1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <link rel="stylesheet" href="llvm.css" type="text/css">
6 <title>LLVM Coding Standards</title>
15 <li><a href="#introduction">Introduction</a></li>
16 <li><a href="#mechanicalissues">Mechanical Source Issues</a>
18 <li><a href="#sourceformating">Source Code Formatting</a>
20 <li><a href="#scf_commenting">Commenting</a></li>
21 <li><a href="#scf_commentformat">Comment Formatting</a></li>
22 <li><a href="#scf_includes"><tt>#include</tt> Style</a></li>
23 <li><a href="#scf_codewidth">Source Code Width</a></li>
24 <li><a href="#scf_spacestabs">Use Spaces Instead of Tabs</a></li>
25 <li><a href="#scf_indentation">Indent Code Consistently</a></li>
27 <li><a href="#compilerissues">Compiler Issues</a>
29 <li><a href="#ci_warningerrors">Treat Compiler Warnings Like
31 <li><a href="#ci_portable_code">Write Portable Code</a></li>
32 <li><a href="#ci_rtti_exceptions">Do not use RTTI or Exceptions</a></li>
33 <li><a href="#ci_class_struct">Use of <tt>class</tt>/<tt>struct</tt> Keywords</a></li>
36 <li><a href="#styleissues">Style Issues</a>
38 <li><a href="#macro">The High-Level Issues</a>
40 <li><a href="#hl_module">A Public Header File <b>is</b> a
42 <li><a href="#hl_dontinclude"><tt>#include</tt> as Little as Possible</a></li>
43 <li><a href="#hl_privateheaders">Keep "internal" Headers
45 <li><a href="#hl_earlyexit">Use Early Exits and <tt>continue</tt> to Simplify
47 <li><a href="#hl_else_after_return">Don't use <tt>else</tt> after a
48 <tt>return</tt></a></li>
49 <li><a href="#hl_predicateloops">Turn Predicate Loops into Predicate
52 <li><a href="#micro">The Low-Level Issues</a>
54 <li><a href="#ll_naming">Name Types, Functions, Variables, and Enumerators Properly</a></li>
55 <li><a href="#ll_assert">Assert Liberally</a></li>
56 <li><a href="#ll_ns_std">Do not use '<tt>using namespace std</tt>'</a></li>
57 <li><a href="#ll_virtual_anch">Provide a virtual method anchor for
58 classes in headers</a></li>
59 <li><a href="#ll_end">Don't evaluate <tt>end()</tt> every time through a
61 <li><a href="#ll_iostream"><tt>#include <iostream></tt> is
62 <em>forbidden</em></a></li>
63 <li><a href="#ll_raw_ostream">Use <tt>raw_ostream</tt></a></li>
64 <li><a href="#ll_avoidendl">Avoid <tt>std::endl</tt></a></li>
67 <li><a href="#nano">Microscopic Details</a>
69 <li><a href="#micro_spaceparen">Spaces Before Parentheses</a></li>
70 <li><a href="#micro_preincrement">Prefer Preincrement</a></li>
71 <li><a href="#micro_namespaceindent">Namespace Indentation</a></li>
72 <li><a href="#micro_anonns">Anonymous Namespaces</a></li>
77 <li><a href="#seealso">See Also</a></li>
80 <div class="doc_author">
81 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
85 <!-- *********************************************************************** -->
87 <a name="introduction">Introduction</a>
89 <!-- *********************************************************************** -->
93 <p>This document attempts to describe a few coding standards that are being used
94 in the LLVM source tree. Although no coding standards should be regarded as
95 absolute requirements to be followed in all instances, coding standards can be
98 <p>This document intentionally does not prescribe fixed standards for religious
99 issues such as brace placement and space usage. For issues like this, follow
104 <p><b><a name="goldenrule">If you are adding a significant body of source to a
105 project, feel free to use whatever style you are most comfortable with. If you
106 are extending, enhancing, or bug fixing already implemented code, use the style
107 that is already being used so that the source is uniform and easy to
112 <p>The ultimate goal of these guidelines is the increase readability and
113 maintainability of our common source base. If you have suggestions for topics to
114 be included, please mail them to <a
115 href="mailto:sabre@nondot.org">Chris</a>.</p>
119 <!-- *********************************************************************** -->
121 <a name="mechanicalissues">Mechanical Source Issues</a>
123 <!-- *********************************************************************** -->
127 <!-- ======================================================================= -->
129 <a name="sourceformating">Source Code Formatting</a>
134 <!-- _______________________________________________________________________ -->
136 <a name="scf_commenting">Commenting</a>
141 <p>Comments are one critical part of readability and maintainability. Everyone
142 knows they should comment, so should you. When writing comments, write them as
143 English prose, which means they should use proper capitalization, punctuation,
144 etc. Although we all should probably
145 comment our code more than we do, there are a few very critical places that
146 documentation is very useful:</p>
148 <h5>File Headers</h5>
152 <p>Every source file should have a header on it that describes the basic
153 purpose of the file. If a file does not have a header, it should not be
154 checked into Subversion. Most source trees will probably have a standard
155 file header format. The standard format for the LLVM source tree looks like
158 <div class="doc_code">
160 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
162 // The LLVM Compiler Infrastructure
164 // This file is distributed under the University of Illinois Open Source
165 // License. See LICENSE.TXT for details.
167 //===----------------------------------------------------------------------===//
169 // This file contains the declaration of the Instruction class, which is the
170 // base class for all of the VM instructions.
172 //===----------------------------------------------------------------------===//
176 <p>A few things to note about this particular format: The "<tt>-*- C++
177 -*-</tt>" string on the first line is there to tell Emacs that the source file
178 is a C++ file, not a C file (Emacs assumes <tt>.h</tt> files are C files by default).
179 Note that this tag is not necessary in <tt>.cpp</tt> files. The name of the file is also
180 on the first line, along with a very short description of the purpose of the
181 file. This is important when printing out code and flipping though lots of
184 <p>The next section in the file is a concise note that defines the license
185 that the file is released under. This makes it perfectly clear what terms the
186 source code can be distributed under and should not be modified in any way.</p>
188 <p>The main body of the description does not have to be very long in most cases.
189 Here it's only two lines. If an algorithm is being implemented or something
190 tricky is going on, a reference to the paper where it is published should be
191 included, as well as any notes or "gotchas" in the code to watch out for.</p>
195 <h5>Class overviews</h5>
197 <p>Classes are one fundamental part of a good object oriented design. As such,
198 a class definition should have a comment block that explains what the class is
199 used for... if it's not obvious. If it's so completely obvious your grandma
200 could figure it out, it's probably safe to leave it out. Naming classes
201 something sane goes a long ways towards avoiding writing documentation.</p>
204 <h5>Method information</h5>
208 <p>Methods defined in a class (as well as any global functions) should also be
209 documented properly. A quick note about what it does and a description of the
210 borderline behaviour is all that is necessary here (unless something
211 particularly tricky or insidious is going on). The hope is that people can
212 figure out how to use your interfaces without reading the code itself... that is
215 <p>Good things to talk about here are what happens when something unexpected
216 happens: does the method return null? Abort? Format your hard disk?</p>
222 <!-- _______________________________________________________________________ -->
224 <a name="scf_commentformat">Comment Formatting</a>
229 <p>In general, prefer C++ style (<tt>//</tt>) comments. They take less space,
230 require less typing, don't have nesting problems, etc. There are a few cases
231 when it is useful to use C style (<tt>/* */</tt>) comments however:</p>
234 <li>When writing C code: Obviously if you are writing C code, use C style
236 <li>When writing a header file that may be <tt>#include</tt>d by a C source
238 <li>When writing a source file that is used by a tool that only accepts C
242 <p>To comment out a large block of code, use <tt>#if 0</tt> and <tt>#endif</tt>.
243 These nest properly and are better behaved in general than C style comments.</p>
247 <!-- _______________________________________________________________________ -->
249 <a name="scf_includes"><tt>#include</tt> Style</a>
254 <p>Immediately after the <a href="#scf_commenting">header file comment</a> (and
255 include guards if working on a header file), the <a
256 href="#hl_dontinclude">minimal</a> list of <tt>#include</tt>s required by the
257 file should be listed. We prefer these <tt>#include</tt>s to be listed in this
261 <li><a href="#mmheader">Main Module Header</a></li>
262 <li><a href="#hl_privateheaders">Local/Private Headers</a></li>
263 <li><tt>llvm/*</tt></li>
264 <li><tt>llvm/Analysis/*</tt></li>
265 <li><tt>llvm/Assembly/*</tt></li>
266 <li><tt>llvm/Bitcode/*</tt></li>
267 <li><tt>llvm/CodeGen/*</tt></li>
269 <li><tt>Support/*</tt></li>
270 <li><tt>Config/*</tt></li>
271 <li>System <tt>#includes</tt></li>
274 <p>and each category should be sorted by name.</p>
276 <p><a name="mmheader">The "Main Module Header"</a> file applies to <tt>.cpp</tt> files
277 which implement an interface defined by a <tt>.h</tt> file. This <tt>#include</tt>
278 should always be included <b>first</b> regardless of where it lives on the file
279 system. By including a header file first in the <tt>.cpp</tt> files that implement the
280 interfaces, we ensure that the header does not have any hidden dependencies
281 which are not explicitly #included in the header, but should be. It is also a
282 form of documentation in the <tt>.cpp</tt> file to indicate where the interfaces it
283 implements are defined.</p>
287 <!-- _______________________________________________________________________ -->
289 <a name="scf_codewidth">Source Code Width</a>
294 <p>Write your code to fit within 80 columns of text. This helps those of us who
295 like to print out code and look at your code in an xterm without resizing
298 <p>The longer answer is that there must be some limit to the width of the code
299 in order to reasonably allow developers to have multiple files side-by-side in
300 windows on a modest display. If you are going to pick a width limit, it is
301 somewhat arbitrary but you might as well pick something standard. Going with
302 90 columns (for example) instead of 80 columns wouldn't add any significant
303 value and would be detrimental to printing out code. Also many other projects
304 have standardized on 80 columns, so some people have already configured their
305 editors for it (vs something else, like 90 columns).</p>
307 <p>This is one of many contentious issues in coding standards, but it is not up
312 <!-- _______________________________________________________________________ -->
314 <a name="scf_spacestabs">Use Spaces Instead of Tabs</a>
319 <p>In all cases, prefer spaces to tabs in source files. People have different
320 preferred indentation levels, and different styles of indentation that they
321 like; this is fine. What isn't fine is that different editors/viewers expand
322 tabs out to different tab stops. This can cause your code to look completely
323 unreadable, and it is not worth dealing with.</p>
325 <p>As always, follow the <a href="#goldenrule">Golden Rule</a> above: follow the
326 style of existing code if you are modifying and extending it. If you like four
327 spaces of indentation, <b>DO NOT</b> do that in the middle of a chunk of code
328 with two spaces of indentation. Also, do not reindent a whole source file: it
329 makes for incredible diffs that are absolutely worthless.</p>
333 <!-- _______________________________________________________________________ -->
335 <a name="scf_indentation">Indent Code Consistently</a>
340 <p>Okay, in your first year of programming you were told that indentation is
341 important. If you didn't believe and internalize this then, now is the time.
348 <!-- ======================================================================= -->
350 <a name="compilerissues">Compiler Issues</a>
355 <!-- _______________________________________________________________________ -->
357 <a name="ci_warningerrors">Treat Compiler Warnings Like Errors</a>
362 <p>If your code has compiler warnings in it, something is wrong — you
363 aren't casting values correctly, your have "questionable" constructs in your
364 code, or you are doing something legitimately wrong. Compiler warnings can
365 cover up legitimate errors in output and make dealing with a translation unit
368 <p>It is not possible to prevent all warnings from all compilers, nor is it
369 desirable. Instead, pick a standard compiler (like <tt>gcc</tt>) that provides
370 a good thorough set of warnings, and stick to it. At least in the case of
371 <tt>gcc</tt>, it is possible to work around any spurious errors by changing the
372 syntax of the code slightly. For example, a warning that annoys me occurs when
373 I write code like this:</p>
375 <div class="doc_code">
377 if (V = getValue()) {
383 <p><tt>gcc</tt> will warn me that I probably want to use the <tt>==</tt>
384 operator, and that I probably mistyped it. In most cases, I haven't, and I
385 really don't want the spurious errors. To fix this particular problem, I
386 rewrite the code like this:</p>
388 <div class="doc_code">
390 if ((V = getValue())) {
396 <p>which shuts <tt>gcc</tt> up. Any <tt>gcc</tt> warning that annoys you can
397 be fixed by massaging the code appropriately.</p>
399 <p>These are the <tt>gcc</tt> warnings that I prefer to enable:</p>
401 <div class="doc_code">
403 -Wall -Winline -W -Wwrite-strings -Wno-unused
409 <!-- _______________________________________________________________________ -->
411 <a name="ci_portable_code">Write Portable Code</a>
416 <p>In almost all cases, it is possible and within reason to write completely
417 portable code. If there are cases where it isn't possible to write portable
418 code, isolate it behind a well defined (and well documented) interface.</p>
420 <p>In practice, this means that you shouldn't assume much about the host
421 compiler, and Visual Studio tends to be the lowest common denominator.
422 If advanced features are used, they should only be an implementation detail of
423 a library which has a simple exposed API, and preferably be buried in
428 <!-- _______________________________________________________________________ -->
430 <a name="ci_rtti_exceptions">Do not use RTTI or Exceptions</a>
434 <p>In an effort to reduce code and executable size, LLVM does not use RTTI
435 (e.g. <tt>dynamic_cast<></tt>) or exceptions. These two language features
436 violate the general C++ principle of <i>"you only pay for what you use"</i>,
437 causing executable bloat even if exceptions are never used in the code base, or
438 if RTTI is never used for a class. Because of this, we turn them off globally
441 <p>That said, LLVM does make extensive use of a hand-rolled form of RTTI that
442 use templates like <a href="ProgrammersManual.html#isa"><tt>isa<></tt>,
443 <tt>cast<></tt>, and <tt>dyn_cast<></tt></a>. This form of RTTI is
444 opt-in and can be added to any class. It is also substantially more efficient
445 than <tt>dynamic_cast<></tt>.</p>
449 <!-- _______________________________________________________________________ -->
451 <a name="ci_class_struct">Use of <tt>class</tt> and <tt>struct</tt> Keywords</a>
455 <p>In C++, the <tt>class</tt> and <tt>struct</tt> keywords can be used almost
456 interchangeably. The only difference is when they are used to declare a class:
457 <tt>class</tt> makes all members private by default while <tt>struct</tt> makes
458 all members public by default.</p>
460 <p>Unfortunately, not all compilers follow the rules and some will generate
461 different symbols based on whether <tt>class</tt> or <tt>struct</tt> was used to
462 declare the symbol. This can lead to problems at link time.</p>
464 <p>So, the rule for LLVM is to always use the <tt>class</tt> keyword, unless
465 <b>all</b> members are public and the type is a C++
466 <a href="http://en.wikipedia.org/wiki/Plain_old_data_structure">POD</a> type, in
467 which case <tt>struct</tt> is allowed.</p>
475 <!-- *********************************************************************** -->
477 <a name="styleissues">Style Issues</a>
479 <!-- *********************************************************************** -->
483 <!-- ======================================================================= -->
485 <a name="macro">The High-Level Issues</a>
487 <!-- ======================================================================= -->
491 <!-- _______________________________________________________________________ -->
493 <a name="hl_module">A Public Header File <b>is</b> a Module</a>
498 <p>C++ doesn't do too well in the modularity department. There is no real
499 encapsulation or data hiding (unless you use expensive protocol classes), but it
500 is what we have to work with. When you write a public header file (in the LLVM
501 source tree, they live in the top level "<tt>include</tt>" directory), you are
502 defining a module of functionality.</p>
504 <p>Ideally, modules should be completely independent of each other, and their
505 header files should only <tt>#include</tt> the absolute minimum number of
506 headers possible. A module is not just a class, a function, or a
507 namespace: <a href="http://www.cuj.com/articles/2000/0002/0002c/0002c.htm">it's
508 a collection of these</a> that defines an interface. This interface may be
509 several functions, classes, or data structures, but the important issue is how
510 they work together.</p>
512 <p>In general, a module should be implemented by one or more <tt>.cpp</tt>
513 files. Each of these <tt>.cpp</tt> files should include the header that defines
514 their interface first. This ensures that all of the dependences of the module
515 header have been properly added to the module header itself, and are not
516 implicit. System headers should be included after user headers for a
517 translation unit.</p>
521 <!-- _______________________________________________________________________ -->
523 <a name="hl_dontinclude"><tt>#include</tt> as Little as Possible</a>
528 <p><tt>#include</tt> hurts compile time performance. Don't do it unless you
529 have to, especially in header files.</p>
531 <p>But wait! Sometimes you need to have the definition of a class to use it, or
532 to inherit from it. In these cases go ahead and <tt>#include</tt> that header
533 file. Be aware however that there are many cases where you don't need to have
534 the full definition of a class. If you are using a pointer or reference to a
535 class, you don't need the header file. If you are simply returning a class
536 instance from a prototyped function or method, you don't need it. In fact, for
537 most cases, you simply don't need the definition of a class. And not
538 <tt>#include</tt>'ing speeds up compilation.</p>
540 <p>It is easy to try to go too overboard on this recommendation, however. You
541 <b>must</b> include all of the header files that you are using — you can
542 include them either directly or indirectly (through another header file). To
543 make sure that you don't accidentally forget to include a header file in your
544 module header, make sure to include your module header <b>first</b> in the
545 implementation file (as mentioned above). This way there won't be any hidden
546 dependencies that you'll find out about later.</p>
550 <!-- _______________________________________________________________________ -->
552 <a name="hl_privateheaders">Keep "Internal" Headers Private</a>
557 <p>Many modules have a complex implementation that causes them to use more than
558 one implementation (<tt>.cpp</tt>) file. It is often tempting to put the
559 internal communication interface (helper classes, extra functions, etc) in the
560 public module header file. Don't do this!</p>
562 <p>If you really need to do something like this, put a private header file in
563 the same directory as the source files, and include it locally. This ensures
564 that your private interface remains private and undisturbed by outsiders.</p>
566 <p>Note however, that it's okay to put extra implementation methods in a public
567 class itself. Just make them private (or protected) and all is well.</p>
571 <!-- _______________________________________________________________________ -->
573 <a name="hl_earlyexit">Use Early Exits and <tt>continue</tt> to Simplify Code</a>
578 <p>When reading code, keep in mind how much state and how many previous
579 decisions have to be remembered by the reader to understand a block of code.
580 Aim to reduce indentation where possible when it doesn't make it more difficult
581 to understand the code. One great way to do this is by making use of early
582 exits and the <tt>continue</tt> keyword in long loops. As an example of using
583 an early exit from a function, consider this "bad" code:</p>
585 <div class="doc_code">
587 Value *DoSomething(Instruction *I) {
588 if (!isa<TerminatorInst>(I) &&
589 I->hasOneUse() && SomeOtherThing(I)) {
590 ... some long code ....
598 <p>This code has several problems if the body of the '<tt>if</tt>' is large.
599 When you're looking at the top of the function, it isn't immediately clear that
600 this <em>only</em> does interesting things with non-terminator instructions, and
601 only applies to things with the other predicates. Second, it is relatively
602 difficult to describe (in comments) why these predicates are important because
603 the <tt>if</tt> statement makes it difficult to lay out the comments. Third,
604 when you're deep within the body of the code, it is indented an extra level.
605 Finally, when reading the top of the function, it isn't clear what the result is
606 if the predicate isn't true; you have to read to the end of the function to know
607 that it returns null.</p>
609 <p>It is much preferred to format the code like this:</p>
611 <div class="doc_code">
613 Value *DoSomething(Instruction *I) {
614 // Terminators never need 'something' done to them because ...
615 if (isa<TerminatorInst>(I))
618 // We conservatively avoid transforming instructions with multiple uses
619 // because goats like cheese.
620 if (!I->hasOneUse())
623 // This is really just here for example.
624 if (!SomeOtherThing(I))
627 ... some long code ....
632 <p>This fixes these problems. A similar problem frequently happens in <tt>for</tt>
633 loops. A silly example is something like this:</p>
635 <div class="doc_code">
637 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
638 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
639 Value *LHS = BO->getOperand(0);
640 Value *RHS = BO->getOperand(1);
649 <p>When you have very, very small loops, this sort of structure is fine. But if
650 it exceeds more than 10-15 lines, it becomes difficult for people to read and
651 understand at a glance. The problem with this sort of code is that it gets very
652 nested very quickly. Meaning that the reader of the code has to keep a lot of
653 context in their brain to remember what is going immediately on in the loop,
654 because they don't know if/when the <tt>if</tt> conditions will have elses etc.
655 It is strongly preferred to structure the loop like this:</p>
657 <div class="doc_code">
659 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
660 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
663 Value *LHS = BO->getOperand(0);
664 Value *RHS = BO->getOperand(1);
665 if (LHS == RHS) continue;
672 <p>This has all the benefits of using early exits for functions: it reduces
673 nesting of the loop, it makes it easier to describe why the conditions are true,
674 and it makes it obvious to the reader that there is no <tt>else</tt> coming up
675 that they have to push context into their brain for. If a loop is large, this
676 can be a big understandability win.</p>
680 <!-- _______________________________________________________________________ -->
682 <a name="hl_else_after_return">Don't use <tt>else</tt> after a <tt>return</tt></a>
687 <p>For similar reasons above (reduction of indentation and easier reading),
688 please do not use '<tt>else</tt>' or '<tt>else if</tt>' after something that
689 interrupts control flow — like <tt>return</tt>, <tt>break</tt>,
690 <tt>continue</tt>, <tt>goto</tt>, etc. For example, this is <em>bad</em>:</p>
692 <div class="doc_code">
696 Type = Context.getsigjmp_bufType();
698 Error = ASTContext::GE_Missing_sigjmp_buf;
704 Type = Context.getjmp_bufType();
706 Error = ASTContext::GE_Missing_jmp_buf;
717 <p>It is better to write it like this:</p>
719 <div class="doc_code">
723 Type = Context.getsigjmp_bufType();
725 Error = ASTContext::GE_Missing_sigjmp_buf;
729 Type = Context.getjmp_bufType();
731 Error = ASTContext::GE_Missing_jmp_buf;
739 <p>Or better yet (in this case) as:</p>
741 <div class="doc_code">
745 Type = Context.getsigjmp_bufType();
747 Type = Context.getjmp_bufType();
750 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
751 ASTContext::GE_Missing_jmp_buf;
758 <p>The idea is to reduce indentation and the amount of code you have to keep
759 track of when reading the code.</p>
763 <!-- _______________________________________________________________________ -->
765 <a name="hl_predicateloops">Turn Predicate Loops into Predicate Functions</a>
770 <p>It is very common to write small loops that just compute a boolean value.
771 There are a number of ways that people commonly write these, but an example of
772 this sort of thing is:</p>
774 <div class="doc_code">
776 <b>bool FoundFoo = false;</b>
777 for (unsigned i = 0, e = BarList.size(); i != e; ++i)
778 if (BarList[i]->isFoo()) {
779 <b>FoundFoo = true;</b>
783 <b>if (FoundFoo) {</b>
789 <p>This sort of code is awkward to write, and is almost always a bad sign.
790 Instead of this sort of loop, we strongly prefer to use a predicate function
791 (which may be <a href="#micro_anonns">static</a>) that uses
792 <a href="#hl_earlyexit">early exits</a> to compute the predicate. We prefer
793 the code to be structured like this:</p>
795 <div class="doc_code">
797 /// ListContainsFoo - Return true if the specified list has an element that is
799 static bool ListContainsFoo(const std::vector<Bar*> &List) {
800 for (unsigned i = 0, e = List.size(); i != e; ++i)
801 if (List[i]->isFoo())
807 <b>if (ListContainsFoo(BarList)) {</b>
813 <p>There are many reasons for doing this: it reduces indentation and factors out
814 code which can often be shared by other code that checks for the same predicate.
815 More importantly, it <em>forces you to pick a name</em> for the function, and
816 forces you to write a comment for it. In this silly example, this doesn't add
817 much value. However, if the condition is complex, this can make it a lot easier
818 for the reader to understand the code that queries for this predicate. Instead
819 of being faced with the in-line details of how we check to see if the BarList
820 contains a foo, we can trust the function name and continue reading with better
827 <!-- ======================================================================= -->
829 <a name="micro">The Low-Level Issues</a>
831 <!-- ======================================================================= -->
835 <!-- _______________________________________________________________________ -->
838 Name Types, Functions, Variables, and Enumerators Properly
844 <p>Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
845 enough how important it is to use <em>descriptive</em> names. Pick names that
846 match the semantics and role of the underlying entities, within reason. Avoid
847 abbreviations unless they are well known. After picking a good name, make sure
848 to use consistent capitalization for the name, as inconsistency requires clients
849 to either memorize the APIs or to look it up to find the exact spelling.</p>
851 <p>In general, names should be in camel case (e.g. <tt>TextFileReader</tt>
852 and <tt>isLValue()</tt>). Different kinds of declarations have different
856 <li><p><b>Type names</b> (including classes, structs, enums, typedefs, etc)
857 should be nouns and start with an upper-case letter (e.g.
858 <tt>TextFileReader</tt>).</p></li>
860 <li><p><b>Variable names</b> should be nouns (as they represent state). The
861 name should be camel case, and start with an upper case letter (e.g.
862 <tt>Leader</tt> or <tt>Boats</tt>).</p></li>
864 <li><p><b>Function names</b> should be verb phrases (as they represent
865 actions), and command-like function should be imperative. The name should
866 be camel case, and start with a lower case letter (e.g. <tt>openFile()</tt>
867 or <tt>isFoo()</tt>).</p></li>
869 <li><p><b>Enum declarations</b> (e.g. <tt>enum Foo {...}</tt>) are types, so
870 they should follow the naming conventions for types. A common use for enums
871 is as a discriminator for a union, or an indicator of a subclass. When an
872 enum is used for something like this, it should have a <tt>Kind</tt> suffix
873 (e.g. <tt>ValueKind</tt>).</p></li>
875 <li><p><b>Enumerators</b> (e.g. <tt>enum { Foo, Bar }</tt>) and <b>public member
876 variables</b> should start with an upper-case letter, just like types.
877 Unless the enumerators are defined in their own small namespace or inside a
878 class, enumerators should have a prefix corresponding to the enum
879 declaration name. For example, <tt>enum ValueKind { ... };</tt> may contain
880 enumerators like <tt>VK_Argument</tt>, <tt>VK_BasicBlock</tt>, etc.
881 Enumerators that are just convenience constants are exempt from the
882 requirement for a prefix. For instance:</p>
884 <div class="doc_code">
896 <p>As an exception, classes that mimic STL classes can have member names in
897 STL's style of lower-case words separated by underscores (e.g. <tt>begin()</tt>,
898 <tt>push_back()</tt>, and <tt>empty()</tt>).</p>
900 <p>Here are some examples of good and bad names:</p>
902 <div class="doc_code">
906 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
907 Factory<Tire> Factory; // Better.
908 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
909 // kind of factories.
912 Vehicle MakeVehicle(VehicleType Type) {
913 VehicleMaker M; // Might be OK if having a short life-span.
914 Tire tmp1 = M.makeTire(); // Bad -- 'tmp1' provides no information.
915 Light headlight = M.makeLight("head"); // Good -- descriptive.
924 <!-- _______________________________________________________________________ -->
926 <a name="ll_assert">Assert Liberally</a>
931 <p>Use the "<tt>assert</tt>" macro to its fullest. Check all of your
932 preconditions and assumptions, you never know when a bug (not necessarily even
933 yours) might be caught early by an assertion, which reduces debugging time
934 dramatically. The "<tt><cassert></tt>" header file is probably already
935 included by the header files you are using, so it doesn't cost anything to use
938 <p>To further assist with debugging, make sure to put some kind of error message
939 in the assertion statement, which is printed if the assertion is tripped. This
940 helps the poor debugger make sense of why an assertion is being made and
941 enforced, and hopefully what to do about it. Here is one complete example:</p>
943 <div class="doc_code">
945 inline Value *getOperand(unsigned i) {
946 assert(i < Operands.size() && "getOperand() out of range!");
952 <p>Here are more examples:</p>
954 <div class="doc_code">
956 assert(Ty->isPointerType() && "Can't allocate a non pointer type!");
958 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
960 assert(idx < getNumSuccessors() && "Successor # out of range!");
962 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
964 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
968 <p>You get the idea.</p>
970 <p>Please be aware that, when adding assert statements, not all compilers are aware of
971 the semantics of the assert. In some places, asserts are used to indicate a piece of
972 code that should not be reached. These are typically of the form:</p>
974 <div class="doc_code">
976 assert(0 && "Some helpful error message");
980 <p>When used in a function that returns a value, they should be followed with a return
981 statement and a comment indicating that this line is never reached. This will prevent
982 a compiler which is unable to deduce that the assert statement never returns from
983 generating a warning.</p>
985 <div class="doc_code">
987 assert(0 && "Some helpful error message");
993 <p>Another issue is that values used only by assertions will produce an "unused
994 value" warning when assertions are disabled. For example, this code will
997 <div class="doc_code">
999 unsigned Size = V.size();
1000 assert(Size > 42 && "Vector smaller than it should be");
1002 bool NewToSet = Myset.insert(Value);
1003 assert(NewToSet && "The value shouldn't be in the set yet");
1007 <p>These are two interesting different cases. In the first case, the call to
1008 V.size() is only useful for the assert, and we don't want it executed when
1009 assertions are disabled. Code like this should move the call into the assert
1010 itself. In the second case, the side effects of the call must happen whether
1011 the assert is enabled or not. In this case, the value should be cast to void to
1012 disable the warning. To be specific, it is preferred to write the code like
1015 <div class="doc_code">
1017 assert(V.size() > 42 && "Vector smaller than it should be");
1019 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1020 assert(NewToSet && "The value shouldn't be in the set yet");
1027 <!-- _______________________________________________________________________ -->
1029 <a name="ll_ns_std">Do Not Use '<tt>using namespace std</tt>'</a>
1034 <p>In LLVM, we prefer to explicitly prefix all identifiers from the standard
1035 namespace with an "<tt>std::</tt>" prefix, rather than rely on
1036 "<tt>using namespace std;</tt>".</p>
1038 <p> In header files, adding a '<tt>using namespace XXX</tt>' directive pollutes
1039 the namespace of any source file that <tt>#include</tt>s the header. This is
1040 clearly a bad thing.</p>
1042 <p>In implementation files (e.g. <tt>.cpp</tt> files), the rule is more of a stylistic
1043 rule, but is still important. Basically, using explicit namespace prefixes
1044 makes the code <b>clearer</b>, because it is immediately obvious what facilities
1045 are being used and where they are coming from. And <b>more portable</b>, because
1046 namespace clashes cannot occur between LLVM code and other namespaces. The
1047 portability rule is important because different standard library implementations
1048 expose different symbols (potentially ones they shouldn't), and future revisions
1049 to the C++ standard will add more symbols to the <tt>std</tt> namespace. As
1050 such, we never use '<tt>using namespace std;</tt>' in LLVM.</p>
1052 <p>The exception to the general rule (i.e. it's not an exception for
1053 the <tt>std</tt> namespace) is for implementation files. For example, all of
1054 the code in the LLVM project implements code that lives in the 'llvm' namespace.
1055 As such, it is ok, and actually clearer, for the <tt>.cpp</tt> files to have a
1056 '<tt>using namespace llvm;</tt>' directive at the top, after the
1057 <tt>#include</tt>s. This reduces indentation in the body of the file for source
1058 editors that indent based on braces, and keeps the conceptual context cleaner.
1059 The general form of this rule is that any <tt>.cpp</tt> file that implements
1060 code in any namespace may use that namespace (and its parents'), but should not
1065 <!-- _______________________________________________________________________ -->
1067 <a name="ll_virtual_anch">
1068 Provide a Virtual Method Anchor for Classes in Headers
1074 <p>If a class is defined in a header file and has a v-table (either it has
1075 virtual methods or it derives from classes with virtual methods), it must
1076 always have at least one out-of-line virtual method in the class. Without
1077 this, the compiler will copy the vtable and RTTI into every <tt>.o</tt> file
1078 that <tt>#include</tt>s the header, bloating <tt>.o</tt> file sizes and
1079 increasing link times.</p>
1083 <!-- _______________________________________________________________________ -->
1085 <a name="ll_end">Don't evaluate <tt>end()</tt> every time through a loop</a>
1090 <p>Because C++ doesn't have a standard "<tt>foreach</tt>" loop (though it can be
1091 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1092 loops that manually iterate from begin to end on a variety of containers or
1093 through other data structures. One common mistake is to write a loop in this
1096 <div class="doc_code">
1098 BasicBlock *BB = ...
1099 for (BasicBlock::iterator I = BB->begin(); I != <b>BB->end()</b>; ++I)
1104 <p>The problem with this construct is that it evaluates "<tt>BB->end()</tt>"
1105 every time through the loop. Instead of writing the loop like this, we strongly
1106 prefer loops to be written so that they evaluate it once before the loop starts.
1107 A convenient way to do this is like so:</p>
1109 <div class="doc_code">
1111 BasicBlock *BB = ...
1112 for (BasicBlock::iterator I = BB->begin(), E = <b>BB->end()</b>; I != E; ++I)
1117 <p>The observant may quickly point out that these two loops may have different
1118 semantics: if the container (a basic block in this case) is being mutated, then
1119 "<tt>BB->end()</tt>" may change its value every time through the loop and the
1120 second loop may not in fact be correct. If you actually do depend on this
1121 behavior, please write the loop in the first form and add a comment indicating
1122 that you did it intentionally.</p>
1124 <p>Why do we prefer the second form (when correct)? Writing the loop in the
1125 first form has two problems. First it may be less efficient than evaluating it
1126 at the start of the loop. In this case, the cost is probably minor — a
1127 few extra loads every time through the loop. However, if the base expression is
1128 more complex, then the cost can rise quickly. I've seen loops where the end
1129 expression was actually something like: "<tt>SomeMap[x]->end()</tt>" and map
1130 lookups really aren't cheap. By writing it in the second form consistently, you
1131 eliminate the issue entirely and don't even have to think about it.</p>
1133 <p>The second (even bigger) issue is that writing the loop in the first form
1134 hints to the reader that the loop is mutating the container (a fact that a
1135 comment would handily confirm!). If you write the loop in the second form, it
1136 is immediately obvious without even looking at the body of the loop that the
1137 container isn't being modified, which makes it easier to read the code and
1138 understand what it does.</p>
1140 <p>While the second form of the loop is a few extra keystrokes, we do strongly
1145 <!-- _______________________________________________________________________ -->
1147 <a name="ll_iostream"><tt>#include <iostream></tt> is Forbidden</a>
1152 <p>The use of <tt>#include <iostream></tt> in library files is
1153 hereby <b><em>forbidden</em></b>. The primary reason for doing this is to
1154 support clients using LLVM libraries as part of larger systems. In particular,
1155 we statically link LLVM into some dynamic libraries. Even if LLVM isn't used,
1156 the static constructors are run whenever an application starts up that uses the
1157 dynamic library. There are two problems with this:</p>
1160 <li>The time to run the static c'tors impacts startup time of applications
1161 — a critical time for GUI apps.</li>
1163 <li>The static c'tors cause the app to pull many extra pages of memory off the
1164 disk: both the code for the static c'tors in each <tt>.o</tt> file and the
1165 small amount of data that gets touched. In addition, touched/dirty pages
1166 put more pressure on the VM system on low-memory machines.</li>
1169 <p>Note that using the other stream headers (<tt><sstream></tt> for
1170 example) is not problematic in this regard —
1171 just <tt><iostream></tt>. However, <tt>raw_ostream</tt> provides various
1172 APIs that are better performing for almost every use than <tt>std::ostream</tt>
1173 style APIs. <b>Therefore new code should always
1174 use <a href="#ll_raw_ostream"><tt>raw_ostream</tt></a> for writing, or
1175 the <tt>llvm::MemoryBuffer</tt> API for reading files.</b></p>
1180 <!-- _______________________________________________________________________ -->
1182 <a name="ll_raw_ostream">Use <tt>raw_ostream</tt></a>
1187 <p>LLVM includes a lightweight, simple, and efficient stream implementation
1188 in <tt>llvm/Support/raw_ostream.h</tt>, which provides all of the common
1189 features of <tt>std::ostream</tt>. All new code should use <tt>raw_ostream</tt>
1190 instead of <tt>ostream</tt>.</p>
1192 <p>Unlike <tt>std::ostream</tt>, <tt>raw_ostream</tt> is not a template and can
1193 be forward declared as <tt>class raw_ostream</tt>. Public headers should
1194 generally not include the <tt>raw_ostream</tt> header, but use forward
1195 declarations and constant references to <tt>raw_ostream</tt> instances.</p>
1200 <!-- _______________________________________________________________________ -->
1202 <a name="ll_avoidendl">Avoid <tt>std::endl</tt></a>
1207 <p>The <tt>std::endl</tt> modifier, when used with <tt>iostreams</tt> outputs a
1208 newline to the output stream specified. In addition to doing this, however, it
1209 also flushes the output stream. In other words, these are equivalent:</p>
1211 <div class="doc_code">
1213 std::cout << std::endl;
1214 std::cout << '\n' << std::flush;
1218 <p>Most of the time, you probably have no reason to flush the output stream, so
1219 it's better to use a literal <tt>'\n'</tt>.</p>
1225 <!-- ======================================================================= -->
1227 <a name="nano">Microscopic Details</a>
1229 <!-- ======================================================================= -->
1233 <p>This section describes preferred low-level formatting guidelines along with
1234 reasoning on why we prefer them.</p>
1236 <!-- _______________________________________________________________________ -->
1238 <a name="micro_spaceparen">Spaces Before Parentheses</a>
1243 <p>We prefer to put a space before an open parenthesis only in control flow
1244 statements, but not in normal function call expressions and function-like
1245 macros. For example, this is good:</p>
1247 <div class="doc_code">
1250 <b>for (</b>i = 0; i != 100; ++i) ...
1251 <b>while (</b>llvm_rocks) ...
1253 <b>somefunc(</b>42);
1254 <b><a href="#ll_assert">assert</a>(</b>3 != 4 && "laws of math are failing me");
1256 a = <b>foo(</b>42, 92) + <b>bar(</b>x);
1260 <p>and this is bad:</p>
1262 <div class="doc_code">
1265 <b>for(</b>i = 0; i != 100; ++i) ...
1266 <b>while(</b>llvm_rocks) ...
1268 <b>somefunc (</b>42);
1269 <b><a href="#ll_assert">assert</a> (</b>3 != 4 && "laws of math are failing me");
1271 a = <b>foo (</b>42, 92) + <b>bar (</b>x);
1275 <p>The reason for doing this is not completely arbitrary. This style makes
1276 control flow operators stand out more, and makes expressions flow better. The
1277 function call operator binds very tightly as a postfix operator. Putting a
1278 space after a function name (as in the last example) makes it appear that the
1279 code might bind the arguments of the left-hand-side of a binary operator with
1280 the argument list of a function and the name of the right side. More
1281 specifically, it is easy to misread the "a" example as:</p>
1283 <div class="doc_code">
1285 a = foo <b>(</b>(42, 92) + bar<b>)</b> (x);
1289 <p>when skimming through the code. By avoiding a space in a function, we avoid
1290 this misinterpretation.</p>
1294 <!-- _______________________________________________________________________ -->
1296 <a name="micro_preincrement">Prefer Preincrement</a>
1301 <p>Hard fast rule: Preincrement (<tt>++X</tt>) may be no slower than
1302 postincrement (<tt>X++</tt>) and could very well be a lot faster than it. Use
1303 preincrementation whenever possible.</p>
1305 <p>The semantics of postincrement include making a copy of the value being
1306 incremented, returning it, and then preincrementing the "work value". For
1307 primitive types, this isn't a big deal... but for iterators, it can be a huge
1308 issue (for example, some iterators contains stack and set objects in them...
1309 copying an iterator could invoke the copy ctor's of these as well). In general,
1310 get in the habit of always using preincrement, and you won't have a problem.</p>
1314 <!-- _______________________________________________________________________ -->
1316 <a name="micro_namespaceindent">Namespace Indentation</a>
1322 In general, we strive to reduce indentation wherever possible. This is useful
1323 because we want code to <a href="#scf_codewidth">fit into 80 columns</a> without
1324 wrapping horribly, but also because it makes it easier to understand the code.
1325 Namespaces are a funny thing: they are often large, and we often desire to put
1326 lots of stuff into them (so they can be large). Other times they are tiny,
1327 because they just hold an enum or something similar. In order to balance this,
1328 we use different approaches for small versus large namespaces.
1332 If a namespace definition is small and <em>easily</em> fits on a screen (say,
1333 less than 35 lines of code), then you should indent its body. Here's an
1337 <div class="doc_code">
1341 /// RelocationType - An enum for the x86 relocation codes. Note that
1342 /// the terminology here doesn't follow x86 convention - word means
1343 /// 32-bit and dword means 64-bit.
1344 enum RelocationType {
1345 /// reloc_pcrel_word - PC relative relocation, add the relocated value to
1346 /// the value already in memory, after we adjust it for where the PC is.
1347 reloc_pcrel_word = 0,
1349 /// reloc_picrel_word - PIC base relative relocation, add the relocated
1350 /// value to the value already in memory, after we adjust it for where the
1352 reloc_picrel_word = 1,
1354 /// reloc_absolute_word, reloc_absolute_dword - Absolute relocation, just
1355 /// add the relocated value to the value already in memory.
1356 reloc_absolute_word = 2,
1357 reloc_absolute_dword = 3
1364 <p>Since the body is small, indenting adds value because it makes it very clear
1365 where the namespace starts and ends, and it is easy to take the whole thing in
1366 in one "gulp" when reading the code. If the blob of code in the namespace is
1367 larger (as it typically is in a header in the <tt>llvm</tt> or <tt>clang</tt> namespaces), do not
1368 indent the code, and add a comment indicating what namespace is being closed.
1371 <div class="doc_code">
1374 namespace knowledge {
1376 /// Grokable - This class represents things that Smith can have an intimate
1377 /// understanding of and contains the data associated with it.
1381 explicit Grokable() { ... }
1382 virtual ~Grokable() = 0;
1388 } // end namespace knowledge
1389 } // end namespace llvm
1393 <p>Because the class is large, we don't expect that the reader can easily
1394 understand the entire concept in a glance, and the end of the file (where the
1395 namespaces end) may be a long ways away from the place they open. As such,
1396 indenting the contents of the namespace doesn't add any value, and detracts from
1397 the readability of the class. In these cases it is best to <em>not</em> indent
1398 the contents of the namespace.</p>
1402 <!-- _______________________________________________________________________ -->
1404 <a name="micro_anonns">Anonymous Namespaces</a>
1409 <p>After talking about namespaces in general, you may be wondering about
1410 anonymous namespaces in particular.
1411 Anonymous namespaces are a great language feature that tells the C++ compiler
1412 that the contents of the namespace are only visible within the current
1413 translation unit, allowing more aggressive optimization and eliminating the
1414 possibility of symbol name collisions. Anonymous namespaces are to C++ as
1415 "static" is to C functions and global variables. While "static" is available
1416 in C++, anonymous namespaces are more general: they can make entire classes
1417 private to a file.</p>
1419 <p>The problem with anonymous namespaces is that they naturally want to
1420 encourage indentation of their body, and they reduce locality of reference: if
1421 you see a random function definition in a C++ file, it is easy to see if it is
1422 marked static, but seeing if it is in an anonymous namespace requires scanning
1423 a big chunk of the file.</p>
1425 <p>Because of this, we have a simple guideline: make anonymous namespaces as
1426 small as possible, and only use them for class declarations. For example, this
1429 <div class="doc_code">
1436 bool operator<(const char *RHS) const;
1438 <b>} // end anonymous namespace</b>
1440 static void Helper() {
1444 bool StringSort::operator<(const char *RHS) const {
1454 <div class="doc_code">
1461 bool operator<(const char *RHS) const;
1468 bool StringSort::operator<(const char *RHS) const {
1472 <b>} // end anonymous namespace</b>
1478 <p>This is bad specifically because if you're looking at "Helper" in the middle
1479 of a large C++ file, that you have no immediate way to tell if it is local to
1480 the file. When it is marked static explicitly, this is immediately obvious.
1481 Also, there is no reason to enclose the definition of "operator<" in the
1482 namespace just because it was declared there.
1491 <!-- *********************************************************************** -->
1493 <a name="seealso">See Also</a>
1495 <!-- *********************************************************************** -->
1499 <p>A lot of these comments and recommendations have been culled for other
1500 sources. Two particularly important books for our work are:</p>
1504 <li><a href="http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876">Effective
1505 C++</a> by Scott Meyers. Also
1506 interesting and useful are "More Effective C++" and "Effective STL" by the same
1509 <li>Large-Scale C++ Software Design by John Lakos</li>
1513 <p>If you get some free time, and you haven't read them: do so, you might learn
1518 <!-- *********************************************************************** -->
1522 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1523 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1524 <a href="http://validator.w3.org/check/referer"><img
1525 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1527 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1528 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1529 Last modified: $Date$