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* ``std::function``, until Clang implements `MSVC-compatible RTTI`_.
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_
119 * ``override`` and ``final``: N2928_, N3206_, N3272_
120 * Atomic operations and the C++11 memory model: N2429_
122 .. _N2118: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2118.html
123 .. _N2439: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2439.htm
124 .. _N1720: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1720.html
125 .. _N1984: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1984.pdf
126 .. _N1737: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1737.pdf
127 .. _N2541: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2541.htm
128 .. _N2927: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2927.pdf
129 .. _N2343: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2343.pdf
130 .. _N1757: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1757.html
131 .. _N1987: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1987.htm
132 .. _N2431: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2431.pdf
133 .. _N2347: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2347.pdf
134 .. _N2764: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2764.pdf
135 .. _N2657: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm
136 .. _N2930: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2930.html
137 .. _N2928: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2928.htm
138 .. _N3206: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3206.htm
139 .. _N3272: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3272.htm
140 .. _N2429: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2429.htm
141 .. _MSVC-compatible RTTI: http://llvm.org/PR18951
143 The supported features in the C++11 standard libraries are less well tracked,
144 but also much greater. Most of the standard libraries implement most of C++11's
145 library. The most likely lowest common denominator is Linux support. For
146 libc++, the support is just poorly tested and undocumented but expected to be
147 largely complete. YMMV. For libstdc++, the support is documented in detail in
148 `the libstdc++ manual`_. There are some very minor missing facilities that are
149 unlikely to be common problems, and there are a few larger gaps that are worth
152 * Not all of the type traits are implemented
153 * No regular expression library.
154 * While most of the atomics library is well implemented, the fences are
155 missing. Fortunately, they are rarely needed.
156 * The locale support is incomplete.
157 * ``std::initializer_list`` (and the constructors and functions that take it as
158 an argument) are not always available, so you cannot (for example) initialize
159 a ``std::vector`` with a braced initializer list.
161 Other than these areas you should assume the standard library is available and
162 working as expected until some build bot tells you otherwise. If you're in an
163 uncertain area of one of the above points, but you cannot test on a Linux
164 system, your best approach is to minimize your use of these features, and watch
165 the Linux build bots to find out if your usage triggered a bug. For example, if
166 you hit a type trait which doesn't work we can then add support to LLVM's
167 traits header to emulate it.
169 .. _the libstdc++ manual:
170 http://gcc.gnu.org/onlinedocs/gcc-4.7.3/libstdc++/manual/manual/status.html#status.iso.2011
172 Mechanical Source Issues
173 ========================
175 Source Code Formatting
176 ----------------------
181 Comments are one critical part of readability and maintainability. Everyone
182 knows they should comment their code, and so should you. When writing comments,
183 write them as English prose, which means they should use proper capitalization,
184 punctuation, etc. Aim to describe what the code is trying to do and why, not
185 *how* it does it at a micro level. Here are a few critical things to document:
187 .. _header file comment:
192 Every source file should have a header on it that describes the basic purpose of
193 the file. If a file does not have a header, it should not be checked into the
194 tree. The standard header looks like this:
198 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
200 // The LLVM Compiler Infrastructure
202 // This file is distributed under the University of Illinois Open Source
203 // License. See LICENSE.TXT for details.
205 //===----------------------------------------------------------------------===//
208 /// \brief This file contains the declaration of the Instruction class, which is
209 /// the base class for all of the VM instructions.
211 //===----------------------------------------------------------------------===//
213 A few things to note about this particular format: The "``-*- C++ -*-``" string
214 on the first line is there to tell Emacs that the source file is a C++ file, not
215 a C file (Emacs assumes ``.h`` files are C files by default).
219 This tag is not necessary in ``.cpp`` files. The name of the file is also
220 on the first line, along with a very short description of the purpose of the
221 file. This is important when printing out code and flipping though lots of
224 The next section in the file is a concise note that defines the license that the
225 file is released under. This makes it perfectly clear what terms the source
226 code can be distributed under and should not be modified in any way.
228 The main body is a ``doxygen`` comment describing the purpose of the file. It
229 should have a ``\brief`` command that describes the file in one or two
230 sentences. Any additional information should be separated by a blank line. If
231 an algorithm is being implemented or something tricky is going on, a reference
232 to the paper where it is published should be included, as well as any notes or
233 *gotchas* in the code to watch out for.
238 Classes are one fundamental part of a good object oriented design. As such, a
239 class definition should have a comment block that explains what the class is
240 used for and how it works. Every non-trivial class is expected to have a
241 ``doxygen`` comment block.
246 Methods defined in a class (as well as any global functions) should also be
247 documented properly. A quick note about what it does and a description of the
248 borderline behaviour is all that is necessary here (unless something
249 particularly tricky or insidious is going on). The hope is that people can
250 figure out how to use your interfaces without reading the code itself.
252 Good things to talk about here are what happens when something unexpected
253 happens: does the method return null? Abort? Format your hard disk?
258 In general, prefer C++ style (``//``) comments. They take less space, require
259 less typing, don't have nesting problems, etc. There are a few cases when it is
260 useful to use C style (``/* */``) comments however:
262 #. When writing C code: Obviously if you are writing C code, use C style
265 #. When writing a header file that may be ``#include``\d by a C source file.
267 #. When writing a source file that is used by a tool that only accepts C style
270 To comment out a large block of code, use ``#if 0`` and ``#endif``. These nest
271 properly and are better behaved in general than C style comments.
273 Doxygen Use in Documentation Comments
274 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
276 Use the ``\file`` command to turn the standard file header into a file-level
279 Include descriptive ``\brief`` paragraphs for all public interfaces (public
280 classes, member and non-member functions). Explain API use and purpose in
281 ``\brief`` paragraphs, don't just restate the information that can be inferred
282 from the API name. Put detailed discussion into separate paragraphs.
284 To refer to parameter names inside a paragraph, use the ``\p name`` command.
285 Don't use the ``\arg name`` command since it starts a new paragraph that
286 contains documentation for the parameter.
288 Wrap non-inline code examples in ``\code ... \endcode``.
290 To document a function parameter, start a new paragraph with the
291 ``\param name`` command. If the parameter is used as an out or an in/out
292 parameter, use the ``\param [out] name`` or ``\param [in,out] name`` command,
295 To describe function return value, start a new paragraph with the ``\returns``
298 A minimal documentation comment:
302 /// \brief Does foo and bar.
303 void fooBar(bool Baz);
305 A documentation comment that uses all Doxygen features in a preferred way:
309 /// \brief Does foo and bar.
311 /// Does not do foo the usual way if \p Baz is true.
315 /// fooBar(false, "quux", Res);
318 /// \param Quux kind of foo to do.
319 /// \param [out] Result filled with bar sequence on foo success.
321 /// \returns true on success.
322 bool fooBar(bool Baz, StringRef Quux, std::vector<int> &Result);
324 Don't duplicate the documentation comment in the header file and in the
325 implementation file. Put the documentation comments for public APIs into the
326 header file. Documentation comments for private APIs can go to the
327 implementation file. In any case, implementation files can include additional
328 comments (not necessarily in Doxygen markup) to explain implementation details
331 Don't duplicate function or class name at the beginning of the comment.
332 For humans it is obvious which function or class is being documented;
333 automatic documentation processing tools are smart enough to bind the comment
334 to the correct declaration.
342 /// Something - An abstraction for some complicated thing.
345 /// fooBar - Does foo and bar.
351 /// fooBar - Does foo and bar.
352 void Something::fooBar() { ... }
360 /// \brief An abstraction for some complicated thing.
363 /// \brief Does foo and bar.
369 // Builds a B-tree in order to do foo. See paper by...
370 void Something::fooBar() { ... }
372 It is not required to use additional Doxygen features, but sometimes it might
373 be a good idea to do so.
377 * adding comments to any narrow namespace containing a collection of
378 related functions or types;
380 * using top-level groups to organize a collection of related functions at
381 namespace scope where the grouping is smaller than the namespace;
383 * using member groups and additional comments attached to member
384 groups to organize within a class.
391 /// \name Functions that do Foo.
402 Immediately after the `header file comment`_ (and include guards if working on a
403 header file), the `minimal list of #includes`_ required by the file should be
404 listed. We prefer these ``#include``\s to be listed in this order:
406 .. _Main Module Header:
407 .. _Local/Private Headers:
409 #. Main Module Header
410 #. Local/Private Headers
412 #. System ``#include``\s
414 and each category should be sorted lexicographically by the full path.
416 The `Main Module Header`_ file applies to ``.cpp`` files which implement an
417 interface defined by a ``.h`` file. This ``#include`` should always be included
418 **first** regardless of where it lives on the file system. By including a
419 header file first in the ``.cpp`` files that implement the interfaces, we ensure
420 that the header does not have any hidden dependencies which are not explicitly
421 ``#include``\d in the header, but should be. It is also a form of documentation
422 in the ``.cpp`` file to indicate where the interfaces it implements are defined.
424 .. _fit into 80 columns:
429 Write your code to fit within 80 columns of text. This helps those of us who
430 like to print out code and look at your code in an ``xterm`` without resizing
433 The longer answer is that there must be some limit to the width of the code in
434 order to reasonably allow developers to have multiple files side-by-side in
435 windows on a modest display. If you are going to pick a width limit, it is
436 somewhat arbitrary but you might as well pick something standard. Going with 90
437 columns (for example) instead of 80 columns wouldn't add any significant value
438 and would be detrimental to printing out code. Also many other projects have
439 standardized on 80 columns, so some people have already configured their editors
440 for it (vs something else, like 90 columns).
442 This is one of many contentious issues in coding standards, but it is not up for
445 Use Spaces Instead of Tabs
446 ^^^^^^^^^^^^^^^^^^^^^^^^^^
448 In all cases, prefer spaces to tabs in source files. People have different
449 preferred indentation levels, and different styles of indentation that they
450 like; this is fine. What isn't fine is that different editors/viewers expand
451 tabs out to different tab stops. This can cause your code to look completely
452 unreadable, and it is not worth dealing with.
454 As always, follow the `Golden Rule`_ above: follow the style of
455 existing code if you are modifying and extending it. If you like four spaces of
456 indentation, **DO NOT** do that in the middle of a chunk of code with two spaces
457 of indentation. Also, do not reindent a whole source file: it makes for
458 incredible diffs that are absolutely worthless.
460 Indent Code Consistently
461 ^^^^^^^^^^^^^^^^^^^^^^^^
463 Okay, in your first year of programming you were told that indentation is
464 important. If you didn't believe and internalize this then, now is the time.
465 Just do it. With the introduction of C++11, there are some new formatting
466 challenges that merit some suggestions to help have consistent, maintainable,
467 and tool-friendly formatting and indentation.
469 Format Lambdas Like Blocks Of Code
470 """"""""""""""""""""""""""""""""""
472 When formatting a multi-line lambda, format it like a block of code, that's
473 what it is. If there is only one multi-line lambda in a statement, and there
474 are no expressions lexically after it in the statement, drop the indent to the
475 standard two space indent for a block of code, as if it were an if-block opened
476 by the preceding part of the statement:
480 std::sort(foo.begin(), foo.end(), [&](Foo a, Foo b) -> bool {
485 return a.bam < b.bam;
488 To take best advantage of this formatting, if you are designing an API which
489 accepts a continuation or single callable argument (be it a functor, or
490 a ``std::function``), it should be the last argument if at all possible.
492 If there are multiple multi-line lambdas in a statement, or there is anything
493 interesting after the lambda in the statement, indent the block two spaces from
494 the indent of the ``[]``:
498 dyn_switch(V->stripPointerCasts(),
502 [] (SelectInst *SI) {
503 // process selects...
508 [] (AllocaInst *AI) {
509 // process allocas...
512 Braced Initializer Lists
513 """"""""""""""""""""""""
515 With C++11, there are significantly more uses of braced lists to perform
516 initialization. These allow you to easily construct aggregate temporaries in
517 expressions among other niceness. They now have a natural way of ending up
518 nested within each other and within function calls in order to build up
519 aggregates (such as option structs) from local variables. To make matters
520 worse, we also have many more uses of braces in an expression context that are
521 *not* performing initialization.
523 The historically common formatting of braced initialization of aggregate
524 variables does not mix cleanly with deep nesting, general expression contexts,
525 function arguments, and lambdas. We suggest new code use a simple rule for
526 formatting braced initialization lists: act as-if the braces were parentheses
527 in a function call. The formatting rules exactly match those already well
528 understood for formatting nested function calls. Examples:
532 foo({a, b, c}, {1, 2, 3});
534 llvm::Constant *Mask[] = {
535 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 0),
536 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 1),
537 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 2)};
539 This formatting scheme also makes it particularly easy to get predictable,
540 consistent, and automatic formatting with tools like `Clang Format`_.
542 .. _Clang Format: http://clang.llvm.org/docs/ClangFormat.html
544 Language and Compiler Issues
545 ----------------------------
547 Treat Compiler Warnings Like Errors
548 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
550 If your code has compiler warnings in it, something is wrong --- you aren't
551 casting values correctly, you have "questionable" constructs in your code, or
552 you are doing something legitimately wrong. Compiler warnings can cover up
553 legitimate errors in output and make dealing with a translation unit difficult.
555 It is not possible to prevent all warnings from all compilers, nor is it
556 desirable. Instead, pick a standard compiler (like ``gcc``) that provides a
557 good thorough set of warnings, and stick to it. At least in the case of
558 ``gcc``, it is possible to work around any spurious errors by changing the
559 syntax of the code slightly. For example, a warning that annoys me occurs when
560 I write code like this:
564 if (V = getValue()) {
568 ``gcc`` will warn me that I probably want to use the ``==`` operator, and that I
569 probably mistyped it. In most cases, I haven't, and I really don't want the
570 spurious errors. To fix this particular problem, I rewrite the code like
575 if ((V = getValue())) {
579 which shuts ``gcc`` up. Any ``gcc`` warning that annoys you can be fixed by
580 massaging the code appropriately.
585 In almost all cases, it is possible and within reason to write completely
586 portable code. If there are cases where it isn't possible to write portable
587 code, isolate it behind a well defined (and well documented) interface.
589 In practice, this means that you shouldn't assume much about the host compiler
590 (and Visual Studio tends to be the lowest common denominator). If advanced
591 features are used, they should only be an implementation detail of a library
592 which has a simple exposed API, and preferably be buried in ``libSystem``.
594 Do not use RTTI or Exceptions
595 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
597 In an effort to reduce code and executable size, LLVM does not use RTTI
598 (e.g. ``dynamic_cast<>;``) or exceptions. These two language features violate
599 the general C++ principle of *"you only pay for what you use"*, causing
600 executable bloat even if exceptions are never used in the code base, or if RTTI
601 is never used for a class. Because of this, we turn them off globally in the
604 That said, LLVM does make extensive use of a hand-rolled form of RTTI that use
605 templates like :ref:`isa\<>, cast\<>, and dyn_cast\<> <isa>`.
606 This form of RTTI is opt-in and can be
607 :doc:`added to any class <HowToSetUpLLVMStyleRTTI>`. It is also
608 substantially more efficient than ``dynamic_cast<>``.
610 .. _static constructor:
612 Do not use Static Constructors
613 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
615 Static constructors and destructors (e.g. global variables whose types have a
616 constructor or destructor) should not be added to the code base, and should be
617 removed wherever possible. Besides `well known problems
618 <http://yosefk.com/c++fqa/ctors.html#fqa-10.12>`_ where the order of
619 initialization is undefined between globals in different source files, the
620 entire concept of static constructors is at odds with the common use case of
621 LLVM as a library linked into a larger application.
623 Consider the use of LLVM as a JIT linked into another application (perhaps for
624 `OpenGL, custom languages <http://llvm.org/Users.html>`_, `shaders in movies
625 <http://llvm.org/devmtg/2010-11/Gritz-OpenShadingLang.pdf>`_, etc). Due to the
626 design of static constructors, they must be executed at startup time of the
627 entire application, regardless of whether or how LLVM is used in that larger
628 application. There are two problems with this:
630 * The time to run the static constructors impacts startup time of applications
631 --- a critical time for GUI apps, among others.
633 * The static constructors cause the app to pull many extra pages of memory off
634 the disk: both the code for the constructor in each ``.o`` file and the small
635 amount of data that gets touched. In addition, touched/dirty pages put more
636 pressure on the VM system on low-memory machines.
638 We would really like for there to be zero cost for linking in an additional LLVM
639 target or other library into an application, but static constructors violate
642 That said, LLVM unfortunately does contain static constructors. It would be a
643 `great project <http://llvm.org/PR11944>`_ for someone to purge all static
644 constructors from LLVM, and then enable the ``-Wglobal-constructors`` warning
645 flag (when building with Clang) to ensure we do not regress in the future.
647 Use of ``class`` and ``struct`` Keywords
648 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
650 In C++, the ``class`` and ``struct`` keywords can be used almost
651 interchangeably. The only difference is when they are used to declare a class:
652 ``class`` makes all members private by default while ``struct`` makes all
653 members public by default.
655 Unfortunately, not all compilers follow the rules and some will generate
656 different symbols based on whether ``class`` or ``struct`` was used to declare
657 the symbol (e.g., MSVC). This can lead to problems at link time.
659 * All declarations and definitions of a given ``class`` or ``struct`` must use
660 the same keyword. For example:
666 // Breaks mangling in MSVC.
667 struct Foo { int Data; };
669 * As a rule of thumb, ``struct`` should be kept to structures where *all*
670 members are declared public.
674 // Foo feels like a class... this is strange.
680 int getData() const { return Data; }
681 void setData(int D) { Data = D; }
684 // Bar isn't POD, but it does look like a struct.
690 Do not use Braced Initializer Lists to Call a Constructor
691 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
693 In C++11 there is a "generalized initialization syntax" which allows calling
694 constructors using braced initializer lists. Do not use these to call
695 constructors with any interesting logic or if you care that you're calling some
696 *particular* constructor. Those should look like function calls using
697 parentheses rather than like aggregate initialization. Similarly, if you need
698 to explicitly name the type and call its constructor to create a temporary,
699 don't use a braced initializer list. Instead, use a braced initializer list
700 (without any type for temporaries) when doing aggregate initialization or
701 something notionally equivalent. Examples:
707 // Construct a Foo by reading data from the disk in the whizbang format, ...
708 Foo(std::string filename);
710 // Construct a Foo by looking up the Nth element of some global data ...
716 // The Foo constructor call is very deliberate, no braces.
717 std::fill(foo.begin(), foo.end(), Foo("name"));
719 // The pair is just being constructed like an aggregate, use braces.
720 bar_map.insert({my_key, my_value});
722 If you use a braced initializer list when initializing a variable, use an equals before the open curly brace:
726 int data[] = {0, 1, 2, 3};
728 Use ``auto`` Type Deduction to Make Code More Readable
729 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
731 Some are advocating a policy of "almost always ``auto``" in C++11, however LLVM
732 uses a more moderate stance. Use ``auto`` if and only if it makes the code more
733 readable or easier to maintain. Don't "almost always" use ``auto``, but do use
734 ``auto`` with initializers like ``cast<Foo>(...)`` or other places where the
735 type is already obvious from the context. Another time when ``auto`` works well
736 for these purposes is when the type would have been abstracted away anyways,
737 often behind a container's typedef such as ``std::vector<T>::iterator``.
739 Beware unnecessary copies with ``auto``
740 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
742 The convenience of ``auto`` makes it easy to forget that its default behavior
743 is a copy. Particularly in range-based ``for`` loops, careless copies are
746 As a rule of thumb, use ``auto &`` unless you need to copy the result, and use
747 ``auto *`` when copying pointers.
751 // Typically there's no reason to copy.
752 for (const auto &Val : Container) { observe(Val); }
753 for (auto &Val : Container) { Val.change(); }
755 // Remove the reference if you really want a new copy.
756 for (auto Val : Container) { Val.change(); saveSomewhere(Val); }
758 // Copy pointers, but make it clear that they're pointers.
759 for (const auto *Ptr : Container) { observe(*Ptr); }
760 for (auto *Ptr : Container) { Ptr->change(); }
765 The High-Level Issues
766 ---------------------
768 A Public Header File **is** a Module
769 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
771 C++ doesn't do too well in the modularity department. There is no real
772 encapsulation or data hiding (unless you use expensive protocol classes), but it
773 is what we have to work with. When you write a public header file (in the LLVM
774 source tree, they live in the top level "``include``" directory), you are
775 defining a module of functionality.
777 Ideally, modules should be completely independent of each other, and their
778 header files should only ``#include`` the absolute minimum number of headers
779 possible. A module is not just a class, a function, or a namespace: it's a
780 collection of these that defines an interface. This interface may be several
781 functions, classes, or data structures, but the important issue is how they work
784 In general, a module should be implemented by one or more ``.cpp`` files. Each
785 of these ``.cpp`` files should include the header that defines their interface
786 first. This ensures that all of the dependences of the module header have been
787 properly added to the module header itself, and are not implicit. System
788 headers should be included after user headers for a translation unit.
790 .. _minimal list of #includes:
792 ``#include`` as Little as Possible
793 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
795 ``#include`` hurts compile time performance. Don't do it unless you have to,
796 especially in header files.
798 But wait! Sometimes you need to have the definition of a class to use it, or to
799 inherit from it. In these cases go ahead and ``#include`` that header file. Be
800 aware however that there are many cases where you don't need to have the full
801 definition of a class. If you are using a pointer or reference to a class, you
802 don't need the header file. If you are simply returning a class instance from a
803 prototyped function or method, you don't need it. In fact, for most cases, you
804 simply don't need the definition of a class. And not ``#include``\ing speeds up
807 It is easy to try to go too overboard on this recommendation, however. You
808 **must** include all of the header files that you are using --- you can include
809 them either directly or indirectly through another header file. To make sure
810 that you don't accidentally forget to include a header file in your module
811 header, make sure to include your module header **first** in the implementation
812 file (as mentioned above). This way there won't be any hidden dependencies that
813 you'll find out about later.
815 Keep "Internal" Headers Private
816 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
818 Many modules have a complex implementation that causes them to use more than one
819 implementation (``.cpp``) file. It is often tempting to put the internal
820 communication interface (helper classes, extra functions, etc) in the public
821 module header file. Don't do this!
823 If you really need to do something like this, put a private header file in the
824 same directory as the source files, and include it locally. This ensures that
825 your private interface remains private and undisturbed by outsiders.
829 It's okay to put extra implementation methods in a public class itself. Just
830 make them private (or protected) and all is well.
834 Use Early Exits and ``continue`` to Simplify Code
835 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
837 When reading code, keep in mind how much state and how many previous decisions
838 have to be remembered by the reader to understand a block of code. Aim to
839 reduce indentation where possible when it doesn't make it more difficult to
840 understand the code. One great way to do this is by making use of early exits
841 and the ``continue`` keyword in long loops. As an example of using an early
842 exit from a function, consider this "bad" code:
846 Value *doSomething(Instruction *I) {
847 if (!isa<TerminatorInst>(I) &&
848 I->hasOneUse() && doOtherThing(I)) {
849 ... some long code ....
855 This code has several problems if the body of the ``'if'`` is large. When
856 you're looking at the top of the function, it isn't immediately clear that this
857 *only* does interesting things with non-terminator instructions, and only
858 applies to things with the other predicates. Second, it is relatively difficult
859 to describe (in comments) why these predicates are important because the ``if``
860 statement makes it difficult to lay out the comments. Third, when you're deep
861 within the body of the code, it is indented an extra level. Finally, when
862 reading the top of the function, it isn't clear what the result is if the
863 predicate isn't true; you have to read to the end of the function to know that
866 It is much preferred to format the code like this:
870 Value *doSomething(Instruction *I) {
871 // Terminators never need 'something' done to them because ...
872 if (isa<TerminatorInst>(I))
875 // We conservatively avoid transforming instructions with multiple uses
876 // because goats like cheese.
880 // This is really just here for example.
881 if (!doOtherThing(I))
884 ... some long code ....
887 This fixes these problems. A similar problem frequently happens in ``for``
888 loops. A silly example is something like this:
892 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
893 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
894 Value *LHS = BO->getOperand(0);
895 Value *RHS = BO->getOperand(1);
902 When you have very, very small loops, this sort of structure is fine. But if it
903 exceeds more than 10-15 lines, it becomes difficult for people to read and
904 understand at a glance. The problem with this sort of code is that it gets very
905 nested very quickly. Meaning that the reader of the code has to keep a lot of
906 context in their brain to remember what is going immediately on in the loop,
907 because they don't know if/when the ``if`` conditions will have ``else``\s etc.
908 It is strongly preferred to structure the loop like this:
912 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
913 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
916 Value *LHS = BO->getOperand(0);
917 Value *RHS = BO->getOperand(1);
918 if (LHS == RHS) continue;
923 This has all the benefits of using early exits for functions: it reduces nesting
924 of the loop, it makes it easier to describe why the conditions are true, and it
925 makes it obvious to the reader that there is no ``else`` coming up that they
926 have to push context into their brain for. If a loop is large, this can be a
927 big understandability win.
929 Don't use ``else`` after a ``return``
930 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
932 For similar reasons above (reduction of indentation and easier reading), please
933 do not use ``'else'`` or ``'else if'`` after something that interrupts control
934 flow --- like ``return``, ``break``, ``continue``, ``goto``, etc. For
935 example, this is *bad*:
941 Type = Context.getsigjmp_bufType();
943 Error = ASTContext::GE_Missing_sigjmp_buf;
949 Type = Context.getjmp_bufType();
951 Error = ASTContext::GE_Missing_jmp_buf;
959 It is better to write it like this:
965 Type = Context.getsigjmp_bufType();
967 Error = ASTContext::GE_Missing_sigjmp_buf;
971 Type = Context.getjmp_bufType();
973 Error = ASTContext::GE_Missing_jmp_buf;
979 Or better yet (in this case) as:
985 Type = Context.getsigjmp_bufType();
987 Type = Context.getjmp_bufType();
990 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
991 ASTContext::GE_Missing_jmp_buf;
996 The idea is to reduce indentation and the amount of code you have to keep track
997 of when reading the code.
999 Turn Predicate Loops into Predicate Functions
1000 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1002 It is very common to write small loops that just compute a boolean value. There
1003 are a number of ways that people commonly write these, but an example of this
1008 bool FoundFoo = false;
1009 for (unsigned I = 0, E = BarList.size(); I != E; ++I)
1010 if (BarList[I]->isFoo()) {
1019 This sort of code is awkward to write, and is almost always a bad sign. Instead
1020 of this sort of loop, we strongly prefer to use a predicate function (which may
1021 be `static`_) that uses `early exits`_ to compute the predicate. We prefer the
1022 code to be structured like this:
1026 /// \returns true if the specified list has an element that is a foo.
1027 static bool containsFoo(const std::vector<Bar*> &List) {
1028 for (unsigned I = 0, E = List.size(); I != E; ++I)
1029 if (List[I]->isFoo())
1035 if (containsFoo(BarList)) {
1039 There are many reasons for doing this: it reduces indentation and factors out
1040 code which can often be shared by other code that checks for the same predicate.
1041 More importantly, it *forces you to pick a name* for the function, and forces
1042 you to write a comment for it. In this silly example, this doesn't add much
1043 value. However, if the condition is complex, this can make it a lot easier for
1044 the reader to understand the code that queries for this predicate. Instead of
1045 being faced with the in-line details of how we check to see if the BarList
1046 contains a foo, we can trust the function name and continue reading with better
1049 The Low-Level Issues
1050 --------------------
1052 Name Types, Functions, Variables, and Enumerators Properly
1053 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1055 Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
1056 enough how important it is to use *descriptive* names. Pick names that match
1057 the semantics and role of the underlying entities, within reason. Avoid
1058 abbreviations unless they are well known. After picking a good name, make sure
1059 to use consistent capitalization for the name, as inconsistency requires clients
1060 to either memorize the APIs or to look it up to find the exact spelling.
1062 In general, names should be in camel case (e.g. ``TextFileReader`` and
1063 ``isLValue()``). Different kinds of declarations have different rules:
1065 * **Type names** (including classes, structs, enums, typedefs, etc) should be
1066 nouns and start with an upper-case letter (e.g. ``TextFileReader``).
1068 * **Variable names** should be nouns (as they represent state). The name should
1069 be camel case, and start with an upper case letter (e.g. ``Leader`` or
1072 * **Function names** should be verb phrases (as they represent actions), and
1073 command-like function should be imperative. The name should be camel case,
1074 and start with a lower case letter (e.g. ``openFile()`` or ``isFoo()``).
1076 * **Enum declarations** (e.g. ``enum Foo {...}``) are types, so they should
1077 follow the naming conventions for types. A common use for enums is as a
1078 discriminator for a union, or an indicator of a subclass. When an enum is
1079 used for something like this, it should have a ``Kind`` suffix
1080 (e.g. ``ValueKind``).
1082 * **Enumerators** (e.g. ``enum { Foo, Bar }``) and **public member variables**
1083 should start with an upper-case letter, just like types. Unless the
1084 enumerators are defined in their own small namespace or inside a class,
1085 enumerators should have a prefix corresponding to the enum declaration name.
1086 For example, ``enum ValueKind { ... };`` may contain enumerators like
1087 ``VK_Argument``, ``VK_BasicBlock``, etc. Enumerators that are just
1088 convenience constants are exempt from the requirement for a prefix. For
1098 As an exception, classes that mimic STL classes can have member names in STL's
1099 style of lower-case words separated by underscores (e.g. ``begin()``,
1100 ``push_back()``, and ``empty()``). Classes that provide multiple
1101 iterators should add a singular prefix to ``begin()`` and ``end()``
1102 (e.g. ``global_begin()`` and ``use_begin()``).
1104 Here are some examples of good and bad names:
1108 class VehicleMaker {
1110 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
1111 Factory<Tire> Factory; // Better.
1112 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
1113 // kind of factories.
1116 Vehicle MakeVehicle(VehicleType Type) {
1117 VehicleMaker M; // Might be OK if having a short life-span.
1118 Tire Tmp1 = M.makeTire(); // Bad -- 'Tmp1' provides no information.
1119 Light Headlight = M.makeLight("head"); // Good -- descriptive.
1126 Use the "``assert``" macro to its fullest. Check all of your preconditions and
1127 assumptions, you never know when a bug (not necessarily even yours) might be
1128 caught early by an assertion, which reduces debugging time dramatically. The
1129 "``<cassert>``" header file is probably already included by the header files you
1130 are using, so it doesn't cost anything to use it.
1132 To further assist with debugging, make sure to put some kind of error message in
1133 the assertion statement, which is printed if the assertion is tripped. This
1134 helps the poor debugger make sense of why an assertion is being made and
1135 enforced, and hopefully what to do about it. Here is one complete example:
1139 inline Value *getOperand(unsigned I) {
1140 assert(I < Operands.size() && "getOperand() out of range!");
1144 Here are more examples:
1148 assert(Ty->isPointerType() && "Can't allocate a non-pointer type!");
1150 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
1152 assert(idx < getNumSuccessors() && "Successor # out of range!");
1154 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
1156 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
1160 In the past, asserts were used to indicate a piece of code that should not be
1161 reached. These were typically of the form:
1165 assert(0 && "Invalid radix for integer literal");
1167 This has a few issues, the main one being that some compilers might not
1168 understand the assertion, or warn about a missing return in builds where
1169 assertions are compiled out.
1171 Today, we have something much better: ``llvm_unreachable``:
1175 llvm_unreachable("Invalid radix for integer literal");
1177 When assertions are enabled, this will print the message if it's ever reached
1178 and then exit the program. When assertions are disabled (i.e. in release
1179 builds), ``llvm_unreachable`` becomes a hint to compilers to skip generating
1180 code for this branch. If the compiler does not support this, it will fall back
1181 to the "abort" implementation.
1183 Another issue is that values used only by assertions will produce an "unused
1184 value" warning when assertions are disabled. For example, this code will warn:
1188 unsigned Size = V.size();
1189 assert(Size > 42 && "Vector smaller than it should be");
1191 bool NewToSet = Myset.insert(Value);
1192 assert(NewToSet && "The value shouldn't be in the set yet");
1194 These are two interesting different cases. In the first case, the call to
1195 ``V.size()`` is only useful for the assert, and we don't want it executed when
1196 assertions are disabled. Code like this should move the call into the assert
1197 itself. In the second case, the side effects of the call must happen whether
1198 the assert is enabled or not. In this case, the value should be cast to void to
1199 disable the warning. To be specific, it is preferred to write the code like
1204 assert(V.size() > 42 && "Vector smaller than it should be");
1206 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1207 assert(NewToSet && "The value shouldn't be in the set yet");
1209 Do Not Use ``using namespace std``
1210 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1212 In LLVM, we prefer to explicitly prefix all identifiers from the standard
1213 namespace with an "``std::``" prefix, rather than rely on "``using namespace
1216 In header files, adding a ``'using namespace XXX'`` directive pollutes the
1217 namespace of any source file that ``#include``\s the header. This is clearly a
1220 In implementation files (e.g. ``.cpp`` files), the rule is more of a stylistic
1221 rule, but is still important. Basically, using explicit namespace prefixes
1222 makes the code **clearer**, because it is immediately obvious what facilities
1223 are being used and where they are coming from. And **more portable**, because
1224 namespace clashes cannot occur between LLVM code and other namespaces. The
1225 portability rule is important because different standard library implementations
1226 expose different symbols (potentially ones they shouldn't), and future revisions
1227 to the C++ standard will add more symbols to the ``std`` namespace. As such, we
1228 never use ``'using namespace std;'`` in LLVM.
1230 The exception to the general rule (i.e. it's not an exception for the ``std``
1231 namespace) is for implementation files. For example, all of the code in the
1232 LLVM project implements code that lives in the 'llvm' namespace. As such, it is
1233 ok, and actually clearer, for the ``.cpp`` files to have a ``'using namespace
1234 llvm;'`` directive at the top, after the ``#include``\s. This reduces
1235 indentation in the body of the file for source editors that indent based on
1236 braces, and keeps the conceptual context cleaner. The general form of this rule
1237 is that any ``.cpp`` file that implements code in any namespace may use that
1238 namespace (and its parents'), but should not use any others.
1240 Provide a Virtual Method Anchor for Classes in Headers
1241 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1243 If a class is defined in a header file and has a vtable (either it has virtual
1244 methods or it derives from classes with virtual methods), it must always have at
1245 least one out-of-line virtual method in the class. Without this, the compiler
1246 will copy the vtable and RTTI into every ``.o`` file that ``#include``\s the
1247 header, bloating ``.o`` file sizes and increasing link times.
1249 Don't use default labels in fully covered switches over enumerations
1250 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1252 ``-Wswitch`` warns if a switch, without a default label, over an enumeration
1253 does not cover every enumeration value. If you write a default label on a fully
1254 covered switch over an enumeration then the ``-Wswitch`` warning won't fire
1255 when new elements are added to that enumeration. To help avoid adding these
1256 kinds of defaults, Clang has the warning ``-Wcovered-switch-default`` which is
1257 off by default but turned on when building LLVM with a version of Clang that
1258 supports the warning.
1260 A knock-on effect of this stylistic requirement is that when building LLVM with
1261 GCC you may get warnings related to "control may reach end of non-void function"
1262 if you return from each case of a covered switch-over-enum because GCC assumes
1263 that the enum expression may take any representable value, not just those of
1264 individual enumerators. To suppress this warning, use ``llvm_unreachable`` after
1267 Use ``LLVM_DELETED_FUNCTION`` to mark uncallable methods
1268 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1270 Prior to C++11, a common pattern to make a class uncopyable was to declare an
1271 unimplemented copy constructor and copy assignment operator and make them
1272 private. This would give a compiler error for accessing a private method or a
1273 linker error because it wasn't implemented.
1275 With C++11, we can mark methods that won't be implemented with ``= delete``.
1276 This will trigger a much better error message and tell the compiler that the
1277 method will never be implemented. This enables other checks like
1278 ``-Wunused-private-field`` to run correctly on classes that contain these
1281 To maintain compatibility with C++03, ``LLVM_DELETED_FUNCTION`` should be used
1282 which will expand to ``= delete`` if the compiler supports it. These methods
1283 should still be declared private. Example of the uncopyable pattern:
1289 DontCopy(const DontCopy&) LLVM_DELETED_FUNCTION;
1290 DontCopy &operator =(const DontCopy&) LLVM_DELETED_FUNCTION;
1295 Don't evaluate ``end()`` every time through a loop
1296 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1298 Because C++ doesn't have a standard "``foreach``" loop (though it can be
1299 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1300 loops that manually iterate from begin to end on a variety of containers or
1301 through other data structures. One common mistake is to write a loop in this
1306 BasicBlock *BB = ...
1307 for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I)
1310 The problem with this construct is that it evaluates "``BB->end()``" every time
1311 through the loop. Instead of writing the loop like this, we strongly prefer
1312 loops to be written so that they evaluate it once before the loop starts. A
1313 convenient way to do this is like so:
1317 BasicBlock *BB = ...
1318 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
1321 The observant may quickly point out that these two loops may have different
1322 semantics: if the container (a basic block in this case) is being mutated, then
1323 "``BB->end()``" may change its value every time through the loop and the second
1324 loop may not in fact be correct. If you actually do depend on this behavior,
1325 please write the loop in the first form and add a comment indicating that you
1326 did it intentionally.
1328 Why do we prefer the second form (when correct)? Writing the loop in the first
1329 form has two problems. First it may be less efficient than evaluating it at the
1330 start of the loop. In this case, the cost is probably minor --- a few extra
1331 loads every time through the loop. However, if the base expression is more
1332 complex, then the cost can rise quickly. I've seen loops where the end
1333 expression was actually something like: "``SomeMap[X]->end()``" and map lookups
1334 really aren't cheap. By writing it in the second form consistently, you
1335 eliminate the issue entirely and don't even have to think about it.
1337 The second (even bigger) issue is that writing the loop in the first form hints
1338 to the reader that the loop is mutating the container (a fact that a comment
1339 would handily confirm!). If you write the loop in the second form, it is
1340 immediately obvious without even looking at the body of the loop that the
1341 container isn't being modified, which makes it easier to read the code and
1342 understand what it does.
1344 While the second form of the loop is a few extra keystrokes, we do strongly
1347 ``#include <iostream>`` is Forbidden
1348 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1350 The use of ``#include <iostream>`` in library files is hereby **forbidden**,
1351 because many common implementations transparently inject a `static constructor`_
1352 into every translation unit that includes it.
1354 Note that using the other stream headers (``<sstream>`` for example) is not
1355 problematic in this regard --- just ``<iostream>``. However, ``raw_ostream``
1356 provides various APIs that are better performing for almost every use than
1357 ``std::ostream`` style APIs.
1361 New code should always use `raw_ostream`_ for writing, or the
1362 ``llvm::MemoryBuffer`` API for reading files.
1369 LLVM includes a lightweight, simple, and efficient stream implementation in
1370 ``llvm/Support/raw_ostream.h``, which provides all of the common features of
1371 ``std::ostream``. All new code should use ``raw_ostream`` instead of
1374 Unlike ``std::ostream``, ``raw_ostream`` is not a template and can be forward
1375 declared as ``class raw_ostream``. Public headers should generally not include
1376 the ``raw_ostream`` header, but use forward declarations and constant references
1377 to ``raw_ostream`` instances.
1382 The ``std::endl`` modifier, when used with ``iostreams`` outputs a newline to
1383 the output stream specified. In addition to doing this, however, it also
1384 flushes the output stream. In other words, these are equivalent:
1388 std::cout << std::endl;
1389 std::cout << '\n' << std::flush;
1391 Most of the time, you probably have no reason to flush the output stream, so
1392 it's better to use a literal ``'\n'``.
1394 Don't use ``inline`` when defining a function in a class definition
1395 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1397 A member function defined in a class definition is implicitly inline, so don't
1398 put the ``inline`` keyword in this case.
1425 This section describes preferred low-level formatting guidelines along with
1426 reasoning on why we prefer them.
1428 Spaces Before Parentheses
1429 ^^^^^^^^^^^^^^^^^^^^^^^^^
1431 We prefer to put a space before an open parenthesis only in control flow
1432 statements, but not in normal function call expressions and function-like
1433 macros. For example, this is good:
1438 for (I = 0; I != 100; ++I) ...
1439 while (LLVMRocks) ...
1442 assert(3 != 4 && "laws of math are failing me");
1444 A = foo(42, 92) + bar(X);
1451 for(I = 0; I != 100; ++I) ...
1452 while(LLVMRocks) ...
1455 assert (3 != 4 && "laws of math are failing me");
1457 A = foo (42, 92) + bar (X);
1459 The reason for doing this is not completely arbitrary. This style makes control
1460 flow operators stand out more, and makes expressions flow better. The function
1461 call operator binds very tightly as a postfix operator. Putting a space after a
1462 function name (as in the last example) makes it appear that the code might bind
1463 the arguments of the left-hand-side of a binary operator with the argument list
1464 of a function and the name of the right side. More specifically, it is easy to
1465 misread the "``A``" example as:
1469 A = foo ((42, 92) + bar) (X);
1471 when skimming through the code. By avoiding a space in a function, we avoid
1472 this misinterpretation.
1477 Hard fast rule: Preincrement (``++X``) may be no slower than postincrement
1478 (``X++``) and could very well be a lot faster than it. Use preincrementation
1481 The semantics of postincrement include making a copy of the value being
1482 incremented, returning it, and then preincrementing the "work value". For
1483 primitive types, this isn't a big deal. But for iterators, it can be a huge
1484 issue (for example, some iterators contains stack and set objects in them...
1485 copying an iterator could invoke the copy ctor's of these as well). In general,
1486 get in the habit of always using preincrement, and you won't have a problem.
1489 Namespace Indentation
1490 ^^^^^^^^^^^^^^^^^^^^^
1492 In general, we strive to reduce indentation wherever possible. This is useful
1493 because we want code to `fit into 80 columns`_ without wrapping horribly, but
1494 also because it makes it easier to understand the code. To facilitate this and
1495 avoid some insanely deep nesting on occasion, don't indent namespaces. If it
1496 helps readability, feel free to add a comment indicating what namespace is
1497 being closed by a ``}``. For example:
1502 namespace knowledge {
1504 /// This class represents things that Smith can have an intimate
1505 /// understanding of and contains the data associated with it.
1509 explicit Grokable() { ... }
1510 virtual ~Grokable() = 0;
1516 } // end namespace knowledge
1517 } // end namespace llvm
1520 Feel free to skip the closing comment when the namespace being closed is
1521 obvious for any reason. For example, the outer-most namespace in a header file
1522 is rarely a source of confusion. But namespaces both anonymous and named in
1523 source files that are being closed half way through the file probably could use
1528 Anonymous Namespaces
1529 ^^^^^^^^^^^^^^^^^^^^
1531 After talking about namespaces in general, you may be wondering about anonymous
1532 namespaces in particular. Anonymous namespaces are a great language feature
1533 that tells the C++ compiler that the contents of the namespace are only visible
1534 within the current translation unit, allowing more aggressive optimization and
1535 eliminating the possibility of symbol name collisions. Anonymous namespaces are
1536 to C++ as "static" is to C functions and global variables. While "``static``"
1537 is available in C++, anonymous namespaces are more general: they can make entire
1538 classes private to a file.
1540 The problem with anonymous namespaces is that they naturally want to encourage
1541 indentation of their body, and they reduce locality of reference: if you see a
1542 random function definition in a C++ file, it is easy to see if it is marked
1543 static, but seeing if it is in an anonymous namespace requires scanning a big
1546 Because of this, we have a simple guideline: make anonymous namespaces as small
1547 as possible, and only use them for class declarations. For example, this is
1557 bool operator<(const char *RHS) const;
1559 } // end anonymous namespace
1561 static void runHelper() {
1565 bool StringSort::operator<(const char *RHS) const {
1579 bool operator<(const char *RHS) const;
1586 bool StringSort::operator<(const char *RHS) const {
1590 } // end anonymous namespace
1592 This is bad specifically because if you're looking at "``runHelper``" in the middle
1593 of a large C++ file, that you have no immediate way to tell if it is local to
1594 the file. When it is marked static explicitly, this is immediately obvious.
1595 Also, there is no reason to enclose the definition of "``operator<``" in the
1596 namespace just because it was declared there.
1601 A lot of these comments and recommendations have been culled from other sources.
1602 Two particularly important books for our work are:
1605 <http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876>`_
1606 by Scott Meyers. Also interesting and useful are "More Effective C++" and
1607 "Effective STL" by the same author.
1609 #. `Large-Scale C++ Software Design
1610 <http://www.amazon.com/Large-Scale-Software-Design-John-Lakos/dp/0201633620/ref=sr_1_1>`_
1613 If you get some free time, and you haven't read them: do so, you might learn