contents of the namespace.</p>
1403 <!-- _______________________________________________________________________ -->
1405 <a name="micro_anonns">Anonymous Namespaces</a>
1410 <p>After talking about namespaces in general, you may be wondering about
1411 anonymous namespaces in particular.
1412 Anonymous namespaces are a great language feature that tells the C++ compiler
1413 that the contents of the namespace are only visible within the current
1414 translation unit, allowing more aggressive optimization and eliminating the
1415 possibility of symbol name collisions. Anonymous namespaces are to C++ as
1416 "static" is to C functions and global variables. While "static" is available
1417 in C++, anonymous namespaces are more general: they can make entire classes
1418 private to a file.</p>
1420 <p>The problem with anonymous namespaces is that they naturally want to
1421 encourage indentation of their body, and they reduce locality of reference: if
1422 you see a random function definition in a C++ file, it is easy to see if it is
1423 marked static, but seeing if it is in an anonymous namespace requires scanning
1424 a big chunk of the file.</p>
1426 <p>Because of this, we have a simple guideline: make anonymous namespaces as
1427 small as possible, and only use them for class declarations. For example, this
1430 <div class="doc_code">
1437 bool operator<(const char *RHS) const;
1439 <b>} // end anonymous namespace</b>
1441 static void Helper() {
1445 bool StringSort::operator<(const char *RHS) const {
1455 <div class="doc_code">
1462 bool operator<(const char *RHS) const;
1469 bool StringSort::operator<(const char *RHS) const {
1473 <b>} // end anonymous namespace</b>
1479 <p>This is bad specifically because if you're looking at "Helper" in the middle
1480 of a large C++ file, that you have no immediate way to tell if it is local to
1481 the file. When it is marked static explicitly, this is immediately obvious.
1482 Also, there is no reason to enclose the definition of "operator<" in the
1483 namespace just because it was declared there.
1492 <!-- *********************************************************************** -->
1494 <a name="seealso">See Also</a>
1496 <!-- *********************************************************************** -->
1500 <p>A lot of these comments and recommendations have been culled for other
1501 sources. Two particularly important books for our work are:</p>
1505 <li><a href="http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876">Effective
1506 C++</a> by Scott Meyers. Also
1507 interesting and useful are "More Effective C++" and "Effective STL" by the same
1510 <li>Large-Scale C++ Software Design by John Lakos</li>
1514 <p>If you get some free time, and you haven't read them: do so, you might learn
1519 <!-- *********************************************************************** -->
1523 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1524 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1525 <a href="http://validator.w3.org/check/referer"><img
1526 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1528 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1529 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1530 Last modified: $Date$