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_else_after_return">Don't use "else" after a
48 <li><a href="#hl_predicateloops">Turn Predicate Loops into Predicate
51 <li><a href="#micro">The Low-Level Issues</a>
53 <li><a href="#ll_assert">Assert Liberally</a></li>
54 <li><a href="#ll_ns_std">Do not use 'using namespace std'</a></li>
55 <li><a href="#ll_virtual_anch">Provide a virtual method anchor for
56 classes in headers</a></li>
57 <li><a href="#ll_end">Don't evaluate end() every time through a
59 <li><a href="#ll_iostream"><tt>#include <iostream></tt> is
60 <em>forbidden</em></a></li>
61 <li><a href="#ll_raw_ostream">Use <tt>raw_ostream</tt></a</li>
62 <li><a href="#ll_avoidendl">Avoid <tt>std::endl</tt></a></li>
65 <li><a href="#nano">Microscopic Details</a>
67 <li><a href="#micro_spaceparen">Spaces Before Parentheses</a></li>
68 <li><a href="#micro_preincrement">Prefer Preincrement</a></li>
69 <li><a href="#micro_namespaceindent">Namespace Indentation</a></li>
70 <li><a href="#micro_anonns">Anonymous Namespaces</a></li>
75 <li><a href="#seealso">See Also</a></li>
78 <div class="doc_author">
79 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
83 <!-- *********************************************************************** -->
84 <div class="doc_section">
85 <a name="introduction">Introduction</a>
87 <!-- *********************************************************************** -->
89 <div class="doc_text">
91 <p>This document attempts to describe a few coding standards that are being used
92 in the LLVM source tree. Although no coding standards should be regarded as
93 absolute requirements to be followed in all instances, coding standards can be
96 <p>This document intentionally does not prescribe fixed standards for religious
97 issues such as brace placement and space usage. For issues like this, follow
102 <p><b><a name="goldenrule">If you are adding a significant body of source to a
103 project, feel free to use whatever style you are most comfortable with. If you
104 are extending, enhancing, or bug fixing already implemented code, use the style
105 that is already being used so that the source is uniform and easy to
110 <p>The ultimate goal of these guidelines is the increase readability and
111 maintainability of our common source base. If you have suggestions for topics to
112 be included, please mail them to <a
113 href="mailto:sabre@nondot.org">Chris</a>.</p>
117 <!-- *********************************************************************** -->
118 <div class="doc_section">
119 <a name="mechanicalissues">Mechanical Source Issues</a>
121 <!-- *********************************************************************** -->
123 <!-- ======================================================================= -->
124 <div class="doc_subsection">
125 <a name="sourceformating">Source Code Formatting</a>
128 <!-- _______________________________________________________________________ -->
129 <div class="doc_subsubsection">
130 <a name="scf_commenting">Commenting</a>
133 <div class="doc_text">
135 <p>Comments are one critical part of readability and maintainability. Everyone
136 knows they should comment, so should you. When writing comments, write them as
137 English prose, which means they should use proper capitalization, punctuation,
138 etc. Although we all should probably
139 comment our code more than we do, there are a few very critical places that
140 documentation is very useful:</p>
144 <p>Every source file should have a header on it that describes the basic
145 purpose of the file. If a file does not have a header, it should not be
146 checked into Subversion. Most source trees will probably have a standard
147 file header format. The standard format for the LLVM source tree looks like
150 <div class="doc_code">
152 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
154 // The LLVM Compiler Infrastructure
156 // This file is distributed under the University of Illinois Open Source
157 // License. See LICENSE.TXT for details.
159 //===----------------------------------------------------------------------===//
161 // This file contains the declaration of the Instruction class, which is the
162 // base class for all of the VM instructions.
164 //===----------------------------------------------------------------------===//
168 <p>A few things to note about this particular format: The "<tt>-*- C++
169 -*-</tt>" string on the first line is there to tell Emacs that the source file
170 is a C++ file, not a C file (Emacs assumes <tt>.h</tt> files are C files by default).
171 Note that this tag is not necessary in <tt>.cpp</tt> files. The name of the file is also
172 on the first line, along with a very short description of the purpose of the
173 file. This is important when printing out code and flipping though lots of
176 <p>The next section in the file is a concise note that defines the license
177 that the file is released under. This makes it perfectly clear what terms the
178 source code can be distributed under and should not be modified in any way.</p>
180 <p>The main body of the description does not have to be very long in most cases.
181 Here it's only two lines. If an algorithm is being implemented or something
182 tricky is going on, a reference to the paper where it is published should be
183 included, as well as any notes or "gotchas" in the code to watch out for.</p>
185 <b>Class overviews</b>
187 <p>Classes are one fundamental part of a good object oriented design. As such,
188 a class definition should have a comment block that explains what the class is
189 used for... if it's not obvious. If it's so completely obvious your grandma
190 could figure it out, it's probably safe to leave it out. Naming classes
191 something sane goes a long ways towards avoiding writing documentation.</p>
194 <b>Method information</b>
196 <p>Methods defined in a class (as well as any global functions) should also be
197 documented properly. A quick note about what it does and a description of the
198 borderline behaviour is all that is necessary here (unless something
199 particularly tricky or insidious is going on). The hope is that people can
200 figure out how to use your interfaces without reading the code itself... that is
203 <p>Good things to talk about here are what happens when something unexpected
204 happens: does the method return null? Abort? Format your hard disk?</p>
208 <!-- _______________________________________________________________________ -->
209 <div class="doc_subsubsection">
210 <a name="scf_commentformat">Comment Formatting</a>
213 <div class="doc_text">
215 <p>In general, prefer C++ style (<tt>//</tt>) comments. They take less space,
216 require less typing, don't have nesting problems, etc. There are a few cases
217 when it is useful to use C style (<tt>/* */</tt>) comments however:</p>
220 <li>When writing a C code: Obviously if you are writing C code, use C style
222 <li>When writing a header file that may be <tt>#include</tt>d by a C source
224 <li>When writing a source file that is used by a tool that only accepts C
228 <p>To comment out a large block of code, use <tt>#if 0</tt> and <tt>#endif</tt>.
229 These nest properly and are better behaved in general than C style comments.</p>
233 <!-- _______________________________________________________________________ -->
234 <div class="doc_subsubsection">
235 <a name="scf_includes"><tt>#include</tt> Style</a>
238 <div class="doc_text">
240 <p>Immediately after the <a href="#scf_commenting">header file comment</a> (and
241 include guards if working on a header file), the <a
242 href="#hl_dontinclude">minimal</a> list of <tt>#include</tt>s required by the
243 file should be listed. We prefer these <tt>#include</tt>s to be listed in this
247 <li><a href="#mmheader">Main Module Header</a></li>
248 <li><a href="#hl_privateheaders">Local/Private Headers</a></li>
249 <li><tt>llvm/*</tt></li>
250 <li><tt>llvm/Analysis/*</tt></li>
251 <li><tt>llvm/Assembly/*</tt></li>
252 <li><tt>llvm/Bitcode/*</tt></li>
253 <li><tt>llvm/CodeGen/*</tt></li>
255 <li><tt>Support/*</tt></li>
256 <li><tt>Config/*</tt></li>
257 <li>System <tt>#includes</tt></li>
260 <p>... and each category should be sorted by name.</p>
262 <p><a name="mmheader">The "Main Module Header"</a> file applies to <tt>.cpp</tt> files
263 which implement an interface defined by a <tt>.h</tt> file. This <tt>#include</tt>
264 should always be included <b>first</b> regardless of where it lives on the file
265 system. By including a header file first in the <tt>.cpp</tt> files that implement the
266 interfaces, we ensure that the header does not have any hidden dependencies
267 which are not explicitly #included in the header, but should be. It is also a
268 form of documentation in the <tt>.cpp</tt> file to indicate where the interfaces it
269 implements are defined.</p>
273 <!-- _______________________________________________________________________ -->
274 <div class="doc_subsubsection">
275 <a name="scf_codewidth">Source Code Width</a>
278 <div class="doc_text">
280 <p>Write your code to fit within 80 columns of text. This helps those of us who
281 like to print out code and look at your code in an xterm without resizing
284 <p>The longer answer is that there must be some limit to the width of the code
285 in order to reasonably allow developers to have multiple files side-by-side in
286 windows on a modest display. If you are going to pick a width limit, it is
287 somewhat arbitrary but you might as well pick something standard. Going with
288 90 columns (for example) instead of 80 columns wouldn't add any significant
289 value and would be detrimental to printing out code. Also many other projects
290 have standardized on 80 columns, so some people have already configured their
291 editors for it (vs something else, like 90 columns).</p>
293 <p>This is one of many contentious issues in coding standards, but is not up
298 <!-- _______________________________________________________________________ -->
299 <div class="doc_subsubsection">
300 <a name="scf_spacestabs">Use Spaces Instead of Tabs</a>
303 <div class="doc_text">
305 <p>In all cases, prefer spaces to tabs in source files. People have different
306 preferred indentation levels, and different styles of indentation that they
307 like... this is fine. What isn't is that different editors/viewers expand tabs
308 out to different tab stops. This can cause your code to look completely
309 unreadable, and it is not worth dealing with.</p>
311 <p>As always, follow the <a href="#goldenrule">Golden Rule</a> above: follow the
312 style of existing code if your are modifying and extending it. If you like four
313 spaces of indentation, <b>DO NOT</b> do that in the middle of a chunk of code
314 with two spaces of indentation. Also, do not reindent a whole source file: it
315 makes for incredible diffs that are absolutely worthless.</p>
319 <!-- _______________________________________________________________________ -->
320 <div class="doc_subsubsection">
321 <a name="scf_indentation">Indent Code Consistently</a>
324 <div class="doc_text">
326 <p>Okay, in your first year of programming you were told that indentation is
327 important. If you didn't believe and internalize this then, now is the time.
333 <!-- ======================================================================= -->
334 <div class="doc_subsection">
335 <a name="compilerissues">Compiler Issues</a>
339 <!-- _______________________________________________________________________ -->
340 <div class="doc_subsubsection">
341 <a name="ci_warningerrors">Treat Compiler Warnings Like Errors</a>
344 <div class="doc_text">
346 <p>If your code has compiler warnings in it, something is wrong: you aren't
347 casting values correctly, your have "questionable" constructs in your code, or
348 you are doing something legitimately wrong. Compiler warnings can cover up
349 legitimate errors in output and make dealing with a translation unit
352 <p>It is not possible to prevent all warnings from all compilers, nor is it
353 desirable. Instead, pick a standard compiler (like <tt>gcc</tt>) that provides
354 a good thorough set of warnings, and stick to them. At least in the case of
355 <tt>gcc</tt>, it is possible to work around any spurious errors by changing the
356 syntax of the code slightly. For example, an warning that annoys me occurs when
357 I write code like this:</p>
359 <div class="doc_code">
361 if (V = getValue()) {
367 <p><tt>gcc</tt> will warn me that I probably want to use the <tt>==</tt>
368 operator, and that I probably mistyped it. In most cases, I haven't, and I
369 really don't want the spurious errors. To fix this particular problem, I
370 rewrite the code like this:</p>
372 <div class="doc_code">
374 if ((V = getValue())) {
380 <p>...which shuts <tt>gcc</tt> up. Any <tt>gcc</tt> warning that annoys you can
381 be fixed by massaging the code appropriately.</p>
383 <p>These are the <tt>gcc</tt> warnings that I prefer to enable: <tt>-Wall
384 -Winline -W -Wwrite-strings -Wno-unused</tt></p>
388 <!-- _______________________________________________________________________ -->
389 <div class="doc_subsubsection">
390 <a name="ci_portable_code">Write Portable Code</a>
393 <div class="doc_text">
395 <p>In almost all cases, it is possible and within reason to write completely
396 portable code. If there are cases where it isn't possible to write portable
397 code, isolate it behind a well defined (and well documented) interface.</p>
399 <p>In practice, this means that you shouldn't assume much about the host
400 compiler, and Visual Studio tends to be the lowest common denominator.
401 If advanced features are used, they should only be an implementation detail of
402 a library which has a simple exposed API, and preferably be buried in
407 <!-- _______________________________________________________________________ -->
408 <div class="doc_subsubsection">
409 <a name="ci_class_struct">Use of <tt>class</tt> and <tt>struct</tt> Keywords</a>
411 <div class="doc_text">
413 <p>In C++, the <tt>class</tt> and <tt>struct</tt> keywords can be used almost
414 interchangeably. The only difference is when they are used to declare a class:
415 <tt>class</tt> makes all members private by default while <tt>struct</tt> makes
416 all members public by default.</p>
418 <p>Unfortunately, not all compilers follow the rules and some will generate
419 different symbols based on whether <tt>class</tt> or <tt>struct</tt> was used to
420 declare the symbol. This can lead to problems at link time.</p>
422 <p>So, the rule for LLVM is to always use the <tt>class</tt> keyword, unless
423 <b>all</b> members are public and the type is a C++ "POD" type, in which case
424 <tt>struct</tt> is allowed.</p>
428 <!-- *********************************************************************** -->
429 <div class="doc_section">
430 <a name="styleissues">Style Issues</a>
432 <!-- *********************************************************************** -->
435 <!-- ======================================================================= -->
436 <div class="doc_subsection">
437 <a name="macro">The High-Level Issues</a>
439 <!-- ======================================================================= -->
442 <!-- _______________________________________________________________________ -->
443 <div class="doc_subsubsection">
444 <a name="hl_module">A Public Header File <b>is</b> a Module</a>
447 <div class="doc_text">
449 <p>C++ doesn't do too well in the modularity department. There is no real
450 encapsulation or data hiding (unless you use expensive protocol classes), but it
451 is what we have to work with. When you write a public header file (in the LLVM
452 source tree, they live in the top level "include" directory), you are defining a
453 module of functionality.</p>
455 <p>Ideally, modules should be completely independent of each other, and their
456 header files should only include the absolute minimum number of headers
457 possible. A module is not just a class, a function, or a namespace: <a
458 href="http://www.cuj.com/articles/2000/0002/0002c/0002c.htm">it's a collection
459 of these</a> that defines an interface. This interface may be several
460 functions, classes or data structures, but the important issue is how they work
463 <p>In general, a module should be implemented with one or more <tt>.cpp</tt>
464 files. Each of these <tt>.cpp</tt> files should include the header that defines
465 their interface first. This ensures that all of the dependences of the module
466 header have been properly added to the module header itself, and are not
467 implicit. System headers should be included after user headers for a
468 translation unit.</p>
472 <!-- _______________________________________________________________________ -->
473 <div class="doc_subsubsection">
474 <a name="hl_dontinclude"><tt>#include</tt> as Little as Possible</a>
477 <div class="doc_text">
479 <p><tt>#include</tt> hurts compile time performance. Don't do it unless you
480 have to, especially in header files.</p>
482 <p>But wait, sometimes you need to have the definition of a class to use it, or
483 to inherit from it. In these cases go ahead and <tt>#include</tt> that header
484 file. Be aware however that there are many cases where you don't need to have
485 the full definition of a class. If you are using a pointer or reference to a
486 class, you don't need the header file. If you are simply returning a class
487 instance from a prototyped function or method, you don't need it. In fact, for
488 most cases, you simply don't need the definition of a class... and not
489 <tt>#include</tt>'ing speeds up compilation.</p>
491 <p>It is easy to try to go too overboard on this recommendation, however. You
492 <b>must</b> include all of the header files that you are using -- you can
493 include them either directly
494 or indirectly (through another header file). To make sure that you don't
495 accidentally forget to include a header file in your module header, make sure to
496 include your module header <b>first</b> in the implementation file (as mentioned
497 above). This way there won't be any hidden dependencies that you'll find out
502 <!-- _______________________________________________________________________ -->
503 <div class="doc_subsubsection">
504 <a name="hl_privateheaders">Keep "internal" Headers Private</a>
507 <div class="doc_text">
509 <p>Many modules have a complex implementation that causes them to use more than
510 one implementation (<tt>.cpp</tt>) file. It is often tempting to put the
511 internal communication interface (helper classes, extra functions, etc) in the
512 public module header file. Don't do this.</p>
514 <p>If you really need to do something like this, put a private header file in
515 the same directory as the source files, and include it locally. This ensures
516 that your private interface remains private and undisturbed by outsiders.</p>
518 <p>Note however, that it's okay to put extra implementation methods a public
519 class itself... just make them private (or protected), and all is well.</p>
523 <!-- _______________________________________________________________________ -->
524 <div class="doc_subsubsection">
525 <a name="hl_earlyexit">Use Early Exits and 'continue' to Simplify Code</a>
528 <div class="doc_text">
530 <p>When reading code, keep in mind how much state and how many previous
531 decisions have to be remembered by the reader to understand a block of code.
532 Aim to reduce indentation where possible when it doesn't make it more difficult
533 to understand the code. One great way to do this is by making use of early
534 exits and the 'continue' keyword in long loops. As an example of using an early
535 exit from a function, consider this "bad" code:</p>
537 <div class="doc_code">
539 Value *DoSomething(Instruction *I) {
540 if (!isa<TerminatorInst>(I) &&
541 I->hasOneUse() && SomeOtherThing(I)) {
542 ... some long code ....
550 <p>This code has several problems if the body of the 'if' is large. When you're
551 looking at the top of the function, it isn't immediately clear that this
552 <em>only</em> does interesting things with non-terminator instructions, and only
553 applies to things with the other predicates. Second, it is relatively difficult
554 to describe (in comments) why these predicates are important because the if
555 statement makes it difficult to lay out the comments. Third, when you're deep
556 within the body of the code, it is indented an extra level. Finally, when
557 reading the top of the function, it isn't clear what the result is if the
558 predicate isn't true, you have to read to the end of the function to know that
561 <p>It is much preferred to format the code like this:</p>
563 <div class="doc_code">
565 Value *DoSomething(Instruction *I) {
566 // Terminators never need 'something' done to them because, ...
567 if (isa<TerminatorInst>(I))
570 // We conservatively avoid transforming instructions with multiple uses
571 // because goats like cheese.
572 if (!I->hasOneUse())
575 // This is really just here for example.
576 if (!SomeOtherThing(I))
579 ... some long code ....
584 <p>This fixes these problems. A similar problem frequently happens in <tt>for</tt>
585 loops. A silly example is something like this:</p>
587 <div class="doc_code">
589 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
590 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
591 Value *LHS = BO->getOperand(0);
592 Value *RHS = BO->getOperand(1);
601 <p>When you have very very small loops, this sort of structure is fine, but if
602 it exceeds more than 10-15 lines, it becomes difficult for people to read and
603 understand at a glance.
604 The problem with this sort of code is that it gets very nested very quickly,
605 meaning that the reader of the code has to keep a lot of context in their brain
606 to remember what is going immediately on in the loop, because they don't know
607 if/when the if conditions will have elses etc. It is strongly preferred to
608 structure the loop like this:</p>
610 <div class="doc_code">
612 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
613 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
616 Value *LHS = BO->getOperand(0);
617 Value *RHS = BO->getOperand(1);
618 if (LHS == RHS) continue;
624 <p>This has all the benefits of using early exits from functions: it reduces
625 nesting of the loop, it makes it easier to describe why the conditions are true,
626 and it makes it obvious to the reader that there is no "else" coming up that
627 they have to push context into their brain for. If a loop is large, this can
628 be a big understandability win.</p>
632 <!-- _______________________________________________________________________ -->
633 <div class="doc_subsubsection">
634 <a name="hl_else_after_return">Don't use "else" after a return</a>
637 <div class="doc_text">
639 <p>For similar reasons above (reduction of indentation and easier reading),
640 please do not use "else" or "else if" after something that interrupts
641 control flow like return, break, continue, goto, etc. For example, this is
644 <div class="doc_code">
648 Type = Context.getsigjmp_bufType();
650 Error = ASTContext::GE_Missing_sigjmp_buf;
656 Type = Context.getjmp_bufType();
658 Error = ASTContext::GE_Missing_jmp_buf;
669 <p>It is better to write this something like:</p>
671 <div class="doc_code">
675 Type = Context.getsigjmp_bufType();
677 Error = ASTContext::GE_Missing_sigjmp_buf;
681 Type = Context.getjmp_bufType();
683 Error = ASTContext::GE_Missing_jmp_buf;
691 <p>Or better yet (in this case), as:</p>
693 <div class="doc_code">
697 Type = Context.getsigjmp_bufType();
699 Type = Context.getjmp_bufType();
702 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
703 ASTContext::GE_Missing_jmp_buf;
710 <p>The idea is to reduce indentation and the amount of code you have to keep
711 track of when reading the code.</p>
715 <!-- _______________________________________________________________________ -->
716 <div class="doc_subsubsection">
717 <a name="hl_predicateloops">Turn Predicate Loops into Predicate Functions</a>
720 <div class="doc_text">
722 <p>It is very common to write small loops that just compute a boolean
723 value. There are a number of ways that people commonly write these, but an
724 example of this sort of thing is:</p>
726 <div class="doc_code">
728 <b>bool FoundFoo = false;</b>
729 for (unsigned i = 0, e = BarList.size(); i != e; ++i)
730 if (BarList[i]->isFoo()) {
731 <b>FoundFoo = true;</b>
735 <b>if (FoundFoo) {</b>
741 <p>This sort of code is awkward to write, and is almost always a bad sign.
742 Instead of this sort of loop, we strongly prefer to use a predicate function
743 (which may be <a href="#micro_anonns">static</a>) that uses
744 <a href="#hl_earlyexit">early exits</a> to compute the predicate. We prefer
745 the code to be structured like this:
749 <div class="doc_code">
751 /// ListContainsFoo - Return true if the specified list has an element that is
753 static bool ListContainsFoo(const std::vector<Bar*> &List) {
754 for (unsigned i = 0, e = List.size(); i != e; ++i)
755 if (List[i]->isFoo())
761 <b>if (ListContainsFoo(BarList)) {</b>
767 <p>There are many reasons for doing this: it reduces indentation and factors out
768 code which can often be shared by other code that checks for the same predicate.
769 More importantly, it <em>forces you to pick a name</em> for the function, and
770 forces you to write a comment for it. In this silly example, this doesn't add
771 much value. However, if the condition is complex, this can make it a lot easier
772 for the reader to understand the code that queries for this predicate. Instead
773 of being faced with the in-line details of how we check to see if the BarList
774 contains a foo, we can trust the function name and continue reading with better
780 <!-- ======================================================================= -->
781 <div class="doc_subsection">
782 <a name="micro">The Low-Level Issues</a>
784 <!-- ======================================================================= -->
787 <!-- _______________________________________________________________________ -->
788 <div class="doc_subsubsection">
789 <a name="ll_assert">Assert Liberally</a>
792 <div class="doc_text">
794 <p>Use the "<tt>assert</tt>" macro to its fullest. Check all of your
795 preconditions and assumptions, you never know when a bug (not necessarily even
796 yours) might be caught early by an assertion, which reduces debugging time
797 dramatically. The "<tt><cassert></tt>" header file is probably already
798 included by the header files you are using, so it doesn't cost anything to use
801 <p>To further assist with debugging, make sure to put some kind of error message
802 in the assertion statement (which is printed if the assertion is tripped). This
803 helps the poor debugger make sense of why an assertion is being made and
804 enforced, and hopefully what to do about it. Here is one complete example:</p>
806 <div class="doc_code">
808 inline Value *getOperand(unsigned i) {
809 assert(i < Operands.size() && "getOperand() out of range!");
815 <p>Here are some examples:</p>
817 <div class="doc_code">
819 assert(Ty->isPointerType() && "Can't allocate a non pointer type!");
821 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
823 assert(idx < getNumSuccessors() && "Successor # out of range!");
825 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
827 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
831 <p>You get the idea...</p>
833 <p>Please be aware when adding assert statements that not all compilers are aware of
834 the semantics of the assert. In some places, asserts are used to indicate a piece of
835 code that should not be reached. These are typically of the form:</p>
837 <div class="doc_code">
839 assert(0 && "Some helpful error message");
843 <p>When used in a function that returns a value, they should be followed with a return
844 statement and a comment indicating that this line is never reached. This will prevent
845 a compiler which is unable to deduce that the assert statement never returns from
846 generating a warning.</p>
848 <div class="doc_code">
850 assert(0 && "Some helpful error message");
856 <p>Another issue is that values used only by assertions will produce an "unused
857 value" warning when assertions are disabled. For example, this code will warn:
860 <div class="doc_code">
862 unsigned Size = V.size();
863 assert(Size > 42 && "Vector smaller than it should be");
865 bool NewToSet = Myset.insert(Value);
866 assert(NewToSet && "The value shouldn't be in the set yet");
870 <p>These are two interesting different cases: in the first case, the call to
871 V.size() is only useful for the assert, and we don't want it executed when
872 assertions are disabled. Code like this should move the call into the assert
873 itself. In the second case, the side effects of the call must happen whether
874 the assert is enabled or not. In this case, the value should be cast to void
875 to disable the warning. To be specific, it is preferred to write the code
878 <div class="doc_code">
880 assert(V.size() > 42 && "Vector smaller than it should be");
882 bool NewToSet = Myset.insert(Value); (void)NewToSet;
883 assert(NewToSet && "The value shouldn't be in the set yet");
890 <!-- _______________________________________________________________________ -->
891 <div class="doc_subsubsection">
892 <a name="ll_ns_std">Do not use '<tt>using namespace std</tt>'</a>
895 <div class="doc_text">
896 <p>In LLVM, we prefer to explicitly prefix all identifiers from the standard
897 namespace with an "<tt>std::</tt>" prefix, rather than rely on
898 "<tt>using namespace std;</tt>".</p>
900 <p> In header files, adding a '<tt>using namespace XXX</tt>' directive pollutes
901 the namespace of any source file that <tt>#include</tt>s the header. This is
902 clearly a bad thing.</p>
904 <p>In implementation files (e.g. <tt>.cpp</tt> files), the rule is more of a stylistic
905 rule, but is still important. Basically, using explicit namespace prefixes
906 makes the code <b>clearer</b>, because it is immediately obvious what facilities
907 are being used and where they are coming from, and <b>more portable</b>, because
908 namespace clashes cannot occur between LLVM code and other namespaces. The
909 portability rule is important because different standard library implementations
910 expose different symbols (potentially ones they shouldn't), and future revisions
911 to the C++ standard will add more symbols to the <tt>std</tt> namespace. As
912 such, we never use '<tt>using namespace std;</tt>' in LLVM.</p>
914 <p>The exception to the general rule (i.e. it's not an exception for
915 the <tt>std</tt> namespace) is for implementation files. For example, all of
916 the code in the LLVM project implements code that lives in the 'llvm' namespace.
917 As such, it is ok, and actually clearer, for the <tt>.cpp</tt> files to have a
918 '<tt>using namespace llvm</tt>' directive at their top, after the
919 <tt>#include</tt>s. This reduces indentation in the body of the file for source
920 editors that indent based on braces, and keeps the conceptual context cleaner.
921 The general form of this rule is that any <tt>.cpp</tt> file that implements
922 code in any namespace may use that namespace (and its parents'), but should not
927 <!-- _______________________________________________________________________ -->
928 <div class="doc_subsubsection">
929 <a name="ll_virtual_anch">Provide a virtual method anchor for classes
933 <div class="doc_text">
935 <p>If a class is defined in a header file and has a v-table (either it has
936 virtual methods or it derives from classes with virtual methods), it must
937 always have at least one out-of-line virtual method in the class. Without
938 this, the compiler will copy the vtable and RTTI into every <tt>.o</tt> file
939 that <tt>#include</tt>s the header, bloating <tt>.o</tt> file sizes and
940 increasing link times.</p>
944 <!-- _______________________________________________________________________ -->
945 <div class="doc_subsubsection">
946 <a name="ll_end">Don't evaluate end() every time through a loop</a>
949 <div class="doc_text">
951 <p>Because C++ doesn't have a standard "foreach" loop (though it can be emulated
952 with macros and may be coming in C++'0x) we end up writing a lot of loops that
953 manually iterate from begin to end on a variety of containers or through other
954 data structures. One common mistake is to write a loop in this style:</p>
956 <div class="doc_code">
959 for (BasicBlock::iterator I = BB->begin(); I != <b>BB->end()</b>; ++I)
964 <p>The problem with this construct is that it evaluates "<tt>BB->end()</tt>"
965 every time through the loop. Instead of writing the loop like this, we strongly
966 prefer loops to be written so that they evaluate it once before the loop starts.
967 A convenient way to do this is like so:</p>
969 <div class="doc_code">
972 for (BasicBlock::iterator I = BB->begin(), E = <b>BB->end()</b>; I != E; ++I)
977 <p>The observant may quickly point out that these two loops may have different
978 semantics: if the container (a basic block in this case) is being mutated, then
979 "<tt>BB->end()</tt>" may change its value every time through the loop and the
980 second loop may not in fact be correct. If you actually do depend on this
981 behavior, please write the loop in the first form and add a comment indicating
982 that you did it intentionally.</p>
984 <p>Why do we prefer the second form (when correct)? Writing the loop in the
985 first form has two problems: First it may be less efficient than evaluating it
986 at the start of the loop. In this case, the cost is probably minor: a few extra
987 loads every time through the loop. However, if the base expression is more
988 complex, then the cost can rise quickly. I've seen loops where the end
989 expression was actually something like: "<tt>SomeMap[x]->end()</tt>" and map
990 lookups really aren't cheap. By writing it in the second form consistently, you
991 eliminate the issue entirely and don't even have to think about it.</p>
993 <p>The second (even bigger) issue is that writing the loop in the first form
994 hints to the reader that the loop is mutating the container (a fact that a
995 comment would handily confirm!). If you write the loop in the second form, it
996 is immediately obvious without even looking at the body of the loop that the
997 container isn't being modified, which makes it easier to read the code and
998 understand what it does.</p>
1000 <p>While the second form of the loop is a few extra keystrokes, we do strongly
1005 <!-- _______________________________________________________________________ -->
1006 <div class="doc_subsubsection">
1007 <a name="ll_iostream"><tt>#include <iostream></tt> is forbidden</a>
1010 <div class="doc_text">
1012 <p>The use of <tt>#include <iostream></tt> in library files is
1013 hereby <b><em>forbidden</em></b>. The primary reason for doing this is to
1014 support clients using LLVM libraries as part of larger systems. In particular,
1015 we statically link LLVM into some dynamic libraries. Even if LLVM isn't used,
1016 the static c'tors are run whenever an application start up that uses the dynamic
1017 library. There are two problems with this:</p>
1020 <li>The time to run the static c'tors impacts startup time of
1021 applications—a critical time for GUI apps.</li>
1022 <li>The static c'tors cause the app to pull many extra pages of memory off the
1023 disk: both the code for the static c'tors in each <tt>.o</tt> file and the
1024 small amount of data that gets touched. In addition, touched/dirty pages
1025 put more pressure on the VM system on low-memory machines.</li>
1028 <p>Note that using the other stream headers (<tt><sstream></tt> for
1029 example) is not problematic in this regard (just <tt><iostream></tt>).
1030 However, <tt>raw_ostream</tt> provides various APIs that are better performing for almost
1031 every use than <tt>std::ostream</tt> style APIs.
1032 <b>Therefore new code should always
1033 use <a href="#ll_raw_ostream"><tt>raw_ostream</tt></a> for writing, or
1034 the <tt>llvm::MemoryBuffer</tt> API for reading files.</b></p>
1039 <!-- _______________________________________________________________________ -->
1040 <div class="doc_subsubsection">
1041 <a name="ll_raw_ostream">Use <tt>raw_ostream</tt></a>
1044 <div class="doc_text">
1046 <p>LLVM includes a lightweight, simple, and efficient stream implementation
1047 in <tt>llvm/Support/raw_ostream.h</tt> which provides all of the common features
1048 of <tt>std::ostream</tt>. All new code should use <tt>raw_ostream</tt> instead
1049 of <tt>ostream</tt>.</p>
1051 <p>Unlike <tt>std::ostream</tt>, <tt>raw_ostream</tt> is not a template and can
1052 be forward declared as <tt>class raw_ostream</tt>. Public headers should
1053 generally not include the <tt>raw_ostream</tt> header, but use forward
1054 declarations and constant references to <tt>raw_ostream</tt> instances.</p>
1059 <!-- _______________________________________________________________________ -->
1060 <div class="doc_subsubsection">
1061 <a name="ll_avoidendl">Avoid <tt>std::endl</tt></a>
1064 <div class="doc_text">
1066 <p>The <tt>std::endl</tt> modifier, when used with iostreams outputs a newline
1067 to the output stream specified. In addition to doing this, however, it also
1068 flushes the output stream. In other words, these are equivalent:</p>
1070 <div class="doc_code">
1072 std::cout << std::endl;
1073 std::cout << '\n' << std::flush;
1077 <p>Most of the time, you probably have no reason to flush the output stream, so
1078 it's better to use a literal <tt>'\n'</tt>.</p>
1083 <!-- ======================================================================= -->
1084 <div class="doc_subsection">
1085 <a name="nano">Microscopic Details</a>
1087 <!-- ======================================================================= -->
1089 <p>This section describes preferred low-level formatting guidelines along with
1090 reasoning on why we prefer them.</p>
1092 <!-- _______________________________________________________________________ -->
1093 <div class="doc_subsubsection">
1094 <a name="micro_spaceparen">Spaces Before Parentheses</a>
1097 <div class="doc_text">
1099 <p>We prefer to put a space before an open parenthesis only in control flow
1100 statements, but not in normal function call expressions and function-like
1101 macros. For example, this is good:</p>
1103 <div class="doc_code">
1106 <b>for (</b>i = 0; i != 100; ++i) ...
1107 <b>while (</b>llvm_rocks) ...
1109 <b>somefunc(</b>42);
1110 <b><a href="#ll_assert">assert</a>(</b>3 != 4 && "laws of math are failing me");
1112 a = <b>foo(</b>42, 92) + <b>bar(</b>x);
1116 <p>... and this is bad:</p>
1118 <div class="doc_code">
1121 <b>for(</b>i = 0; i != 100; ++i) ...
1122 <b>while(</b>llvm_rocks) ...
1124 <b>somefunc (</b>42);
1125 <b><a href="#ll_assert">assert</a> (</b>3 != 4 && "laws of math are failing me");
1127 a = <b>foo (</b>42, 92) + <b>bar (</b>x);
1131 <p>The reason for doing this is not completely arbitrary. This style makes
1132 control flow operators stand out more, and makes expressions flow better. The
1133 function call operator binds very tightly as a postfix operator. Putting
1134 a space after a function name (as in the last example) makes it appear that
1135 the code might bind the arguments of the left-hand-side of a binary operator
1136 with the argument list of a function and the name of the right side. More
1137 specifically, it is easy to misread the "a" example as:</p>
1139 <div class="doc_code">
1141 a = foo <b>(</b>(42, 92) + bar<b>)</b> (x);
1145 <p>... when skimming through the code. By avoiding a space in a function, we
1146 avoid this misinterpretation.</p>
1150 <!-- _______________________________________________________________________ -->
1151 <div class="doc_subsubsection">
1152 <a name="micro_preincrement">Prefer Preincrement</a>
1155 <div class="doc_text">
1157 <p>Hard fast rule: Preincrement (<tt>++X</tt>) may be no slower than
1158 postincrement (<tt>X++</tt>) and could very well be a lot faster than it. Use
1159 preincrementation whenever possible.</p>
1161 <p>The semantics of postincrement include making a copy of the value being
1162 incremented, returning it, and then preincrementing the "work value". For
1163 primitive types, this isn't a big deal... but for iterators, it can be a huge
1164 issue (for example, some iterators contains stack and set objects in them...
1165 copying an iterator could invoke the copy ctor's of these as well). In general,
1166 get in the habit of always using preincrement, and you won't have a problem.</p>
1170 <!-- _______________________________________________________________________ -->
1171 <div class="doc_subsubsection">
1172 <a name="micro_namespaceindent">Namespace Indentation</a>
1175 <div class="doc_text">
1178 In general, we strive to reduce indentation wherever possible. This is useful
1179 because we want code to <a href="#scf_codewidth">fit into 80 columns</a> without
1180 wrapping horribly, but also because it makes it easier to understand the code.
1181 Namespaces are a funny thing: they are often large, and we often desire to put
1182 lots of stuff into them (so they can be large). Other times they are tiny,
1183 because they just hold an enum or something similar. In order to balance this,
1184 we use different approaches for small versus large namespaces.
1188 If a namespace definition is small and <em>easily</em> fits on a screen (say,
1189 less than 35 lines of code), then you should indent its body. Here's an
1193 <div class="doc_code">
1197 /// RelocationType - An enum for the x86 relocation codes. Note that
1198 /// the terminology here doesn't follow x86 convention - word means
1199 /// 32-bit and dword means 64-bit.
1200 enum RelocationType {
1201 /// reloc_pcrel_word - PC relative relocation, add the relocated value to
1202 /// the value already in memory, after we adjust it for where the PC is.
1203 reloc_pcrel_word = 0,
1205 /// reloc_picrel_word - PIC base relative relocation, add the relocated
1206 /// value to the value already in memory, after we adjust it for where the
1208 reloc_picrel_word = 1,
1210 /// reloc_absolute_word, reloc_absolute_dword - Absolute relocation, just
1211 /// add the relocated value to the value already in memory.
1212 reloc_absolute_word = 2,
1213 reloc_absolute_dword = 3
1220 <p>Since the body is small, indenting adds value because it makes it very clear
1221 where the namespace starts and ends, and it is easy to take the whole thing in
1222 in one "gulp" when reading the code. If the blob of code in the namespace is
1223 larger (as it typically is in a header in the <tt>llvm</tt> or <tt>clang</tt> namespaces), do not
1224 indent the code, and add a comment indicating what namespace is being closed.
1227 <div class="doc_code">
1230 namespace knowledge {
1232 /// Grokable - This class represents things that Smith can have an intimate
1233 /// understanding of and contains the data associated with it.
1237 explicit Grokable() { ... }
1238 virtual ~Grokable() = 0;
1244 } // end namespace knowledge
1245 } // end namespace llvm
1249 <p>Because the class is large, we don't expect that the reader can easily
1250 understand the entire concept in a glance, and the end of the file (where the
1251 namespaces end) may be a long ways away from the place they open. As such,
1252 indenting the contents of the namespace doesn't add any value, and detracts from
1253 the readability of the class. In these cases it is best to <em>not</em> indent
1254 the contents of the namespace.</p>
1258 <!-- _______________________________________________________________________ -->
1259 <div class="doc_subsubsection">
1260 <a name="micro_anonns">Anonymous Namespaces</a>
1263 <div class="doc_text">
1265 <p>After talking about namespaces in general, you may be wondering about
1266 anonymous namespaces in particular.
1267 Anonymous namespaces are a great language feature that tells the C++ compiler
1268 that the contents of the namespace are only visible within the current
1269 translation unit, allowing more aggressive optimization and eliminating the
1270 possibility of symbol name collisions. Anonymous namespaces are to C++ as
1271 "static" is to C functions and global variables. While "static" is available
1272 in C++, anonymous namespaces are more general: they can make entire classes
1273 private to a file.</p>
1275 <p>The problem with anonymous namespaces is that they naturally want to
1276 encourage indentation of their body, and they reduce locality of reference: if
1277 you see a random function definition in a C++ file, it is easy to see if it is
1278 marked static, but seeing if it is in an anonymous namespace requires scanning
1279 a big chunk of the file.</p>
1281 <p>Because of this, we have a simple guideline: make anonymous namespaces as
1282 small as possible, and only use them for class declarations. For example, this
1285 <div class="doc_code">
1292 bool operator<(const char *RHS) const;
1294 <b>} // end anonymous namespace</b>
1296 static void Helper() {
1300 bool StringSort::operator<(const char *RHS) const {
1310 <div class="doc_code">
1317 bool operator<(const char *RHS) const;
1324 bool StringSort::operator<(const char *RHS) const {
1328 <b>} // end anonymous namespace</b>
1334 <p>This is bad specifically because if you're looking at "Helper" in the middle
1335 of a large C++ file, that you have no immediate way to tell if it is local to
1336 the file. When it is marked static explicitly, this is immediately obvious.
1337 Also, there is no reason to enclose the definition of "operator<" in the
1338 namespace just because it was declared there.
1345 <!-- *********************************************************************** -->
1346 <div class="doc_section">
1347 <a name="seealso">See Also</a>
1349 <!-- *********************************************************************** -->
1351 <div class="doc_text">
1353 <p>A lot of these comments and recommendations have been culled for other
1354 sources. Two particularly important books for our work are:</p>
1358 <li><a href="http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876">Effective
1359 C++</a> by Scott Meyers. Also
1360 interesting and useful are "More Effective C++" and "Effective STL" by the same
1363 <li>Large-Scale C++ Software Design by John Lakos</li>
1367 <p>If you get some free time, and you haven't read them: do so, you might learn
1372 <!-- *********************************************************************** -->
1376 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1377 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1378 <a href="http://validator.w3.org/check/referer"><img
1379 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1381 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1382 <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
1383 Last modified: $Date$