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 `Programmer's Manual`_.
81 .. _Programmer's Manual:
82 http://llvm.org/docs/ProgrammersManual.html
84 Supported C++11 Language and Library Features
85 ---------------------------------------------
87 While LLVM, Clang, and LLD use C++11, not all features are available in all of
88 the toolchains which we support. The set of features supported for use in LLVM
89 is the intersection of those supported in MSVC 2012, GCC 4.7, and Clang 3.1.
90 The ultimate definition of this set is what build bots with those respective
91 toolchains accept. Don't argue with the build bots. However, we have some
92 guidance below to help you know what to expect.
94 Each toolchain provides a good reference for what it accepts:
96 * Clang: http://clang.llvm.org/cxx_status.html
97 * GCC: http://gcc.gnu.org/projects/cxx0x.html
98 * MSVC: http://msdn.microsoft.com/en-us/library/hh567368.aspx
100 In most cases, the MSVC list will be the dominating factor. Here is a summary
101 of the features that are expected to work. Features not on this list are
102 unlikely to be supported by our host compilers.
104 * Rvalue references: N2118_
106 * But *not* Rvalue references for ``*this`` or member qualifiers (N2439_)
108 * Static assert: N1720_
109 * ``auto`` type deduction: N1984_, N1737_
110 * Trailing return types: N2541_
113 * But *not* ``std::function``, until Clang implements `MSVC-compatible RTTI`_.
115 * ``decltype``: N2343_
116 * Nested closing right angle brackets: N1757_
117 * Extern templates: N1987_
118 * ``nullptr``: N2431_
119 * Strongly-typed and forward declarable enums: N2347_, N2764_
120 * Local and unnamed types as template arguments: N2657_
121 * Range-based for-loop: N2930_
122 * ``override`` and ``final``: N2928_, N3206_, N3272_
123 * Atomic operations and the C++11 memory model: N2429_
125 .. _N2118: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2118.html
126 .. _N2439: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2439.htm
127 .. _N1720: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1720.html
128 .. _N1984: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1984.pdf
129 .. _N1737: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1737.pdf
130 .. _N2541: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2541.htm
131 .. _N2927: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2927.pdf
132 .. _N2343: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2343.pdf
133 .. _N1757: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1757.html
134 .. _N1987: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1987.htm
135 .. _N2431: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2431.pdf
136 .. _N2347: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2347.pdf
137 .. _N2764: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2764.pdf
138 .. _N2657: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm
139 .. _N2930: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2930.html
140 .. _N2928: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2928.htm
141 .. _N3206: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3206.htm
142 .. _N3272: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3272.htm
143 .. _N2429: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2429.htm
144 .. _MSVC-compatible RTTI: http://llvm.org/PR18951
146 The supported features in the C++11 standard libraries are less well tracked,
147 but also much greater. Most of the standard libraries implement most of C++11's
148 library. The most likely lowest common denominator is Linux support. For
149 libc++, the support is just poorly tested and undocumented but expected to be
150 largely complete. YMMV. For libstdc++, the support is documented in detail in
151 `the libstdc++ manual`_. There are some very minor missing facilities that are
152 unlikely to be common problems, and there are a few larger gaps that are worth
155 * Not all of the type traits are implemented
156 * No regular expression library.
157 * While most of the atomics library is well implemented, the fences are
158 missing. Fortunately, they are rarely needed.
159 * The locale support is incomplete.
160 * ``std::initializer_list`` (and the constructors and functions that take it as
161 an argument) are not always available, so you cannot (for example) initialize
162 a ``std::vector`` with a braced initializer list.
164 Other than these areas you should assume the standard library is available and
165 working as expected until some build bot tells you otherwise. If you're in an
166 uncertain area of one of the above points, but you cannot test on a Linux
167 system, your best approach is to minimize your use of these features, and watch
168 the Linux build bots to find out if your usage triggered a bug. For example, if
169 you hit a type trait which doesn't work we can then add support to LLVM's
170 traits header to emulate it.
172 .. _the libstdc++ manual:
173 http://gcc.gnu.org/onlinedocs/gcc-4.7.3/libstdc++/manual/manual/status.html#status.iso.2011
175 Mechanical Source Issues
176 ========================
178 Source Code Formatting
179 ----------------------
184 Comments are one critical part of readability and maintainability. Everyone
185 knows they should comment their code, and so should you. When writing comments,
186 write them as English prose, which means they should use proper capitalization,
187 punctuation, etc. Aim to describe what the code is trying to do and why, not
188 *how* it does it at a micro level. Here are a few critical things to document:
190 .. _header file comment:
195 Every source file should have a header on it that describes the basic purpose of
196 the file. If a file does not have a header, it should not be checked into the
197 tree. The standard header looks like this:
201 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
203 // The LLVM Compiler Infrastructure
205 // This file is distributed under the University of Illinois Open Source
206 // License. See LICENSE.TXT for details.
208 //===----------------------------------------------------------------------===//
211 /// \brief This file contains the declaration of the Instruction class, which is
212 /// the base class for all of the VM instructions.
214 //===----------------------------------------------------------------------===//
216 A few things to note about this particular format: The "``-*- C++ -*-``" string
217 on the first line is there to tell Emacs that the source file is a C++ file, not
218 a C file (Emacs assumes ``.h`` files are C files by default).
222 This tag is not necessary in ``.cpp`` files. The name of the file is also
223 on the first line, along with a very short description of the purpose of the
224 file. This is important when printing out code and flipping though lots of
227 The next section in the file is a concise note that defines the license that the
228 file is released under. This makes it perfectly clear what terms the source
229 code can be distributed under and should not be modified in any way.
231 The main body is a ``doxygen`` comment describing the purpose of the file. It
232 should have a ``\brief`` command that describes the file in one or two
233 sentences. Any additional information should be separated by a blank line. If
234 an algorithm is being implemented or something tricky is going on, a reference
235 to the paper where it is published should be included, as well as any notes or
236 *gotchas* in the code to watch out for.
241 Classes are one fundamental part of a good object oriented design. As such, a
242 class definition should have a comment block that explains what the class is
243 used for and how it works. Every non-trivial class is expected to have a
244 ``doxygen`` comment block.
249 Methods defined in a class (as well as any global functions) should also be
250 documented properly. A quick note about what it does and a description of the
251 borderline behaviour is all that is necessary here (unless something
252 particularly tricky or insidious is going on). The hope is that people can
253 figure out how to use your interfaces without reading the code itself.
255 Good things to talk about here are what happens when something unexpected
256 happens: does the method return null? Abort? Format your hard disk?
261 In general, prefer C++ style (``//``) comments. They take less space, require
262 less typing, don't have nesting problems, etc. There are a few cases when it is
263 useful to use C style (``/* */``) comments however:
265 #. When writing C code: Obviously if you are writing C code, use C style
268 #. When writing a header file that may be ``#include``\d by a C source file.
270 #. When writing a source file that is used by a tool that only accepts C style
273 To comment out a large block of code, use ``#if 0`` and ``#endif``. These nest
274 properly and are better behaved in general than C style comments.
276 Doxygen Use in Documentation Comments
277 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
279 Use the ``\file`` command to turn the standard file header into a file-level
282 Include descriptive ``\brief`` paragraphs for all public interfaces (public
283 classes, member and non-member functions). Explain API use and purpose in
284 ``\brief`` paragraphs, don't just restate the information that can be inferred
285 from the API name. Put detailed discussion into separate paragraphs.
287 To refer to parameter names inside a paragraph, use the ``\p name`` command.
288 Don't use the ``\arg name`` command since it starts a new paragraph that
289 contains documentation for the parameter.
291 Wrap non-inline code examples in ``\code ... \endcode``.
293 To document a function parameter, start a new paragraph with the
294 ``\param name`` command. If the parameter is used as an out or an in/out
295 parameter, use the ``\param [out] name`` or ``\param [in,out] name`` command,
298 To describe function return value, start a new paragraph with the ``\returns``
301 A minimal documentation comment:
305 /// \brief Does foo and bar.
306 void fooBar(bool Baz);
308 A documentation comment that uses all Doxygen features in a preferred way:
312 /// \brief Does foo and bar.
314 /// Does not do foo the usual way if \p Baz is true.
318 /// fooBar(false, "quux", Res);
321 /// \param Quux kind of foo to do.
322 /// \param [out] Result filled with bar sequence on foo success.
324 /// \returns true on success.
325 bool fooBar(bool Baz, StringRef Quux, std::vector<int> &Result);
327 Don't duplicate the documentation comment in the header file and in the
328 implementation file. Put the documentation comments for public APIs into the
329 header file. Documentation comments for private APIs can go to the
330 implementation file. In any case, implementation files can include additional
331 comments (not necessarily in Doxygen markup) to explain implementation details
334 Don't duplicate function or class name at the beginning of the comment.
335 For humans it is obvious which function or class is being documented;
336 automatic documentation processing tools are smart enough to bind the comment
337 to the correct declaration.
345 /// Something - An abstraction for some complicated thing.
348 /// fooBar - Does foo and bar.
354 /// fooBar - Does foo and bar.
355 void Something::fooBar() { ... }
363 /// \brief An abstraction for some complicated thing.
366 /// \brief Does foo and bar.
372 // Builds a B-tree in order to do foo. See paper by...
373 void Something::fooBar() { ... }
375 It is not required to use additional Doxygen features, but sometimes it might
376 be a good idea to do so.
380 * adding comments to any narrow namespace containing a collection of
381 related functions or types;
383 * using top-level groups to organize a collection of related functions at
384 namespace scope where the grouping is smaller than the namespace;
386 * using member groups and additional comments attached to member
387 groups to organize within a class.
394 /// \name Functions that do Foo.
405 Immediately after the `header file comment`_ (and include guards if working on a
406 header file), the `minimal list of #includes`_ required by the file should be
407 listed. We prefer these ``#include``\s to be listed in this order:
409 .. _Main Module Header:
410 .. _Local/Private Headers:
412 #. Main Module Header
413 #. Local/Private Headers
415 #. System ``#include``\s
417 and each category should be sorted lexicographically by the full path.
419 The `Main Module Header`_ file applies to ``.cpp`` files which implement an
420 interface defined by a ``.h`` file. This ``#include`` should always be included
421 **first** regardless of where it lives on the file system. By including a
422 header file first in the ``.cpp`` files that implement the interfaces, we ensure
423 that the header does not have any hidden dependencies which are not explicitly
424 ``#include``\d in the header, but should be. It is also a form of documentation
425 in the ``.cpp`` file to indicate where the interfaces it implements are defined.
427 .. _fit into 80 columns:
432 Write your code to fit within 80 columns of text. This helps those of us who
433 like to print out code and look at your code in an ``xterm`` without resizing
436 The longer answer is that there must be some limit to the width of the code in
437 order to reasonably allow developers to have multiple files side-by-side in
438 windows on a modest display. If you are going to pick a width limit, it is
439 somewhat arbitrary but you might as well pick something standard. Going with 90
440 columns (for example) instead of 80 columns wouldn't add any significant value
441 and would be detrimental to printing out code. Also many other projects have
442 standardized on 80 columns, so some people have already configured their editors
443 for it (vs something else, like 90 columns).
445 This is one of many contentious issues in coding standards, but it is not up for
448 Use Spaces Instead of Tabs
449 ^^^^^^^^^^^^^^^^^^^^^^^^^^
451 In all cases, prefer spaces to tabs in source files. People have different
452 preferred indentation levels, and different styles of indentation that they
453 like; this is fine. What isn't fine is that different editors/viewers expand
454 tabs out to different tab stops. This can cause your code to look completely
455 unreadable, and it is not worth dealing with.
457 As always, follow the `Golden Rule`_ above: follow the style of
458 existing code if you are modifying and extending it. If you like four spaces of
459 indentation, **DO NOT** do that in the middle of a chunk of code with two spaces
460 of indentation. Also, do not reindent a whole source file: it makes for
461 incredible diffs that are absolutely worthless.
463 Indent Code Consistently
464 ^^^^^^^^^^^^^^^^^^^^^^^^
466 Okay, in your first year of programming you were told that indentation is
467 important. If you didn't believe and internalize this then, now is the time.
468 Just do it. With the introduction of C++11, there are some new formatting
469 challenges that merit some suggestions to help have consistent, maintainable,
470 and tool-friendly formatting and indentation.
472 Format Lambdas Like Blocks Of Code
473 """"""""""""""""""""""""""""""""""
475 When formatting a multi-line lambda, format it like a block of code, that's
476 what it is. If there is only one multi-line lambda in a statement, and there
477 are no expressions lexically after it in the statement, drop the indent to the
478 standard two space indent for a block of code, as if it were an if-block opened
479 by the preceding part of the statement:
483 std::sort(foo.begin(), foo.end(), [&](Foo a, Foo b) -> bool {
488 return a.bam < b.bam;
491 To take best advantage of this formatting, if you are designing an API which
492 accepts a continuation or single callable argument (be it a functor, or
493 a ``std::function``), it should be the last argument if at all possible.
495 If there are multiple multi-line lambdas in a statement, or there is anything
496 interesting after the lambda in the statement, indent the block two spaces from
497 the indent of the ``[]``:
501 dyn_switch(V->stripPointerCasts(),
505 [] (SelectInst *SI) {
506 // process selects...
511 [] (AllocaInst *AI) {
512 // process allocas...
515 Braced Initializer Lists
516 """"""""""""""""""""""""
518 With C++11, there are significantly more uses of braced lists to perform
519 initialization. These allow you to easily construct aggregate temporaries in
520 expressions among other niceness. They now have a natural way of ending up
521 nested within each other and within function calls in order to build up
522 aggregates (such as option structs) from local variables. To make matters
523 worse, we also have many more uses of braces in an expression context that are
524 *not* performing initialization.
526 The historically common formatting of braced initialization of aggregate
527 variables does not mix cleanly with deep nesting, general expression contexts,
528 function arguments, and lambdas. We suggest new code use a simple rule for
529 formatting braced initialization lists: act as-if the braces were parentheses
530 in a function call. The formatting rules exactly match those already well
531 understood for formatting nested function calls. Examples:
535 foo({a, b, c}, {1, 2, 3});
537 llvm::Constant *Mask[] = {
538 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 0),
539 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 1),
540 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 2)};
542 This formatting scheme also makes it particularly easy to get predictable,
543 consistent, and automatic formatting with tools like `Clang Format`_.
545 .. _Clang Format: http://clang.llvm.org/docs/ClangFormat.html
547 Language and Compiler Issues
548 ----------------------------
550 Treat Compiler Warnings Like Errors
551 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
553 If your code has compiler warnings in it, something is wrong --- you aren't
554 casting values correctly, you have "questionable" constructs in your code, or
555 you are doing something legitimately wrong. Compiler warnings can cover up
556 legitimate errors in output and make dealing with a translation unit difficult.
558 It is not possible to prevent all warnings from all compilers, nor is it
559 desirable. Instead, pick a standard compiler (like ``gcc``) that provides a
560 good thorough set of warnings, and stick to it. At least in the case of
561 ``gcc``, it is possible to work around any spurious errors by changing the
562 syntax of the code slightly. For example, a warning that annoys me occurs when
563 I write code like this:
567 if (V = getValue()) {
571 ``gcc`` will warn me that I probably want to use the ``==`` operator, and that I
572 probably mistyped it. In most cases, I haven't, and I really don't want the
573 spurious errors. To fix this particular problem, I rewrite the code like
578 if ((V = getValue())) {
582 which shuts ``gcc`` up. Any ``gcc`` warning that annoys you can be fixed by
583 massaging the code appropriately.
588 In almost all cases, it is possible and within reason to write completely
589 portable code. If there are cases where it isn't possible to write portable
590 code, isolate it behind a well defined (and well documented) interface.
592 In practice, this means that you shouldn't assume much about the host compiler
593 (and Visual Studio tends to be the lowest common denominator). If advanced
594 features are used, they should only be an implementation detail of a library
595 which has a simple exposed API, and preferably be buried in ``libSystem``.
597 Do not use RTTI or Exceptions
598 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
600 In an effort to reduce code and executable size, LLVM does not use RTTI
601 (e.g. ``dynamic_cast<>;``) or exceptions. These two language features violate
602 the general C++ principle of *"you only pay for what you use"*, causing
603 executable bloat even if exceptions are never used in the code base, or if RTTI
604 is never used for a class. Because of this, we turn them off globally in the
607 That said, LLVM does make extensive use of a hand-rolled form of RTTI that use
608 templates like `isa<>, cast<>, and dyn_cast<> <ProgrammersManual.html#isa>`_.
609 This form of RTTI is opt-in and can be
610 :doc:`added to any class <HowToSetUpLLVMStyleRTTI>`. It is also
611 substantially more efficient than ``dynamic_cast<>``.
613 .. _static constructor:
615 Do not use Static Constructors
616 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
618 Static constructors and destructors (e.g. global variables whose types have a
619 constructor or destructor) should not be added to the code base, and should be
620 removed wherever possible. Besides `well known problems
621 <http://yosefk.com/c++fqa/ctors.html#fqa-10.12>`_ where the order of
622 initialization is undefined between globals in different source files, the
623 entire concept of static constructors is at odds with the common use case of
624 LLVM as a library linked into a larger application.
626 Consider the use of LLVM as a JIT linked into another application (perhaps for
627 `OpenGL, custom languages <http://llvm.org/Users.html>`_, `shaders in movies
628 <http://llvm.org/devmtg/2010-11/Gritz-OpenShadingLang.pdf>`_, etc). Due to the
629 design of static constructors, they must be executed at startup time of the
630 entire application, regardless of whether or how LLVM is used in that larger
631 application. There are two problems with this:
633 * The time to run the static constructors impacts startup time of applications
634 --- a critical time for GUI apps, among others.
636 * The static constructors cause the app to pull many extra pages of memory off
637 the disk: both the code for the constructor in each ``.o`` file and the small
638 amount of data that gets touched. In addition, touched/dirty pages put more
639 pressure on the VM system on low-memory machines.
641 We would really like for there to be zero cost for linking in an additional LLVM
642 target or other library into an application, but static constructors violate
645 That said, LLVM unfortunately does contain static constructors. It would be a
646 `great project <http://llvm.org/PR11944>`_ for someone to purge all static
647 constructors from LLVM, and then enable the ``-Wglobal-constructors`` warning
648 flag (when building with Clang) to ensure we do not regress in the future.
650 Use of ``class`` and ``struct`` Keywords
651 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
653 In C++, the ``class`` and ``struct`` keywords can be used almost
654 interchangeably. The only difference is when they are used to declare a class:
655 ``class`` makes all members private by default while ``struct`` makes all
656 members public by default.
658 Unfortunately, not all compilers follow the rules and some will generate
659 different symbols based on whether ``class`` or ``struct`` was used to declare
660 the symbol (e.g., MSVC). This can lead to problems at link time.
662 * All declarations and definitions of a given ``class`` or ``struct`` must use
663 the same keyword. For example:
669 // Breaks mangling in MSVC.
670 struct Foo { int Data; };
672 * As a rule of thumb, ``struct`` should be kept to structures where *all*
673 members are declared public.
677 // Foo feels like a class... this is strange.
683 int getData() const { return Data; }
684 void setData(int D) { Data = D; }
687 // Bar isn't POD, but it does look like a struct.
693 Do not use Braced Initializer Lists to Call a Constructor
694 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
696 In C++11 there is a "generalized initialization syntax" which allows calling
697 constructors using braced initializer lists. Do not use these to call
698 constructors with any interesting logic or if you care that you're calling some
699 *particular* constructor. Those should look like function calls using
700 parentheses rather than like aggregate initialization. Similarly, if you need
701 to explicitly name the type and call its constructor to create a temporary,
702 don't use a braced initializer list. Instead, use a braced initializer list
703 (without any type for temporaries) when doing aggregate initialization or
704 something notionally equivalent. Examples:
710 // Construct a Foo by reading data from the disk in the whizbang format, ...
711 Foo(std::string filename);
713 // Construct a Foo by looking up the Nth element of some global data ...
719 // The Foo constructor call is very deliberate, no braces.
720 std::fill(foo.begin(), foo.end(), Foo("name"));
722 // The pair is just being constructed like an aggregate, use braces.
723 bar_map.insert({my_key, my_value});
725 If you use a braced initializer list when initializing a variable, use an equals before the open curly brace:
729 int data[] = {0, 1, 2, 3};
731 Use ``auto`` Type Deduction to Make Code More Readable
732 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
734 Some are advocating a policy of "almost always ``auto``" in C++11, however LLVM
735 uses a more moderate stance. Use ``auto`` if and only if it makes the code more
736 readable or easier to maintain. Don't "almost always" use ``auto``, but do use
737 ``auto`` with initializers like ``cast<Foo>(...)`` or other places where the
738 type is already obvious from the context. Another time when ``auto`` works well
739 for these purposes is when the type would have been abstracted away anyways,
740 often behind a container's typedef such as ``std::vector<T>::iterator``.
742 Beware unnecessary copies with ``auto``
743 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
745 The convenience of ``auto`` makes it easy to forget that its default behavior
746 is a copy. Particularly in range-based ``for`` loops, careless copies are
749 As a rule of thumb, use ``const auto &`` unless you need to mutate or copy the
754 // Typically there's no reason to mutate or modify Val.
755 for (const auto &Val : Container) { observe(Val); }
757 // Remove the const if you need to modify Val.
758 for (auto &Val : Container) { Val.change(); }
760 // Remove the reference if you really want a new copy.
761 for (auto Val : Container) { Val.change(); saveSomewhere(Val); }
766 The High-Level Issues
767 ---------------------
769 A Public Header File **is** a Module
770 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
772 C++ doesn't do too well in the modularity department. There is no real
773 encapsulation or data hiding (unless you use expensive protocol classes), but it
774 is what we have to work with. When you write a public header file (in the LLVM
775 source tree, they live in the top level "``include``" directory), you are
776 defining a module of functionality.
778 Ideally, modules should be completely independent of each other, and their
779 header files should only ``#include`` the absolute minimum number of headers
780 possible. A module is not just a class, a function, or a namespace: it's a
781 collection of these that defines an interface. This interface may be several
782 functions, classes, or data structures, but the important issue is how they work
785 In general, a module should be implemented by one or more ``.cpp`` files. Each
786 of these ``.cpp`` files should include the header that defines their interface
787 first. This ensures that all of the dependences of the module header have been
788 properly added to the module header itself, and are not implicit. System
789 headers should be included after user headers for a translation unit.
791 .. _minimal list of #includes:
793 ``#include`` as Little as Possible
794 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
796 ``#include`` hurts compile time performance. Don't do it unless you have to,
797 especially in header files.
799 But wait! Sometimes you need to have the definition of a class to use it, or to
800 inherit from it. In these cases go ahead and ``#include`` that header file. Be
801 aware however that there are many cases where you don't need to have the full
802 definition of a class. If you are using a pointer or reference to a class, you
803 don't need the header file. If you are simply returning a class instance from a
804 prototyped function or method, you don't need it. In fact, for most cases, you
805 simply don't need the definition of a class. And not ``#include``\ing speeds up
808 It is easy to try to go too overboard on this recommendation, however. You
809 **must** include all of the header files that you are using --- you can include
810 them either directly or indirectly through another header file. To make sure
811 that you don't accidentally forget to include a header file in your module
812 header, make sure to include your module header **first** in the implementation
813 file (as mentioned above). This way there won't be any hidden dependencies that
814 you'll find out about later.
816 Keep "Internal" Headers Private
817 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
819 Many modules have a complex implementation that causes them to use more than one
820 implementation (``.cpp``) file. It is often tempting to put the internal
821 communication interface (helper classes, extra functions, etc) in the public
822 module header file. Don't do this!
824 If you really need to do something like this, put a private header file in the
825 same directory as the source files, and include it locally. This ensures that
826 your private interface remains private and undisturbed by outsiders.
830 It's okay to put extra implementation methods in a public class itself. Just
831 make them private (or protected) and all is well.
835 Use Early Exits and ``continue`` to Simplify Code
836 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
838 When reading code, keep in mind how much state and how many previous decisions
839 have to be remembered by the reader to understand a block of code. Aim to
840 reduce indentation where possible when it doesn't make it more difficult to
841 understand the code. One great way to do this is by making use of early exits
842 and the ``continue`` keyword in long loops. As an example of using an early
843 exit from a function, consider this "bad" code:
847 Value *doSomething(Instruction *I) {
848 if (!isa<TerminatorInst>(I) &&
849 I->hasOneUse() && doOtherThing(I)) {
850 ... some long code ....
856 This code has several problems if the body of the ``'if'`` is large. When
857 you're looking at the top of the function, it isn't immediately clear that this
858 *only* does interesting things with non-terminator instructions, and only
859 applies to things with the other predicates. Second, it is relatively difficult
860 to describe (in comments) why these predicates are important because the ``if``
861 statement makes it difficult to lay out the comments. Third, when you're deep
862 within the body of the code, it is indented an extra level. Finally, when
863 reading the top of the function, it isn't clear what the result is if the
864 predicate isn't true; you have to read to the end of the function to know that
867 It is much preferred to format the code like this:
871 Value *doSomething(Instruction *I) {
872 // Terminators never need 'something' done to them because ...
873 if (isa<TerminatorInst>(I))
876 // We conservatively avoid transforming instructions with multiple uses
877 // because goats like cheese.
881 // This is really just here for example.
882 if (!doOtherThing(I))
885 ... some long code ....
888 This fixes these problems. A similar problem frequently happens in ``for``
889 loops. A silly example is something like this:
893 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
894 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
895 Value *LHS = BO->getOperand(0);
896 Value *RHS = BO->getOperand(1);
903 When you have very, very small loops, this sort of structure is fine. But if it
904 exceeds more than 10-15 lines, it becomes difficult for people to read and
905 understand at a glance. The problem with this sort of code is that it gets very
906 nested very quickly. Meaning that the reader of the code has to keep a lot of
907 context in their brain to remember what is going immediately on in the loop,
908 because they don't know if/when the ``if`` conditions will have ``else``\s etc.
909 It is strongly preferred to structure the loop like this:
913 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
914 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
917 Value *LHS = BO->getOperand(0);
918 Value *RHS = BO->getOperand(1);
919 if (LHS == RHS) continue;
924 This has all the benefits of using early exits for functions: it reduces nesting
925 of the loop, it makes it easier to describe why the conditions are true, and it
926 makes it obvious to the reader that there is no ``else`` coming up that they
927 have to push context into their brain for. If a loop is large, this can be a
928 big understandability win.
930 Don't use ``else`` after a ``return``
931 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
933 For similar reasons above (reduction of indentation and easier reading), please
934 do not use ``'else'`` or ``'else if'`` after something that interrupts control
935 flow --- like ``return``, ``break``, ``continue``, ``goto``, etc. For
936 example, this is *bad*:
942 Type = Context.getsigjmp_bufType();
944 Error = ASTContext::GE_Missing_sigjmp_buf;
950 Type = Context.getjmp_bufType();
952 Error = ASTContext::GE_Missing_jmp_buf;
960 It is better to write it like this:
966 Type = Context.getsigjmp_bufType();
968 Error = ASTContext::GE_Missing_sigjmp_buf;
972 Type = Context.getjmp_bufType();
974 Error = ASTContext::GE_Missing_jmp_buf;
980 Or better yet (in this case) as:
986 Type = Context.getsigjmp_bufType();
988 Type = Context.getjmp_bufType();
991 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
992 ASTContext::GE_Missing_jmp_buf;
997 The idea is to reduce indentation and the amount of code you have to keep track
998 of when reading the code.
1000 Turn Predicate Loops into Predicate Functions
1001 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1003 It is very common to write small loops that just compute a boolean value. There
1004 are a number of ways that people commonly write these, but an example of this
1009 bool FoundFoo = false;
1010 for (unsigned I = 0, E = BarList.size(); I != E; ++I)
1011 if (BarList[I]->isFoo()) {
1020 This sort of code is awkward to write, and is almost always a bad sign. Instead
1021 of this sort of loop, we strongly prefer to use a predicate function (which may
1022 be `static`_) that uses `early exits`_ to compute the predicate. We prefer the
1023 code to be structured like this:
1027 /// \returns true if the specified list has an element that is a foo.
1028 static bool containsFoo(const std::vector<Bar*> &List) {
1029 for (unsigned I = 0, E = List.size(); I != E; ++I)
1030 if (List[I]->isFoo())
1036 if (containsFoo(BarList)) {
1040 There are many reasons for doing this: it reduces indentation and factors out
1041 code which can often be shared by other code that checks for the same predicate.
1042 More importantly, it *forces you to pick a name* for the function, and forces
1043 you to write a comment for it. In this silly example, this doesn't add much
1044 value. However, if the condition is complex, this can make it a lot easier for
1045 the reader to understand the code that queries for this predicate. Instead of
1046 being faced with the in-line details of how we check to see if the BarList
1047 contains a foo, we can trust the function name and continue reading with better
1050 The Low-Level Issues
1051 --------------------
1053 Name Types, Functions, Variables, and Enumerators Properly
1054 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1056 Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
1057 enough how important it is to use *descriptive* names. Pick names that match
1058 the semantics and role of the underlying entities, within reason. Avoid
1059 abbreviations unless they are well known. After picking a good name, make sure
1060 to use consistent capitalization for the name, as inconsistency requires clients
1061 to either memorize the APIs or to look it up to find the exact spelling.
1063 In general, names should be in camel case (e.g. ``TextFileReader`` and
1064 ``isLValue()``). Different kinds of declarations have different rules:
1066 * **Type names** (including classes, structs, enums, typedefs, etc) should be
1067 nouns and start with an upper-case letter (e.g. ``TextFileReader``).
1069 * **Variable names** should be nouns (as they represent state). The name should
1070 be camel case, and start with an upper case letter (e.g. ``Leader`` or
1073 * **Function names** should be verb phrases (as they represent actions), and
1074 command-like function should be imperative. The name should be camel case,
1075 and start with a lower case letter (e.g. ``openFile()`` or ``isFoo()``).
1077 * **Enum declarations** (e.g. ``enum Foo {...}``) are types, so they should
1078 follow the naming conventions for types. A common use for enums is as a
1079 discriminator for a union, or an indicator of a subclass. When an enum is
1080 used for something like this, it should have a ``Kind`` suffix
1081 (e.g. ``ValueKind``).
1083 * **Enumerators** (e.g. ``enum { Foo, Bar }``) and **public member variables**
1084 should start with an upper-case letter, just like types. Unless the
1085 enumerators are defined in their own small namespace or inside a class,
1086 enumerators should have a prefix corresponding to the enum declaration name.
1087 For example, ``enum ValueKind { ... };`` may contain enumerators like
1088 ``VK_Argument``, ``VK_BasicBlock``, etc. Enumerators that are just
1089 convenience constants are exempt from the requirement for a prefix. For
1099 As an exception, classes that mimic STL classes can have member names in STL's
1100 style of lower-case words separated by underscores (e.g. ``begin()``,
1101 ``push_back()``, and ``empty()``). Classes that provide multiple
1102 iterators should add a singular prefix to ``begin()`` and ``end()``
1103 (e.g. ``global_begin()`` and ``use_begin()``).
1105 Here are some examples of good and bad names:
1109 class VehicleMaker {
1111 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
1112 Factory<Tire> Factory; // Better.
1113 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
1114 // kind of factories.
1117 Vehicle MakeVehicle(VehicleType Type) {
1118 VehicleMaker M; // Might be OK if having a short life-span.
1119 Tire Tmp1 = M.makeTire(); // Bad -- 'Tmp1' provides no information.
1120 Light Headlight = M.makeLight("head"); // Good -- descriptive.
1127 Use the "``assert``" macro to its fullest. Check all of your preconditions and
1128 assumptions, you never know when a bug (not necessarily even yours) might be
1129 caught early by an assertion, which reduces debugging time dramatically. The
1130 "``<cassert>``" header file is probably already included by the header files you
1131 are using, so it doesn't cost anything to use it.
1133 To further assist with debugging, make sure to put some kind of error message in
1134 the assertion statement, which is printed if the assertion is tripped. This
1135 helps the poor debugger make sense of why an assertion is being made and
1136 enforced, and hopefully what to do about it. Here is one complete example:
1140 inline Value *getOperand(unsigned I) {
1141 assert(I < Operands.size() && "getOperand() out of range!");
1145 Here are more examples:
1149 assert(Ty->isPointerType() && "Can't allocate a non-pointer type!");
1151 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
1153 assert(idx < getNumSuccessors() && "Successor # out of range!");
1155 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
1157 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
1161 In the past, asserts were used to indicate a piece of code that should not be
1162 reached. These were typically of the form:
1166 assert(0 && "Invalid radix for integer literal");
1168 This has a few issues, the main one being that some compilers might not
1169 understand the assertion, or warn about a missing return in builds where
1170 assertions are compiled out.
1172 Today, we have something much better: ``llvm_unreachable``:
1176 llvm_unreachable("Invalid radix for integer literal");
1178 When assertions are enabled, this will print the message if it's ever reached
1179 and then exit the program. When assertions are disabled (i.e. in release
1180 builds), ``llvm_unreachable`` becomes a hint to compilers to skip generating
1181 code for this branch. If the compiler does not support this, it will fall back
1182 to the "abort" implementation.
1184 Another issue is that values used only by assertions will produce an "unused
1185 value" warning when assertions are disabled. For example, this code will warn:
1189 unsigned Size = V.size();
1190 assert(Size > 42 && "Vector smaller than it should be");
1192 bool NewToSet = Myset.insert(Value);
1193 assert(NewToSet && "The value shouldn't be in the set yet");
1195 These are two interesting different cases. In the first case, the call to
1196 ``V.size()`` is only useful for the assert, and we don't want it executed when
1197 assertions are disabled. Code like this should move the call into the assert
1198 itself. In the second case, the side effects of the call must happen whether
1199 the assert is enabled or not. In this case, the value should be cast to void to
1200 disable the warning. To be specific, it is preferred to write the code like
1205 assert(V.size() > 42 && "Vector smaller than it should be");
1207 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1208 assert(NewToSet && "The value shouldn't be in the set yet");
1210 Do Not Use ``using namespace std``
1211 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1213 In LLVM, we prefer to explicitly prefix all identifiers from the standard
1214 namespace with an "``std::``" prefix, rather than rely on "``using namespace
1217 In header files, adding a ``'using namespace XXX'`` directive pollutes the
1218 namespace of any source file that ``#include``\s the header. This is clearly a
1221 In implementation files (e.g. ``.cpp`` files), the rule is more of a stylistic
1222 rule, but is still important. Basically, using explicit namespace prefixes
1223 makes the code **clearer**, because it is immediately obvious what facilities
1224 are being used and where they are coming from. And **more portable**, because
1225 namespace clashes cannot occur between LLVM code and other namespaces. The
1226 portability rule is important because different standard library implementations
1227 expose different symbols (potentially ones they shouldn't), and future revisions
1228 to the C++ standard will add more symbols to the ``std`` namespace. As such, we
1229 never use ``'using namespace std;'`` in LLVM.
1231 The exception to the general rule (i.e. it's not an exception for the ``std``
1232 namespace) is for implementation files. For example, all of the code in the
1233 LLVM project implements code that lives in the 'llvm' namespace. As such, it is
1234 ok, and actually clearer, for the ``.cpp`` files to have a ``'using namespace
1235 llvm;'`` directive at the top, after the ``#include``\s. This reduces
1236 indentation in the body of the file for source editors that indent based on
1237 braces, and keeps the conceptual context cleaner. The general form of this rule
1238 is that any ``.cpp`` file that implements code in any namespace may use that
1239 namespace (and its parents'), but should not use any others.
1241 Provide a Virtual Method Anchor for Classes in Headers
1242 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1244 If a class is defined in a header file and has a vtable (either it has virtual
1245 methods or it derives from classes with virtual methods), it must always have at
1246 least one out-of-line virtual method in the class. Without this, the compiler
1247 will copy the vtable and RTTI into every ``.o`` file that ``#include``\s the
1248 header, bloating ``.o`` file sizes and increasing link times.
1250 Don't use default labels in fully covered switches over enumerations
1251 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1253 ``-Wswitch`` warns if a switch, without a default label, over an enumeration
1254 does not cover every enumeration value. If you write a default label on a fully
1255 covered switch over an enumeration then the ``-Wswitch`` warning won't fire
1256 when new elements are added to that enumeration. To help avoid adding these
1257 kinds of defaults, Clang has the warning ``-Wcovered-switch-default`` which is
1258 off by default but turned on when building LLVM with a version of Clang that
1259 supports the warning.
1261 A knock-on effect of this stylistic requirement is that when building LLVM with
1262 GCC you may get warnings related to "control may reach end of non-void function"
1263 if you return from each case of a covered switch-over-enum because GCC assumes
1264 that the enum expression may take any representable value, not just those of
1265 individual enumerators. To suppress this warning, use ``llvm_unreachable`` after
1268 Use ``LLVM_DELETED_FUNCTION`` to mark uncallable methods
1269 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1271 Prior to C++11, a common pattern to make a class uncopyable was to declare an
1272 unimplemented copy constructor and copy assignment operator and make them
1273 private. This would give a compiler error for accessing a private method or a
1274 linker error because it wasn't implemented.
1276 With C++11, we can mark methods that won't be implemented with ``= delete``.
1277 This will trigger a much better error message and tell the compiler that the
1278 method will never be implemented. This enables other checks like
1279 ``-Wunused-private-field`` to run correctly on classes that contain these
1282 To maintain compatibility with C++03, ``LLVM_DELETED_FUNCTION`` should be used
1283 which will expand to ``= delete`` if the compiler supports it. These methods
1284 should still be declared private. Example of the uncopyable pattern:
1290 DontCopy(const DontCopy&) LLVM_DELETED_FUNCTION;
1291 DontCopy &operator =(const DontCopy&) LLVM_DELETED_FUNCTION;
1296 Don't evaluate ``end()`` every time through a loop
1297 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1299 Because C++ doesn't have a standard "``foreach``" loop (though it can be
1300 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1301 loops that manually iterate from begin to end on a variety of containers or
1302 through other data structures. One common mistake is to write a loop in this
1307 BasicBlock *BB = ...
1308 for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I)
1311 The problem with this construct is that it evaluates "``BB->end()``" every time
1312 through the loop. Instead of writing the loop like this, we strongly prefer
1313 loops to be written so that they evaluate it once before the loop starts. A
1314 convenient way to do this is like so:
1318 BasicBlock *BB = ...
1319 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
1322 The observant may quickly point out that these two loops may have different
1323 semantics: if the container (a basic block in this case) is being mutated, then
1324 "``BB->end()``" may change its value every time through the loop and the second
1325 loop may not in fact be correct. If you actually do depend on this behavior,
1326 please write the loop in the first form and add a comment indicating that you
1327 did it intentionally.
1329 Why do we prefer the second form (when correct)? Writing the loop in the first
1330 form has two problems. First it may be less efficient than evaluating it at the
1331 start of the loop. In this case, the cost is probably minor --- a few extra
1332 loads every time through the loop. However, if the base expression is more
1333 complex, then the cost can rise quickly. I've seen loops where the end
1334 expression was actually something like: "``SomeMap[X]->end()``" and map lookups
1335 really aren't cheap. By writing it in the second form consistently, you
1336 eliminate the issue entirely and don't even have to think about it.
1338 The second (even bigger) issue is that writing the loop in the first form hints
1339 to the reader that the loop is mutating the container (a fact that a comment
1340 would handily confirm!). If you write the loop in the second form, it is
1341 immediately obvious without even looking at the body of the loop that the
1342 container isn't being modified, which makes it easier to read the code and
1343 understand what it does.
1345 While the second form of the loop is a few extra keystrokes, we do strongly
1348 ``#include <iostream>`` is Forbidden
1349 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1351 The use of ``#include <iostream>`` in library files is hereby **forbidden**,
1352 because many common implementations transparently inject a `static constructor`_
1353 into every translation unit that includes it.
1355 Note that using the other stream headers (``<sstream>`` for example) is not
1356 problematic in this regard --- just ``<iostream>``. However, ``raw_ostream``
1357 provides various APIs that are better performing for almost every use than
1358 ``std::ostream`` style APIs.
1362 New code should always use `raw_ostream`_ for writing, or the
1363 ``llvm::MemoryBuffer`` API for reading files.
1370 LLVM includes a lightweight, simple, and efficient stream implementation in
1371 ``llvm/Support/raw_ostream.h``, which provides all of the common features of
1372 ``std::ostream``. All new code should use ``raw_ostream`` instead of
1375 Unlike ``std::ostream``, ``raw_ostream`` is not a template and can be forward
1376 declared as ``class raw_ostream``. Public headers should generally not include
1377 the ``raw_ostream`` header, but use forward declarations and constant references
1378 to ``raw_ostream`` instances.
1383 The ``std::endl`` modifier, when used with ``iostreams`` outputs a newline to
1384 the output stream specified. In addition to doing this, however, it also
1385 flushes the output stream. In other words, these are equivalent:
1389 std::cout << std::endl;
1390 std::cout << '\n' << std::flush;
1392 Most of the time, you probably have no reason to flush the output stream, so
1393 it's better to use a literal ``'\n'``.
1395 Don't use ``inline`` when defining a function in a class definition
1396 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1398 A member function defined in a class definition is implicitly inline, so don't
1399 put the ``inline`` keyword in this case.
1426 This section describes preferred low-level formatting guidelines along with
1427 reasoning on why we prefer them.
1429 Spaces Before Parentheses
1430 ^^^^^^^^^^^^^^^^^^^^^^^^^
1432 We prefer to put a space before an open parenthesis only in control flow
1433 statements, but not in normal function call expressions and function-like
1434 macros. For example, this is good:
1439 for (I = 0; I != 100; ++I) ...
1440 while (LLVMRocks) ...
1443 assert(3 != 4 && "laws of math are failing me");
1445 A = foo(42, 92) + bar(X);
1452 for(I = 0; I != 100; ++I) ...
1453 while(LLVMRocks) ...
1456 assert (3 != 4 && "laws of math are failing me");
1458 A = foo (42, 92) + bar (X);
1460 The reason for doing this is not completely arbitrary. This style makes control
1461 flow operators stand out more, and makes expressions flow better. The function
1462 call operator binds very tightly as a postfix operator. Putting a space after a
1463 function name (as in the last example) makes it appear that the code might bind
1464 the arguments of the left-hand-side of a binary operator with the argument list
1465 of a function and the name of the right side. More specifically, it is easy to
1466 misread the "``A``" example as:
1470 A = foo ((42, 92) + bar) (X);
1472 when skimming through the code. By avoiding a space in a function, we avoid
1473 this misinterpretation.
1478 Hard fast rule: Preincrement (``++X``) may be no slower than postincrement
1479 (``X++``) and could very well be a lot faster than it. Use preincrementation
1482 The semantics of postincrement include making a copy of the value being
1483 incremented, returning it, and then preincrementing the "work value". For
1484 primitive types, this isn't a big deal. But for iterators, it can be a huge
1485 issue (for example, some iterators contains stack and set objects in them...
1486 copying an iterator could invoke the copy ctor's of these as well). In general,
1487 get in the habit of always using preincrement, and you won't have a problem.
1490 Namespace Indentation
1491 ^^^^^^^^^^^^^^^^^^^^^
1493 In general, we strive to reduce indentation wherever possible. This is useful
1494 because we want code to `fit into 80 columns`_ without wrapping horribly, but
1495 also because it makes it easier to understand the code. To facilitate this and
1496 avoid some insanely deep nesting on occasion, don't indent namespaces. If it
1497 helps readability, feel free to add a comment indicating what namespace is
1498 being closed by a ``}``. For example:
1503 namespace knowledge {
1505 /// This class represents things that Smith can have an intimate
1506 /// understanding of and contains the data associated with it.
1510 explicit Grokable() { ... }
1511 virtual ~Grokable() = 0;
1517 } // end namespace knowledge
1518 } // end namespace llvm
1521 Feel free to skip the closing comment when the namespace being closed is
1522 obvious for any reason. For example, the outer-most namespace in a header file
1523 is rarely a source of confusion. But namespaces both anonymous and named in
1524 source files that are being closed half way through the file probably could use
1529 Anonymous Namespaces
1530 ^^^^^^^^^^^^^^^^^^^^
1532 After talking about namespaces in general, you may be wondering about anonymous
1533 namespaces in particular. Anonymous namespaces are a great language feature
1534 that tells the C++ compiler that the contents of the namespace are only visible
1535 within the current translation unit, allowing more aggressive optimization and
1536 eliminating the possibility of symbol name collisions. Anonymous namespaces are
1537 to C++ as "static" is to C functions and global variables. While "``static``"
1538 is available in C++, anonymous namespaces are more general: they can make entire
1539 classes private to a file.
1541 The problem with anonymous namespaces is that they naturally want to encourage
1542 indentation of their body, and they reduce locality of reference: if you see a
1543 random function definition in a C++ file, it is easy to see if it is marked
1544 static, but seeing if it is in an anonymous namespace requires scanning a big
1547 Because of this, we have a simple guideline: make anonymous namespaces as small
1548 as possible, and only use them for class declarations. For example, this is
1558 bool operator<(const char *RHS) const;
1560 } // end anonymous namespace
1562 static void runHelper() {
1566 bool StringSort::operator<(const char *RHS) const {
1580 bool operator<(const char *RHS) const;
1587 bool StringSort::operator<(const char *RHS) const {
1591 } // end anonymous namespace
1593 This is bad specifically because if you're looking at "``runHelper``" in the middle
1594 of a large C++ file, that you have no immediate way to tell if it is local to
1595 the file. When it is marked static explicitly, this is immediately obvious.
1596 Also, there is no reason to enclose the definition of "``operator<``" in the
1597 namespace just because it was declared there.
1602 A lot of these comments and recommendations have been culled from other sources.
1603 Two particularly important books for our work are:
1606 <http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876>`_
1607 by Scott Meyers. Also interesting and useful are "More Effective C++" and
1608 "Effective STL" by the same author.
1610 #. `Large-Scale C++ Software Design
1611 <http://www.amazon.com/Large-Scale-Software-Design-John-Lakos/dp/0201633620/ref=sr_1_1>`_
1614 If you get some free time, and you haven't read them: do so, you might learn