11 This document attempts to describe a few coding standards that are being used in
12 the LLVM source tree. Although no coding standards should be regarded as
13 absolute requirements to be followed in all instances, coding standards are
14 particularly important for large-scale code bases that follow a library-based
17 While this document may provide guidance for some mechanical formatting issues,
18 whitespace, or other "microscopic details", these are not fixed standards.
19 Always follow the golden rule:
23 **If you are extending, enhancing, or bug fixing already implemented code,
24 use the style that is already being used so that the source is uniform and
27 Note that some code bases (e.g. ``libc++``) have really good reasons to deviate
28 from the coding standards. In the case of ``libc++``, this is because the
29 naming and other conventions are dictated by the C++ standard. If you think
30 there is a specific good reason to deviate from the standards here, please bring
31 it up on the LLVMdev mailing list.
33 There are some conventions that are not uniformly followed in the code base
34 (e.g. the naming convention). This is because they are relatively new, and a
35 lot of code was written before they were put in place. Our long term goal is
36 for the entire codebase to follow the convention, but we explicitly *do not*
37 want patches that do large-scale reformating of existing code. On the other
38 hand, it is reasonable to rename the methods of a class if you're about to
39 change it in some other way. Just do the reformating as a separate commit from
40 the functionality change.
42 The ultimate goal of these guidelines is the increase readability and
43 maintainability of our common source base. If you have suggestions for topics to
44 be included, please mail them to `Chris <mailto:sabre@nondot.org>`_.
46 Languages, Libraries, and Standards
47 ===================================
49 Most source code in LLVM and other LLVM projects using these coding standards
50 is C++ code. There are some places where C code is used either due to
51 environment restrictions, historical restrictions, or due to third-party source
52 code imported into the tree. Generally, our preference is for standards
53 conforming, modern, and portable C++ code as the implementation language of
59 LLVM, Clang, and LLD are currently written using C++11 conforming code,
60 although we restrict ourselves to features which are available in the major
61 toolchains supported as host compilers. The LLDB project is even more
62 aggressive in the set of host compilers supported and thus uses still more
63 features. Regardless of the supported features, code is expected to (when
64 reasonable) be standard, portable, and modern C++11 code. We avoid unnecessary
65 vendor-specific extensions, etc.
70 Use the C++ standard library facilities whenever they are available for
71 a particular task. LLVM and related projects emphasize and rely on the standard
72 library facilities for as much as possible. Common support libraries providing
73 functionality missing from the standard library for which there are standard
74 interfaces or active work on adding standard interfaces will often be
75 implemented in the LLVM namespace following the expected standard interface.
77 There are some exceptions such as the standard I/O streams library which are
78 avoided. Also, there is much more detailed information on these subjects in the
79 :doc:`ProgrammersManual`.
81 Supported C++11 Language and Library Features
82 ---------------------------------------------
84 While LLVM, Clang, and LLD use C++11, not all features are available in all of
85 the toolchains which we support. The set of features supported for use in LLVM
86 is the intersection of those supported in MSVC 2012, GCC 4.7, and Clang 3.1.
87 The ultimate definition of this set is what build bots with those respective
88 toolchains accept. Don't argue with the build bots. However, we have some
89 guidance below to help you know what to expect.
91 Each toolchain provides a good reference for what it accepts:
93 * Clang: http://clang.llvm.org/cxx_status.html
94 * GCC: http://gcc.gnu.org/projects/cxx0x.html
95 * MSVC: http://msdn.microsoft.com/en-us/library/hh567368.aspx
97 In most cases, the MSVC list will be the dominating factor. Here is a summary
98 of the features that are expected to work. Features not on this list are
99 unlikely to be supported by our host compilers.
101 * Rvalue references: N2118_
103 * But *not* Rvalue references for ``*this`` or member qualifiers (N2439_)
105 * Static assert: N1720_
106 * ``auto`` type deduction: N1984_, N1737_
107 * Trailing return types: N2541_
110 * But *not* lambdas with default arguments.
112 * ``decltype``: N2343_
113 * Nested closing right angle brackets: N1757_
114 * Extern templates: N1987_
115 * ``nullptr``: N2431_
116 * Strongly-typed and forward declarable enums: N2347_, N2764_
117 * Local and unnamed types as template arguments: N2657_
118 * Range-based for-loop: N2930_
120 * But ``{}`` are required around inner ``do {} while()`` loops. As a result,
121 ``{}`` are required around function-like macros inside range-based for
124 * ``override`` and ``final``: N2928_, N3206_, N3272_
125 * Atomic operations and the C++11 memory model: N2429_
127 .. _N2118: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2118.html
128 .. _N2439: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2439.htm
129 .. _N1720: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1720.html
130 .. _N1984: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1984.pdf
131 .. _N1737: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1737.pdf
132 .. _N2541: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2541.htm
133 .. _N2927: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2927.pdf
134 .. _N2343: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2343.pdf
135 .. _N1757: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1757.html
136 .. _N1987: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1987.htm
137 .. _N2431: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2431.pdf
138 .. _N2347: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2347.pdf
139 .. _N2764: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2764.pdf
140 .. _N2657: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm
141 .. _N2930: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2930.html
142 .. _N2928: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2928.htm
143 .. _N3206: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3206.htm
144 .. _N3272: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3272.htm
145 .. _N2429: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2429.htm
146 .. _MSVC-compatible RTTI: http://llvm.org/PR18951
148 The supported features in the C++11 standard libraries are less well tracked,
149 but also much greater. Most of the standard libraries implement most of C++11's
150 library. The most likely lowest common denominator is Linux support. For
151 libc++, the support is just poorly tested and undocumented but expected to be
152 largely complete. YMMV. For libstdc++, the support is documented in detail in
153 `the libstdc++ manual`_. There are some very minor missing facilities that are
154 unlikely to be common problems, and there are a few larger gaps that are worth
157 * Not all of the type traits are implemented
158 * No regular expression library.
159 * While most of the atomics library is well implemented, the fences are
160 missing. Fortunately, they are rarely needed.
161 * The locale support is incomplete.
162 * ``std::initializer_list`` (and the constructors and functions that take it as
163 an argument) are not always available, so you cannot (for example) initialize
164 a ``std::vector`` with a braced initializer list.
165 * ``std::equal()`` (and other algorithms) incorrectly assert in MSVC when given
166 ``nullptr`` as an iterator.
168 Other than these areas you should assume the standard library is available and
169 working as expected until some build bot tells you otherwise. If you're in an
170 uncertain area of one of the above points, but you cannot test on a Linux
171 system, your best approach is to minimize your use of these features, and watch
172 the Linux build bots to find out if your usage triggered a bug. For example, if
173 you hit a type trait which doesn't work we can then add support to LLVM's
174 traits header to emulate it.
176 .. _the libstdc++ manual:
177 http://gcc.gnu.org/onlinedocs/gcc-4.7.3/libstdc++/manual/manual/status.html#status.iso.2011
179 Mechanical Source Issues
180 ========================
182 Source Code Formatting
183 ----------------------
188 Comments are one critical part of readability and maintainability. Everyone
189 knows they should comment their code, and so should you. When writing comments,
190 write them as English prose, which means they should use proper capitalization,
191 punctuation, etc. Aim to describe what the code is trying to do and why, not
192 *how* it does it at a micro level. Here are a few critical things to document:
194 .. _header file comment:
199 Every source file should have a header on it that describes the basic purpose of
200 the file. If a file does not have a header, it should not be checked into the
201 tree. The standard header looks like this:
205 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
207 // The LLVM Compiler Infrastructure
209 // This file is distributed under the University of Illinois Open Source
210 // License. See LICENSE.TXT for details.
212 //===----------------------------------------------------------------------===//
215 /// \brief This file contains the declaration of the Instruction class, which is
216 /// the base class for all of the VM instructions.
218 //===----------------------------------------------------------------------===//
220 A few things to note about this particular format: The "``-*- C++ -*-``" string
221 on the first line is there to tell Emacs that the source file is a C++ file, not
222 a C file (Emacs assumes ``.h`` files are C files by default).
226 This tag is not necessary in ``.cpp`` files. The name of the file is also
227 on the first line, along with a very short description of the purpose of the
228 file. This is important when printing out code and flipping though lots of
231 The next section in the file is a concise note that defines the license that the
232 file is released under. This makes it perfectly clear what terms the source
233 code can be distributed under and should not be modified in any way.
235 The main body is a ``doxygen`` comment describing the purpose of the file. It
236 should have a ``\brief`` command that describes the file in one or two
237 sentences. Any additional information should be separated by a blank line. If
238 an algorithm is being implemented or something tricky is going on, a reference
239 to the paper where it is published should be included, as well as any notes or
240 *gotchas* in the code to watch out for.
245 Classes are one fundamental part of a good object oriented design. As such, a
246 class definition should have a comment block that explains what the class is
247 used for and how it works. Every non-trivial class is expected to have a
248 ``doxygen`` comment block.
253 Methods defined in a class (as well as any global functions) should also be
254 documented properly. A quick note about what it does and a description of the
255 borderline behaviour is all that is necessary here (unless something
256 particularly tricky or insidious is going on). The hope is that people can
257 figure out how to use your interfaces without reading the code itself.
259 Good things to talk about here are what happens when something unexpected
260 happens: does the method return null? Abort? Format your hard disk?
265 In general, prefer C++ style (``//``) comments. They take less space, require
266 less typing, don't have nesting problems, etc. There are a few cases when it is
267 useful to use C style (``/* */``) comments however:
269 #. When writing C code: Obviously if you are writing C code, use C style
272 #. When writing a header file that may be ``#include``\d by a C source file.
274 #. When writing a source file that is used by a tool that only accepts C style
277 To comment out a large block of code, use ``#if 0`` and ``#endif``. These nest
278 properly and are better behaved in general than C style comments.
280 Doxygen Use in Documentation Comments
281 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
283 Use the ``\file`` command to turn the standard file header into a file-level
286 Include descriptive ``\brief`` paragraphs for all public interfaces (public
287 classes, member and non-member functions). Explain API use and purpose in
288 ``\brief`` paragraphs, don't just restate the information that can be inferred
289 from the API name. Put detailed discussion into separate paragraphs.
291 To refer to parameter names inside a paragraph, use the ``\p name`` command.
292 Don't use the ``\arg name`` command since it starts a new paragraph that
293 contains documentation for the parameter.
295 Wrap non-inline code examples in ``\code ... \endcode``.
297 To document a function parameter, start a new paragraph with the
298 ``\param name`` command. If the parameter is used as an out or an in/out
299 parameter, use the ``\param [out] name`` or ``\param [in,out] name`` command,
302 To describe function return value, start a new paragraph with the ``\returns``
305 A minimal documentation comment:
309 /// \brief Does foo and bar.
310 void fooBar(bool Baz);
312 A documentation comment that uses all Doxygen features in a preferred way:
316 /// \brief Does foo and bar.
318 /// Does not do foo the usual way if \p Baz is true.
322 /// fooBar(false, "quux", Res);
325 /// \param Quux kind of foo to do.
326 /// \param [out] Result filled with bar sequence on foo success.
328 /// \returns true on success.
329 bool fooBar(bool Baz, StringRef Quux, std::vector<int> &Result);
331 Don't duplicate the documentation comment in the header file and in the
332 implementation file. Put the documentation comments for public APIs into the
333 header file. Documentation comments for private APIs can go to the
334 implementation file. In any case, implementation files can include additional
335 comments (not necessarily in Doxygen markup) to explain implementation details
338 Don't duplicate function or class name at the beginning of the comment.
339 For humans it is obvious which function or class is being documented;
340 automatic documentation processing tools are smart enough to bind the comment
341 to the correct declaration.
349 /// Something - An abstraction for some complicated thing.
352 /// fooBar - Does foo and bar.
358 /// fooBar - Does foo and bar.
359 void Something::fooBar() { ... }
367 /// \brief An abstraction for some complicated thing.
370 /// \brief Does foo and bar.
376 // Builds a B-tree in order to do foo. See paper by...
377 void Something::fooBar() { ... }
379 It is not required to use additional Doxygen features, but sometimes it might
380 be a good idea to do so.
384 * adding comments to any narrow namespace containing a collection of
385 related functions or types;
387 * using top-level groups to organize a collection of related functions at
388 namespace scope where the grouping is smaller than the namespace;
390 * using member groups and additional comments attached to member
391 groups to organize within a class.
398 /// \name Functions that do Foo.
409 Immediately after the `header file comment`_ (and include guards if working on a
410 header file), the `minimal list of #includes`_ required by the file should be
411 listed. We prefer these ``#include``\s to be listed in this order:
413 .. _Main Module Header:
414 .. _Local/Private Headers:
416 #. Main Module Header
417 #. Local/Private Headers
419 #. System ``#include``\s
421 and each category should be sorted lexicographically by the full path.
423 The `Main Module Header`_ file applies to ``.cpp`` files which implement an
424 interface defined by a ``.h`` file. This ``#include`` should always be included
425 **first** regardless of where it lives on the file system. By including a
426 header file first in the ``.cpp`` files that implement the interfaces, we ensure
427 that the header does not have any hidden dependencies which are not explicitly
428 ``#include``\d in the header, but should be. It is also a form of documentation
429 in the ``.cpp`` file to indicate where the interfaces it implements are defined.
431 .. _fit into 80 columns:
436 Write your code to fit within 80 columns of text. This helps those of us who
437 like to print out code and look at your code in an ``xterm`` without resizing
440 The longer answer is that there must be some limit to the width of the code in
441 order to reasonably allow developers to have multiple files side-by-side in
442 windows on a modest display. If you are going to pick a width limit, it is
443 somewhat arbitrary but you might as well pick something standard. Going with 90
444 columns (for example) instead of 80 columns wouldn't add any significant value
445 and would be detrimental to printing out code. Also many other projects have
446 standardized on 80 columns, so some people have already configured their editors
447 for it (vs something else, like 90 columns).
449 This is one of many contentious issues in coding standards, but it is not up for
452 Use Spaces Instead of Tabs
453 ^^^^^^^^^^^^^^^^^^^^^^^^^^
455 In all cases, prefer spaces to tabs in source files. People have different
456 preferred indentation levels, and different styles of indentation that they
457 like; this is fine. What isn't fine is that different editors/viewers expand
458 tabs out to different tab stops. This can cause your code to look completely
459 unreadable, and it is not worth dealing with.
461 As always, follow the `Golden Rule`_ above: follow the style of
462 existing code if you are modifying and extending it. If you like four spaces of
463 indentation, **DO NOT** do that in the middle of a chunk of code with two spaces
464 of indentation. Also, do not reindent a whole source file: it makes for
465 incredible diffs that are absolutely worthless.
467 Indent Code Consistently
468 ^^^^^^^^^^^^^^^^^^^^^^^^
470 Okay, in your first year of programming you were told that indentation is
471 important. If you didn't believe and internalize this then, now is the time.
472 Just do it. With the introduction of C++11, there are some new formatting
473 challenges that merit some suggestions to help have consistent, maintainable,
474 and tool-friendly formatting and indentation.
476 Format Lambdas Like Blocks Of Code
477 """"""""""""""""""""""""""""""""""
479 When formatting a multi-line lambda, format it like a block of code, that's
480 what it is. If there is only one multi-line lambda in a statement, and there
481 are no expressions lexically after it in the statement, drop the indent to the
482 standard two space indent for a block of code, as if it were an if-block opened
483 by the preceding part of the statement:
487 std::sort(foo.begin(), foo.end(), [&](Foo a, Foo b) -> bool {
492 return a.bam < b.bam;
495 To take best advantage of this formatting, if you are designing an API which
496 accepts a continuation or single callable argument (be it a functor, or
497 a ``std::function``), it should be the last argument if at all possible.
499 If there are multiple multi-line lambdas in a statement, or there is anything
500 interesting after the lambda in the statement, indent the block two spaces from
501 the indent of the ``[]``:
505 dyn_switch(V->stripPointerCasts(),
509 [] (SelectInst *SI) {
510 // process selects...
515 [] (AllocaInst *AI) {
516 // process allocas...
519 Braced Initializer Lists
520 """"""""""""""""""""""""
522 With C++11, there are significantly more uses of braced lists to perform
523 initialization. These allow you to easily construct aggregate temporaries in
524 expressions among other niceness. They now have a natural way of ending up
525 nested within each other and within function calls in order to build up
526 aggregates (such as option structs) from local variables. To make matters
527 worse, we also have many more uses of braces in an expression context that are
528 *not* performing initialization.
530 The historically common formatting of braced initialization of aggregate
531 variables does not mix cleanly with deep nesting, general expression contexts,
532 function arguments, and lambdas. We suggest new code use a simple rule for
533 formatting braced initialization lists: act as-if the braces were parentheses
534 in a function call. The formatting rules exactly match those already well
535 understood for formatting nested function calls. Examples:
539 foo({a, b, c}, {1, 2, 3});
541 llvm::Constant *Mask[] = {
542 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 0),
543 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 1),
544 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 2)};
546 This formatting scheme also makes it particularly easy to get predictable,
547 consistent, and automatic formatting with tools like `Clang Format`_.
549 .. _Clang Format: http://clang.llvm.org/docs/ClangFormat.html
551 Language and Compiler Issues
552 ----------------------------
554 Treat Compiler Warnings Like Errors
555 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
557 If your code has compiler warnings in it, something is wrong --- you aren't
558 casting values correctly, you have "questionable" constructs in your code, or
559 you are doing something legitimately wrong. Compiler warnings can cover up
560 legitimate errors in output and make dealing with a translation unit difficult.
562 It is not possible to prevent all warnings from all compilers, nor is it
563 desirable. Instead, pick a standard compiler (like ``gcc``) that provides a
564 good thorough set of warnings, and stick to it. At least in the case of
565 ``gcc``, it is possible to work around any spurious errors by changing the
566 syntax of the code slightly. For example, a warning that annoys me occurs when
567 I write code like this:
571 if (V = getValue()) {
575 ``gcc`` will warn me that I probably want to use the ``==`` operator, and that I
576 probably mistyped it. In most cases, I haven't, and I really don't want the
577 spurious errors. To fix this particular problem, I rewrite the code like
582 if ((V = getValue())) {
586 which shuts ``gcc`` up. Any ``gcc`` warning that annoys you can be fixed by
587 massaging the code appropriately.
592 In almost all cases, it is possible and within reason to write completely
593 portable code. If there are cases where it isn't possible to write portable
594 code, isolate it behind a well defined (and well documented) interface.
596 In practice, this means that you shouldn't assume much about the host compiler
597 (and Visual Studio tends to be the lowest common denominator). If advanced
598 features are used, they should only be an implementation detail of a library
599 which has a simple exposed API, and preferably be buried in ``libSystem``.
601 Do not use RTTI or Exceptions
602 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
604 In an effort to reduce code and executable size, LLVM does not use RTTI
605 (e.g. ``dynamic_cast<>;``) or exceptions. These two language features violate
606 the general C++ principle of *"you only pay for what you use"*, causing
607 executable bloat even if exceptions are never used in the code base, or if RTTI
608 is never used for a class. Because of this, we turn them off globally in the
611 That said, LLVM does make extensive use of a hand-rolled form of RTTI that use
612 templates like :ref:`isa\<>, cast\<>, and dyn_cast\<> <isa>`.
613 This form of RTTI is opt-in and can be
614 :doc:`added to any class <HowToSetUpLLVMStyleRTTI>`. It is also
615 substantially more efficient than ``dynamic_cast<>``.
617 .. _static constructor:
619 Do not use Static Constructors
620 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
622 Static constructors and destructors (e.g. global variables whose types have a
623 constructor or destructor) should not be added to the code base, and should be
624 removed wherever possible. Besides `well known problems
625 <http://yosefk.com/c++fqa/ctors.html#fqa-10.12>`_ where the order of
626 initialization is undefined between globals in different source files, the
627 entire concept of static constructors is at odds with the common use case of
628 LLVM as a library linked into a larger application.
630 Consider the use of LLVM as a JIT linked into another application (perhaps for
631 `OpenGL, custom languages <http://llvm.org/Users.html>`_, `shaders in movies
632 <http://llvm.org/devmtg/2010-11/Gritz-OpenShadingLang.pdf>`_, etc). Due to the
633 design of static constructors, they must be executed at startup time of the
634 entire application, regardless of whether or how LLVM is used in that larger
635 application. There are two problems with this:
637 * The time to run the static constructors impacts startup time of applications
638 --- a critical time for GUI apps, among others.
640 * The static constructors cause the app to pull many extra pages of memory off
641 the disk: both the code for the constructor in each ``.o`` file and the small
642 amount of data that gets touched. In addition, touched/dirty pages put more
643 pressure on the VM system on low-memory machines.
645 We would really like for there to be zero cost for linking in an additional LLVM
646 target or other library into an application, but static constructors violate
649 That said, LLVM unfortunately does contain static constructors. It would be a
650 `great project <http://llvm.org/PR11944>`_ for someone to purge all static
651 constructors from LLVM, and then enable the ``-Wglobal-constructors`` warning
652 flag (when building with Clang) to ensure we do not regress in the future.
654 Use of ``class`` and ``struct`` Keywords
655 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
657 In C++, the ``class`` and ``struct`` keywords can be used almost
658 interchangeably. The only difference is when they are used to declare a class:
659 ``class`` makes all members private by default while ``struct`` makes all
660 members public by default.
662 Unfortunately, not all compilers follow the rules and some will generate
663 different symbols based on whether ``class`` or ``struct`` was used to declare
664 the symbol (e.g., MSVC). This can lead to problems at link time.
666 * All declarations and definitions of a given ``class`` or ``struct`` must use
667 the same keyword. For example:
673 // Breaks mangling in MSVC.
674 struct Foo { int Data; };
676 * As a rule of thumb, ``struct`` should be kept to structures where *all*
677 members are declared public.
681 // Foo feels like a class... this is strange.
687 int getData() const { return Data; }
688 void setData(int D) { Data = D; }
691 // Bar isn't POD, but it does look like a struct.
697 Do not use Braced Initializer Lists to Call a Constructor
698 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
700 In C++11 there is a "generalized initialization syntax" which allows calling
701 constructors using braced initializer lists. Do not use these to call
702 constructors with any interesting logic or if you care that you're calling some
703 *particular* constructor. Those should look like function calls using
704 parentheses rather than like aggregate initialization. Similarly, if you need
705 to explicitly name the type and call its constructor to create a temporary,
706 don't use a braced initializer list. Instead, use a braced initializer list
707 (without any type for temporaries) when doing aggregate initialization or
708 something notionally equivalent. Examples:
714 // Construct a Foo by reading data from the disk in the whizbang format, ...
715 Foo(std::string filename);
717 // Construct a Foo by looking up the Nth element of some global data ...
723 // The Foo constructor call is very deliberate, no braces.
724 std::fill(foo.begin(), foo.end(), Foo("name"));
726 // The pair is just being constructed like an aggregate, use braces.
727 bar_map.insert({my_key, my_value});
729 If you use a braced initializer list when initializing a variable, use an equals before the open curly brace:
733 int data[] = {0, 1, 2, 3};
735 Use ``auto`` Type Deduction to Make Code More Readable
736 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
738 Some are advocating a policy of "almost always ``auto``" in C++11, however LLVM
739 uses a more moderate stance. Use ``auto`` if and only if it makes the code more
740 readable or easier to maintain. Don't "almost always" use ``auto``, but do use
741 ``auto`` with initializers like ``cast<Foo>(...)`` or other places where the
742 type is already obvious from the context. Another time when ``auto`` works well
743 for these purposes is when the type would have been abstracted away anyways,
744 often behind a container's typedef such as ``std::vector<T>::iterator``.
746 Beware unnecessary copies with ``auto``
747 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
749 The convenience of ``auto`` makes it easy to forget that its default behavior
750 is a copy. Particularly in range-based ``for`` loops, careless copies are
753 As a rule of thumb, use ``auto &`` unless you need to copy the result, and use
754 ``auto *`` when copying pointers.
758 // Typically there's no reason to copy.
759 for (const auto &Val : Container) { observe(Val); }
760 for (auto &Val : Container) { Val.change(); }
762 // Remove the reference if you really want a new copy.
763 for (auto Val : Container) { Val.change(); saveSomewhere(Val); }
765 // Copy pointers, but make it clear that they're pointers.
766 for (const auto *Ptr : Container) { observe(*Ptr); }
767 for (auto *Ptr : Container) { Ptr->change(); }
772 The High-Level Issues
773 ---------------------
775 A Public Header File **is** a Module
776 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
778 C++ doesn't do too well in the modularity department. There is no real
779 encapsulation or data hiding (unless you use expensive protocol classes), but it
780 is what we have to work with. When you write a public header file (in the LLVM
781 source tree, they live in the top level "``include``" directory), you are
782 defining a module of functionality.
784 Ideally, modules should be completely independent of each other, and their
785 header files should only ``#include`` the absolute minimum number of headers
786 possible. A module is not just a class, a function, or a namespace: it's a
787 collection of these that defines an interface. This interface may be several
788 functions, classes, or data structures, but the important issue is how they work
791 In general, a module should be implemented by one or more ``.cpp`` files. Each
792 of these ``.cpp`` files should include the header that defines their interface
793 first. This ensures that all of the dependences of the module header have been
794 properly added to the module header itself, and are not implicit. System
795 headers should be included after user headers for a translation unit.
797 .. _minimal list of #includes:
799 ``#include`` as Little as Possible
800 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
802 ``#include`` hurts compile time performance. Don't do it unless you have to,
803 especially in header files.
805 But wait! Sometimes you need to have the definition of a class to use it, or to
806 inherit from it. In these cases go ahead and ``#include`` that header file. Be
807 aware however that there are many cases where you don't need to have the full
808 definition of a class. If you are using a pointer or reference to a class, you
809 don't need the header file. If you are simply returning a class instance from a
810 prototyped function or method, you don't need it. In fact, for most cases, you
811 simply don't need the definition of a class. And not ``#include``\ing speeds up
814 It is easy to try to go too overboard on this recommendation, however. You
815 **must** include all of the header files that you are using --- you can include
816 them either directly or indirectly through another header file. To make sure
817 that you don't accidentally forget to include a header file in your module
818 header, make sure to include your module header **first** in the implementation
819 file (as mentioned above). This way there won't be any hidden dependencies that
820 you'll find out about later.
822 Keep "Internal" Headers Private
823 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
825 Many modules have a complex implementation that causes them to use more than one
826 implementation (``.cpp``) file. It is often tempting to put the internal
827 communication interface (helper classes, extra functions, etc) in the public
828 module header file. Don't do this!
830 If you really need to do something like this, put a private header file in the
831 same directory as the source files, and include it locally. This ensures that
832 your private interface remains private and undisturbed by outsiders.
836 It's okay to put extra implementation methods in a public class itself. Just
837 make them private (or protected) and all is well.
841 Use Early Exits and ``continue`` to Simplify Code
842 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
844 When reading code, keep in mind how much state and how many previous decisions
845 have to be remembered by the reader to understand a block of code. Aim to
846 reduce indentation where possible when it doesn't make it more difficult to
847 understand the code. One great way to do this is by making use of early exits
848 and the ``continue`` keyword in long loops. As an example of using an early
849 exit from a function, consider this "bad" code:
853 Value *doSomething(Instruction *I) {
854 if (!isa<TerminatorInst>(I) &&
855 I->hasOneUse() && doOtherThing(I)) {
856 ... some long code ....
862 This code has several problems if the body of the ``'if'`` is large. When
863 you're looking at the top of the function, it isn't immediately clear that this
864 *only* does interesting things with non-terminator instructions, and only
865 applies to things with the other predicates. Second, it is relatively difficult
866 to describe (in comments) why these predicates are important because the ``if``
867 statement makes it difficult to lay out the comments. Third, when you're deep
868 within the body of the code, it is indented an extra level. Finally, when
869 reading the top of the function, it isn't clear what the result is if the
870 predicate isn't true; you have to read to the end of the function to know that
873 It is much preferred to format the code like this:
877 Value *doSomething(Instruction *I) {
878 // Terminators never need 'something' done to them because ...
879 if (isa<TerminatorInst>(I))
882 // We conservatively avoid transforming instructions with multiple uses
883 // because goats like cheese.
887 // This is really just here for example.
888 if (!doOtherThing(I))
891 ... some long code ....
894 This fixes these problems. A similar problem frequently happens in ``for``
895 loops. A silly example is something like this:
899 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
900 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
901 Value *LHS = BO->getOperand(0);
902 Value *RHS = BO->getOperand(1);
909 When you have very, very small loops, this sort of structure is fine. But if it
910 exceeds more than 10-15 lines, it becomes difficult for people to read and
911 understand at a glance. The problem with this sort of code is that it gets very
912 nested very quickly. Meaning that the reader of the code has to keep a lot of
913 context in their brain to remember what is going immediately on in the loop,
914 because they don't know if/when the ``if`` conditions will have ``else``\s etc.
915 It is strongly preferred to structure the loop like this:
919 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
920 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
923 Value *LHS = BO->getOperand(0);
924 Value *RHS = BO->getOperand(1);
925 if (LHS == RHS) continue;
930 This has all the benefits of using early exits for functions: it reduces nesting
931 of the loop, it makes it easier to describe why the conditions are true, and it
932 makes it obvious to the reader that there is no ``else`` coming up that they
933 have to push context into their brain for. If a loop is large, this can be a
934 big understandability win.
936 Don't use ``else`` after a ``return``
937 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
939 For similar reasons above (reduction of indentation and easier reading), please
940 do not use ``'else'`` or ``'else if'`` after something that interrupts control
941 flow --- like ``return``, ``break``, ``continue``, ``goto``, etc. For
942 example, this is *bad*:
948 Type = Context.getsigjmp_bufType();
950 Error = ASTContext::GE_Missing_sigjmp_buf;
956 Type = Context.getjmp_bufType();
958 Error = ASTContext::GE_Missing_jmp_buf;
966 It is better to write it like this:
972 Type = Context.getsigjmp_bufType();
974 Error = ASTContext::GE_Missing_sigjmp_buf;
978 Type = Context.getjmp_bufType();
980 Error = ASTContext::GE_Missing_jmp_buf;
986 Or better yet (in this case) as:
992 Type = Context.getsigjmp_bufType();
994 Type = Context.getjmp_bufType();
997 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
998 ASTContext::GE_Missing_jmp_buf;
1003 The idea is to reduce indentation and the amount of code you have to keep track
1004 of when reading the code.
1006 Turn Predicate Loops into Predicate Functions
1007 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1009 It is very common to write small loops that just compute a boolean value. There
1010 are a number of ways that people commonly write these, but an example of this
1015 bool FoundFoo = false;
1016 for (unsigned I = 0, E = BarList.size(); I != E; ++I)
1017 if (BarList[I]->isFoo()) {
1026 This sort of code is awkward to write, and is almost always a bad sign. Instead
1027 of this sort of loop, we strongly prefer to use a predicate function (which may
1028 be `static`_) that uses `early exits`_ to compute the predicate. We prefer the
1029 code to be structured like this:
1033 /// \returns true if the specified list has an element that is a foo.
1034 static bool containsFoo(const std::vector<Bar*> &List) {
1035 for (unsigned I = 0, E = List.size(); I != E; ++I)
1036 if (List[I]->isFoo())
1042 if (containsFoo(BarList)) {
1046 There are many reasons for doing this: it reduces indentation and factors out
1047 code which can often be shared by other code that checks for the same predicate.
1048 More importantly, it *forces you to pick a name* for the function, and forces
1049 you to write a comment for it. In this silly example, this doesn't add much
1050 value. However, if the condition is complex, this can make it a lot easier for
1051 the reader to understand the code that queries for this predicate. Instead of
1052 being faced with the in-line details of how we check to see if the BarList
1053 contains a foo, we can trust the function name and continue reading with better
1056 The Low-Level Issues
1057 --------------------
1059 Name Types, Functions, Variables, and Enumerators Properly
1060 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1062 Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
1063 enough how important it is to use *descriptive* names. Pick names that match
1064 the semantics and role of the underlying entities, within reason. Avoid
1065 abbreviations unless they are well known. After picking a good name, make sure
1066 to use consistent capitalization for the name, as inconsistency requires clients
1067 to either memorize the APIs or to look it up to find the exact spelling.
1069 In general, names should be in camel case (e.g. ``TextFileReader`` and
1070 ``isLValue()``). Different kinds of declarations have different rules:
1072 * **Type names** (including classes, structs, enums, typedefs, etc) should be
1073 nouns and start with an upper-case letter (e.g. ``TextFileReader``).
1075 * **Variable names** should be nouns (as they represent state). The name should
1076 be camel case, and start with an upper case letter (e.g. ``Leader`` or
1079 * **Function names** should be verb phrases (as they represent actions), and
1080 command-like function should be imperative. The name should be camel case,
1081 and start with a lower case letter (e.g. ``openFile()`` or ``isFoo()``).
1083 * **Enum declarations** (e.g. ``enum Foo {...}``) are types, so they should
1084 follow the naming conventions for types. A common use for enums is as a
1085 discriminator for a union, or an indicator of a subclass. When an enum is
1086 used for something like this, it should have a ``Kind`` suffix
1087 (e.g. ``ValueKind``).
1089 * **Enumerators** (e.g. ``enum { Foo, Bar }``) and **public member variables**
1090 should start with an upper-case letter, just like types. Unless the
1091 enumerators are defined in their own small namespace or inside a class,
1092 enumerators should have a prefix corresponding to the enum declaration name.
1093 For example, ``enum ValueKind { ... };`` may contain enumerators like
1094 ``VK_Argument``, ``VK_BasicBlock``, etc. Enumerators that are just
1095 convenience constants are exempt from the requirement for a prefix. For
1105 As an exception, classes that mimic STL classes can have member names in STL's
1106 style of lower-case words separated by underscores (e.g. ``begin()``,
1107 ``push_back()``, and ``empty()``). Classes that provide multiple
1108 iterators should add a singular prefix to ``begin()`` and ``end()``
1109 (e.g. ``global_begin()`` and ``use_begin()``).
1111 Here are some examples of good and bad names:
1115 class VehicleMaker {
1117 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
1118 Factory<Tire> Factory; // Better.
1119 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
1120 // kind of factories.
1123 Vehicle MakeVehicle(VehicleType Type) {
1124 VehicleMaker M; // Might be OK if having a short life-span.
1125 Tire Tmp1 = M.makeTire(); // Bad -- 'Tmp1' provides no information.
1126 Light Headlight = M.makeLight("head"); // Good -- descriptive.
1133 Use the "``assert``" macro to its fullest. Check all of your preconditions and
1134 assumptions, you never know when a bug (not necessarily even yours) might be
1135 caught early by an assertion, which reduces debugging time dramatically. The
1136 "``<cassert>``" header file is probably already included by the header files you
1137 are using, so it doesn't cost anything to use it.
1139 To further assist with debugging, make sure to put some kind of error message in
1140 the assertion statement, which is printed if the assertion is tripped. This
1141 helps the poor debugger make sense of why an assertion is being made and
1142 enforced, and hopefully what to do about it. Here is one complete example:
1146 inline Value *getOperand(unsigned I) {
1147 assert(I < Operands.size() && "getOperand() out of range!");
1151 Here are more examples:
1155 assert(Ty->isPointerType() && "Can't allocate a non-pointer type!");
1157 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
1159 assert(idx < getNumSuccessors() && "Successor # out of range!");
1161 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
1163 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
1167 In the past, asserts were used to indicate a piece of code that should not be
1168 reached. These were typically of the form:
1172 assert(0 && "Invalid radix for integer literal");
1174 This has a few issues, the main one being that some compilers might not
1175 understand the assertion, or warn about a missing return in builds where
1176 assertions are compiled out.
1178 Today, we have something much better: ``llvm_unreachable``:
1182 llvm_unreachable("Invalid radix for integer literal");
1184 When assertions are enabled, this will print the message if it's ever reached
1185 and then exit the program. When assertions are disabled (i.e. in release
1186 builds), ``llvm_unreachable`` becomes a hint to compilers to skip generating
1187 code for this branch. If the compiler does not support this, it will fall back
1188 to the "abort" implementation.
1190 Another issue is that values used only by assertions will produce an "unused
1191 value" warning when assertions are disabled. For example, this code will warn:
1195 unsigned Size = V.size();
1196 assert(Size > 42 && "Vector smaller than it should be");
1198 bool NewToSet = Myset.insert(Value);
1199 assert(NewToSet && "The value shouldn't be in the set yet");
1201 These are two interesting different cases. In the first case, the call to
1202 ``V.size()`` is only useful for the assert, and we don't want it executed when
1203 assertions are disabled. Code like this should move the call into the assert
1204 itself. In the second case, the side effects of the call must happen whether
1205 the assert is enabled or not. In this case, the value should be cast to void to
1206 disable the warning. To be specific, it is preferred to write the code like
1211 assert(V.size() > 42 && "Vector smaller than it should be");
1213 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1214 assert(NewToSet && "The value shouldn't be in the set yet");
1216 Do Not Use ``using namespace std``
1217 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1219 In LLVM, we prefer to explicitly prefix all identifiers from the standard
1220 namespace with an "``std::``" prefix, rather than rely on "``using namespace
1223 In header files, adding a ``'using namespace XXX'`` directive pollutes the
1224 namespace of any source file that ``#include``\s the header. This is clearly a
1227 In implementation files (e.g. ``.cpp`` files), the rule is more of a stylistic
1228 rule, but is still important. Basically, using explicit namespace prefixes
1229 makes the code **clearer**, because it is immediately obvious what facilities
1230 are being used and where they are coming from. And **more portable**, because
1231 namespace clashes cannot occur between LLVM code and other namespaces. The
1232 portability rule is important because different standard library implementations
1233 expose different symbols (potentially ones they shouldn't), and future revisions
1234 to the C++ standard will add more symbols to the ``std`` namespace. As such, we
1235 never use ``'using namespace std;'`` in LLVM.
1237 The exception to the general rule (i.e. it's not an exception for the ``std``
1238 namespace) is for implementation files. For example, all of the code in the
1239 LLVM project implements code that lives in the 'llvm' namespace. As such, it is
1240 ok, and actually clearer, for the ``.cpp`` files to have a ``'using namespace
1241 llvm;'`` directive at the top, after the ``#include``\s. This reduces
1242 indentation in the body of the file for source editors that indent based on
1243 braces, and keeps the conceptual context cleaner. The general form of this rule
1244 is that any ``.cpp`` file that implements code in any namespace may use that
1245 namespace (and its parents'), but should not use any others.
1247 Provide a Virtual Method Anchor for Classes in Headers
1248 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1250 If a class is defined in a header file and has a vtable (either it has virtual
1251 methods or it derives from classes with virtual methods), it must always have at
1252 least one out-of-line virtual method in the class. Without this, the compiler
1253 will copy the vtable and RTTI into every ``.o`` file that ``#include``\s the
1254 header, bloating ``.o`` file sizes and increasing link times.
1256 Don't use default labels in fully covered switches over enumerations
1257 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1259 ``-Wswitch`` warns if a switch, without a default label, over an enumeration
1260 does not cover every enumeration value. If you write a default label on a fully
1261 covered switch over an enumeration then the ``-Wswitch`` warning won't fire
1262 when new elements are added to that enumeration. To help avoid adding these
1263 kinds of defaults, Clang has the warning ``-Wcovered-switch-default`` which is
1264 off by default but turned on when building LLVM with a version of Clang that
1265 supports the warning.
1267 A knock-on effect of this stylistic requirement is that when building LLVM with
1268 GCC you may get warnings related to "control may reach end of non-void function"
1269 if you return from each case of a covered switch-over-enum because GCC assumes
1270 that the enum expression may take any representable value, not just those of
1271 individual enumerators. To suppress this warning, use ``llvm_unreachable`` after
1274 Use ``LLVM_DELETED_FUNCTION`` to mark uncallable methods
1275 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1277 Prior to C++11, a common pattern to make a class uncopyable was to declare an
1278 unimplemented copy constructor and copy assignment operator and make them
1279 private. This would give a compiler error for accessing a private method or a
1280 linker error because it wasn't implemented.
1282 With C++11, we can mark methods that won't be implemented with ``= delete``.
1283 This will trigger a much better error message and tell the compiler that the
1284 method will never be implemented. This enables other checks like
1285 ``-Wunused-private-field`` to run correctly on classes that contain these
1288 For compatibility with MSVC, ``LLVM_DELETED_FUNCTION`` should be used which
1289 will expand to ``= delete`` on compilers that support it. These methods should
1290 still be declared private. Example of the uncopyable pattern:
1296 DontCopy(const DontCopy&) LLVM_DELETED_FUNCTION;
1297 DontCopy &operator =(const DontCopy&) LLVM_DELETED_FUNCTION;
1302 Don't evaluate ``end()`` every time through a loop
1303 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1305 Because C++ doesn't have a standard "``foreach``" loop (though it can be
1306 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1307 loops that manually iterate from begin to end on a variety of containers or
1308 through other data structures. One common mistake is to write a loop in this
1313 BasicBlock *BB = ...
1314 for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I)
1317 The problem with this construct is that it evaluates "``BB->end()``" every time
1318 through the loop. Instead of writing the loop like this, we strongly prefer
1319 loops to be written so that they evaluate it once before the loop starts. A
1320 convenient way to do this is like so:
1324 BasicBlock *BB = ...
1325 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
1328 The observant may quickly point out that these two loops may have different
1329 semantics: if the container (a basic block in this case) is being mutated, then
1330 "``BB->end()``" may change its value every time through the loop and the second
1331 loop may not in fact be correct. If you actually do depend on this behavior,
1332 please write the loop in the first form and add a comment indicating that you
1333 did it intentionally.
1335 Why do we prefer the second form (when correct)? Writing the loop in the first
1336 form has two problems. First it may be less efficient than evaluating it at the
1337 start of the loop. In this case, the cost is probably minor --- a few extra
1338 loads every time through the loop. However, if the base expression is more
1339 complex, then the cost can rise quickly. I've seen loops where the end
1340 expression was actually something like: "``SomeMap[X]->end()``" and map lookups
1341 really aren't cheap. By writing it in the second form consistently, you
1342 eliminate the issue entirely and don't even have to think about it.
1344 The second (even bigger) issue is that writing the loop in the first form hints
1345 to the reader that the loop is mutating the container (a fact that a comment
1346 would handily confirm!). If you write the loop in the second form, it is
1347 immediately obvious without even looking at the body of the loop that the
1348 container isn't being modified, which makes it easier to read the code and
1349 understand what it does.
1351 While the second form of the loop is a few extra keystrokes, we do strongly
1354 ``#include <iostream>`` is Forbidden
1355 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1357 The use of ``#include <iostream>`` in library files is hereby **forbidden**,
1358 because many common implementations transparently inject a `static constructor`_
1359 into every translation unit that includes it.
1361 Note that using the other stream headers (``<sstream>`` for example) is not
1362 problematic in this regard --- just ``<iostream>``. However, ``raw_ostream``
1363 provides various APIs that are better performing for almost every use than
1364 ``std::ostream`` style APIs.
1368 New code should always use `raw_ostream`_ for writing, or the
1369 ``llvm::MemoryBuffer`` API for reading files.
1376 LLVM includes a lightweight, simple, and efficient stream implementation in
1377 ``llvm/Support/raw_ostream.h``, which provides all of the common features of
1378 ``std::ostream``. All new code should use ``raw_ostream`` instead of
1381 Unlike ``std::ostream``, ``raw_ostream`` is not a template and can be forward
1382 declared as ``class raw_ostream``. Public headers should generally not include
1383 the ``raw_ostream`` header, but use forward declarations and constant references
1384 to ``raw_ostream`` instances.
1389 The ``std::endl`` modifier, when used with ``iostreams`` outputs a newline to
1390 the output stream specified. In addition to doing this, however, it also
1391 flushes the output stream. In other words, these are equivalent:
1395 std::cout << std::endl;
1396 std::cout << '\n' << std::flush;
1398 Most of the time, you probably have no reason to flush the output stream, so
1399 it's better to use a literal ``'\n'``.
1401 Don't use ``inline`` when defining a function in a class definition
1402 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1404 A member function defined in a class definition is implicitly inline, so don't
1405 put the ``inline`` keyword in this case.
1432 This section describes preferred low-level formatting guidelines along with
1433 reasoning on why we prefer them.
1435 Spaces Before Parentheses
1436 ^^^^^^^^^^^^^^^^^^^^^^^^^
1438 We prefer to put a space before an open parenthesis only in control flow
1439 statements, but not in normal function call expressions and function-like
1440 macros. For example, this is good:
1445 for (I = 0; I != 100; ++I) ...
1446 while (LLVMRocks) ...
1449 assert(3 != 4 && "laws of math are failing me");
1451 A = foo(42, 92) + bar(X);
1458 for(I = 0; I != 100; ++I) ...
1459 while(LLVMRocks) ...
1462 assert (3 != 4 && "laws of math are failing me");
1464 A = foo (42, 92) + bar (X);
1466 The reason for doing this is not completely arbitrary. This style makes control
1467 flow operators stand out more, and makes expressions flow better. The function
1468 call operator binds very tightly as a postfix operator. Putting a space after a
1469 function name (as in the last example) makes it appear that the code might bind
1470 the arguments of the left-hand-side of a binary operator with the argument list
1471 of a function and the name of the right side. More specifically, it is easy to
1472 misread the "``A``" example as:
1476 A = foo ((42, 92) + bar) (X);
1478 when skimming through the code. By avoiding a space in a function, we avoid
1479 this misinterpretation.
1484 Hard fast rule: Preincrement (``++X``) may be no slower than postincrement
1485 (``X++``) and could very well be a lot faster than it. Use preincrementation
1488 The semantics of postincrement include making a copy of the value being
1489 incremented, returning it, and then preincrementing the "work value". For
1490 primitive types, this isn't a big deal. But for iterators, it can be a huge
1491 issue (for example, some iterators contains stack and set objects in them...
1492 copying an iterator could invoke the copy ctor's of these as well). In general,
1493 get in the habit of always using preincrement, and you won't have a problem.
1496 Namespace Indentation
1497 ^^^^^^^^^^^^^^^^^^^^^
1499 In general, we strive to reduce indentation wherever possible. This is useful
1500 because we want code to `fit into 80 columns`_ without wrapping horribly, but
1501 also because it makes it easier to understand the code. To facilitate this and
1502 avoid some insanely deep nesting on occasion, don't indent namespaces. If it
1503 helps readability, feel free to add a comment indicating what namespace is
1504 being closed by a ``}``. For example:
1509 namespace knowledge {
1511 /// This class represents things that Smith can have an intimate
1512 /// understanding of and contains the data associated with it.
1516 explicit Grokable() { ... }
1517 virtual ~Grokable() = 0;
1523 } // end namespace knowledge
1524 } // end namespace llvm
1527 Feel free to skip the closing comment when the namespace being closed is
1528 obvious for any reason. For example, the outer-most namespace in a header file
1529 is rarely a source of confusion. But namespaces both anonymous and named in
1530 source files that are being closed half way through the file probably could use
1535 Anonymous Namespaces
1536 ^^^^^^^^^^^^^^^^^^^^
1538 After talking about namespaces in general, you may be wondering about anonymous
1539 namespaces in particular. Anonymous namespaces are a great language feature
1540 that tells the C++ compiler that the contents of the namespace are only visible
1541 within the current translation unit, allowing more aggressive optimization and
1542 eliminating the possibility of symbol name collisions. Anonymous namespaces are
1543 to C++ as "static" is to C functions and global variables. While "``static``"
1544 is available in C++, anonymous namespaces are more general: they can make entire
1545 classes private to a file.
1547 The problem with anonymous namespaces is that they naturally want to encourage
1548 indentation of their body, and they reduce locality of reference: if you see a
1549 random function definition in a C++ file, it is easy to see if it is marked
1550 static, but seeing if it is in an anonymous namespace requires scanning a big
1553 Because of this, we have a simple guideline: make anonymous namespaces as small
1554 as possible, and only use them for class declarations. For example, this is
1564 bool operator<(const char *RHS) const;
1566 } // end anonymous namespace
1568 static void runHelper() {
1572 bool StringSort::operator<(const char *RHS) const {
1586 bool operator<(const char *RHS) const;
1593 bool StringSort::operator<(const char *RHS) const {
1597 } // end anonymous namespace
1599 This is bad specifically because if you're looking at "``runHelper``" in the middle
1600 of a large C++ file, that you have no immediate way to tell if it is local to
1601 the file. When it is marked static explicitly, this is immediately obvious.
1602 Also, there is no reason to enclose the definition of "``operator<``" in the
1603 namespace just because it was declared there.
1608 A lot of these comments and recommendations have been culled from other sources.
1609 Two particularly important books for our work are:
1612 <http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876>`_
1613 by Scott Meyers. Also interesting and useful are "More Effective C++" and
1614 "Effective STL" by the same author.
1616 #. `Large-Scale C++ Software Design
1617 <http://www.amazon.com/Large-Scale-Software-Design-John-Lakos/dp/0201633620/ref=sr_1_1>`_
1620 If you get some free time, and you haven't read them: do so, you might learn