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>
10 <div class="doc_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_class_struct">Use of class/struct Keywords</a></li>
35 <li><a href="#styleissues">Style Issues</a>
37 <li><a href="#macro">The High Level Issues</a>
39 <li><a href="#hl_module">A Public Header File <b>is</b> a
41 <li><a href="#hl_dontinclude">#include as Little as Possible</a></li>
42 <li><a href="#hl_privateheaders">Keep "internal" Headers
44 <li><a href="#hl_earlyexit">Use Early Exits and 'continue' to Simplify
46 <li><a href="#hl_predicateloops">Turn Predicate Loops into Predicate
49 <li><a href="#micro">The Low Level Issues</a>
51 <li><a href="#ll_assert">Assert Liberally</a></li>
52 <li><a href="#ll_ns_std">Do not use 'using namespace std'</a></li>
53 <li><a href="#ll_virtual_anch">Provide a virtual method anchor for
54 classes in headers</a></li>
55 <li><a href="#ll_end">Don't evaluate end() every time through a
57 <li><a href="#ll_iostream"><tt>#include <iostream></tt> is
58 <em>forbidden</em></a></li>
59 <li><a href="#ll_avoidendl">Avoid <tt>std::endl</tt></a></li>
62 <li><a href="#nano">Microscopic Details</a>
64 <li><a href="#micro_spaceparen">Spaces Before Parentheses</a></li>
65 <li><a href="#micro_preincrement">Prefer Preincrement</a></li>
66 <li><a href="#micro_namespaceindent">Namespace Indentation</a></li>
67 <li><a href="#micro_anonns">Anonymous Namespaces</a></li>
72 <li><a href="#seealso">See Also</a></li>
75 <div class="doc_author">
76 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
80 <!-- *********************************************************************** -->
81 <div class="doc_section">
82 <a name="introduction">Introduction</a>
84 <!-- *********************************************************************** -->
86 <div class="doc_text">
88 <p>This document attempts to describe a few coding standards that are being used
89 in the LLVM source tree. Although no coding standards should be regarded as
90 absolute requirements to be followed in all instances, coding standards can be
93 <p>This document intentionally does not prescribe fixed standards for religious
94 issues such as brace placement and space usage. For issues like this, follow
99 <p><b><a name="goldenrule">If you are adding a significant body of source to a
100 project, feel free to use whatever style you are most comfortable with. If you
101 are extending, enhancing, or bug fixing already implemented code, use the style
102 that is already being used so that the source is uniform and easy to
107 <p>The ultimate goal of these guidelines is the increase readability and
108 maintainability of our common source base. If you have suggestions for topics to
109 be included, please mail them to <a
110 href="mailto:sabre@nondot.org">Chris</a>.</p>
114 <!-- *********************************************************************** -->
115 <div class="doc_section">
116 <a name="mechanicalissues">Mechanical Source Issues</a>
118 <!-- *********************************************************************** -->
120 <!-- ======================================================================= -->
121 <div class="doc_subsection">
122 <a name="sourceformating">Source Code Formatting</a>
125 <!-- _______________________________________________________________________ -->
126 <div class="doc_subsubsection">
127 <a name="scf_commenting">Commenting</a>
130 <div class="doc_text">
132 <p>Comments are one critical part of readability and maintainability. Everyone
133 knows they should comment, so should you. When writing comments, write them as
134 English prose, which means they should use proper capitalization, punctuation,
135 etc. Although we all should probably
136 comment our code more than we do, there are a few very critical places that
137 documentation is very useful:</p>
141 <p>Every source file should have a header on it that describes the basic
142 purpose of the file. If a file does not have a header, it should not be
143 checked into Subversion. Most source trees will probably have a standard
144 file header format. The standard format for the LLVM source tree looks like
147 <div class="doc_code">
149 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
151 // The LLVM Compiler Infrastructure
153 // This file is distributed under the University of Illinois Open Source
154 // License. See LICENSE.TXT for details.
156 //===----------------------------------------------------------------------===//
158 // This file contains the declaration of the Instruction class, which is the
159 // base class for all of the VM instructions.
161 //===----------------------------------------------------------------------===//
165 <p>A few things to note about this particular format: The "<tt>-*- C++
166 -*-</tt>" string on the first line is there to tell Emacs that the source file
167 is a C++ file, not a C file (Emacs assumes .h files are C files by default).
168 Note that this tag is not necessary in .cpp files. The name of the file is also
169 on the first line, along with a very short description of the purpose of the
170 file. This is important when printing out code and flipping though lots of
173 <p>The next section in the file is a concise note that defines the license
174 that the file is released under. This makes it perfectly clear what terms the
175 source code can be distributed under and should not be modified in any way.</p>
177 <p>The main body of the description does not have to be very long in most cases.
178 Here it's only two lines. If an algorithm is being implemented or something
179 tricky is going on, a reference to the paper where it is published should be
180 included, as well as any notes or "gotchas" in the code to watch out for.</p>
182 <b>Class overviews</b>
184 <p>Classes are one fundamental part of a good object oriented design. As such,
185 a class definition should have a comment block that explains what the class is
186 used for... if it's not obvious. If it's so completely obvious your grandma
187 could figure it out, it's probably safe to leave it out. Naming classes
188 something sane goes a long ways towards avoiding writing documentation.</p>
191 <b>Method information</b>
193 <p>Methods defined in a class (as well as any global functions) should also be
194 documented properly. A quick note about what it does any a description of the
195 borderline behaviour is all that is necessary here (unless something
196 particularly tricky or insideous is going on). The hope is that people can
197 figure out how to use your interfaces without reading the code itself... that is
200 <p>Good things to talk about here are what happens when something unexpected
201 happens: does the method return null? Abort? Format your hard disk?</p>
205 <!-- _______________________________________________________________________ -->
206 <div class="doc_subsubsection">
207 <a name="scf_commentformat">Comment Formatting</a>
210 <div class="doc_text">
212 <p>In general, prefer C++ style (<tt>//</tt>) comments. They take less space,
213 require less typing, don't have nesting problems, etc. There are a few cases
214 when it is useful to use C style (<tt>/* */</tt>) comments however:</p>
217 <li>When writing a C code: Obviously if you are writing C code, use C style
219 <li>When writing a header file that may be <tt>#include</tt>d by a C source
221 <li>When writing a source file that is used by a tool that only accepts C
225 <p>To comment out a large block of code, use <tt>#if 0</tt> and <tt>#endif</tt>.
226 These nest properly and are better behaved in general than C style comments.</p>
230 <!-- _______________________________________________________________________ -->
231 <div class="doc_subsubsection">
232 <a name="scf_includes"><tt>#include</tt> Style</a>
235 <div class="doc_text">
237 <p>Immediately after the <a href="#scf_commenting">header file comment</a> (and
238 include guards if working on a header file), the <a
239 href="#hl_dontinclude">minimal</a> list of <tt>#include</tt>s required by the
240 file should be listed. We prefer these <tt>#include</tt>s to be listed in this
244 <li><a href="#mmheader">Main Module header</a></li>
245 <li><a href="#hl_privateheaders">Local/Private Headers</a></li>
246 <li><tt>llvm/*</tt></li>
247 <li><tt>llvm/Analysis/*</tt></li>
248 <li><tt>llvm/Assembly/*</tt></li>
249 <li><tt>llvm/Bytecode/*</tt></li>
250 <li><tt>llvm/CodeGen/*</tt></li>
252 <li><tt>Support/*</tt></li>
253 <li><tt>Config/*</tt></li>
254 <li>System <tt>#includes</tt></li>
257 <p>... and each category should be sorted by name.</p>
259 <p><a name="mmheader">The "Main Module Header"</a> file applies to .cpp file
260 which implement an interface defined by a .h file. This <tt>#include</tt>
261 should always be included <b>first</b> regardless of where it lives on the file
262 system. By including a header file first in the .cpp files that implement the
263 interfaces, we ensure that the header does not have any hidden dependencies
264 which are not explicitly #included in the header, but should be. It is also a
265 form of documentation in the .cpp file to indicate where the interfaces it
266 implements are defined.</p>
270 <!-- _______________________________________________________________________ -->
271 <div class="doc_subsubsection">
272 <a name="scf_codewidth">Source Code Width</a>
275 <div class="doc_text">
277 <p>Write your code to fit within 80 columns of text. This helps those of us who
278 like to print out code and look at your code in an xterm without resizing
281 <p>The longer answer is that there must be some limit to the width of the code
282 in order to reasonably allow developers to have multiple files side-by-side in
283 windows on a modest display. If you are going to pick a width limit, it is
284 somewhat arbitrary but you might as well pick something standard. Going with
285 90 columns (for example) instead of 80 columns wouldn't add any significant
286 value and would be detrimental to printing out code. Also many other projects
287 have standardized on 80 columns, so some people have already configured their
288 editors for it (vs something else, like 90 columns).</p>
290 <p>This is one of many contentious issues in coding standards, but is not up
295 <!-- _______________________________________________________________________ -->
296 <div class="doc_subsubsection">
297 <a name="scf_spacestabs">Use Spaces Instead of Tabs</a>
300 <div class="doc_text">
302 <p>In all cases, prefer spaces to tabs in source files. People have different
303 prefered indentation levels, and different styles of indentation that they
304 like... this is fine. What isn't is that different editors/viewers expand tabs
305 out to different tab stops. This can cause your code to look completely
306 unreadable, and it is not worth dealing with.</p>
308 <p>As always, follow the <a href="#goldenrule">Golden Rule</a> above: follow the
309 style of existing code if your are modifying and extending it. If you like four
310 spaces of indentation, <b>DO NOT</b> do that in the middle of a chunk of code
311 with two spaces of indentation. Also, do not reindent a whole source file: it
312 makes for incredible diffs that are absolutely worthless.</p>
316 <!-- _______________________________________________________________________ -->
317 <div class="doc_subsubsection">
318 <a name="scf_indentation">Indent Code Consistently</a>
321 <div class="doc_text">
323 <p>Okay, your first year of programming you were told that indentation is
324 important. If you didn't believe and internalize this then, now is the time.
330 <!-- ======================================================================= -->
331 <div class="doc_subsection">
332 <a name="compilerissues">Compiler Issues</a>
336 <!-- _______________________________________________________________________ -->
337 <div class="doc_subsubsection">
338 <a name="ci_warningerrors">Treat Compiler Warnings Like Errors</a>
341 <div class="doc_text">
343 <p>If your code has compiler warnings in it, something is wrong: you aren't
344 casting values correctly, your have "questionable" constructs in your code, or
345 you are doing something legitimately wrong. Compiler warnings can cover up
346 legitimate errors in output and make dealing with a translation unit
349 <p>It is not possible to prevent all warnings from all compilers, nor is it
350 desirable. Instead, pick a standard compiler (like <tt>gcc</tt>) that provides
351 a good thorough set of warnings, and stick to them. At least in the case of
352 <tt>gcc</tt>, it is possible to work around any spurious errors by changing the
353 syntax of the code slightly. For example, an warning that annoys me occurs when
354 I write code like this:</p>
356 <div class="doc_code">
358 if (V = getValue()) {
364 <p><tt>gcc</tt> will warn me that I probably want to use the <tt>==</tt>
365 operator, and that I probably mistyped it. In most cases, I haven't, and I
366 really don't want the spurious errors. To fix this particular problem, I
367 rewrite the code like this:</p>
369 <div class="doc_code">
371 if ((V = getValue())) {
377 <p>...which shuts <tt>gcc</tt> up. Any <tt>gcc</tt> warning that annoys you can
378 be fixed by massaging the code appropriately.</p>
380 <p>These are the <tt>gcc</tt> warnings that I prefer to enable: <tt>-Wall
381 -Winline -W -Wwrite-strings -Wno-unused</tt></p>
385 <!-- _______________________________________________________________________ -->
386 <div class="doc_subsubsection">
387 <a name="ci_portable_code">Write Portable Code</a>
390 <div class="doc_text">
392 <p>In almost all cases, it is possible and within reason to write completely
393 portable code. If there are cases where it isn't possible to write portable
394 code, isolate it behind a well defined (and well documented) interface.</p>
396 <p>In practice, this means that you shouldn't assume much about the host
397 compiler, including its support for "high tech" features like partial
398 specialization of templates. If these features are used, they should only be
399 an implementation detail of a library which has a simple exposed API.</p>
403 <!-- _______________________________________________________________________ -->
404 <div class="doc_subsubsection">
405 <a name="ci_class_struct">Use of <tt>class</tt> and <tt>struct</tt> Keywords</a>
407 <div class="doc_text">
409 <p>In C++, the <tt>class</tt> and <tt>struct</tt> keywords can be used almost
410 interchangeably. The only difference is when they are used to declare a class:
411 <tt>class</tt> makes all members private by default while <tt>struct</tt> makes
412 all members public by default.</p>
414 <p>Unfortunately, not all compilers follow the rules and some will generate
415 different symbols based on whether <tt>class</tt> or <tt>struct</tt> was used to
416 declare the symbol. This can lead to problems at link time.</p>
418 <p>So, the rule for LLVM is to always use the <tt>class</tt> keyword, unless
419 <b>all</b> members are public, in which case <tt>struct</tt> is allowed.</p>
423 <!-- *********************************************************************** -->
424 <div class="doc_section">
425 <a name="styleissues">Style Issues</a>
427 <!-- *********************************************************************** -->
430 <!-- ======================================================================= -->
431 <div class="doc_subsection">
432 <a name="macro">The High Level Issues</a>
434 <!-- ======================================================================= -->
437 <!-- _______________________________________________________________________ -->
438 <div class="doc_subsubsection">
439 <a name="hl_module">A Public Header File <b>is</b> a Module</a>
442 <div class="doc_text">
444 <p>C++ doesn't do too well in the modularity department. There is no real
445 encapsulation or data hiding (unless you use expensive protocol classes), but it
446 is what we have to work with. When you write a public header file (in the LLVM
447 source tree, they live in the top level "include" directory), you are defining a
448 module of functionality.</p>
450 <p>Ideally, modules should be completely independent of each other, and their
451 header files should only include the absolute minimum number of headers
452 possible. A module is not just a class, a function, or a namespace: <a
453 href="http://www.cuj.com/articles/2000/0002/0002c/0002c.htm">it's a collection
454 of these</a> that defines an interface. This interface may be several
455 functions, classes or data structures, but the important issue is how they work
458 <p>In general, a module should be implemented with one or more <tt>.cpp</tt>
459 files. Each of these <tt>.cpp</tt> files should include the header that defines
460 their interface first. This ensure that all of the dependences of the module
461 header have been properly added to the module header itself, and are not
462 implicit. System headers should be included after user headers for a
463 translation unit.</p>
467 <!-- _______________________________________________________________________ -->
468 <div class="doc_subsubsection">
469 <a name="hl_dontinclude"><tt>#include</tt> as Little as Possible</a>
472 <div class="doc_text">
474 <p><tt>#include</tt> hurts compile time performance. Don't do it unless you
475 have to, especially in header files.</p>
477 <p>But wait, sometimes you need to have the definition of a class to use it, or
478 to inherit from it. In these cases go ahead and <tt>#include</tt> that header
479 file. Be aware however that there are many cases where you don't need to have
480 the full definition of a class. If you are using a pointer or reference to a
481 class, you don't need the header file. If you are simply returning a class
482 instance from a prototyped function or method, you don't need it. In fact, for
483 most cases, you simply don't need the definition of a class... and not
484 <tt>#include</tt>'ing speeds up compilation.</p>
486 <p>It is easy to try to go too overboard on this recommendation, however. You
487 <b>must</b> include all of the header files that you are using -- you can
488 include them either directly
489 or indirectly (through another header file). To make sure that you don't
490 accidently forget to include a header file in your module header, make sure to
491 include your module header <b>first</b> in the implementation file (as mentioned
492 above). This way there won't be any hidden dependencies that you'll find out
497 <!-- _______________________________________________________________________ -->
498 <div class="doc_subsubsection">
499 <a name="hl_privateheaders">Keep "internal" Headers Private</a>
502 <div class="doc_text">
504 <p>Many modules have a complex implementation that causes them to use more than
505 one implementation (<tt>.cpp</tt>) file. It is often tempting to put the
506 internal communication interface (helper classes, extra functions, etc) in the
507 public module header file. Don't do this.</p>
509 <p>If you really need to do something like this, put a private header file in
510 the same directory as the source files, and include it locally. This ensures
511 that your private interface remains private and undisturbed by outsiders.</p>
513 <p>Note however, that it's okay to put extra implementation methods a public
514 class itself... just make them private (or protected), and all is well.</p>
518 <!-- _______________________________________________________________________ -->
519 <div class="doc_subsubsection">
520 <a name="hl_earlyexit">Use Early Exits and 'continue' to Simplify Code</a>
523 <div class="doc_text">
525 <p>When reading code, keep in mind how much state and how many previous
526 decisions have to be remembered by the reader to understand a block of code.
527 Aim to reduce indentation where possible when it doesn't make it more difficult
528 to understand the code. One great way to do this is by making use of early
529 exits and the 'continue' keyword in long loops. As an example of using an early
530 exit from a function, consider this "bad" code:</p>
532 <div class="doc_code">
534 Value *DoSomething(Instruction *I) {
535 if (!isa<TerminatorInst>(I) &&
536 I->hasOneUse() && SomeOtherThing(I)) {
537 ... some long code ....
545 <p>This code has several problems if the body of the 'if' is large. When you're
546 looking at the top of the function, it isn't immediately clear that this
547 <em>only</em> does interesting things with non-terminator instructions, and only
548 applies to things with the other predicates. Second, it is relatively difficult
549 to describe (in comments) why these predicates are important because the if
550 statement makes it difficult to lay out the comments. Third, when you're deep
551 within the body of the code, it is indented an extra level. Finally, when
552 reading the top of the function, it isn't clear what the result is if the
553 predicate isn't true, you have to read to the end of the function to know that
556 <p>It is much preferred to format the code like this:</p>
558 <div class="doc_code">
560 Value *DoSomething(Instruction *I) {
561 // Terminators never need 'something' done to them because, ...
562 if (isa<TerminatorInst>(I))
565 // We conservatively avoid transforming instructions with multiple uses
566 // because goats like cheese.
567 if (!I->hasOneUse())
570 // This is really just here for example.
571 if (!SomeOtherThing(I))
574 ... some long code ....
579 <p>This fixes these problems. A similar problem frequently happens in for
580 loops. A silly example is something like this:</p>
582 <div class="doc_code">
584 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
585 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
586 Value *LHS = BO->getOperand(0);
587 Value *RHS = BO->getOperand(1);
596 <p>When you have very very small loops, this sort of structure is fine, but if
597 it exceeds more than 10-15 lines, it becomes difficult for people to read and
598 understand at a glance.
599 The problem with this sort of code is that it gets very nested very quickly,
600 meaning that the reader of the code has to keep a lot of context in their brain
601 to remember what is going immediately on in the loop, because they don't know
602 if/when the if conditions will have elses etc. It is strongly preferred to
603 structure the loop like this:</p>
605 <div class="doc_code">
607 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
608 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
611 Value *LHS = BO->getOperand(0);
612 Value *RHS = BO->getOperand(1);
613 if (LHS == RHS) continue;
618 <p>This has all the benefits of using early exits from functions: it reduces
619 nesting of the loop, it makes it easier to describe why the conditions are true,
620 and it makes it obvious to the reader that there is no "else" coming up that
621 they have to push context into their brain for. If a loop is large, this can
622 be a big understandability win.</p>
627 <!-- _______________________________________________________________________ -->
628 <div class="doc_subsubsection">
629 <a name="hl_predicateloops">Turn Predicate Loops into Predicate Functions</a>
632 <div class="doc_text">
634 <p>It is very common to write small loops that just compute a boolean
635 value. There are a number of ways that people commonly write these, but an
636 example of this sort of thing is:</p>
638 <div class="doc_code">
640 <b>bool FoundFoo = false;</b>
641 for (unsigned i = 0, e = BarList.size(); i != e; ++i)
642 if (BarList[i]->isFoo()) {
643 <b>FoundFoo = true;</b>
647 <b>if (FoundFoo) {</b>
653 <p>This sort of code is awkward to write, and is almost always a bad sign.
654 Instead of this sort of loop, we strongly prefer to use a predicate function
655 (which may be <a href="#micro_anonns">static</a>) that uses
656 <a href="#hl_earlyexit">early exits</a> to compute the predicate. We prefer
657 the code to be structured like this:
661 <div class="doc_code">
663 /// ListContainsFoo - Return true if the specified list has an element that is
665 static bool ListContainsFoo(const std::vector<Bar*> &List) {
666 for (unsigned i = 0, e = List.size(); i != e; ++i)
667 if (List[i]->isFoo())
673 <b>if (ListContainsFoo(BarList)) {</b>
679 <p>There are many reasons for doing this: it reduces indentation and factors out
680 code which can often be shared by other code that checks for the same predicate.
681 More importantly, it <em>forces you to pick a name</em> for the function, and
682 forces you to write a comment for it. In this silly example, this doesn't add
683 much value. However, if the condition is complex, this can make it a lot easier
684 for the reader to understand the code that queries for this predicate. Instead
685 of being faced with the in-line details of how we check to see if the BarList
686 contains a foo, we can trust the function name and continue reading with better
692 <!-- ======================================================================= -->
693 <div class="doc_subsection">
694 <a name="micro">The Low Level Issues</a>
696 <!-- ======================================================================= -->
699 <!-- _______________________________________________________________________ -->
700 <div class="doc_subsubsection">
701 <a name="ll_assert">Assert Liberally</a>
704 <div class="doc_text">
706 <p>Use the "<tt>assert</tt>" function to its fullest. Check all of your
707 preconditions and assumptions, you never know when a bug (not neccesarily even
708 yours) might be caught early by an assertion, which reduces debugging time
709 dramatically. The "<tt><cassert></tt>" header file is probably already
710 included by the header files you are using, so it doesn't cost anything to use
713 <p>To further assist with debugging, make sure to put some kind of error message
714 in the assertion statement (which is printed if the assertion is tripped). This
715 helps the poor debugging make sense of why an assertion is being made and
716 enforced, and hopefully what to do about it. Here is one complete example:</p>
718 <div class="doc_code">
720 inline Value *getOperand(unsigned i) {
721 assert(i < Operands.size() && "getOperand() out of range!");
727 <p>Here are some examples:</p>
729 <div class="doc_code">
731 assert(Ty->isPointerType() && "Can't allocate a non pointer type!");
733 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
735 assert(idx < getNumSuccessors() && "Successor # out of range!");
737 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
739 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
743 <p>You get the idea...</p>
745 <p>Please be aware when adding assert statements that not all compilers are aware of
746 the semantics of the assert. In some places, asserts are used to indicate a piece of
747 code that should not be reached. These are typically of the form:</p>
749 <div class="doc_code">
751 assert(0 && "Some helpful error message");
755 <p>When used in a function that returns a value, they should be followed with a return
756 statement and a comment indicating that this line is never reached. This will prevent
757 a compiler which is unable to deduce that the assert statement never returns from
758 generating a warning.</p>
760 <div class="doc_code">
762 assert(0 && "Some helpful error message");
770 <!-- _______________________________________________________________________ -->
771 <div class="doc_subsubsection">
772 <a name="ll_ns_std">Do not use '<tt>using namespace std</tt>'</a>
775 <div class="doc_text">
776 <p>In LLVM, we prefer to explicitly prefix all identifiers from the standard
777 namespace with an "<tt>std::</tt>" prefix, rather than rely on
778 "<tt>using namespace std;</tt>".</p>
780 <p> In header files, adding a '<tt>using namespace XXX</tt>' directive pollutes
781 the namespace of any source file that <tt>#include</tt>s the header. This is
782 clearly a bad thing.</p>
784 <p>In implementation files (e.g. .cpp files), the rule is more of a stylistic
785 rule, but is still important. Basically, using explicit namespace prefixes
786 makes the code <b>clearer</b>, because it is immediately obvious what facilities
787 are being used and where they are coming from, and <b>more portable</b>, because
788 namespace clashes cannot occur between LLVM code and other namespaces. The
789 portability rule is important because different standard library implementations
790 expose different symbols (potentially ones they shouldn't), and future revisions
791 to the C++ standard will add more symbols to the <tt>std</tt> namespace. As
792 such, we never use '<tt>using namespace std;</tt>' in LLVM.</p>
794 <p>The exception to the general rule (i.e. it's not an exception for
795 the <tt>std</tt> namespace) is for implementation files. For example, all of
796 the code in the LLVM project implements code that lives in the 'llvm' namespace.
797 As such, it is ok, and actually clearer, for the .cpp files to have a '<tt>using
798 namespace llvm</tt>' directive at their top, after the <tt>#include</tt>s. The
799 general form of this rule is that any .cpp file that implements code in any
800 namespace may use that namespace (and its parents'), but should not use any
805 <!-- _______________________________________________________________________ -->
806 <div class="doc_subsubsection">
807 <a name="ll_virtual_anch">Provide a virtual method anchor for classes
811 <div class="doc_text">
813 <p>If a class is defined in a header file and has a v-table (either it has
814 virtual methods or it derives from classes with virtual methods), it must
815 always have at least one out-of-line virtual method in the class. Without
816 this, the compiler will copy the vtable and RTTI into every <tt>.o</tt> file
817 that <tt>#include</tt>s the header, bloating <tt>.o</tt> file sizes and
818 increasing link times.</p>
822 <!-- _______________________________________________________________________ -->
823 <div class="doc_subsubsection">
824 <a name="ll_end">Don't evaluate end() every time through a loop</a>
827 <div class="doc_text">
829 <p>Because C++ doesn't have a standard "foreach" loop (though it can be emulated
830 with macros and may be coming in C++'0x) we end up writing a lot of loops that
831 manually iterate from begin to end on a variety of containers or through other
832 data structures. One common mistake is to write a loop in this style:</p>
834 <div class="doc_code">
837 for (BasicBlock::iterator I = BB->begin(); I != <b>BB->end()</b>; ++I)
842 <p>The problem with this construct is that it evaluates "<tt>BB->end()</tt>"
843 every time through the loop. Instead of writing the loop like this, we strongly
844 prefer loops to be written so that they evaluate it once before the loop starts.
845 A convenient way to do this is like so:</p>
847 <div class="doc_code">
850 for (BasicBlock::iterator I = BB->begin(), E = <b>BB->end()</b>; I != E; ++I)
855 <p>The observant may quickly point out that these two loops may have different
856 semantics: if the container (a basic block in this case) is being mutated, then
857 "<tt>BB->end()</tt>" may change its value every time through the loop and the
858 second loop may not in fact be correct. If you actually do depend on this
859 behavior, please write the loop in the first form and add a comment indicating
860 that you did it intentionally.</p>
862 <p>Why do we prefer the second form (when correct)? Writing the loop in the
863 first form has two problems: First it may be less efficient than evaluating it
864 at the start of the loop. In this case, the cost is probably minor: a few extra
865 loads every time through the loop. However, if the base expression is more
866 complex, then the cost can rise quickly. I've seen loops where the end
867 expression was actually something like: "<tt>SomeMap[x]->end()</tt>" and map
868 lookups really aren't cheap. By writing it in the second form consistently, you
869 eliminate the issue entirely and don't even have to think about it.</p>
871 <p>The second (even bigger) issue is that writing the loop in the first form
872 hints to the reader that the loop is mutating the container (a fact that a
873 comment would handily confirm!). If you write the loop in the second form, it
874 is immediately obvious without even looking at the body of the loop that the
875 container isn't being modified, which makes it easier to read the code and
876 understand what it does.</p>
878 <p>While the second form of the loop is a few extra keystrokes, we do strongly
883 <!-- _______________________________________________________________________ -->
884 <div class="doc_subsubsection">
885 <a name="ll_iostream"><tt>#include <iostream></tt> is forbidden</a>
888 <div class="doc_text">
890 <p>The use of <tt>#include <iostream></tt> in library files is
891 hereby <b><em>forbidden</em></b>. The primary reason for doing this is to
892 support clients using LLVM libraries as part of larger systems. In particular,
893 we statically link LLVM into some dynamic libraries. Even if LLVM isn't used,
894 the static c'tors are run whenever an application start up that uses the dynamic
895 library. There are two problems with this:</p>
898 <li>The time to run the static c'tors impacts startup time of
899 applications—a critical time for GUI apps.</li>
900 <li>The static c'tors cause the app to pull many extra pages of memory off the
901 disk: both the code for the static c'tors in each <tt>.o</tt> file and the
902 small amount of data that gets touched. In addition, touched/dirty pages
903 put more pressure on the VM system on low-memory machines.</li>
906 <p>Note that using the other stream headers (<tt><sstream></tt> for
907 example) is allowed normally, it is just <tt><iostream></tt> that is
908 causing problems.</p>
910 <p>The preferred replacement for stream functionality is the
911 <tt>llvm::raw_ostream</tt> class (for writing to output streams of various
912 sorts) and the <tt>llvm::MemoryBuffer</tt> API (for reading in files).</p>
917 <!-- _______________________________________________________________________ -->
918 <div class="doc_subsubsection">
919 <a name="ll_avoidendl">Avoid <tt>std::endl</tt></a>
922 <div class="doc_text">
924 <p>The <tt>std::endl</tt> modifier, when used with iostreams outputs a newline
925 to the output stream specified. In addition to doing this, however, it also
926 flushes the output stream. In other words, these are equivalent:</p>
928 <div class="doc_code">
930 std::cout << std::endl;
931 std::cout << '\n' << std::flush;
935 <p>Most of the time, you probably have no reason to flush the output stream, so
936 it's better to use a literal <tt>'\n'</tt>.</p>
941 <!-- ======================================================================= -->
942 <div class="doc_subsection">
943 <a name="nano">Microscopic Details</a>
945 <!-- ======================================================================= -->
947 <p>This section describes preferred low-level formatting guidelines along with
948 reasoning on why we prefer them.</p>
950 <!-- _______________________________________________________________________ -->
951 <div class="doc_subsubsection">
952 <a name="micro_spaceparen">Spaces Before Parentheses</a>
955 <div class="doc_text">
957 <p>We prefer to put a space before a parentheses only in control flow
958 statements, but not in normal function call expressions and function-like
959 macros. For example, this is good:</p>
961 <div class="doc_code">
964 <b>for (</b>i = 0; i != 100; ++i) ...
965 <b>while (</b>llvm_rocks) ...
968 <b><a href="#ll_assert">assert</a>(</b>3 != 4 && "laws of math are failing me");
970 a = <b>foo(</b>42, 92) + <b>bar(</b>x);
974 <p>... and this is bad:</p>
976 <div class="doc_code">
979 <b>for(</b>i = 0; i != 100; ++i) ...
980 <b>while(</b>llvm_rocks) ...
982 <b>somefunc (</b>42);
983 <b><a href="#ll_assert">assert</a> (</b>3 != 4 && "laws of math are failing me");
985 a = <b>foo (</b>42, 92) + <b>bar (</b>x);
989 <p>The reason for doing this is not completely arbitrary. This style makes
990 control flow operators stand out more, and makes expressions flow better. The
991 function call operator binds very tightly as a postfix operator. Putting
992 a space after a function name (as in the last example) makes it appear that
993 the code might bind the arguments of the left-hand-side of a binary operator
994 with the argument list of a function and the name of the right side. More
995 specifically, it is easy to misread the "a" example as:</p>
997 <div class="doc_code">
999 a = foo <b>(</b>(42, 92) + bar<b>)</b> (x);
1003 <p>... when skimming through the code. By avoiding a space in a function, we
1004 avoid this misinterpretation.</p>
1008 <!-- _______________________________________________________________________ -->
1009 <div class="doc_subsubsection">
1010 <a name="micro_preincrement">Prefer Preincrement</a>
1013 <div class="doc_text">
1015 <p>Hard fast rule: Preincrement (<tt>++X</tt>) may be no slower than
1016 postincrement (<tt>X++</tt>) and could very well be a lot faster than it. Use
1017 preincrementation whenever possible.</p>
1019 <p>The semantics of postincrement include making a copy of the value being
1020 incremented, returning it, and then preincrementing the "work value". For
1021 primitive types, this isn't a big deal... but for iterators, it can be a huge
1022 issue (for example, some iterators contains stack and set objects in them...
1023 copying an iterator could invoke the copy ctor's of these as well). In general,
1024 get in the habit of always using preincrement, and you won't have a problem.</p>
1028 <!-- _______________________________________________________________________ -->
1029 <div class="doc_subsubsection">
1030 <a name="micro_namespaceindent">Namespace Indentation</a>
1033 <div class="doc_text">
1036 In general, we strive to reduce indentation where ever possible. This is useful
1037 because we want code to <a href="#scf_codewidth">fit into 80 columns</a> without
1038 wrapping horribly, but also because it makes it easier to understand the code.
1039 Namespaces are a funny thing: they are often large, and we often desire to put
1040 lots of stuff into them (so they can be large). Other times they are tiny,
1041 because they just hold an enum or something similar. In order to balance this,
1042 we use different approaches for small versus large namespaces.
1046 If a namespace definition is small and <em>easily</em> fits on a screen (say,
1047 less than 35 lines of code), then you should indent its body. Here's an
1051 <div class="doc_code">
1055 /// RelocationType - An enum for the x86 relocation codes. Note that
1056 /// the terminology here doesn't follow x86 convention - word means
1057 /// 32-bit and dword means 64-bit.
1058 enum RelocationType {
1059 /// reloc_pcrel_word - PC relative relocation, add the relocated value to
1060 /// the value already in memory, after we adjust it for where the PC is.
1061 reloc_pcrel_word = 0,
1063 /// reloc_picrel_word - PIC base relative relocation, add the relocated
1064 /// value to the value already in memory, after we adjust it for where the
1066 reloc_picrel_word = 1,
1068 /// reloc_absolute_word, reloc_absolute_dword - Absolute relocation, just
1069 /// add the relocated value to the value already in memory.
1070 reloc_absolute_word = 2,
1071 reloc_absolute_dword = 3
1078 <p>Since the body is small, indenting adds value because it makes it very clear
1079 where the namespace starts and ends, and it is easy to take the whole thing in
1080 in one "gulp" when reading the code. If the blob of code in the namespace is
1081 larger (as it typically is in a header in the llvm or clang namespaces), do not
1082 indent the code, and add a comment indicating what namespace is being closed.
1085 <div class="doc_code">
1088 namespace knowledge {
1090 /// Grokable - This class represents things that Smith can have an intimate
1091 /// understanding of and contains the data associated with it.
1095 explicit Grokable() { ... }
1096 virtual ~Grokable() = 0;
1102 } // end namespace knowledge
1103 } // end namespace llvm
1107 <p>Because the class is large, we don't expect that the reader can easily
1108 understand the entire concept in a glance, and the end of the file (where the
1109 namespaces end) may be a long ways away from the place they open. As such,
1110 indenting the contents of the namespace doesn't add any value, and detracts from
1111 the readability of the class. In these cases it is best to <em>not</em> indent
1112 the contents of the namespace.</p>
1116 <!-- _______________________________________________________________________ -->
1117 <div class="doc_subsubsection">
1118 <a name="micro_anonns">Anonymous Namespaces</a>
1121 <div class="doc_text">
1123 <p>After talking about namespaces in general, you may be wondering about
1124 anonymous namespaces in particular.
1125 Anonymous namespaces are a great language feature that tells the C++ compiler
1126 that the contents of the namespace are only visible within the current
1127 translation unit, allowing more aggressive optimization and eliminating the
1128 possibility of symbol name collisions. Anonymous namespaces are to C++ as
1129 "static" is to C functions and global variables. While "static" is available
1130 in C++, anonymous namespaces are more general: they can make entire classes
1131 private to a file.</p>
1133 <p>The problem with anonymous namespaces is that they naturally want to
1134 encourage indentation of their body, and they reduce locality of reference: if
1135 you see a random function definition in a C++ file, it is easy to see if it is
1136 marked static, but seeing if it is in an anonymous namespace requires scanning
1137 a big chunk of the file.</p>
1139 <p>Because of this, we have a simple guideline: make anonymous namespaces as
1140 small as possible, and only use them for class declarations. For example, this
1143 <div class="doc_code">
1150 bool operator<(const char *RHS) const;
1152 <b>} // end anonymous namespace</b>
1154 static void Helper() {
1158 bool StringSort::operator<(const char *RHS) const {
1168 <div class="doc_code">
1175 bool operator<(const char *RHS) const;
1182 bool StringSort::operator<(const char *RHS) const {
1186 <b>} // end anonymous namespace</b>
1192 <p>This is bad specifically because if you're looking at "Helper" in the middle
1193 of a large C++ file, that you have no immediate way to tell if it is local to
1194 the file. When it is marked static explicitly, this is immediately obvious.
1195 Also, there is no reason to enclose the definition of "operator<" in the
1196 namespace just because it was declared there.
1203 <!-- *********************************************************************** -->
1204 <div class="doc_section">
1205 <a name="seealso">See Also</a>
1207 <!-- *********************************************************************** -->
1209 <div class="doc_text">
1211 <p>A lot of these comments and recommendations have been culled for other
1212 sources. Two particularly important books for our work are:</p>
1216 <li><a href="http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876">Effective
1217 C++</a> by Scott Meyers. Also
1218 interesting and useful are "More Effective C++" and "Effective STL" by the same
1221 <li>Large-Scale C++ Software Design by John Lakos</li>
1225 <p>If you get some free time, and you haven't read them: do so, you might learn
1230 <!-- *********************************************************************** -->
1234 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1235 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1236 <a href="http://validator.w3.org/check/referer"><img
1237 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1239 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1240 <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
1241 Last modified: $Date$