11 This document attempts to describe a few coding standards that are being used in
12 the LLVM source tree. Although no coding standards should be regarded as
13 absolute requirements to be followed in all instances, coding standards are
14 particularly important for large-scale code bases that follow a library-based
17 While this document may provide guidance for some mechanical formatting issues,
18 whitespace, or other "microscopic details", these are not fixed standards.
19 Always follow the golden rule:
23 **If you are extending, enhancing, or bug fixing already implemented code,
24 use the style that is already being used so that the source is uniform and
27 Note that some code bases (e.g. ``libc++``) have really good reasons to deviate
28 from the coding standards. In the case of ``libc++``, this is because the
29 naming and other conventions are dictated by the C++ standard. If you think
30 there is a specific good reason to deviate from the standards here, please bring
31 it up on the LLVMdev mailing list.
33 There are some conventions that are not uniformly followed in the code base
34 (e.g. the naming convention). This is because they are relatively new, and a
35 lot of code was written before they were put in place. Our long term goal is
36 for the entire codebase to follow the convention, but we explicitly *do not*
37 want patches that do large-scale reformating of existing code. On the other
38 hand, it is reasonable to rename the methods of a class if you're about to
39 change it in some other way. Just do the reformating as a separate commit from
40 the functionality change.
42 The ultimate goal of these guidelines is the increase readability and
43 maintainability of our common source base. If you have suggestions for topics to
44 be included, please mail them to `Chris <mailto:sabre@nondot.org>`_.
46 Languages, Libraries, and Standards
47 ===================================
49 Most source code in LLVM and other LLVM projects using these coding standards
50 is C++ code. There are some places where C code is used either due to
51 environment restrictions, historical restrictions, or due to third-party source
52 code imported into the tree. Generally, our preference is for standards
53 conforming, modern, and portable C++ code as the implementation language of
59 LLVM, Clang, and LLD are currently written using C++11 conforming code,
60 although we restrict ourselves to features which are available in the major
61 toolchains supported as host compilers. The LLDB project is even more
62 aggressive in the set of host compilers supported and thus uses still more
63 features. Regardless of the supported features, code is expected to (when
64 reasonable) be standard, portable, and modern C++11 code. We avoid unnecessary
65 vendor-specific extensions, etc.
70 Use the C++ standard library facilities whenever they are available for
71 a particular task. LLVM and related projects emphasize and rely on the standard
72 library facilities for as much as possible. Common support libraries providing
73 functionality missing from the standard library for which there are standard
74 interfaces or active work on adding standard interfaces will often be
75 implemented in the LLVM namespace following the expected standard interface.
77 There are some exceptions such as the standard I/O streams library which are
78 avoided. Also, there is much more detailed information on these subjects in the
79 :doc:`ProgrammersManual`.
81 Supported C++11 Language and Library Features
82 ---------------------------------------------
84 While LLVM, Clang, and LLD use C++11, not all features are available in all of
85 the toolchains which we support. The set of features supported for use in LLVM
86 is the intersection of those supported in MSVC 2012, GCC 4.7, and Clang 3.1.
87 The ultimate definition of this set is what build bots with those respective
88 toolchains accept. Don't argue with the build bots. However, we have some
89 guidance below to help you know what to expect.
91 Each toolchain provides a good reference for what it accepts:
93 * Clang: http://clang.llvm.org/cxx_status.html
94 * GCC: http://gcc.gnu.org/projects/cxx0x.html
95 * MSVC: http://msdn.microsoft.com/en-us/library/hh567368.aspx
97 In most cases, the MSVC list will be the dominating factor. Here is a summary
98 of the features that are expected to work. Features not on this list are
99 unlikely to be supported by our host compilers.
101 * Rvalue references: N2118_
103 * But *not* Rvalue references for ``*this`` or member qualifiers (N2439_)
105 * Static assert: N1720_
106 * ``auto`` type deduction: N1984_, N1737_
107 * Trailing return types: N2541_
110 * But *not* lambdas with default arguments.
112 * ``decltype``: N2343_
113 * Nested closing right angle brackets: N1757_
114 * Extern templates: N1987_
115 * ``nullptr``: N2431_
116 * Strongly-typed and forward declarable enums: N2347_, N2764_
117 * Local and unnamed types as template arguments: N2657_
118 * Range-based for-loop: N2930_
120 * But ``{}`` are required around inner ``do {} while()`` loops. As a result,
121 ``{}`` are required around function-like macros inside range-based for
124 * ``override`` and ``final``: N2928_, N3206_, N3272_
125 * Atomic operations and the C++11 memory model: N2429_
127 .. _N2118: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2118.html
128 .. _N2439: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2439.htm
129 .. _N1720: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1720.html
130 .. _N1984: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1984.pdf
131 .. _N1737: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1737.pdf
132 .. _N2541: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2541.htm
133 .. _N2927: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2927.pdf
134 .. _N2343: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2343.pdf
135 .. _N1757: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1757.html
136 .. _N1987: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1987.htm
137 .. _N2431: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2431.pdf
138 .. _N2347: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2347.pdf
139 .. _N2764: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2764.pdf
140 .. _N2657: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm
141 .. _N2930: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2930.html
142 .. _N2928: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2928.htm
143 .. _N3206: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3206.htm
144 .. _N3272: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3272.htm
145 .. _N2429: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2429.htm
146 .. _MSVC-compatible RTTI: http://llvm.org/PR18951
148 The supported features in the C++11 standard libraries are less well tracked,
149 but also much greater. Most of the standard libraries implement most of C++11's
150 library. The most likely lowest common denominator is Linux support. For
151 libc++, the support is just poorly tested and undocumented but expected to be
152 largely complete. YMMV. For libstdc++, the support is documented in detail in
153 `the libstdc++ manual`_. There are some very minor missing facilities that are
154 unlikely to be common problems, and there are a few larger gaps that are worth
157 * Not all of the type traits are implemented
158 * No regular expression library.
159 * While most of the atomics library is well implemented, the fences are
160 missing. Fortunately, they are rarely needed.
161 * The locale support is incomplete.
162 * ``std::initializer_list`` (and the constructors and functions that take it as
163 an argument) are not always available, so you cannot (for example) initialize
164 a ``std::vector`` with a braced initializer list.
165 * ``std::equal()`` (and other algorithms) incorrectly assert in MSVC when given
166 ``nullptr`` as an iterator.
168 Other than these areas you should assume the standard library is available and
169 working as expected until some build bot tells you otherwise. If you're in an
170 uncertain area of one of the above points, but you cannot test on a Linux
171 system, your best approach is to minimize your use of these features, and watch
172 the Linux build bots to find out if your usage triggered a bug. For example, if
173 you hit a type trait which doesn't work we can then add support to LLVM's
174 traits header to emulate it.
176 .. _the libstdc++ manual:
177 http://gcc.gnu.org/onlinedocs/gcc-4.7.3/libstdc++/manual/manual/status.html#status.iso.2011
182 Any code written in the Go programming language is not subject to the
183 formatting rules below. Instead, we adopt the formatting rules enforced by
186 Go code should strive to be idiomatic. Two good sets of guidelines for what
187 this means are `Effective Go`_ and `Go Code Review Comments`_.
190 https://golang.org/cmd/gofmt/
193 https://golang.org/doc/effective_go.html
195 .. _Go Code Review Comments:
196 https://code.google.com/p/go-wiki/wiki/CodeReviewComments
198 Mechanical Source Issues
199 ========================
201 Source Code Formatting
202 ----------------------
207 Comments are one critical part of readability and maintainability. Everyone
208 knows they should comment their code, and so should you. When writing comments,
209 write them as English prose, which means they should use proper capitalization,
210 punctuation, etc. Aim to describe what the code is trying to do and why, not
211 *how* it does it at a micro level. Here are a few critical things to document:
213 .. _header file comment:
218 Every source file should have a header on it that describes the basic purpose of
219 the file. If a file does not have a header, it should not be checked into the
220 tree. The standard header looks like this:
224 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
226 // The LLVM Compiler Infrastructure
228 // This file is distributed under the University of Illinois Open Source
229 // License. See LICENSE.TXT for details.
231 //===----------------------------------------------------------------------===//
234 /// \brief This file contains the declaration of the Instruction class, which is
235 /// the base class for all of the VM instructions.
237 //===----------------------------------------------------------------------===//
239 A few things to note about this particular format: The "``-*- C++ -*-``" string
240 on the first line is there to tell Emacs that the source file is a C++ file, not
241 a C file (Emacs assumes ``.h`` files are C files by default).
245 This tag is not necessary in ``.cpp`` files. The name of the file is also
246 on the first line, along with a very short description of the purpose of the
247 file. This is important when printing out code and flipping though lots of
250 The next section in the file is a concise note that defines the license that the
251 file is released under. This makes it perfectly clear what terms the source
252 code can be distributed under and should not be modified in any way.
254 The main body is a ``doxygen`` comment describing the purpose of the file. It
255 should have a ``\brief`` command that describes the file in one or two
256 sentences. Any additional information should be separated by a blank line. If
257 an algorithm is being implemented or something tricky is going on, a reference
258 to the paper where it is published should be included, as well as any notes or
259 *gotchas* in the code to watch out for.
264 Classes are one fundamental part of a good object oriented design. As such, a
265 class definition should have a comment block that explains what the class is
266 used for and how it works. Every non-trivial class is expected to have a
267 ``doxygen`` comment block.
272 Methods defined in a class (as well as any global functions) should also be
273 documented properly. A quick note about what it does and a description of the
274 borderline behaviour is all that is necessary here (unless something
275 particularly tricky or insidious is going on). The hope is that people can
276 figure out how to use your interfaces without reading the code itself.
278 Good things to talk about here are what happens when something unexpected
279 happens: does the method return null? Abort? Format your hard disk?
284 In general, prefer C++ style (``//``) comments. They take less space, require
285 less typing, don't have nesting problems, etc. There are a few cases when it is
286 useful to use C style (``/* */``) comments however:
288 #. When writing C code: Obviously if you are writing C code, use C style
291 #. When writing a header file that may be ``#include``\d by a C source file.
293 #. When writing a source file that is used by a tool that only accepts C style
296 To comment out a large block of code, use ``#if 0`` and ``#endif``. These nest
297 properly and are better behaved in general than C style comments.
299 Doxygen Use in Documentation Comments
300 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
302 Use the ``\file`` command to turn the standard file header into a file-level
305 Include descriptive ``\brief`` paragraphs for all public interfaces (public
306 classes, member and non-member functions). Explain API use and purpose in
307 ``\brief`` paragraphs, don't just restate the information that can be inferred
308 from the API name. Put detailed discussion into separate paragraphs.
310 To refer to parameter names inside a paragraph, use the ``\p name`` command.
311 Don't use the ``\arg name`` command since it starts a new paragraph that
312 contains documentation for the parameter.
314 Wrap non-inline code examples in ``\code ... \endcode``.
316 To document a function parameter, start a new paragraph with the
317 ``\param name`` command. If the parameter is used as an out or an in/out
318 parameter, use the ``\param [out] name`` or ``\param [in,out] name`` command,
321 To describe function return value, start a new paragraph with the ``\returns``
324 A minimal documentation comment:
328 /// \brief Does foo and bar.
329 void fooBar(bool Baz);
331 A documentation comment that uses all Doxygen features in a preferred way:
335 /// \brief Does foo and bar.
337 /// Does not do foo the usual way if \p Baz is true.
341 /// fooBar(false, "quux", Res);
344 /// \param Quux kind of foo to do.
345 /// \param [out] Result filled with bar sequence on foo success.
347 /// \returns true on success.
348 bool fooBar(bool Baz, StringRef Quux, std::vector<int> &Result);
350 Don't duplicate the documentation comment in the header file and in the
351 implementation file. Put the documentation comments for public APIs into the
352 header file. Documentation comments for private APIs can go to the
353 implementation file. In any case, implementation files can include additional
354 comments (not necessarily in Doxygen markup) to explain implementation details
357 Don't duplicate function or class name at the beginning of the comment.
358 For humans it is obvious which function or class is being documented;
359 automatic documentation processing tools are smart enough to bind the comment
360 to the correct declaration.
368 /// Something - An abstraction for some complicated thing.
371 /// fooBar - Does foo and bar.
377 /// fooBar - Does foo and bar.
378 void Something::fooBar() { ... }
386 /// \brief An abstraction for some complicated thing.
389 /// \brief Does foo and bar.
395 // Builds a B-tree in order to do foo. See paper by...
396 void Something::fooBar() { ... }
398 It is not required to use additional Doxygen features, but sometimes it might
399 be a good idea to do so.
403 * adding comments to any narrow namespace containing a collection of
404 related functions or types;
406 * using top-level groups to organize a collection of related functions at
407 namespace scope where the grouping is smaller than the namespace;
409 * using member groups and additional comments attached to member
410 groups to organize within a class.
417 /// \name Functions that do Foo.
428 Immediately after the `header file comment`_ (and include guards if working on a
429 header file), the `minimal list of #includes`_ required by the file should be
430 listed. We prefer these ``#include``\s to be listed in this order:
432 .. _Main Module Header:
433 .. _Local/Private Headers:
435 #. Main Module Header
436 #. Local/Private Headers
438 #. System ``#include``\s
440 and each category should be sorted lexicographically by the full path.
442 The `Main Module Header`_ file applies to ``.cpp`` files which implement an
443 interface defined by a ``.h`` file. This ``#include`` should always be included
444 **first** regardless of where it lives on the file system. By including a
445 header file first in the ``.cpp`` files that implement the interfaces, we ensure
446 that the header does not have any hidden dependencies which are not explicitly
447 ``#include``\d in the header, but should be. It is also a form of documentation
448 in the ``.cpp`` file to indicate where the interfaces it implements are defined.
450 .. _fit into 80 columns:
455 Write your code to fit within 80 columns of text. This helps those of us who
456 like to print out code and look at your code in an ``xterm`` without resizing
459 The longer answer is that there must be some limit to the width of the code in
460 order to reasonably allow developers to have multiple files side-by-side in
461 windows on a modest display. If you are going to pick a width limit, it is
462 somewhat arbitrary but you might as well pick something standard. Going with 90
463 columns (for example) instead of 80 columns wouldn't add any significant value
464 and would be detrimental to printing out code. Also many other projects have
465 standardized on 80 columns, so some people have already configured their editors
466 for it (vs something else, like 90 columns).
468 This is one of many contentious issues in coding standards, but it is not up for
471 Use Spaces Instead of Tabs
472 ^^^^^^^^^^^^^^^^^^^^^^^^^^
474 In all cases, prefer spaces to tabs in source files. People have different
475 preferred indentation levels, and different styles of indentation that they
476 like; this is fine. What isn't fine is that different editors/viewers expand
477 tabs out to different tab stops. This can cause your code to look completely
478 unreadable, and it is not worth dealing with.
480 As always, follow the `Golden Rule`_ above: follow the style of
481 existing code if you are modifying and extending it. If you like four spaces of
482 indentation, **DO NOT** do that in the middle of a chunk of code with two spaces
483 of indentation. Also, do not reindent a whole source file: it makes for
484 incredible diffs that are absolutely worthless.
486 Indent Code Consistently
487 ^^^^^^^^^^^^^^^^^^^^^^^^
489 Okay, in your first year of programming you were told that indentation is
490 important. If you didn't believe and internalize this then, now is the time.
491 Just do it. With the introduction of C++11, there are some new formatting
492 challenges that merit some suggestions to help have consistent, maintainable,
493 and tool-friendly formatting and indentation.
495 Format Lambdas Like Blocks Of Code
496 """"""""""""""""""""""""""""""""""
498 When formatting a multi-line lambda, format it like a block of code, that's
499 what it is. If there is only one multi-line lambda in a statement, and there
500 are no expressions lexically after it in the statement, drop the indent to the
501 standard two space indent for a block of code, as if it were an if-block opened
502 by the preceding part of the statement:
506 std::sort(foo.begin(), foo.end(), [&](Foo a, Foo b) -> bool {
511 return a.bam < b.bam;
514 To take best advantage of this formatting, if you are designing an API which
515 accepts a continuation or single callable argument (be it a functor, or
516 a ``std::function``), it should be the last argument if at all possible.
518 If there are multiple multi-line lambdas in a statement, or there is anything
519 interesting after the lambda in the statement, indent the block two spaces from
520 the indent of the ``[]``:
524 dyn_switch(V->stripPointerCasts(),
528 [] (SelectInst *SI) {
529 // process selects...
534 [] (AllocaInst *AI) {
535 // process allocas...
538 Braced Initializer Lists
539 """"""""""""""""""""""""
541 With C++11, there are significantly more uses of braced lists to perform
542 initialization. These allow you to easily construct aggregate temporaries in
543 expressions among other niceness. They now have a natural way of ending up
544 nested within each other and within function calls in order to build up
545 aggregates (such as option structs) from local variables. To make matters
546 worse, we also have many more uses of braces in an expression context that are
547 *not* performing initialization.
549 The historically common formatting of braced initialization of aggregate
550 variables does not mix cleanly with deep nesting, general expression contexts,
551 function arguments, and lambdas. We suggest new code use a simple rule for
552 formatting braced initialization lists: act as-if the braces were parentheses
553 in a function call. The formatting rules exactly match those already well
554 understood for formatting nested function calls. Examples:
558 foo({a, b, c}, {1, 2, 3});
560 llvm::Constant *Mask[] = {
561 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 0),
562 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 1),
563 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 2)};
565 This formatting scheme also makes it particularly easy to get predictable,
566 consistent, and automatic formatting with tools like `Clang Format`_.
568 .. _Clang Format: http://clang.llvm.org/docs/ClangFormat.html
570 Language and Compiler Issues
571 ----------------------------
573 Treat Compiler Warnings Like Errors
574 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
576 If your code has compiler warnings in it, something is wrong --- you aren't
577 casting values correctly, you have "questionable" constructs in your code, or
578 you are doing something legitimately wrong. Compiler warnings can cover up
579 legitimate errors in output and make dealing with a translation unit difficult.
581 It is not possible to prevent all warnings from all compilers, nor is it
582 desirable. Instead, pick a standard compiler (like ``gcc``) that provides a
583 good thorough set of warnings, and stick to it. At least in the case of
584 ``gcc``, it is possible to work around any spurious errors by changing the
585 syntax of the code slightly. For example, a warning that annoys me occurs when
586 I write code like this:
590 if (V = getValue()) {
594 ``gcc`` will warn me that I probably want to use the ``==`` operator, and that I
595 probably mistyped it. In most cases, I haven't, and I really don't want the
596 spurious errors. To fix this particular problem, I rewrite the code like
601 if ((V = getValue())) {
605 which shuts ``gcc`` up. Any ``gcc`` warning that annoys you can be fixed by
606 massaging the code appropriately.
611 In almost all cases, it is possible and within reason to write completely
612 portable code. If there are cases where it isn't possible to write portable
613 code, isolate it behind a well defined (and well documented) interface.
615 In practice, this means that you shouldn't assume much about the host compiler
616 (and Visual Studio tends to be the lowest common denominator). If advanced
617 features are used, they should only be an implementation detail of a library
618 which has a simple exposed API, and preferably be buried in ``libSystem``.
620 Do not use RTTI or Exceptions
621 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
623 In an effort to reduce code and executable size, LLVM does not use RTTI
624 (e.g. ``dynamic_cast<>;``) or exceptions. These two language features violate
625 the general C++ principle of *"you only pay for what you use"*, causing
626 executable bloat even if exceptions are never used in the code base, or if RTTI
627 is never used for a class. Because of this, we turn them off globally in the
630 That said, LLVM does make extensive use of a hand-rolled form of RTTI that use
631 templates like :ref:`isa\<>, cast\<>, and dyn_cast\<> <isa>`.
632 This form of RTTI is opt-in and can be
633 :doc:`added to any class <HowToSetUpLLVMStyleRTTI>`. It is also
634 substantially more efficient than ``dynamic_cast<>``.
636 .. _static constructor:
638 Do not use Static Constructors
639 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
641 Static constructors and destructors (e.g. global variables whose types have a
642 constructor or destructor) should not be added to the code base, and should be
643 removed wherever possible. Besides `well known problems
644 <http://yosefk.com/c++fqa/ctors.html#fqa-10.12>`_ where the order of
645 initialization is undefined between globals in different source files, the
646 entire concept of static constructors is at odds with the common use case of
647 LLVM as a library linked into a larger application.
649 Consider the use of LLVM as a JIT linked into another application (perhaps for
650 `OpenGL, custom languages <http://llvm.org/Users.html>`_, `shaders in movies
651 <http://llvm.org/devmtg/2010-11/Gritz-OpenShadingLang.pdf>`_, etc). Due to the
652 design of static constructors, they must be executed at startup time of the
653 entire application, regardless of whether or how LLVM is used in that larger
654 application. There are two problems with this:
656 * The time to run the static constructors impacts startup time of applications
657 --- a critical time for GUI apps, among others.
659 * The static constructors cause the app to pull many extra pages of memory off
660 the disk: both the code for the constructor in each ``.o`` file and the small
661 amount of data that gets touched. In addition, touched/dirty pages put more
662 pressure on the VM system on low-memory machines.
664 We would really like for there to be zero cost for linking in an additional LLVM
665 target or other library into an application, but static constructors violate
668 That said, LLVM unfortunately does contain static constructors. It would be a
669 `great project <http://llvm.org/PR11944>`_ for someone to purge all static
670 constructors from LLVM, and then enable the ``-Wglobal-constructors`` warning
671 flag (when building with Clang) to ensure we do not regress in the future.
673 Use of ``class`` and ``struct`` Keywords
674 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
676 In C++, the ``class`` and ``struct`` keywords can be used almost
677 interchangeably. The only difference is when they are used to declare a class:
678 ``class`` makes all members private by default while ``struct`` makes all
679 members public by default.
681 Unfortunately, not all compilers follow the rules and some will generate
682 different symbols based on whether ``class`` or ``struct`` was used to declare
683 the symbol (e.g., MSVC). This can lead to problems at link time.
685 * All declarations and definitions of a given ``class`` or ``struct`` must use
686 the same keyword. For example:
692 // Breaks mangling in MSVC.
693 struct Foo { int Data; };
695 * As a rule of thumb, ``struct`` should be kept to structures where *all*
696 members are declared public.
700 // Foo feels like a class... this is strange.
706 int getData() const { return Data; }
707 void setData(int D) { Data = D; }
710 // Bar isn't POD, but it does look like a struct.
716 Do not use Braced Initializer Lists to Call a Constructor
717 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
719 In C++11 there is a "generalized initialization syntax" which allows calling
720 constructors using braced initializer lists. Do not use these to call
721 constructors with any interesting logic or if you care that you're calling some
722 *particular* constructor. Those should look like function calls using
723 parentheses rather than like aggregate initialization. Similarly, if you need
724 to explicitly name the type and call its constructor to create a temporary,
725 don't use a braced initializer list. Instead, use a braced initializer list
726 (without any type for temporaries) when doing aggregate initialization or
727 something notionally equivalent. Examples:
733 // Construct a Foo by reading data from the disk in the whizbang format, ...
734 Foo(std::string filename);
736 // Construct a Foo by looking up the Nth element of some global data ...
742 // The Foo constructor call is very deliberate, no braces.
743 std::fill(foo.begin(), foo.end(), Foo("name"));
745 // The pair is just being constructed like an aggregate, use braces.
746 bar_map.insert({my_key, my_value});
748 If you use a braced initializer list when initializing a variable, use an equals before the open curly brace:
752 int data[] = {0, 1, 2, 3};
754 Use ``auto`` Type Deduction to Make Code More Readable
755 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
757 Some are advocating a policy of "almost always ``auto``" in C++11, however LLVM
758 uses a more moderate stance. Use ``auto`` if and only if it makes the code more
759 readable or easier to maintain. Don't "almost always" use ``auto``, but do use
760 ``auto`` with initializers like ``cast<Foo>(...)`` or other places where the
761 type is already obvious from the context. Another time when ``auto`` works well
762 for these purposes is when the type would have been abstracted away anyways,
763 often behind a container's typedef such as ``std::vector<T>::iterator``.
765 Beware unnecessary copies with ``auto``
766 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
768 The convenience of ``auto`` makes it easy to forget that its default behavior
769 is a copy. Particularly in range-based ``for`` loops, careless copies are
772 As a rule of thumb, use ``auto &`` unless you need to copy the result, and use
773 ``auto *`` when copying pointers.
777 // Typically there's no reason to copy.
778 for (const auto &Val : Container) { observe(Val); }
779 for (auto &Val : Container) { Val.change(); }
781 // Remove the reference if you really want a new copy.
782 for (auto Val : Container) { Val.change(); saveSomewhere(Val); }
784 // Copy pointers, but make it clear that they're pointers.
785 for (const auto *Ptr : Container) { observe(*Ptr); }
786 for (auto *Ptr : Container) { Ptr->change(); }
791 The High-Level Issues
792 ---------------------
794 A Public Header File **is** a Module
795 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
797 C++ doesn't do too well in the modularity department. There is no real
798 encapsulation or data hiding (unless you use expensive protocol classes), but it
799 is what we have to work with. When you write a public header file (in the LLVM
800 source tree, they live in the top level "``include``" directory), you are
801 defining a module of functionality.
803 Ideally, modules should be completely independent of each other, and their
804 header files should only ``#include`` the absolute minimum number of headers
805 possible. A module is not just a class, a function, or a namespace: it's a
806 collection of these that defines an interface. This interface may be several
807 functions, classes, or data structures, but the important issue is how they work
810 In general, a module should be implemented by one or more ``.cpp`` files. Each
811 of these ``.cpp`` files should include the header that defines their interface
812 first. This ensures that all of the dependences of the module header have been
813 properly added to the module header itself, and are not implicit. System
814 headers should be included after user headers for a translation unit.
816 .. _minimal list of #includes:
818 ``#include`` as Little as Possible
819 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
821 ``#include`` hurts compile time performance. Don't do it unless you have to,
822 especially in header files.
824 But wait! Sometimes you need to have the definition of a class to use it, or to
825 inherit from it. In these cases go ahead and ``#include`` that header file. Be
826 aware however that there are many cases where you don't need to have the full
827 definition of a class. If you are using a pointer or reference to a class, you
828 don't need the header file. If you are simply returning a class instance from a
829 prototyped function or method, you don't need it. In fact, for most cases, you
830 simply don't need the definition of a class. And not ``#include``\ing speeds up
833 It is easy to try to go too overboard on this recommendation, however. You
834 **must** include all of the header files that you are using --- you can include
835 them either directly or indirectly through another header file. To make sure
836 that you don't accidentally forget to include a header file in your module
837 header, make sure to include your module header **first** in the implementation
838 file (as mentioned above). This way there won't be any hidden dependencies that
839 you'll find out about later.
841 Keep "Internal" Headers Private
842 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
844 Many modules have a complex implementation that causes them to use more than one
845 implementation (``.cpp``) file. It is often tempting to put the internal
846 communication interface (helper classes, extra functions, etc) in the public
847 module header file. Don't do this!
849 If you really need to do something like this, put a private header file in the
850 same directory as the source files, and include it locally. This ensures that
851 your private interface remains private and undisturbed by outsiders.
855 It's okay to put extra implementation methods in a public class itself. Just
856 make them private (or protected) and all is well.
860 Use Early Exits and ``continue`` to Simplify Code
861 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
863 When reading code, keep in mind how much state and how many previous decisions
864 have to be remembered by the reader to understand a block of code. Aim to
865 reduce indentation where possible when it doesn't make it more difficult to
866 understand the code. One great way to do this is by making use of early exits
867 and the ``continue`` keyword in long loops. As an example of using an early
868 exit from a function, consider this "bad" code:
872 Value *doSomething(Instruction *I) {
873 if (!isa<TerminatorInst>(I) &&
874 I->hasOneUse() && doOtherThing(I)) {
875 ... some long code ....
881 This code has several problems if the body of the ``'if'`` is large. When
882 you're looking at the top of the function, it isn't immediately clear that this
883 *only* does interesting things with non-terminator instructions, and only
884 applies to things with the other predicates. Second, it is relatively difficult
885 to describe (in comments) why these predicates are important because the ``if``
886 statement makes it difficult to lay out the comments. Third, when you're deep
887 within the body of the code, it is indented an extra level. Finally, when
888 reading the top of the function, it isn't clear what the result is if the
889 predicate isn't true; you have to read to the end of the function to know that
892 It is much preferred to format the code like this:
896 Value *doSomething(Instruction *I) {
897 // Terminators never need 'something' done to them because ...
898 if (isa<TerminatorInst>(I))
901 // We conservatively avoid transforming instructions with multiple uses
902 // because goats like cheese.
906 // This is really just here for example.
907 if (!doOtherThing(I))
910 ... some long code ....
913 This fixes these problems. A similar problem frequently happens in ``for``
914 loops. A silly example is something like this:
918 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
919 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
920 Value *LHS = BO->getOperand(0);
921 Value *RHS = BO->getOperand(1);
928 When you have very, very small loops, this sort of structure is fine. But if it
929 exceeds more than 10-15 lines, it becomes difficult for people to read and
930 understand at a glance. The problem with this sort of code is that it gets very
931 nested very quickly. Meaning that the reader of the code has to keep a lot of
932 context in their brain to remember what is going immediately on in the loop,
933 because they don't know if/when the ``if`` conditions will have ``else``\s etc.
934 It is strongly preferred to structure the loop like this:
938 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
939 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
942 Value *LHS = BO->getOperand(0);
943 Value *RHS = BO->getOperand(1);
944 if (LHS == RHS) continue;
949 This has all the benefits of using early exits for functions: it reduces nesting
950 of the loop, it makes it easier to describe why the conditions are true, and it
951 makes it obvious to the reader that there is no ``else`` coming up that they
952 have to push context into their brain for. If a loop is large, this can be a
953 big understandability win.
955 Don't use ``else`` after a ``return``
956 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
958 For similar reasons above (reduction of indentation and easier reading), please
959 do not use ``'else'`` or ``'else if'`` after something that interrupts control
960 flow --- like ``return``, ``break``, ``continue``, ``goto``, etc. For
961 example, this is *bad*:
967 Type = Context.getsigjmp_bufType();
969 Error = ASTContext::GE_Missing_sigjmp_buf;
975 Type = Context.getjmp_bufType();
977 Error = ASTContext::GE_Missing_jmp_buf;
985 It is better to write it like this:
991 Type = Context.getsigjmp_bufType();
993 Error = ASTContext::GE_Missing_sigjmp_buf;
997 Type = Context.getjmp_bufType();
999 Error = ASTContext::GE_Missing_jmp_buf;
1005 Or better yet (in this case) as:
1011 Type = Context.getsigjmp_bufType();
1013 Type = Context.getjmp_bufType();
1015 if (Type.isNull()) {
1016 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
1017 ASTContext::GE_Missing_jmp_buf;
1022 The idea is to reduce indentation and the amount of code you have to keep track
1023 of when reading the code.
1025 Turn Predicate Loops into Predicate Functions
1026 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1028 It is very common to write small loops that just compute a boolean value. There
1029 are a number of ways that people commonly write these, but an example of this
1034 bool FoundFoo = false;
1035 for (unsigned I = 0, E = BarList.size(); I != E; ++I)
1036 if (BarList[I]->isFoo()) {
1045 This sort of code is awkward to write, and is almost always a bad sign. Instead
1046 of this sort of loop, we strongly prefer to use a predicate function (which may
1047 be `static`_) that uses `early exits`_ to compute the predicate. We prefer the
1048 code to be structured like this:
1052 /// \returns true if the specified list has an element that is a foo.
1053 static bool containsFoo(const std::vector<Bar*> &List) {
1054 for (unsigned I = 0, E = List.size(); I != E; ++I)
1055 if (List[I]->isFoo())
1061 if (containsFoo(BarList)) {
1065 There are many reasons for doing this: it reduces indentation and factors out
1066 code which can often be shared by other code that checks for the same predicate.
1067 More importantly, it *forces you to pick a name* for the function, and forces
1068 you to write a comment for it. In this silly example, this doesn't add much
1069 value. However, if the condition is complex, this can make it a lot easier for
1070 the reader to understand the code that queries for this predicate. Instead of
1071 being faced with the in-line details of how we check to see if the BarList
1072 contains a foo, we can trust the function name and continue reading with better
1075 The Low-Level Issues
1076 --------------------
1078 Name Types, Functions, Variables, and Enumerators Properly
1079 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1081 Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
1082 enough how important it is to use *descriptive* names. Pick names that match
1083 the semantics and role of the underlying entities, within reason. Avoid
1084 abbreviations unless they are well known. After picking a good name, make sure
1085 to use consistent capitalization for the name, as inconsistency requires clients
1086 to either memorize the APIs or to look it up to find the exact spelling.
1088 In general, names should be in camel case (e.g. ``TextFileReader`` and
1089 ``isLValue()``). Different kinds of declarations have different rules:
1091 * **Type names** (including classes, structs, enums, typedefs, etc) should be
1092 nouns and start with an upper-case letter (e.g. ``TextFileReader``).
1094 * **Variable names** should be nouns (as they represent state). The name should
1095 be camel case, and start with an upper case letter (e.g. ``Leader`` or
1098 * **Function names** should be verb phrases (as they represent actions), and
1099 command-like function should be imperative. The name should be camel case,
1100 and start with a lower case letter (e.g. ``openFile()`` or ``isFoo()``).
1102 * **Enum declarations** (e.g. ``enum Foo {...}``) are types, so they should
1103 follow the naming conventions for types. A common use for enums is as a
1104 discriminator for a union, or an indicator of a subclass. When an enum is
1105 used for something like this, it should have a ``Kind`` suffix
1106 (e.g. ``ValueKind``).
1108 * **Enumerators** (e.g. ``enum { Foo, Bar }``) and **public member variables**
1109 should start with an upper-case letter, just like types. Unless the
1110 enumerators are defined in their own small namespace or inside a class,
1111 enumerators should have a prefix corresponding to the enum declaration name.
1112 For example, ``enum ValueKind { ... };`` may contain enumerators like
1113 ``VK_Argument``, ``VK_BasicBlock``, etc. Enumerators that are just
1114 convenience constants are exempt from the requirement for a prefix. For
1124 As an exception, classes that mimic STL classes can have member names in STL's
1125 style of lower-case words separated by underscores (e.g. ``begin()``,
1126 ``push_back()``, and ``empty()``). Classes that provide multiple
1127 iterators should add a singular prefix to ``begin()`` and ``end()``
1128 (e.g. ``global_begin()`` and ``use_begin()``).
1130 Here are some examples of good and bad names:
1134 class VehicleMaker {
1136 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
1137 Factory<Tire> Factory; // Better.
1138 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
1139 // kind of factories.
1142 Vehicle MakeVehicle(VehicleType Type) {
1143 VehicleMaker M; // Might be OK if having a short life-span.
1144 Tire Tmp1 = M.makeTire(); // Bad -- 'Tmp1' provides no information.
1145 Light Headlight = M.makeLight("head"); // Good -- descriptive.
1152 Use the "``assert``" macro to its fullest. Check all of your preconditions and
1153 assumptions, you never know when a bug (not necessarily even yours) might be
1154 caught early by an assertion, which reduces debugging time dramatically. The
1155 "``<cassert>``" header file is probably already included by the header files you
1156 are using, so it doesn't cost anything to use it.
1158 To further assist with debugging, make sure to put some kind of error message in
1159 the assertion statement, which is printed if the assertion is tripped. This
1160 helps the poor debugger make sense of why an assertion is being made and
1161 enforced, and hopefully what to do about it. Here is one complete example:
1165 inline Value *getOperand(unsigned I) {
1166 assert(I < Operands.size() && "getOperand() out of range!");
1170 Here are more examples:
1174 assert(Ty->isPointerType() && "Can't allocate a non-pointer type!");
1176 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
1178 assert(idx < getNumSuccessors() && "Successor # out of range!");
1180 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
1182 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
1186 In the past, asserts were used to indicate a piece of code that should not be
1187 reached. These were typically of the form:
1191 assert(0 && "Invalid radix for integer literal");
1193 This has a few issues, the main one being that some compilers might not
1194 understand the assertion, or warn about a missing return in builds where
1195 assertions are compiled out.
1197 Today, we have something much better: ``llvm_unreachable``:
1201 llvm_unreachable("Invalid radix for integer literal");
1203 When assertions are enabled, this will print the message if it's ever reached
1204 and then exit the program. When assertions are disabled (i.e. in release
1205 builds), ``llvm_unreachable`` becomes a hint to compilers to skip generating
1206 code for this branch. If the compiler does not support this, it will fall back
1207 to the "abort" implementation.
1209 Another issue is that values used only by assertions will produce an "unused
1210 value" warning when assertions are disabled. For example, this code will warn:
1214 unsigned Size = V.size();
1215 assert(Size > 42 && "Vector smaller than it should be");
1217 bool NewToSet = Myset.insert(Value);
1218 assert(NewToSet && "The value shouldn't be in the set yet");
1220 These are two interesting different cases. In the first case, the call to
1221 ``V.size()`` is only useful for the assert, and we don't want it executed when
1222 assertions are disabled. Code like this should move the call into the assert
1223 itself. In the second case, the side effects of the call must happen whether
1224 the assert is enabled or not. In this case, the value should be cast to void to
1225 disable the warning. To be specific, it is preferred to write the code like
1230 assert(V.size() > 42 && "Vector smaller than it should be");
1232 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1233 assert(NewToSet && "The value shouldn't be in the set yet");
1235 Do Not Use ``using namespace std``
1236 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1238 In LLVM, we prefer to explicitly prefix all identifiers from the standard
1239 namespace with an "``std::``" prefix, rather than rely on "``using namespace
1242 In header files, adding a ``'using namespace XXX'`` directive pollutes the
1243 namespace of any source file that ``#include``\s the header. This is clearly a
1246 In implementation files (e.g. ``.cpp`` files), the rule is more of a stylistic
1247 rule, but is still important. Basically, using explicit namespace prefixes
1248 makes the code **clearer**, because it is immediately obvious what facilities
1249 are being used and where they are coming from. And **more portable**, because
1250 namespace clashes cannot occur between LLVM code and other namespaces. The
1251 portability rule is important because different standard library implementations
1252 expose different symbols (potentially ones they shouldn't), and future revisions
1253 to the C++ standard will add more symbols to the ``std`` namespace. As such, we
1254 never use ``'using namespace std;'`` in LLVM.
1256 The exception to the general rule (i.e. it's not an exception for the ``std``
1257 namespace) is for implementation files. For example, all of the code in the
1258 LLVM project implements code that lives in the 'llvm' namespace. As such, it is
1259 ok, and actually clearer, for the ``.cpp`` files to have a ``'using namespace
1260 llvm;'`` directive at the top, after the ``#include``\s. This reduces
1261 indentation in the body of the file for source editors that indent based on
1262 braces, and keeps the conceptual context cleaner. The general form of this rule
1263 is that any ``.cpp`` file that implements code in any namespace may use that
1264 namespace (and its parents'), but should not use any others.
1266 Provide a Virtual Method Anchor for Classes in Headers
1267 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1269 If a class is defined in a header file and has a vtable (either it has virtual
1270 methods or it derives from classes with virtual methods), it must always have at
1271 least one out-of-line virtual method in the class. Without this, the compiler
1272 will copy the vtable and RTTI into every ``.o`` file that ``#include``\s the
1273 header, bloating ``.o`` file sizes and increasing link times.
1275 Don't use default labels in fully covered switches over enumerations
1276 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1278 ``-Wswitch`` warns if a switch, without a default label, over an enumeration
1279 does not cover every enumeration value. If you write a default label on a fully
1280 covered switch over an enumeration then the ``-Wswitch`` warning won't fire
1281 when new elements are added to that enumeration. To help avoid adding these
1282 kinds of defaults, Clang has the warning ``-Wcovered-switch-default`` which is
1283 off by default but turned on when building LLVM with a version of Clang that
1284 supports the warning.
1286 A knock-on effect of this stylistic requirement is that when building LLVM with
1287 GCC you may get warnings related to "control may reach end of non-void function"
1288 if you return from each case of a covered switch-over-enum because GCC assumes
1289 that the enum expression may take any representable value, not just those of
1290 individual enumerators. To suppress this warning, use ``llvm_unreachable`` after
1293 Use ``LLVM_DELETED_FUNCTION`` to mark uncallable methods
1294 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1296 Prior to C++11, a common pattern to make a class uncopyable was to declare an
1297 unimplemented copy constructor and copy assignment operator and make them
1298 private. This would give a compiler error for accessing a private method or a
1299 linker error because it wasn't implemented.
1301 With C++11, we can mark methods that won't be implemented with ``= delete``.
1302 This will trigger a much better error message and tell the compiler that the
1303 method will never be implemented. This enables other checks like
1304 ``-Wunused-private-field`` to run correctly on classes that contain these
1307 For compatibility with MSVC, ``LLVM_DELETED_FUNCTION`` should be used which
1308 will expand to ``= delete`` on compilers that support it. These methods should
1309 still be declared private. Example of the uncopyable pattern:
1315 DontCopy(const DontCopy&) LLVM_DELETED_FUNCTION;
1316 DontCopy &operator =(const DontCopy&) LLVM_DELETED_FUNCTION;
1321 Don't evaluate ``end()`` every time through a loop
1322 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1324 Because C++ doesn't have a standard "``foreach``" loop (though it can be
1325 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1326 loops that manually iterate from begin to end on a variety of containers or
1327 through other data structures. One common mistake is to write a loop in this
1332 BasicBlock *BB = ...
1333 for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I)
1336 The problem with this construct is that it evaluates "``BB->end()``" every time
1337 through the loop. Instead of writing the loop like this, we strongly prefer
1338 loops to be written so that they evaluate it once before the loop starts. A
1339 convenient way to do this is like so:
1343 BasicBlock *BB = ...
1344 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
1347 The observant may quickly point out that these two loops may have different
1348 semantics: if the container (a basic block in this case) is being mutated, then
1349 "``BB->end()``" may change its value every time through the loop and the second
1350 loop may not in fact be correct. If you actually do depend on this behavior,
1351 please write the loop in the first form and add a comment indicating that you
1352 did it intentionally.
1354 Why do we prefer the second form (when correct)? Writing the loop in the first
1355 form has two problems. First it may be less efficient than evaluating it at the
1356 start of the loop. In this case, the cost is probably minor --- a few extra
1357 loads every time through the loop. However, if the base expression is more
1358 complex, then the cost can rise quickly. I've seen loops where the end
1359 expression was actually something like: "``SomeMap[X]->end()``" and map lookups
1360 really aren't cheap. By writing it in the second form consistently, you
1361 eliminate the issue entirely and don't even have to think about it.
1363 The second (even bigger) issue is that writing the loop in the first form hints
1364 to the reader that the loop is mutating the container (a fact that a comment
1365 would handily confirm!). If you write the loop in the second form, it is
1366 immediately obvious without even looking at the body of the loop that the
1367 container isn't being modified, which makes it easier to read the code and
1368 understand what it does.
1370 While the second form of the loop is a few extra keystrokes, we do strongly
1373 ``#include <iostream>`` is Forbidden
1374 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1376 The use of ``#include <iostream>`` in library files is hereby **forbidden**,
1377 because many common implementations transparently inject a `static constructor`_
1378 into every translation unit that includes it.
1380 Note that using the other stream headers (``<sstream>`` for example) is not
1381 problematic in this regard --- just ``<iostream>``. However, ``raw_ostream``
1382 provides various APIs that are better performing for almost every use than
1383 ``std::ostream`` style APIs.
1387 New code should always use `raw_ostream`_ for writing, or the
1388 ``llvm::MemoryBuffer`` API for reading files.
1395 LLVM includes a lightweight, simple, and efficient stream implementation in
1396 ``llvm/Support/raw_ostream.h``, which provides all of the common features of
1397 ``std::ostream``. All new code should use ``raw_ostream`` instead of
1400 Unlike ``std::ostream``, ``raw_ostream`` is not a template and can be forward
1401 declared as ``class raw_ostream``. Public headers should generally not include
1402 the ``raw_ostream`` header, but use forward declarations and constant references
1403 to ``raw_ostream`` instances.
1408 The ``std::endl`` modifier, when used with ``iostreams`` outputs a newline to
1409 the output stream specified. In addition to doing this, however, it also
1410 flushes the output stream. In other words, these are equivalent:
1414 std::cout << std::endl;
1415 std::cout << '\n' << std::flush;
1417 Most of the time, you probably have no reason to flush the output stream, so
1418 it's better to use a literal ``'\n'``.
1420 Don't use ``inline`` when defining a function in a class definition
1421 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1423 A member function defined in a class definition is implicitly inline, so don't
1424 put the ``inline`` keyword in this case.
1451 This section describes preferred low-level formatting guidelines along with
1452 reasoning on why we prefer them.
1454 Spaces Before Parentheses
1455 ^^^^^^^^^^^^^^^^^^^^^^^^^
1457 We prefer to put a space before an open parenthesis only in control flow
1458 statements, but not in normal function call expressions and function-like
1459 macros. For example, this is good:
1464 for (I = 0; I != 100; ++I) ...
1465 while (LLVMRocks) ...
1468 assert(3 != 4 && "laws of math are failing me");
1470 A = foo(42, 92) + bar(X);
1477 for(I = 0; I != 100; ++I) ...
1478 while(LLVMRocks) ...
1481 assert (3 != 4 && "laws of math are failing me");
1483 A = foo (42, 92) + bar (X);
1485 The reason for doing this is not completely arbitrary. This style makes control
1486 flow operators stand out more, and makes expressions flow better. The function
1487 call operator binds very tightly as a postfix operator. Putting a space after a
1488 function name (as in the last example) makes it appear that the code might bind
1489 the arguments of the left-hand-side of a binary operator with the argument list
1490 of a function and the name of the right side. More specifically, it is easy to
1491 misread the "``A``" example as:
1495 A = foo ((42, 92) + bar) (X);
1497 when skimming through the code. By avoiding a space in a function, we avoid
1498 this misinterpretation.
1503 Hard fast rule: Preincrement (``++X``) may be no slower than postincrement
1504 (``X++``) and could very well be a lot faster than it. Use preincrementation
1507 The semantics of postincrement include making a copy of the value being
1508 incremented, returning it, and then preincrementing the "work value". For
1509 primitive types, this isn't a big deal. But for iterators, it can be a huge
1510 issue (for example, some iterators contains stack and set objects in them...
1511 copying an iterator could invoke the copy ctor's of these as well). In general,
1512 get in the habit of always using preincrement, and you won't have a problem.
1515 Namespace Indentation
1516 ^^^^^^^^^^^^^^^^^^^^^
1518 In general, we strive to reduce indentation wherever possible. This is useful
1519 because we want code to `fit into 80 columns`_ without wrapping horribly, but
1520 also because it makes it easier to understand the code. To facilitate this and
1521 avoid some insanely deep nesting on occasion, don't indent namespaces. If it
1522 helps readability, feel free to add a comment indicating what namespace is
1523 being closed by a ``}``. For example:
1528 namespace knowledge {
1530 /// This class represents things that Smith can have an intimate
1531 /// understanding of and contains the data associated with it.
1535 explicit Grokable() { ... }
1536 virtual ~Grokable() = 0;
1542 } // end namespace knowledge
1543 } // end namespace llvm
1546 Feel free to skip the closing comment when the namespace being closed is
1547 obvious for any reason. For example, the outer-most namespace in a header file
1548 is rarely a source of confusion. But namespaces both anonymous and named in
1549 source files that are being closed half way through the file probably could use
1554 Anonymous Namespaces
1555 ^^^^^^^^^^^^^^^^^^^^
1557 After talking about namespaces in general, you may be wondering about anonymous
1558 namespaces in particular. Anonymous namespaces are a great language feature
1559 that tells the C++ compiler that the contents of the namespace are only visible
1560 within the current translation unit, allowing more aggressive optimization and
1561 eliminating the possibility of symbol name collisions. Anonymous namespaces are
1562 to C++ as "static" is to C functions and global variables. While "``static``"
1563 is available in C++, anonymous namespaces are more general: they can make entire
1564 classes private to a file.
1566 The problem with anonymous namespaces is that they naturally want to encourage
1567 indentation of their body, and they reduce locality of reference: if you see a
1568 random function definition in a C++ file, it is easy to see if it is marked
1569 static, but seeing if it is in an anonymous namespace requires scanning a big
1572 Because of this, we have a simple guideline: make anonymous namespaces as small
1573 as possible, and only use them for class declarations. For example, this is
1583 bool operator<(const char *RHS) const;
1585 } // end anonymous namespace
1587 static void runHelper() {
1591 bool StringSort::operator<(const char *RHS) const {
1605 bool operator<(const char *RHS) const;
1612 bool StringSort::operator<(const char *RHS) const {
1616 } // end anonymous namespace
1618 This is bad specifically because if you're looking at "``runHelper``" in the middle
1619 of a large C++ file, that you have no immediate way to tell if it is local to
1620 the file. When it is marked static explicitly, this is immediately obvious.
1621 Also, there is no reason to enclose the definition of "``operator<``" in the
1622 namespace just because it was declared there.
1627 A lot of these comments and recommendations have been culled from other sources.
1628 Two particularly important books for our work are:
1631 <http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876>`_
1632 by Scott Meyers. Also interesting and useful are "More Effective C++" and
1633 "Effective STL" by the same author.
1635 #. `Large-Scale C++ Software Design
1636 <http://www.amazon.com/Large-Scale-Software-Design-John-Lakos/dp/0201633620/ref=sr_1_1>`_
1639 If you get some free time, and you haven't read them: do so, you might learn