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.
166 Other than these areas you should assume the standard library is available and
167 working as expected until some build bot tells you otherwise. If you're in an
168 uncertain area of one of the above points, but you cannot test on a Linux
169 system, your best approach is to minimize your use of these features, and watch
170 the Linux build bots to find out if your usage triggered a bug. For example, if
171 you hit a type trait which doesn't work we can then add support to LLVM's
172 traits header to emulate it.
174 .. _the libstdc++ manual:
175 http://gcc.gnu.org/onlinedocs/gcc-4.7.3/libstdc++/manual/manual/status.html#status.iso.2011
177 Mechanical Source Issues
178 ========================
180 Source Code Formatting
181 ----------------------
186 Comments are one critical part of readability and maintainability. Everyone
187 knows they should comment their code, and so should you. When writing comments,
188 write them as English prose, which means they should use proper capitalization,
189 punctuation, etc. Aim to describe what the code is trying to do and why, not
190 *how* it does it at a micro level. Here are a few critical things to document:
192 .. _header file comment:
197 Every source file should have a header on it that describes the basic purpose of
198 the file. If a file does not have a header, it should not be checked into the
199 tree. The standard header looks like this:
203 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
205 // The LLVM Compiler Infrastructure
207 // This file is distributed under the University of Illinois Open Source
208 // License. See LICENSE.TXT for details.
210 //===----------------------------------------------------------------------===//
213 /// \brief This file contains the declaration of the Instruction class, which is
214 /// the base class for all of the VM instructions.
216 //===----------------------------------------------------------------------===//
218 A few things to note about this particular format: The "``-*- C++ -*-``" string
219 on the first line is there to tell Emacs that the source file is a C++ file, not
220 a C file (Emacs assumes ``.h`` files are C files by default).
224 This tag is not necessary in ``.cpp`` files. The name of the file is also
225 on the first line, along with a very short description of the purpose of the
226 file. This is important when printing out code and flipping though lots of
229 The next section in the file is a concise note that defines the license that the
230 file is released under. This makes it perfectly clear what terms the source
231 code can be distributed under and should not be modified in any way.
233 The main body is a ``doxygen`` comment describing the purpose of the file. It
234 should have a ``\brief`` command that describes the file in one or two
235 sentences. Any additional information should be separated by a blank line. If
236 an algorithm is being implemented or something tricky is going on, a reference
237 to the paper where it is published should be included, as well as any notes or
238 *gotchas* in the code to watch out for.
243 Classes are one fundamental part of a good object oriented design. As such, a
244 class definition should have a comment block that explains what the class is
245 used for and how it works. Every non-trivial class is expected to have a
246 ``doxygen`` comment block.
251 Methods defined in a class (as well as any global functions) should also be
252 documented properly. A quick note about what it does and a description of the
253 borderline behaviour is all that is necessary here (unless something
254 particularly tricky or insidious is going on). The hope is that people can
255 figure out how to use your interfaces without reading the code itself.
257 Good things to talk about here are what happens when something unexpected
258 happens: does the method return null? Abort? Format your hard disk?
263 In general, prefer C++ style (``//``) comments. They take less space, require
264 less typing, don't have nesting problems, etc. There are a few cases when it is
265 useful to use C style (``/* */``) comments however:
267 #. When writing C code: Obviously if you are writing C code, use C style
270 #. When writing a header file that may be ``#include``\d by a C source file.
272 #. When writing a source file that is used by a tool that only accepts C style
275 To comment out a large block of code, use ``#if 0`` and ``#endif``. These nest
276 properly and are better behaved in general than C style comments.
278 Doxygen Use in Documentation Comments
279 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
281 Use the ``\file`` command to turn the standard file header into a file-level
284 Include descriptive ``\brief`` paragraphs for all public interfaces (public
285 classes, member and non-member functions). Explain API use and purpose in
286 ``\brief`` paragraphs, don't just restate the information that can be inferred
287 from the API name. Put detailed discussion into separate paragraphs.
289 To refer to parameter names inside a paragraph, use the ``\p name`` command.
290 Don't use the ``\arg name`` command since it starts a new paragraph that
291 contains documentation for the parameter.
293 Wrap non-inline code examples in ``\code ... \endcode``.
295 To document a function parameter, start a new paragraph with the
296 ``\param name`` command. If the parameter is used as an out or an in/out
297 parameter, use the ``\param [out] name`` or ``\param [in,out] name`` command,
300 To describe function return value, start a new paragraph with the ``\returns``
303 A minimal documentation comment:
307 /// \brief Does foo and bar.
308 void fooBar(bool Baz);
310 A documentation comment that uses all Doxygen features in a preferred way:
314 /// \brief Does foo and bar.
316 /// Does not do foo the usual way if \p Baz is true.
320 /// fooBar(false, "quux", Res);
323 /// \param Quux kind of foo to do.
324 /// \param [out] Result filled with bar sequence on foo success.
326 /// \returns true on success.
327 bool fooBar(bool Baz, StringRef Quux, std::vector<int> &Result);
329 Don't duplicate the documentation comment in the header file and in the
330 implementation file. Put the documentation comments for public APIs into the
331 header file. Documentation comments for private APIs can go to the
332 implementation file. In any case, implementation files can include additional
333 comments (not necessarily in Doxygen markup) to explain implementation details
336 Don't duplicate function or class name at the beginning of the comment.
337 For humans it is obvious which function or class is being documented;
338 automatic documentation processing tools are smart enough to bind the comment
339 to the correct declaration.
347 /// Something - An abstraction for some complicated thing.
350 /// fooBar - Does foo and bar.
356 /// fooBar - Does foo and bar.
357 void Something::fooBar() { ... }
365 /// \brief An abstraction for some complicated thing.
368 /// \brief Does foo and bar.
374 // Builds a B-tree in order to do foo. See paper by...
375 void Something::fooBar() { ... }
377 It is not required to use additional Doxygen features, but sometimes it might
378 be a good idea to do so.
382 * adding comments to any narrow namespace containing a collection of
383 related functions or types;
385 * using top-level groups to organize a collection of related functions at
386 namespace scope where the grouping is smaller than the namespace;
388 * using member groups and additional comments attached to member
389 groups to organize within a class.
396 /// \name Functions that do Foo.
407 Immediately after the `header file comment`_ (and include guards if working on a
408 header file), the `minimal list of #includes`_ required by the file should be
409 listed. We prefer these ``#include``\s to be listed in this order:
411 .. _Main Module Header:
412 .. _Local/Private Headers:
414 #. Main Module Header
415 #. Local/Private Headers
417 #. System ``#include``\s
419 and each category should be sorted lexicographically by the full path.
421 The `Main Module Header`_ file applies to ``.cpp`` files which implement an
422 interface defined by a ``.h`` file. This ``#include`` should always be included
423 **first** regardless of where it lives on the file system. By including a
424 header file first in the ``.cpp`` files that implement the interfaces, we ensure
425 that the header does not have any hidden dependencies which are not explicitly
426 ``#include``\d in the header, but should be. It is also a form of documentation
427 in the ``.cpp`` file to indicate where the interfaces it implements are defined.
429 .. _fit into 80 columns:
434 Write your code to fit within 80 columns of text. This helps those of us who
435 like to print out code and look at your code in an ``xterm`` without resizing
438 The longer answer is that there must be some limit to the width of the code in
439 order to reasonably allow developers to have multiple files side-by-side in
440 windows on a modest display. If you are going to pick a width limit, it is
441 somewhat arbitrary but you might as well pick something standard. Going with 90
442 columns (for example) instead of 80 columns wouldn't add any significant value
443 and would be detrimental to printing out code. Also many other projects have
444 standardized on 80 columns, so some people have already configured their editors
445 for it (vs something else, like 90 columns).
447 This is one of many contentious issues in coding standards, but it is not up for
450 Use Spaces Instead of Tabs
451 ^^^^^^^^^^^^^^^^^^^^^^^^^^
453 In all cases, prefer spaces to tabs in source files. People have different
454 preferred indentation levels, and different styles of indentation that they
455 like; this is fine. What isn't fine is that different editors/viewers expand
456 tabs out to different tab stops. This can cause your code to look completely
457 unreadable, and it is not worth dealing with.
459 As always, follow the `Golden Rule`_ above: follow the style of
460 existing code if you are modifying and extending it. If you like four spaces of
461 indentation, **DO NOT** do that in the middle of a chunk of code with two spaces
462 of indentation. Also, do not reindent a whole source file: it makes for
463 incredible diffs that are absolutely worthless.
465 Indent Code Consistently
466 ^^^^^^^^^^^^^^^^^^^^^^^^
468 Okay, in your first year of programming you were told that indentation is
469 important. If you didn't believe and internalize this then, now is the time.
470 Just do it. With the introduction of C++11, there are some new formatting
471 challenges that merit some suggestions to help have consistent, maintainable,
472 and tool-friendly formatting and indentation.
474 Format Lambdas Like Blocks Of Code
475 """"""""""""""""""""""""""""""""""
477 When formatting a multi-line lambda, format it like a block of code, that's
478 what it is. If there is only one multi-line lambda in a statement, and there
479 are no expressions lexically after it in the statement, drop the indent to the
480 standard two space indent for a block of code, as if it were an if-block opened
481 by the preceding part of the statement:
485 std::sort(foo.begin(), foo.end(), [&](Foo a, Foo b) -> bool {
490 return a.bam < b.bam;
493 To take best advantage of this formatting, if you are designing an API which
494 accepts a continuation or single callable argument (be it a functor, or
495 a ``std::function``), it should be the last argument if at all possible.
497 If there are multiple multi-line lambdas in a statement, or there is anything
498 interesting after the lambda in the statement, indent the block two spaces from
499 the indent of the ``[]``:
503 dyn_switch(V->stripPointerCasts(),
507 [] (SelectInst *SI) {
508 // process selects...
513 [] (AllocaInst *AI) {
514 // process allocas...
517 Braced Initializer Lists
518 """"""""""""""""""""""""
520 With C++11, there are significantly more uses of braced lists to perform
521 initialization. These allow you to easily construct aggregate temporaries in
522 expressions among other niceness. They now have a natural way of ending up
523 nested within each other and within function calls in order to build up
524 aggregates (such as option structs) from local variables. To make matters
525 worse, we also have many more uses of braces in an expression context that are
526 *not* performing initialization.
528 The historically common formatting of braced initialization of aggregate
529 variables does not mix cleanly with deep nesting, general expression contexts,
530 function arguments, and lambdas. We suggest new code use a simple rule for
531 formatting braced initialization lists: act as-if the braces were parentheses
532 in a function call. The formatting rules exactly match those already well
533 understood for formatting nested function calls. Examples:
537 foo({a, b, c}, {1, 2, 3});
539 llvm::Constant *Mask[] = {
540 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 0),
541 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 1),
542 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 2)};
544 This formatting scheme also makes it particularly easy to get predictable,
545 consistent, and automatic formatting with tools like `Clang Format`_.
547 .. _Clang Format: http://clang.llvm.org/docs/ClangFormat.html
549 Language and Compiler Issues
550 ----------------------------
552 Treat Compiler Warnings Like Errors
553 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
555 If your code has compiler warnings in it, something is wrong --- you aren't
556 casting values correctly, you have "questionable" constructs in your code, or
557 you are doing something legitimately wrong. Compiler warnings can cover up
558 legitimate errors in output and make dealing with a translation unit difficult.
560 It is not possible to prevent all warnings from all compilers, nor is it
561 desirable. Instead, pick a standard compiler (like ``gcc``) that provides a
562 good thorough set of warnings, and stick to it. At least in the case of
563 ``gcc``, it is possible to work around any spurious errors by changing the
564 syntax of the code slightly. For example, a warning that annoys me occurs when
565 I write code like this:
569 if (V = getValue()) {
573 ``gcc`` will warn me that I probably want to use the ``==`` operator, and that I
574 probably mistyped it. In most cases, I haven't, and I really don't want the
575 spurious errors. To fix this particular problem, I rewrite the code like
580 if ((V = getValue())) {
584 which shuts ``gcc`` up. Any ``gcc`` warning that annoys you can be fixed by
585 massaging the code appropriately.
590 In almost all cases, it is possible and within reason to write completely
591 portable code. If there are cases where it isn't possible to write portable
592 code, isolate it behind a well defined (and well documented) interface.
594 In practice, this means that you shouldn't assume much about the host compiler
595 (and Visual Studio tends to be the lowest common denominator). If advanced
596 features are used, they should only be an implementation detail of a library
597 which has a simple exposed API, and preferably be buried in ``libSystem``.
599 Do not use RTTI or Exceptions
600 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
602 In an effort to reduce code and executable size, LLVM does not use RTTI
603 (e.g. ``dynamic_cast<>;``) or exceptions. These two language features violate
604 the general C++ principle of *"you only pay for what you use"*, causing
605 executable bloat even if exceptions are never used in the code base, or if RTTI
606 is never used for a class. Because of this, we turn them off globally in the
609 That said, LLVM does make extensive use of a hand-rolled form of RTTI that use
610 templates like :ref:`isa\<>, cast\<>, and dyn_cast\<> <isa>`.
611 This form of RTTI is opt-in and can be
612 :doc:`added to any class <HowToSetUpLLVMStyleRTTI>`. It is also
613 substantially more efficient than ``dynamic_cast<>``.
615 .. _static constructor:
617 Do not use Static Constructors
618 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
620 Static constructors and destructors (e.g. global variables whose types have a
621 constructor or destructor) should not be added to the code base, and should be
622 removed wherever possible. Besides `well known problems
623 <http://yosefk.com/c++fqa/ctors.html#fqa-10.12>`_ where the order of
624 initialization is undefined between globals in different source files, the
625 entire concept of static constructors is at odds with the common use case of
626 LLVM as a library linked into a larger application.
628 Consider the use of LLVM as a JIT linked into another application (perhaps for
629 `OpenGL, custom languages <http://llvm.org/Users.html>`_, `shaders in movies
630 <http://llvm.org/devmtg/2010-11/Gritz-OpenShadingLang.pdf>`_, etc). Due to the
631 design of static constructors, they must be executed at startup time of the
632 entire application, regardless of whether or how LLVM is used in that larger
633 application. There are two problems with this:
635 * The time to run the static constructors impacts startup time of applications
636 --- a critical time for GUI apps, among others.
638 * The static constructors cause the app to pull many extra pages of memory off
639 the disk: both the code for the constructor in each ``.o`` file and the small
640 amount of data that gets touched. In addition, touched/dirty pages put more
641 pressure on the VM system on low-memory machines.
643 We would really like for there to be zero cost for linking in an additional LLVM
644 target or other library into an application, but static constructors violate
647 That said, LLVM unfortunately does contain static constructors. It would be a
648 `great project <http://llvm.org/PR11944>`_ for someone to purge all static
649 constructors from LLVM, and then enable the ``-Wglobal-constructors`` warning
650 flag (when building with Clang) to ensure we do not regress in the future.
652 Use of ``class`` and ``struct`` Keywords
653 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
655 In C++, the ``class`` and ``struct`` keywords can be used almost
656 interchangeably. The only difference is when they are used to declare a class:
657 ``class`` makes all members private by default while ``struct`` makes all
658 members public by default.
660 Unfortunately, not all compilers follow the rules and some will generate
661 different symbols based on whether ``class`` or ``struct`` was used to declare
662 the symbol (e.g., MSVC). This can lead to problems at link time.
664 * All declarations and definitions of a given ``class`` or ``struct`` must use
665 the same keyword. For example:
671 // Breaks mangling in MSVC.
672 struct Foo { int Data; };
674 * As a rule of thumb, ``struct`` should be kept to structures where *all*
675 members are declared public.
679 // Foo feels like a class... this is strange.
685 int getData() const { return Data; }
686 void setData(int D) { Data = D; }
689 // Bar isn't POD, but it does look like a struct.
695 Do not use Braced Initializer Lists to Call a Constructor
696 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
698 In C++11 there is a "generalized initialization syntax" which allows calling
699 constructors using braced initializer lists. Do not use these to call
700 constructors with any interesting logic or if you care that you're calling some
701 *particular* constructor. Those should look like function calls using
702 parentheses rather than like aggregate initialization. Similarly, if you need
703 to explicitly name the type and call its constructor to create a temporary,
704 don't use a braced initializer list. Instead, use a braced initializer list
705 (without any type for temporaries) when doing aggregate initialization or
706 something notionally equivalent. Examples:
712 // Construct a Foo by reading data from the disk in the whizbang format, ...
713 Foo(std::string filename);
715 // Construct a Foo by looking up the Nth element of some global data ...
721 // The Foo constructor call is very deliberate, no braces.
722 std::fill(foo.begin(), foo.end(), Foo("name"));
724 // The pair is just being constructed like an aggregate, use braces.
725 bar_map.insert({my_key, my_value});
727 If you use a braced initializer list when initializing a variable, use an equals before the open curly brace:
731 int data[] = {0, 1, 2, 3};
733 Use ``auto`` Type Deduction to Make Code More Readable
734 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
736 Some are advocating a policy of "almost always ``auto``" in C++11, however LLVM
737 uses a more moderate stance. Use ``auto`` if and only if it makes the code more
738 readable or easier to maintain. Don't "almost always" use ``auto``, but do use
739 ``auto`` with initializers like ``cast<Foo>(...)`` or other places where the
740 type is already obvious from the context. Another time when ``auto`` works well
741 for these purposes is when the type would have been abstracted away anyways,
742 often behind a container's typedef such as ``std::vector<T>::iterator``.
744 Beware unnecessary copies with ``auto``
745 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
747 The convenience of ``auto`` makes it easy to forget that its default behavior
748 is a copy. Particularly in range-based ``for`` loops, careless copies are
751 As a rule of thumb, use ``auto &`` unless you need to copy the result, and use
752 ``auto *`` when copying pointers.
756 // Typically there's no reason to copy.
757 for (const auto &Val : Container) { observe(Val); }
758 for (auto &Val : Container) { Val.change(); }
760 // Remove the reference if you really want a new copy.
761 for (auto Val : Container) { Val.change(); saveSomewhere(Val); }
763 // Copy pointers, but make it clear that they're pointers.
764 for (const auto *Ptr : Container) { observe(*Ptr); }
765 for (auto *Ptr : Container) { Ptr->change(); }
770 The High-Level Issues
771 ---------------------
773 A Public Header File **is** a Module
774 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
776 C++ doesn't do too well in the modularity department. There is no real
777 encapsulation or data hiding (unless you use expensive protocol classes), but it
778 is what we have to work with. When you write a public header file (in the LLVM
779 source tree, they live in the top level "``include``" directory), you are
780 defining a module of functionality.
782 Ideally, modules should be completely independent of each other, and their
783 header files should only ``#include`` the absolute minimum number of headers
784 possible. A module is not just a class, a function, or a namespace: it's a
785 collection of these that defines an interface. This interface may be several
786 functions, classes, or data structures, but the important issue is how they work
789 In general, a module should be implemented by one or more ``.cpp`` files. Each
790 of these ``.cpp`` files should include the header that defines their interface
791 first. This ensures that all of the dependences of the module header have been
792 properly added to the module header itself, and are not implicit. System
793 headers should be included after user headers for a translation unit.
795 .. _minimal list of #includes:
797 ``#include`` as Little as Possible
798 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
800 ``#include`` hurts compile time performance. Don't do it unless you have to,
801 especially in header files.
803 But wait! Sometimes you need to have the definition of a class to use it, or to
804 inherit from it. In these cases go ahead and ``#include`` that header file. Be
805 aware however that there are many cases where you don't need to have the full
806 definition of a class. If you are using a pointer or reference to a class, you
807 don't need the header file. If you are simply returning a class instance from a
808 prototyped function or method, you don't need it. In fact, for most cases, you
809 simply don't need the definition of a class. And not ``#include``\ing speeds up
812 It is easy to try to go too overboard on this recommendation, however. You
813 **must** include all of the header files that you are using --- you can include
814 them either directly or indirectly through another header file. To make sure
815 that you don't accidentally forget to include a header file in your module
816 header, make sure to include your module header **first** in the implementation
817 file (as mentioned above). This way there won't be any hidden dependencies that
818 you'll find out about later.
820 Keep "Internal" Headers Private
821 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
823 Many modules have a complex implementation that causes them to use more than one
824 implementation (``.cpp``) file. It is often tempting to put the internal
825 communication interface (helper classes, extra functions, etc) in the public
826 module header file. Don't do this!
828 If you really need to do something like this, put a private header file in the
829 same directory as the source files, and include it locally. This ensures that
830 your private interface remains private and undisturbed by outsiders.
834 It's okay to put extra implementation methods in a public class itself. Just
835 make them private (or protected) and all is well.
839 Use Early Exits and ``continue`` to Simplify Code
840 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
842 When reading code, keep in mind how much state and how many previous decisions
843 have to be remembered by the reader to understand a block of code. Aim to
844 reduce indentation where possible when it doesn't make it more difficult to
845 understand the code. One great way to do this is by making use of early exits
846 and the ``continue`` keyword in long loops. As an example of using an early
847 exit from a function, consider this "bad" code:
851 Value *doSomething(Instruction *I) {
852 if (!isa<TerminatorInst>(I) &&
853 I->hasOneUse() && doOtherThing(I)) {
854 ... some long code ....
860 This code has several problems if the body of the ``'if'`` is large. When
861 you're looking at the top of the function, it isn't immediately clear that this
862 *only* does interesting things with non-terminator instructions, and only
863 applies to things with the other predicates. Second, it is relatively difficult
864 to describe (in comments) why these predicates are important because the ``if``
865 statement makes it difficult to lay out the comments. Third, when you're deep
866 within the body of the code, it is indented an extra level. Finally, when
867 reading the top of the function, it isn't clear what the result is if the
868 predicate isn't true; you have to read to the end of the function to know that
871 It is much preferred to format the code like this:
875 Value *doSomething(Instruction *I) {
876 // Terminators never need 'something' done to them because ...
877 if (isa<TerminatorInst>(I))
880 // We conservatively avoid transforming instructions with multiple uses
881 // because goats like cheese.
885 // This is really just here for example.
886 if (!doOtherThing(I))
889 ... some long code ....
892 This fixes these problems. A similar problem frequently happens in ``for``
893 loops. A silly example is something like this:
897 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
898 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
899 Value *LHS = BO->getOperand(0);
900 Value *RHS = BO->getOperand(1);
907 When you have very, very small loops, this sort of structure is fine. But if it
908 exceeds more than 10-15 lines, it becomes difficult for people to read and
909 understand at a glance. The problem with this sort of code is that it gets very
910 nested very quickly. Meaning that the reader of the code has to keep a lot of
911 context in their brain to remember what is going immediately on in the loop,
912 because they don't know if/when the ``if`` conditions will have ``else``\s etc.
913 It is strongly preferred to structure the loop like this:
917 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
918 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
921 Value *LHS = BO->getOperand(0);
922 Value *RHS = BO->getOperand(1);
923 if (LHS == RHS) continue;
928 This has all the benefits of using early exits for functions: it reduces nesting
929 of the loop, it makes it easier to describe why the conditions are true, and it
930 makes it obvious to the reader that there is no ``else`` coming up that they
931 have to push context into their brain for. If a loop is large, this can be a
932 big understandability win.
934 Don't use ``else`` after a ``return``
935 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
937 For similar reasons above (reduction of indentation and easier reading), please
938 do not use ``'else'`` or ``'else if'`` after something that interrupts control
939 flow --- like ``return``, ``break``, ``continue``, ``goto``, etc. For
940 example, this is *bad*:
946 Type = Context.getsigjmp_bufType();
948 Error = ASTContext::GE_Missing_sigjmp_buf;
954 Type = Context.getjmp_bufType();
956 Error = ASTContext::GE_Missing_jmp_buf;
964 It is better to write it like this:
970 Type = Context.getsigjmp_bufType();
972 Error = ASTContext::GE_Missing_sigjmp_buf;
976 Type = Context.getjmp_bufType();
978 Error = ASTContext::GE_Missing_jmp_buf;
984 Or better yet (in this case) as:
990 Type = Context.getsigjmp_bufType();
992 Type = Context.getjmp_bufType();
995 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
996 ASTContext::GE_Missing_jmp_buf;
1001 The idea is to reduce indentation and the amount of code you have to keep track
1002 of when reading the code.
1004 Turn Predicate Loops into Predicate Functions
1005 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1007 It is very common to write small loops that just compute a boolean value. There
1008 are a number of ways that people commonly write these, but an example of this
1013 bool FoundFoo = false;
1014 for (unsigned I = 0, E = BarList.size(); I != E; ++I)
1015 if (BarList[I]->isFoo()) {
1024 This sort of code is awkward to write, and is almost always a bad sign. Instead
1025 of this sort of loop, we strongly prefer to use a predicate function (which may
1026 be `static`_) that uses `early exits`_ to compute the predicate. We prefer the
1027 code to be structured like this:
1031 /// \returns true if the specified list has an element that is a foo.
1032 static bool containsFoo(const std::vector<Bar*> &List) {
1033 for (unsigned I = 0, E = List.size(); I != E; ++I)
1034 if (List[I]->isFoo())
1040 if (containsFoo(BarList)) {
1044 There are many reasons for doing this: it reduces indentation and factors out
1045 code which can often be shared by other code that checks for the same predicate.
1046 More importantly, it *forces you to pick a name* for the function, and forces
1047 you to write a comment for it. In this silly example, this doesn't add much
1048 value. However, if the condition is complex, this can make it a lot easier for
1049 the reader to understand the code that queries for this predicate. Instead of
1050 being faced with the in-line details of how we check to see if the BarList
1051 contains a foo, we can trust the function name and continue reading with better
1054 The Low-Level Issues
1055 --------------------
1057 Name Types, Functions, Variables, and Enumerators Properly
1058 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1060 Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
1061 enough how important it is to use *descriptive* names. Pick names that match
1062 the semantics and role of the underlying entities, within reason. Avoid
1063 abbreviations unless they are well known. After picking a good name, make sure
1064 to use consistent capitalization for the name, as inconsistency requires clients
1065 to either memorize the APIs or to look it up to find the exact spelling.
1067 In general, names should be in camel case (e.g. ``TextFileReader`` and
1068 ``isLValue()``). Different kinds of declarations have different rules:
1070 * **Type names** (including classes, structs, enums, typedefs, etc) should be
1071 nouns and start with an upper-case letter (e.g. ``TextFileReader``).
1073 * **Variable names** should be nouns (as they represent state). The name should
1074 be camel case, and start with an upper case letter (e.g. ``Leader`` or
1077 * **Function names** should be verb phrases (as they represent actions), and
1078 command-like function should be imperative. The name should be camel case,
1079 and start with a lower case letter (e.g. ``openFile()`` or ``isFoo()``).
1081 * **Enum declarations** (e.g. ``enum Foo {...}``) are types, so they should
1082 follow the naming conventions for types. A common use for enums is as a
1083 discriminator for a union, or an indicator of a subclass. When an enum is
1084 used for something like this, it should have a ``Kind`` suffix
1085 (e.g. ``ValueKind``).
1087 * **Enumerators** (e.g. ``enum { Foo, Bar }``) and **public member variables**
1088 should start with an upper-case letter, just like types. Unless the
1089 enumerators are defined in their own small namespace or inside a class,
1090 enumerators should have a prefix corresponding to the enum declaration name.
1091 For example, ``enum ValueKind { ... };`` may contain enumerators like
1092 ``VK_Argument``, ``VK_BasicBlock``, etc. Enumerators that are just
1093 convenience constants are exempt from the requirement for a prefix. For
1103 As an exception, classes that mimic STL classes can have member names in STL's
1104 style of lower-case words separated by underscores (e.g. ``begin()``,
1105 ``push_back()``, and ``empty()``). Classes that provide multiple
1106 iterators should add a singular prefix to ``begin()`` and ``end()``
1107 (e.g. ``global_begin()`` and ``use_begin()``).
1109 Here are some examples of good and bad names:
1113 class VehicleMaker {
1115 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
1116 Factory<Tire> Factory; // Better.
1117 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
1118 // kind of factories.
1121 Vehicle MakeVehicle(VehicleType Type) {
1122 VehicleMaker M; // Might be OK if having a short life-span.
1123 Tire Tmp1 = M.makeTire(); // Bad -- 'Tmp1' provides no information.
1124 Light Headlight = M.makeLight("head"); // Good -- descriptive.
1131 Use the "``assert``" macro to its fullest. Check all of your preconditions and
1132 assumptions, you never know when a bug (not necessarily even yours) might be
1133 caught early by an assertion, which reduces debugging time dramatically. The
1134 "``<cassert>``" header file is probably already included by the header files you
1135 are using, so it doesn't cost anything to use it.
1137 To further assist with debugging, make sure to put some kind of error message in
1138 the assertion statement, which is printed if the assertion is tripped. This
1139 helps the poor debugger make sense of why an assertion is being made and
1140 enforced, and hopefully what to do about it. Here is one complete example:
1144 inline Value *getOperand(unsigned I) {
1145 assert(I < Operands.size() && "getOperand() out of range!");
1149 Here are more examples:
1153 assert(Ty->isPointerType() && "Can't allocate a non-pointer type!");
1155 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
1157 assert(idx < getNumSuccessors() && "Successor # out of range!");
1159 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
1161 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
1165 In the past, asserts were used to indicate a piece of code that should not be
1166 reached. These were typically of the form:
1170 assert(0 && "Invalid radix for integer literal");
1172 This has a few issues, the main one being that some compilers might not
1173 understand the assertion, or warn about a missing return in builds where
1174 assertions are compiled out.
1176 Today, we have something much better: ``llvm_unreachable``:
1180 llvm_unreachable("Invalid radix for integer literal");
1182 When assertions are enabled, this will print the message if it's ever reached
1183 and then exit the program. When assertions are disabled (i.e. in release
1184 builds), ``llvm_unreachable`` becomes a hint to compilers to skip generating
1185 code for this branch. If the compiler does not support this, it will fall back
1186 to the "abort" implementation.
1188 Another issue is that values used only by assertions will produce an "unused
1189 value" warning when assertions are disabled. For example, this code will warn:
1193 unsigned Size = V.size();
1194 assert(Size > 42 && "Vector smaller than it should be");
1196 bool NewToSet = Myset.insert(Value);
1197 assert(NewToSet && "The value shouldn't be in the set yet");
1199 These are two interesting different cases. In the first case, the call to
1200 ``V.size()`` is only useful for the assert, and we don't want it executed when
1201 assertions are disabled. Code like this should move the call into the assert
1202 itself. In the second case, the side effects of the call must happen whether
1203 the assert is enabled or not. In this case, the value should be cast to void to
1204 disable the warning. To be specific, it is preferred to write the code like
1209 assert(V.size() > 42 && "Vector smaller than it should be");
1211 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1212 assert(NewToSet && "The value shouldn't be in the set yet");
1214 Do Not Use ``using namespace std``
1215 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1217 In LLVM, we prefer to explicitly prefix all identifiers from the standard
1218 namespace with an "``std::``" prefix, rather than rely on "``using namespace
1221 In header files, adding a ``'using namespace XXX'`` directive pollutes the
1222 namespace of any source file that ``#include``\s the header. This is clearly a
1225 In implementation files (e.g. ``.cpp`` files), the rule is more of a stylistic
1226 rule, but is still important. Basically, using explicit namespace prefixes
1227 makes the code **clearer**, because it is immediately obvious what facilities
1228 are being used and where they are coming from. And **more portable**, because
1229 namespace clashes cannot occur between LLVM code and other namespaces. The
1230 portability rule is important because different standard library implementations
1231 expose different symbols (potentially ones they shouldn't), and future revisions
1232 to the C++ standard will add more symbols to the ``std`` namespace. As such, we
1233 never use ``'using namespace std;'`` in LLVM.
1235 The exception to the general rule (i.e. it's not an exception for the ``std``
1236 namespace) is for implementation files. For example, all of the code in the
1237 LLVM project implements code that lives in the 'llvm' namespace. As such, it is
1238 ok, and actually clearer, for the ``.cpp`` files to have a ``'using namespace
1239 llvm;'`` directive at the top, after the ``#include``\s. This reduces
1240 indentation in the body of the file for source editors that indent based on
1241 braces, and keeps the conceptual context cleaner. The general form of this rule
1242 is that any ``.cpp`` file that implements code in any namespace may use that
1243 namespace (and its parents'), but should not use any others.
1245 Provide a Virtual Method Anchor for Classes in Headers
1246 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1248 If a class is defined in a header file and has a vtable (either it has virtual
1249 methods or it derives from classes with virtual methods), it must always have at
1250 least one out-of-line virtual method in the class. Without this, the compiler
1251 will copy the vtable and RTTI into every ``.o`` file that ``#include``\s the
1252 header, bloating ``.o`` file sizes and increasing link times.
1254 Don't use default labels in fully covered switches over enumerations
1255 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1257 ``-Wswitch`` warns if a switch, without a default label, over an enumeration
1258 does not cover every enumeration value. If you write a default label on a fully
1259 covered switch over an enumeration then the ``-Wswitch`` warning won't fire
1260 when new elements are added to that enumeration. To help avoid adding these
1261 kinds of defaults, Clang has the warning ``-Wcovered-switch-default`` which is
1262 off by default but turned on when building LLVM with a version of Clang that
1263 supports the warning.
1265 A knock-on effect of this stylistic requirement is that when building LLVM with
1266 GCC you may get warnings related to "control may reach end of non-void function"
1267 if you return from each case of a covered switch-over-enum because GCC assumes
1268 that the enum expression may take any representable value, not just those of
1269 individual enumerators. To suppress this warning, use ``llvm_unreachable`` after
1272 Use ``LLVM_DELETED_FUNCTION`` to mark uncallable methods
1273 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1275 Prior to C++11, a common pattern to make a class uncopyable was to declare an
1276 unimplemented copy constructor and copy assignment operator and make them
1277 private. This would give a compiler error for accessing a private method or a
1278 linker error because it wasn't implemented.
1280 With C++11, we can mark methods that won't be implemented with ``= delete``.
1281 This will trigger a much better error message and tell the compiler that the
1282 method will never be implemented. This enables other checks like
1283 ``-Wunused-private-field`` to run correctly on classes that contain these
1286 For compatibility with MSVC, ``LLVM_DELETED_FUNCTION`` should be used which
1287 will expand to ``= delete`` on compilers that support it. These methods should
1288 still be declared private. Example of the uncopyable pattern:
1294 DontCopy(const DontCopy&) LLVM_DELETED_FUNCTION;
1295 DontCopy &operator =(const DontCopy&) LLVM_DELETED_FUNCTION;
1300 Don't evaluate ``end()`` every time through a loop
1301 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1303 Because C++ doesn't have a standard "``foreach``" loop (though it can be
1304 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1305 loops that manually iterate from begin to end on a variety of containers or
1306 through other data structures. One common mistake is to write a loop in this
1311 BasicBlock *BB = ...
1312 for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I)
1315 The problem with this construct is that it evaluates "``BB->end()``" every time
1316 through the loop. Instead of writing the loop like this, we strongly prefer
1317 loops to be written so that they evaluate it once before the loop starts. A
1318 convenient way to do this is like so:
1322 BasicBlock *BB = ...
1323 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
1326 The observant may quickly point out that these two loops may have different
1327 semantics: if the container (a basic block in this case) is being mutated, then
1328 "``BB->end()``" may change its value every time through the loop and the second
1329 loop may not in fact be correct. If you actually do depend on this behavior,
1330 please write the loop in the first form and add a comment indicating that you
1331 did it intentionally.
1333 Why do we prefer the second form (when correct)? Writing the loop in the first
1334 form has two problems. First it may be less efficient than evaluating it at the
1335 start of the loop. In this case, the cost is probably minor --- a few extra
1336 loads every time through the loop. However, if the base expression is more
1337 complex, then the cost can rise quickly. I've seen loops where the end
1338 expression was actually something like: "``SomeMap[X]->end()``" and map lookups
1339 really aren't cheap. By writing it in the second form consistently, you
1340 eliminate the issue entirely and don't even have to think about it.
1342 The second (even bigger) issue is that writing the loop in the first form hints
1343 to the reader that the loop is mutating the container (a fact that a comment
1344 would handily confirm!). If you write the loop in the second form, it is
1345 immediately obvious without even looking at the body of the loop that the
1346 container isn't being modified, which makes it easier to read the code and
1347 understand what it does.
1349 While the second form of the loop is a few extra keystrokes, we do strongly
1352 ``#include <iostream>`` is Forbidden
1353 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1355 The use of ``#include <iostream>`` in library files is hereby **forbidden**,
1356 because many common implementations transparently inject a `static constructor`_
1357 into every translation unit that includes it.
1359 Note that using the other stream headers (``<sstream>`` for example) is not
1360 problematic in this regard --- just ``<iostream>``. However, ``raw_ostream``
1361 provides various APIs that are better performing for almost every use than
1362 ``std::ostream`` style APIs.
1366 New code should always use `raw_ostream`_ for writing, or the
1367 ``llvm::MemoryBuffer`` API for reading files.
1374 LLVM includes a lightweight, simple, and efficient stream implementation in
1375 ``llvm/Support/raw_ostream.h``, which provides all of the common features of
1376 ``std::ostream``. All new code should use ``raw_ostream`` instead of
1379 Unlike ``std::ostream``, ``raw_ostream`` is not a template and can be forward
1380 declared as ``class raw_ostream``. Public headers should generally not include
1381 the ``raw_ostream`` header, but use forward declarations and constant references
1382 to ``raw_ostream`` instances.
1387 The ``std::endl`` modifier, when used with ``iostreams`` outputs a newline to
1388 the output stream specified. In addition to doing this, however, it also
1389 flushes the output stream. In other words, these are equivalent:
1393 std::cout << std::endl;
1394 std::cout << '\n' << std::flush;
1396 Most of the time, you probably have no reason to flush the output stream, so
1397 it's better to use a literal ``'\n'``.
1399 Don't use ``inline`` when defining a function in a class definition
1400 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1402 A member function defined in a class definition is implicitly inline, so don't
1403 put the ``inline`` keyword in this case.
1430 This section describes preferred low-level formatting guidelines along with
1431 reasoning on why we prefer them.
1433 Spaces Before Parentheses
1434 ^^^^^^^^^^^^^^^^^^^^^^^^^
1436 We prefer to put a space before an open parenthesis only in control flow
1437 statements, but not in normal function call expressions and function-like
1438 macros. For example, this is good:
1443 for (I = 0; I != 100; ++I) ...
1444 while (LLVMRocks) ...
1447 assert(3 != 4 && "laws of math are failing me");
1449 A = foo(42, 92) + bar(X);
1456 for(I = 0; I != 100; ++I) ...
1457 while(LLVMRocks) ...
1460 assert (3 != 4 && "laws of math are failing me");
1462 A = foo (42, 92) + bar (X);
1464 The reason for doing this is not completely arbitrary. This style makes control
1465 flow operators stand out more, and makes expressions flow better. The function
1466 call operator binds very tightly as a postfix operator. Putting a space after a
1467 function name (as in the last example) makes it appear that the code might bind
1468 the arguments of the left-hand-side of a binary operator with the argument list
1469 of a function and the name of the right side. More specifically, it is easy to
1470 misread the "``A``" example as:
1474 A = foo ((42, 92) + bar) (X);
1476 when skimming through the code. By avoiding a space in a function, we avoid
1477 this misinterpretation.
1482 Hard fast rule: Preincrement (``++X``) may be no slower than postincrement
1483 (``X++``) and could very well be a lot faster than it. Use preincrementation
1486 The semantics of postincrement include making a copy of the value being
1487 incremented, returning it, and then preincrementing the "work value". For
1488 primitive types, this isn't a big deal. But for iterators, it can be a huge
1489 issue (for example, some iterators contains stack and set objects in them...
1490 copying an iterator could invoke the copy ctor's of these as well). In general,
1491 get in the habit of always using preincrement, and you won't have a problem.
1494 Namespace Indentation
1495 ^^^^^^^^^^^^^^^^^^^^^
1497 In general, we strive to reduce indentation wherever possible. This is useful
1498 because we want code to `fit into 80 columns`_ without wrapping horribly, but
1499 also because it makes it easier to understand the code. To facilitate this and
1500 avoid some insanely deep nesting on occasion, don't indent namespaces. If it
1501 helps readability, feel free to add a comment indicating what namespace is
1502 being closed by a ``}``. For example:
1507 namespace knowledge {
1509 /// This class represents things that Smith can have an intimate
1510 /// understanding of and contains the data associated with it.
1514 explicit Grokable() { ... }
1515 virtual ~Grokable() = 0;
1521 } // end namespace knowledge
1522 } // end namespace llvm
1525 Feel free to skip the closing comment when the namespace being closed is
1526 obvious for any reason. For example, the outer-most namespace in a header file
1527 is rarely a source of confusion. But namespaces both anonymous and named in
1528 source files that are being closed half way through the file probably could use
1533 Anonymous Namespaces
1534 ^^^^^^^^^^^^^^^^^^^^
1536 After talking about namespaces in general, you may be wondering about anonymous
1537 namespaces in particular. Anonymous namespaces are a great language feature
1538 that tells the C++ compiler that the contents of the namespace are only visible
1539 within the current translation unit, allowing more aggressive optimization and
1540 eliminating the possibility of symbol name collisions. Anonymous namespaces are
1541 to C++ as "static" is to C functions and global variables. While "``static``"
1542 is available in C++, anonymous namespaces are more general: they can make entire
1543 classes private to a file.
1545 The problem with anonymous namespaces is that they naturally want to encourage
1546 indentation of their body, and they reduce locality of reference: if you see a
1547 random function definition in a C++ file, it is easy to see if it is marked
1548 static, but seeing if it is in an anonymous namespace requires scanning a big
1551 Because of this, we have a simple guideline: make anonymous namespaces as small
1552 as possible, and only use them for class declarations. For example, this is
1562 bool operator<(const char *RHS) const;
1564 } // end anonymous namespace
1566 static void runHelper() {
1570 bool StringSort::operator<(const char *RHS) const {
1584 bool operator<(const char *RHS) const;
1591 bool StringSort::operator<(const char *RHS) const {
1595 } // end anonymous namespace
1597 This is bad specifically because if you're looking at "``runHelper``" in the middle
1598 of a large C++ file, that you have no immediate way to tell if it is local to
1599 the file. When it is marked static explicitly, this is immediately obvious.
1600 Also, there is no reason to enclose the definition of "``operator<``" in the
1601 namespace just because it was declared there.
1606 A lot of these comments and recommendations have been culled from other sources.
1607 Two particularly important books for our work are:
1610 <http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876>`_
1611 by Scott Meyers. Also interesting and useful are "More Effective C++" and
1612 "Effective STL" by the same author.
1614 #. `Large-Scale C++ Software Design
1615 <http://www.amazon.com/Large-Scale-Software-Design-John-Lakos/dp/0201633620/ref=sr_1_1>`_
1618 If you get some free time, and you haven't read them: do so, you might learn