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>
60 <li><a href="#ll_raw_ostream">Use <tt>raw_ostream</tt></a</li>
63 <li><a href="#nano">Microscopic Details</a>
65 <li><a href="#micro_spaceparen">Spaces Before Parentheses</a></li>
66 <li><a href="#micro_preincrement">Prefer Preincrement</a></li>
67 <li><a href="#micro_namespaceindent">Namespace Indentation</a></li>
68 <li><a href="#micro_anonns">Anonymous Namespaces</a></li>
73 <li><a href="#seealso">See Also</a></li>
76 <div class="doc_author">
77 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
81 <!-- *********************************************************************** -->
82 <div class="doc_section">
83 <a name="introduction">Introduction</a>
85 <!-- *********************************************************************** -->
87 <div class="doc_text">
89 <p>This document attempts to describe a few coding standards that are being used
90 in the LLVM source tree. Although no coding standards should be regarded as
91 absolute requirements to be followed in all instances, coding standards can be
94 <p>This document intentionally does not prescribe fixed standards for religious
95 issues such as brace placement and space usage. For issues like this, follow
100 <p><b><a name="goldenrule">If you are adding a significant body of source to a
101 project, feel free to use whatever style you are most comfortable with. If you
102 are extending, enhancing, or bug fixing already implemented code, use the style
103 that is already being used so that the source is uniform and easy to
108 <p>The ultimate goal of these guidelines is the increase readability and
109 maintainability of our common source base. If you have suggestions for topics to
110 be included, please mail them to <a
111 href="mailto:sabre@nondot.org">Chris</a>.</p>
115 <!-- *********************************************************************** -->
116 <div class="doc_section">
117 <a name="mechanicalissues">Mechanical Source Issues</a>
119 <!-- *********************************************************************** -->
121 <!-- ======================================================================= -->
122 <div class="doc_subsection">
123 <a name="sourceformating">Source Code Formatting</a>
126 <!-- _______________________________________________________________________ -->
127 <div class="doc_subsubsection">
128 <a name="scf_commenting">Commenting</a>
131 <div class="doc_text">
133 <p>Comments are one critical part of readability and maintainability. Everyone
134 knows they should comment, so should you. When writing comments, write them as
135 English prose, which means they should use proper capitalization, punctuation,
136 etc. Although we all should probably
137 comment our code more than we do, there are a few very critical places that
138 documentation is very useful:</p>
142 <p>Every source file should have a header on it that describes the basic
143 purpose of the file. If a file does not have a header, it should not be
144 checked into Subversion. Most source trees will probably have a standard
145 file header format. The standard format for the LLVM source tree looks like
148 <div class="doc_code">
150 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
152 // The LLVM Compiler Infrastructure
154 // This file is distributed under the University of Illinois Open Source
155 // License. See LICENSE.TXT for details.
157 //===----------------------------------------------------------------------===//
159 // This file contains the declaration of the Instruction class, which is the
160 // base class for all of the VM instructions.
162 //===----------------------------------------------------------------------===//
166 <p>A few things to note about this particular format: The "<tt>-*- C++
167 -*-</tt>" string on the first line is there to tell Emacs that the source file
168 is a C++ file, not a C file (Emacs assumes .h files are C files by default).
169 Note that this tag is not necessary in .cpp files. The name of the file is also
170 on the first line, along with a very short description of the purpose of the
171 file. This is important when printing out code and flipping though lots of
174 <p>The next section in the file is a concise note that defines the license
175 that the file is released under. This makes it perfectly clear what terms the
176 source code can be distributed under and should not be modified in any way.</p>
178 <p>The main body of the description does not have to be very long in most cases.
179 Here it's only two lines. If an algorithm is being implemented or something
180 tricky is going on, a reference to the paper where it is published should be
181 included, as well as any notes or "gotchas" in the code to watch out for.</p>
183 <b>Class overviews</b>
185 <p>Classes are one fundamental part of a good object oriented design. As such,
186 a class definition should have a comment block that explains what the class is
187 used for... if it's not obvious. If it's so completely obvious your grandma
188 could figure it out, it's probably safe to leave it out. Naming classes
189 something sane goes a long ways towards avoiding writing documentation.</p>
192 <b>Method information</b>
194 <p>Methods defined in a class (as well as any global functions) should also be
195 documented properly. A quick note about what it does any a description of the
196 borderline behaviour is all that is necessary here (unless something
197 particularly tricky or insideous is going on). The hope is that people can
198 figure out how to use your interfaces without reading the code itself... that is
201 <p>Good things to talk about here are what happens when something unexpected
202 happens: does the method return null? Abort? Format your hard disk?</p>
206 <!-- _______________________________________________________________________ -->
207 <div class="doc_subsubsection">
208 <a name="scf_commentformat">Comment Formatting</a>
211 <div class="doc_text">
213 <p>In general, prefer C++ style (<tt>//</tt>) comments. They take less space,
214 require less typing, don't have nesting problems, etc. There are a few cases
215 when it is useful to use C style (<tt>/* */</tt>) comments however:</p>
218 <li>When writing a C code: Obviously if you are writing C code, use C style
220 <li>When writing a header file that may be <tt>#include</tt>d by a C source
222 <li>When writing a source file that is used by a tool that only accepts C
226 <p>To comment out a large block of code, use <tt>#if 0</tt> and <tt>#endif</tt>.
227 These nest properly and are better behaved in general than C style comments.</p>
231 <!-- _______________________________________________________________________ -->
232 <div class="doc_subsubsection">
233 <a name="scf_includes"><tt>#include</tt> Style</a>
236 <div class="doc_text">
238 <p>Immediately after the <a href="#scf_commenting">header file comment</a> (and
239 include guards if working on a header file), the <a
240 href="#hl_dontinclude">minimal</a> list of <tt>#include</tt>s required by the
241 file should be listed. We prefer these <tt>#include</tt>s to be listed in this
245 <li><a href="#mmheader">Main Module header</a></li>
246 <li><a href="#hl_privateheaders">Local/Private Headers</a></li>
247 <li><tt>llvm/*</tt></li>
248 <li><tt>llvm/Analysis/*</tt></li>
249 <li><tt>llvm/Assembly/*</tt></li>
250 <li><tt>llvm/Bytecode/*</tt></li>
251 <li><tt>llvm/CodeGen/*</tt></li>
253 <li><tt>Support/*</tt></li>
254 <li><tt>Config/*</tt></li>
255 <li>System <tt>#includes</tt></li>
258 <p>... and each category should be sorted by name.</p>
260 <p><a name="mmheader">The "Main Module Header"</a> file applies to .cpp file
261 which implement an interface defined by a .h file. This <tt>#include</tt>
262 should always be included <b>first</b> regardless of where it lives on the file
263 system. By including a header file first in the .cpp files that implement the
264 interfaces, we ensure that the header does not have any hidden dependencies
265 which are not explicitly #included in the header, but should be. It is also a
266 form of documentation in the .cpp file to indicate where the interfaces it
267 implements are defined.</p>
271 <!-- _______________________________________________________________________ -->
272 <div class="doc_subsubsection">
273 <a name="scf_codewidth">Source Code Width</a>
276 <div class="doc_text">
278 <p>Write your code to fit within 80 columns of text. This helps those of us who
279 like to print out code and look at your code in an xterm without resizing
282 <p>The longer answer is that there must be some limit to the width of the code
283 in order to reasonably allow developers to have multiple files side-by-side in
284 windows on a modest display. If you are going to pick a width limit, it is
285 somewhat arbitrary but you might as well pick something standard. Going with
286 90 columns (for example) instead of 80 columns wouldn't add any significant
287 value and would be detrimental to printing out code. Also many other projects
288 have standardized on 80 columns, so some people have already configured their
289 editors for it (vs something else, like 90 columns).</p>
291 <p>This is one of many contentious issues in coding standards, but is not up
296 <!-- _______________________________________________________________________ -->
297 <div class="doc_subsubsection">
298 <a name="scf_spacestabs">Use Spaces Instead of Tabs</a>
301 <div class="doc_text">
303 <p>In all cases, prefer spaces to tabs in source files. People have different
304 prefered indentation levels, and different styles of indentation that they
305 like... this is fine. What isn't is that different editors/viewers expand tabs
306 out to different tab stops. This can cause your code to look completely
307 unreadable, and it is not worth dealing with.</p>
309 <p>As always, follow the <a href="#goldenrule">Golden Rule</a> above: follow the
310 style of existing code if your are modifying and extending it. If you like four
311 spaces of indentation, <b>DO NOT</b> do that in the middle of a chunk of code
312 with two spaces of indentation. Also, do not reindent a whole source file: it
313 makes for incredible diffs that are absolutely worthless.</p>
317 <!-- _______________________________________________________________________ -->
318 <div class="doc_subsubsection">
319 <a name="scf_indentation">Indent Code Consistently</a>
322 <div class="doc_text">
324 <p>Okay, your first year of programming you were told that indentation is
325 important. If you didn't believe and internalize this then, now is the time.
331 <!-- ======================================================================= -->
332 <div class="doc_subsection">
333 <a name="compilerissues">Compiler Issues</a>
337 <!-- _______________________________________________________________________ -->
338 <div class="doc_subsubsection">
339 <a name="ci_warningerrors">Treat Compiler Warnings Like Errors</a>
342 <div class="doc_text">
344 <p>If your code has compiler warnings in it, something is wrong: you aren't
345 casting values correctly, your have "questionable" constructs in your code, or
346 you are doing something legitimately wrong. Compiler warnings can cover up
347 legitimate errors in output and make dealing with a translation unit
350 <p>It is not possible to prevent all warnings from all compilers, nor is it
351 desirable. Instead, pick a standard compiler (like <tt>gcc</tt>) that provides
352 a good thorough set of warnings, and stick to them. At least in the case of
353 <tt>gcc</tt>, it is possible to work around any spurious errors by changing the
354 syntax of the code slightly. For example, an warning that annoys me occurs when
355 I write code like this:</p>
357 <div class="doc_code">
359 if (V = getValue()) {
365 <p><tt>gcc</tt> will warn me that I probably want to use the <tt>==</tt>
366 operator, and that I probably mistyped it. In most cases, I haven't, and I
367 really don't want the spurious errors. To fix this particular problem, I
368 rewrite the code like this:</p>
370 <div class="doc_code">
372 if ((V = getValue())) {
378 <p>...which shuts <tt>gcc</tt> up. Any <tt>gcc</tt> warning that annoys you can
379 be fixed by massaging the code appropriately.</p>
381 <p>These are the <tt>gcc</tt> warnings that I prefer to enable: <tt>-Wall
382 -Winline -W -Wwrite-strings -Wno-unused</tt></p>
386 <!-- _______________________________________________________________________ -->
387 <div class="doc_subsubsection">
388 <a name="ci_portable_code">Write Portable Code</a>
391 <div class="doc_text">
393 <p>In almost all cases, it is possible and within reason to write completely
394 portable code. If there are cases where it isn't possible to write portable
395 code, isolate it behind a well defined (and well documented) interface.</p>
397 <p>In practice, this means that you shouldn't assume much about the host
398 compiler, including its support for "high tech" features like partial
399 specialization of templates. If these features are used, they should only be
400 an implementation detail of a library which has a simple exposed API.</p>
404 <!-- _______________________________________________________________________ -->
405 <div class="doc_subsubsection">
406 <a name="ci_class_struct">Use of <tt>class</tt> and <tt>struct</tt> Keywords</a>
408 <div class="doc_text">
410 <p>In C++, the <tt>class</tt> and <tt>struct</tt> keywords can be used almost
411 interchangeably. The only difference is when they are used to declare a class:
412 <tt>class</tt> makes all members private by default while <tt>struct</tt> makes
413 all members public by default.</p>
415 <p>Unfortunately, not all compilers follow the rules and some will generate
416 different symbols based on whether <tt>class</tt> or <tt>struct</tt> was used to
417 declare the symbol. This can lead to problems at link time.</p>
419 <p>So, the rule for LLVM is to always use the <tt>class</tt> keyword, unless
420 <b>all</b> members are public, in which case <tt>struct</tt> is allowed.</p>
424 <!-- *********************************************************************** -->
425 <div class="doc_section">
426 <a name="styleissues">Style Issues</a>
428 <!-- *********************************************************************** -->
431 <!-- ======================================================================= -->
432 <div class="doc_subsection">
433 <a name="macro">The High Level Issues</a>
435 <!-- ======================================================================= -->
438 <!-- _______________________________________________________________________ -->
439 <div class="doc_subsubsection">
440 <a name="hl_module">A Public Header File <b>is</b> a Module</a>
443 <div class="doc_text">
445 <p>C++ doesn't do too well in the modularity department. There is no real
446 encapsulation or data hiding (unless you use expensive protocol classes), but it
447 is what we have to work with. When you write a public header file (in the LLVM
448 source tree, they live in the top level "include" directory), you are defining a
449 module of functionality.</p>
451 <p>Ideally, modules should be completely independent of each other, and their
452 header files should only include the absolute minimum number of headers
453 possible. A module is not just a class, a function, or a namespace: <a
454 href="http://www.cuj.com/articles/2000/0002/0002c/0002c.htm">it's a collection
455 of these</a> that defines an interface. This interface may be several
456 functions, classes or data structures, but the important issue is how they work
459 <p>In general, a module should be implemented with one or more <tt>.cpp</tt>
460 files. Each of these <tt>.cpp</tt> files should include the header that defines
461 their interface first. This ensure that all of the dependences of the module
462 header have been properly added to the module header itself, and are not
463 implicit. System headers should be included after user headers for a
464 translation unit.</p>
468 <!-- _______________________________________________________________________ -->
469 <div class="doc_subsubsection">
470 <a name="hl_dontinclude"><tt>#include</tt> as Little as Possible</a>
473 <div class="doc_text">
475 <p><tt>#include</tt> hurts compile time performance. Don't do it unless you
476 have to, especially in header files.</p>
478 <p>But wait, sometimes you need to have the definition of a class to use it, or
479 to inherit from it. In these cases go ahead and <tt>#include</tt> that header
480 file. Be aware however that there are many cases where you don't need to have
481 the full definition of a class. If you are using a pointer or reference to a
482 class, you don't need the header file. If you are simply returning a class
483 instance from a prototyped function or method, you don't need it. In fact, for
484 most cases, you simply don't need the definition of a class... and not
485 <tt>#include</tt>'ing speeds up compilation.</p>
487 <p>It is easy to try to go too overboard on this recommendation, however. You
488 <b>must</b> include all of the header files that you are using -- you can
489 include them either directly
490 or indirectly (through another header file). To make sure that you don't
491 accidently forget to include a header file in your module header, make sure to
492 include your module header <b>first</b> in the implementation file (as mentioned
493 above). This way there won't be any hidden dependencies that you'll find out
498 <!-- _______________________________________________________________________ -->
499 <div class="doc_subsubsection">
500 <a name="hl_privateheaders">Keep "internal" Headers Private</a>
503 <div class="doc_text">
505 <p>Many modules have a complex implementation that causes them to use more than
506 one implementation (<tt>.cpp</tt>) file. It is often tempting to put the
507 internal communication interface (helper classes, extra functions, etc) in the
508 public module header file. Don't do this.</p>
510 <p>If you really need to do something like this, put a private header file in
511 the same directory as the source files, and include it locally. This ensures
512 that your private interface remains private and undisturbed by outsiders.</p>
514 <p>Note however, that it's okay to put extra implementation methods a public
515 class itself... just make them private (or protected), and all is well.</p>
519 <!-- _______________________________________________________________________ -->
520 <div class="doc_subsubsection">
521 <a name="hl_earlyexit">Use Early Exits and 'continue' to Simplify Code</a>
524 <div class="doc_text">
526 <p>When reading code, keep in mind how much state and how many previous
527 decisions have to be remembered by the reader to understand a block of code.
528 Aim to reduce indentation where possible when it doesn't make it more difficult
529 to understand the code. One great way to do this is by making use of early
530 exits and the 'continue' keyword in long loops. As an example of using an early
531 exit from a function, consider this "bad" code:</p>
533 <div class="doc_code">
535 Value *DoSomething(Instruction *I) {
536 if (!isa<TerminatorInst>(I) &&
537 I->hasOneUse() && SomeOtherThing(I)) {
538 ... some long code ....
546 <p>This code has several problems if the body of the 'if' is large. When you're
547 looking at the top of the function, it isn't immediately clear that this
548 <em>only</em> does interesting things with non-terminator instructions, and only
549 applies to things with the other predicates. Second, it is relatively difficult
550 to describe (in comments) why these predicates are important because the if
551 statement makes it difficult to lay out the comments. Third, when you're deep
552 within the body of the code, it is indented an extra level. Finally, when
553 reading the top of the function, it isn't clear what the result is if the
554 predicate isn't true, you have to read to the end of the function to know that
557 <p>It is much preferred to format the code like this:</p>
559 <div class="doc_code">
561 Value *DoSomething(Instruction *I) {
562 // Terminators never need 'something' done to them because, ...
563 if (isa<TerminatorInst>(I))
566 // We conservatively avoid transforming instructions with multiple uses
567 // because goats like cheese.
568 if (!I->hasOneUse())
571 // This is really just here for example.
572 if (!SomeOtherThing(I))
575 ... some long code ....
580 <p>This fixes these problems. A similar problem frequently happens in for
581 loops. A silly example is something like this:</p>
583 <div class="doc_code">
585 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
586 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
587 Value *LHS = BO->getOperand(0);
588 Value *RHS = BO->getOperand(1);
597 <p>When you have very very small loops, this sort of structure is fine, but if
598 it exceeds more than 10-15 lines, it becomes difficult for people to read and
599 understand at a glance.
600 The problem with this sort of code is that it gets very nested very quickly,
601 meaning that the reader of the code has to keep a lot of context in their brain
602 to remember what is going immediately on in the loop, because they don't know
603 if/when the if conditions will have elses etc. It is strongly preferred to
604 structure the loop like this:</p>
606 <div class="doc_code">
608 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
609 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
612 Value *LHS = BO->getOperand(0);
613 Value *RHS = BO->getOperand(1);
614 if (LHS == RHS) continue;
619 <p>This has all the benefits of using early exits from functions: it reduces
620 nesting of the loop, it makes it easier to describe why the conditions are true,
621 and it makes it obvious to the reader that there is no "else" coming up that
622 they have to push context into their brain for. If a loop is large, this can
623 be a big understandability win.</p>
628 <!-- _______________________________________________________________________ -->
629 <div class="doc_subsubsection">
630 <a name="hl_predicateloops">Turn Predicate Loops into Predicate Functions</a>
633 <div class="doc_text">
635 <p>It is very common to write small loops that just compute a boolean
636 value. There are a number of ways that people commonly write these, but an
637 example of this sort of thing is:</p>
639 <div class="doc_code">
641 <b>bool FoundFoo = false;</b>
642 for (unsigned i = 0, e = BarList.size(); i != e; ++i)
643 if (BarList[i]->isFoo()) {
644 <b>FoundFoo = true;</b>
648 <b>if (FoundFoo) {</b>
654 <p>This sort of code is awkward to write, and is almost always a bad sign.
655 Instead of this sort of loop, we strongly prefer to use a predicate function
656 (which may be <a href="#micro_anonns">static</a>) that uses
657 <a href="#hl_earlyexit">early exits</a> to compute the predicate. We prefer
658 the code to be structured like this:
662 <div class="doc_code">
664 /// ListContainsFoo - Return true if the specified list has an element that is
666 static bool ListContainsFoo(const std::vector<Bar*> &List) {
667 for (unsigned i = 0, e = List.size(); i != e; ++i)
668 if (List[i]->isFoo())
674 <b>if (ListContainsFoo(BarList)) {</b>
680 <p>There are many reasons for doing this: it reduces indentation and factors out
681 code which can often be shared by other code that checks for the same predicate.
682 More importantly, it <em>forces you to pick a name</em> for the function, and
683 forces you to write a comment for it. In this silly example, this doesn't add
684 much value. However, if the condition is complex, this can make it a lot easier
685 for the reader to understand the code that queries for this predicate. Instead
686 of being faced with the in-line details of how we check to see if the BarList
687 contains a foo, we can trust the function name and continue reading with better
693 <!-- ======================================================================= -->
694 <div class="doc_subsection">
695 <a name="micro">The Low Level Issues</a>
697 <!-- ======================================================================= -->
700 <!-- _______________________________________________________________________ -->
701 <div class="doc_subsubsection">
702 <a name="ll_assert">Assert Liberally</a>
705 <div class="doc_text">
707 <p>Use the "<tt>assert</tt>" function to its fullest. Check all of your
708 preconditions and assumptions, you never know when a bug (not neccesarily even
709 yours) might be caught early by an assertion, which reduces debugging time
710 dramatically. The "<tt><cassert></tt>" header file is probably already
711 included by the header files you are using, so it doesn't cost anything to use
714 <p>To further assist with debugging, make sure to put some kind of error message
715 in the assertion statement (which is printed if the assertion is tripped). This
716 helps the poor debugging make sense of why an assertion is being made and
717 enforced, and hopefully what to do about it. Here is one complete example:</p>
719 <div class="doc_code">
721 inline Value *getOperand(unsigned i) {
722 assert(i < Operands.size() && "getOperand() out of range!");
728 <p>Here are some examples:</p>
730 <div class="doc_code">
732 assert(Ty->isPointerType() && "Can't allocate a non pointer type!");
734 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
736 assert(idx < getNumSuccessors() && "Successor # out of range!");
738 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
740 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
744 <p>You get the idea...</p>
746 <p>Please be aware when adding assert statements that not all compilers are aware of
747 the semantics of the assert. In some places, asserts are used to indicate a piece of
748 code that should not be reached. These are typically of the form:</p>
750 <div class="doc_code">
752 assert(0 && "Some helpful error message");
756 <p>When used in a function that returns a value, they should be followed with a return
757 statement and a comment indicating that this line is never reached. This will prevent
758 a compiler which is unable to deduce that the assert statement never returns from
759 generating a warning.</p>
761 <div class="doc_code">
763 assert(0 && "Some helpful error message");
771 <!-- _______________________________________________________________________ -->
772 <div class="doc_subsubsection">
773 <a name="ll_ns_std">Do not use '<tt>using namespace std</tt>'</a>
776 <div class="doc_text">
777 <p>In LLVM, we prefer to explicitly prefix all identifiers from the standard
778 namespace with an "<tt>std::</tt>" prefix, rather than rely on
779 "<tt>using namespace std;</tt>".</p>
781 <p> In header files, adding a '<tt>using namespace XXX</tt>' directive pollutes
782 the namespace of any source file that <tt>#include</tt>s the header. This is
783 clearly a bad thing.</p>
785 <p>In implementation files (e.g. .cpp files), the rule is more of a stylistic
786 rule, but is still important. Basically, using explicit namespace prefixes
787 makes the code <b>clearer</b>, because it is immediately obvious what facilities
788 are being used and where they are coming from, and <b>more portable</b>, because
789 namespace clashes cannot occur between LLVM code and other namespaces. The
790 portability rule is important because different standard library implementations
791 expose different symbols (potentially ones they shouldn't), and future revisions
792 to the C++ standard will add more symbols to the <tt>std</tt> namespace. As
793 such, we never use '<tt>using namespace std;</tt>' in LLVM.</p>
795 <p>The exception to the general rule (i.e. it's not an exception for
796 the <tt>std</tt> namespace) is for implementation files. For example, all of
797 the code in the LLVM project implements code that lives in the 'llvm' namespace.
798 As such, it is ok, and actually clearer, for the .cpp files to have a '<tt>using
799 namespace llvm</tt>' directive at their top, after the <tt>#include</tt>s. The
800 general form of this rule is that any .cpp file that implements code in any
801 namespace may use that namespace (and its parents'), but should not use any
806 <!-- _______________________________________________________________________ -->
807 <div class="doc_subsubsection">
808 <a name="ll_virtual_anch">Provide a virtual method anchor for classes
812 <div class="doc_text">
814 <p>If a class is defined in a header file and has a v-table (either it has
815 virtual methods or it derives from classes with virtual methods), it must
816 always have at least one out-of-line virtual method in the class. Without
817 this, the compiler will copy the vtable and RTTI into every <tt>.o</tt> file
818 that <tt>#include</tt>s the header, bloating <tt>.o</tt> file sizes and
819 increasing link times.</p>
823 <!-- _______________________________________________________________________ -->
824 <div class="doc_subsubsection">
825 <a name="ll_end">Don't evaluate end() every time through a loop</a>
828 <div class="doc_text">
830 <p>Because C++ doesn't have a standard "foreach" loop (though it can be emulated
831 with macros and may be coming in C++'0x) we end up writing a lot of loops that
832 manually iterate from begin to end on a variety of containers or through other
833 data structures. One common mistake is to write a loop in this style:</p>
835 <div class="doc_code">
838 for (BasicBlock::iterator I = BB->begin(); I != <b>BB->end()</b>; ++I)
843 <p>The problem with this construct is that it evaluates "<tt>BB->end()</tt>"
844 every time through the loop. Instead of writing the loop like this, we strongly
845 prefer loops to be written so that they evaluate it once before the loop starts.
846 A convenient way to do this is like so:</p>
848 <div class="doc_code">
851 for (BasicBlock::iterator I = BB->begin(), E = <b>BB->end()</b>; I != E; ++I)
856 <p>The observant may quickly point out that these two loops may have different
857 semantics: if the container (a basic block in this case) is being mutated, then
858 "<tt>BB->end()</tt>" may change its value every time through the loop and the
859 second loop may not in fact be correct. If you actually do depend on this
860 behavior, please write the loop in the first form and add a comment indicating
861 that you did it intentionally.</p>
863 <p>Why do we prefer the second form (when correct)? Writing the loop in the
864 first form has two problems: First it may be less efficient than evaluating it
865 at the start of the loop. In this case, the cost is probably minor: a few extra
866 loads every time through the loop. However, if the base expression is more
867 complex, then the cost can rise quickly. I've seen loops where the end
868 expression was actually something like: "<tt>SomeMap[x]->end()</tt>" and map
869 lookups really aren't cheap. By writing it in the second form consistently, you
870 eliminate the issue entirely and don't even have to think about it.</p>
872 <p>The second (even bigger) issue is that writing the loop in the first form
873 hints to the reader that the loop is mutating the container (a fact that a
874 comment would handily confirm!). If you write the loop in the second form, it
875 is immediately obvious without even looking at the body of the loop that the
876 container isn't being modified, which makes it easier to read the code and
877 understand what it does.</p>
879 <p>While the second form of the loop is a few extra keystrokes, we do strongly
884 <!-- _______________________________________________________________________ -->
885 <div class="doc_subsubsection">
886 <a name="ll_iostream"><tt>#include <iostream></tt> is forbidden</a>
889 <div class="doc_text">
891 <p>The use of <tt>#include <iostream></tt> in library files is
892 hereby <b><em>forbidden</em></b>. The primary reason for doing this is to
893 support clients using LLVM libraries as part of larger systems. In particular,
894 we statically link LLVM into some dynamic libraries. Even if LLVM isn't used,
895 the static c'tors are run whenever an application start up that uses the dynamic
896 library. There are two problems with this:</p>
899 <li>The time to run the static c'tors impacts startup time of
900 applications—a critical time for GUI apps.</li>
901 <li>The static c'tors cause the app to pull many extra pages of memory off the
902 disk: both the code for the static c'tors in each <tt>.o</tt> file and the
903 small amount of data that gets touched. In addition, touched/dirty pages
904 put more pressure on the VM system on low-memory machines.</li>
907 <p>Note that using the other stream headers (<tt><sstream></tt> for
908 example) is allowed normally, it is just <tt><iostream></tt> that is
909 causing problems.</p>
911 <p>In addition, new code should always
912 use <a href="#ll_raw_ostream"><tt>raw_ostream</tt></a> or
913 the <tt>llvm::MemoryBuffer</tt> API (for reading in files).</p>
918 <!-- _______________________________________________________________________ -->
919 <div class="doc_subsubsection">
920 <a name="ll_avoidendl">Avoid <tt>std::endl</tt></a>
923 <div class="doc_text">
925 <p>The <tt>std::endl</tt> modifier, when used with iostreams outputs a newline
926 to the output stream specified. In addition to doing this, however, it also
927 flushes the output stream. In other words, these are equivalent:</p>
929 <div class="doc_code">
931 std::cout << std::endl;
932 std::cout << '\n' << std::flush;
936 <p>Most of the time, you probably have no reason to flush the output stream, so
937 it's better to use a literal <tt>'\n'</tt>.</p>
942 <!-- _______________________________________________________________________ -->
943 <div class="doc_subsubsection">
944 <a name="ll_raw_ostream">Use <tt>raw_ostream</tt></a>
947 <div class="doc_text">
949 <p>LLVM includes a lightweight, simple, and efficient stream implementation
950 in <tt>llvm/Support/raw_ostream.h</tt> which provides all of the common features
951 of <tt>std::ostream</tt>. All new code should use <tt>raw_ostream</tt> instead
952 of <tt>ostream</tt>.</p>
954 <p>Unlike <tt>std::ostream</tt>, <tt>raw_ostream</tt> is not a template and can
955 be forward declared as <tt>class raw_ostream</tt>. Public headers should
956 generally not include the <tt>raw_ostream</tt> header, but use forward
957 declarations and constant references to <tt>raw_ostream</tt> instances.</p>
962 <!-- ======================================================================= -->
963 <div class="doc_subsection">
964 <a name="nano">Microscopic Details</a>
966 <!-- ======================================================================= -->
968 <p>This section describes preferred low-level formatting guidelines along with
969 reasoning on why we prefer them.</p>
971 <!-- _______________________________________________________________________ -->
972 <div class="doc_subsubsection">
973 <a name="micro_spaceparen">Spaces Before Parentheses</a>
976 <div class="doc_text">
978 <p>We prefer to put a space before a parentheses only in control flow
979 statements, but not in normal function call expressions and function-like
980 macros. For example, this is good:</p>
982 <div class="doc_code">
985 <b>for (</b>i = 0; i != 100; ++i) ...
986 <b>while (</b>llvm_rocks) ...
989 <b><a href="#ll_assert">assert</a>(</b>3 != 4 && "laws of math are failing me");
991 a = <b>foo(</b>42, 92) + <b>bar(</b>x);
995 <p>... and this is bad:</p>
997 <div class="doc_code">
1000 <b>for(</b>i = 0; i != 100; ++i) ...
1001 <b>while(</b>llvm_rocks) ...
1003 <b>somefunc (</b>42);
1004 <b><a href="#ll_assert">assert</a> (</b>3 != 4 && "laws of math are failing me");
1006 a = <b>foo (</b>42, 92) + <b>bar (</b>x);
1010 <p>The reason for doing this is not completely arbitrary. This style makes
1011 control flow operators stand out more, and makes expressions flow better. The
1012 function call operator binds very tightly as a postfix operator. Putting
1013 a space after a function name (as in the last example) makes it appear that
1014 the code might bind the arguments of the left-hand-side of a binary operator
1015 with the argument list of a function and the name of the right side. More
1016 specifically, it is easy to misread the "a" example as:</p>
1018 <div class="doc_code">
1020 a = foo <b>(</b>(42, 92) + bar<b>)</b> (x);
1024 <p>... when skimming through the code. By avoiding a space in a function, we
1025 avoid this misinterpretation.</p>
1029 <!-- _______________________________________________________________________ -->
1030 <div class="doc_subsubsection">
1031 <a name="micro_preincrement">Prefer Preincrement</a>
1034 <div class="doc_text">
1036 <p>Hard fast rule: Preincrement (<tt>++X</tt>) may be no slower than
1037 postincrement (<tt>X++</tt>) and could very well be a lot faster than it. Use
1038 preincrementation whenever possible.</p>
1040 <p>The semantics of postincrement include making a copy of the value being
1041 incremented, returning it, and then preincrementing the "work value". For
1042 primitive types, this isn't a big deal... but for iterators, it can be a huge
1043 issue (for example, some iterators contains stack and set objects in them...
1044 copying an iterator could invoke the copy ctor's of these as well). In general,
1045 get in the habit of always using preincrement, and you won't have a problem.</p>
1049 <!-- _______________________________________________________________________ -->
1050 <div class="doc_subsubsection">
1051 <a name="micro_namespaceindent">Namespace Indentation</a>
1054 <div class="doc_text">
1057 In general, we strive to reduce indentation where ever possible. This is useful
1058 because we want code to <a href="#scf_codewidth">fit into 80 columns</a> without
1059 wrapping horribly, but also because it makes it easier to understand the code.
1060 Namespaces are a funny thing: they are often large, and we often desire to put
1061 lots of stuff into them (so they can be large). Other times they are tiny,
1062 because they just hold an enum or something similar. In order to balance this,
1063 we use different approaches for small versus large namespaces.
1067 If a namespace definition is small and <em>easily</em> fits on a screen (say,
1068 less than 35 lines of code), then you should indent its body. Here's an
1072 <div class="doc_code">
1076 /// RelocationType - An enum for the x86 relocation codes. Note that
1077 /// the terminology here doesn't follow x86 convention - word means
1078 /// 32-bit and dword means 64-bit.
1079 enum RelocationType {
1080 /// reloc_pcrel_word - PC relative relocation, add the relocated value to
1081 /// the value already in memory, after we adjust it for where the PC is.
1082 reloc_pcrel_word = 0,
1084 /// reloc_picrel_word - PIC base relative relocation, add the relocated
1085 /// value to the value already in memory, after we adjust it for where the
1087 reloc_picrel_word = 1,
1089 /// reloc_absolute_word, reloc_absolute_dword - Absolute relocation, just
1090 /// add the relocated value to the value already in memory.
1091 reloc_absolute_word = 2,
1092 reloc_absolute_dword = 3
1099 <p>Since the body is small, indenting adds value because it makes it very clear
1100 where the namespace starts and ends, and it is easy to take the whole thing in
1101 in one "gulp" when reading the code. If the blob of code in the namespace is
1102 larger (as it typically is in a header in the llvm or clang namespaces), do not
1103 indent the code, and add a comment indicating what namespace is being closed.
1106 <div class="doc_code">
1109 namespace knowledge {
1111 /// Grokable - This class represents things that Smith can have an intimate
1112 /// understanding of and contains the data associated with it.
1116 explicit Grokable() { ... }
1117 virtual ~Grokable() = 0;
1123 } // end namespace knowledge
1124 } // end namespace llvm
1128 <p>Because the class is large, we don't expect that the reader can easily
1129 understand the entire concept in a glance, and the end of the file (where the
1130 namespaces end) may be a long ways away from the place they open. As such,
1131 indenting the contents of the namespace doesn't add any value, and detracts from
1132 the readability of the class. In these cases it is best to <em>not</em> indent
1133 the contents of the namespace.</p>
1137 <!-- _______________________________________________________________________ -->
1138 <div class="doc_subsubsection">
1139 <a name="micro_anonns">Anonymous Namespaces</a>
1142 <div class="doc_text">
1144 <p>After talking about namespaces in general, you may be wondering about
1145 anonymous namespaces in particular.
1146 Anonymous namespaces are a great language feature that tells the C++ compiler
1147 that the contents of the namespace are only visible within the current
1148 translation unit, allowing more aggressive optimization and eliminating the
1149 possibility of symbol name collisions. Anonymous namespaces are to C++ as
1150 "static" is to C functions and global variables. While "static" is available
1151 in C++, anonymous namespaces are more general: they can make entire classes
1152 private to a file.</p>
1154 <p>The problem with anonymous namespaces is that they naturally want to
1155 encourage indentation of their body, and they reduce locality of reference: if
1156 you see a random function definition in a C++ file, it is easy to see if it is
1157 marked static, but seeing if it is in an anonymous namespace requires scanning
1158 a big chunk of the file.</p>
1160 <p>Because of this, we have a simple guideline: make anonymous namespaces as
1161 small as possible, and only use them for class declarations. For example, this
1164 <div class="doc_code">
1171 bool operator<(const char *RHS) const;
1173 <b>} // end anonymous namespace</b>
1175 static void Helper() {
1179 bool StringSort::operator<(const char *RHS) const {
1189 <div class="doc_code">
1196 bool operator<(const char *RHS) const;
1203 bool StringSort::operator<(const char *RHS) const {
1207 <b>} // end anonymous namespace</b>
1213 <p>This is bad specifically because if you're looking at "Helper" in the middle
1214 of a large C++ file, that you have no immediate way to tell if it is local to
1215 the file. When it is marked static explicitly, this is immediately obvious.
1216 Also, there is no reason to enclose the definition of "operator<" in the
1217 namespace just because it was declared there.
1224 <!-- *********************************************************************** -->
1225 <div class="doc_section">
1226 <a name="seealso">See Also</a>
1228 <!-- *********************************************************************** -->
1230 <div class="doc_text">
1232 <p>A lot of these comments and recommendations have been culled for other
1233 sources. Two particularly important books for our work are:</p>
1237 <li><a href="http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876">Effective
1238 C++</a> by Scott Meyers. Also
1239 interesting and useful are "More Effective C++" and "Effective STL" by the same
1242 <li>Large-Scale C++ Software Design by John Lakos</li>
1246 <p>If you get some free time, and you haven't read them: do so, you might learn
1251 <!-- *********************************************************************** -->
1255 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1256 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1257 <a href="http://validator.w3.org/check/referer"><img
1258 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1260 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1261 <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
1262 Last modified: $Date$