11 This document attempts to describe a few coding standards that are being used in
12 the LLVM source tree. Although no coding standards should be regarded as
13 absolute requirements to be followed in all instances, coding standards are
14 particularly important for large-scale code bases that follow a library-based
17 While this document may provide guidance for some mechanical formatting issues,
18 whitespace, or other "microscopic details", these are not fixed standards.
19 Always follow the golden rule:
23 **If you are extending, enhancing, or bug fixing already implemented code,
24 use the style that is already being used so that the source is uniform and
27 Note that some code bases (e.g. ``libc++``) have really good reasons to deviate
28 from the coding standards. In the case of ``libc++``, this is because the
29 naming and other conventions are dictated by the C++ standard. If you think
30 there is a specific good reason to deviate from the standards here, please bring
31 it up on the LLVMdev mailing list.
33 There are some conventions that are not uniformly followed in the code base
34 (e.g. the naming convention). This is because they are relatively new, and a
35 lot of code was written before they were put in place. Our long term goal is
36 for the entire codebase to follow the convention, but we explicitly *do not*
37 want patches that do large-scale reformating of existing code. On the other
38 hand, it is reasonable to rename the methods of a class if you're about to
39 change it in some other way. Just do the reformating as a separate commit from
40 the functionality change.
42 The ultimate goal of these guidelines is the increase readability and
43 maintainability of our common source base. If you have suggestions for topics to
44 be included, please mail them to `Chris <mailto:sabre@nondot.org>`_.
46 Languages, Libraries, and Standards
47 ===================================
49 Most source code in LLVM and other LLVM projects using these coding standards
50 is C++ code. There are some places where C code is used either due to
51 environment restrictions, historical restrictions, or due to third-party source
52 code imported into the tree. Generally, our preference is for standards
53 conforming, modern, and portable C++ code as the implementation language of
59 LLVM, Clang, and LLD are currently written using C++11 conforming code,
60 although we restrict ourselves to features which are available in the major
61 toolchains supported as host compilers. The LLDB project is even more
62 aggressive in the set of host compilers supported and thus uses still more
63 features. Regardless of the supported features, code is expected to (when
64 reasonable) be standard, portable, and modern C++11 code. We avoid unnecessary
65 vendor-specific extensions, etc.
70 Use the C++ standard library facilities whenever they are available for
71 a particular task. LLVM and related projects emphasize and rely on the standard
72 library facilities for as much as possible. Common support libraries providing
73 functionality missing from the standard library for which there are standard
74 interfaces or active work on adding standard interfaces will often be
75 implemented in the LLVM namespace following the expected standard interface.
77 There are some exceptions such as the standard I/O streams library which are
78 avoided. Also, there is much more detailed information on these subjects in the
79 `Programmer's Manual`_.
81 .. _Programmer's Manual:
82 http://llvm.org/docs/ProgrammersManual.html
84 Supported C++11 Language and Library Features
85 ---------------------------------------------
87 While LLVM, Clang, and LLD use C++11, not all features are available in all of
88 the toolchains which we support. The set of features supported for use in LLVM
89 is the intersection of those supported in MSVC 2012, GCC 4.7, and Clang 3.1.
90 The ultimate definition of this set is what build bots with those respective
91 toolchains accept. Don't argue with the build bots. However, we have some
92 guidance below to help you know what to expect.
94 Each toolchain provides a good reference for what it accepts:
96 * Clang: http://clang.llvm.org/cxx_status.html
97 * GCC: http://gcc.gnu.org/projects/cxx0x.html
98 * MSVC: http://msdn.microsoft.com/en-us/library/hh567368.aspx
100 In most cases, the MSVC list will be the dominating factor. Here is a summary
101 of the features that are expected to work. Features not on this list are
102 unlikely to be supported by our host compilers.
104 * Rvalue references: N2118_
106 * But *not* Rvalue references for ``*this`` or member qualifiers (N2439_)
108 * Static assert: N1720_
109 * ``auto`` type deduction: N1984_, N1737_
110 * Trailing return types: N2541_
112 * ``decltype``: N2343_
113 * Nested closing right angle brackets: N1757_
114 * Extern templates: N1987_
115 * ``nullptr``: N2431_
116 * Strongly-typed and forward declarable enums: N2347_, N2764_
117 * Local and unnamed types as template arguments: N2657_
118 * Range-based for-loop: N2930_
119 * ``override`` and ``final``: N2928_, N3206_, N3272_
120 * Atomic operations and the C++11 memory model: N2429_
122 .. _N2118: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2118.html
123 .. _N2439: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2439.htm
124 .. _N1720: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1720.html
125 .. _N1984: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1984.pdf
126 .. _N1737: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1737.pdf
127 .. _N2541: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2541.htm
128 .. _N2927: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2927.pdf
129 .. _N2343: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2343.pdf
130 .. _N1757: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1757.html
131 .. _N1987: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1987.htm
132 .. _N2431: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2431.pdf
133 .. _N2347: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2347.pdf
134 .. _N2764: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2764.pdf
135 .. _N2657: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm
136 .. _N2930: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2930.html
137 .. _N2928: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2928.htm
138 .. _N3206: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3206.htm
139 .. _N3272: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3272.htm
140 .. _N2429: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2429.htm
142 The supported features in the C++11 standard libraries are less well tracked,
143 but also much greater. Most of the standard libraries implement most of C++11's
144 library. The most likely lowest common denominator is Linux support. For
145 libc++, the support is just poorly tested and undocumented but expected to be
146 largely complete. YMMV. For libstdc++, the support is documented in detail in
147 `the libstdc++ manual`_. There are some very minor missing facilities that are
148 unlikely to be common problems, and there are a few larger gaps that are worth
151 * Not all of the type traits are implemented
152 * No regular expression library.
153 * While most of the atomics library is well implemented, the fences are
154 missing. Fortunately, they are rarely needed.
155 * The locale support is incomplete.
156 * ``std::initializer_list`` (and the constructors and functions that take it as
157 an argument) are not always available, so you cannot (for example) initialize
158 a ``std::vector`` with a braced initializer list.
160 Other than these areas you should assume the standard library is available and
161 working as expected until some build bot tells you otherwise. If you're in an
162 uncertain area of one of the above points, but you cannot test on a Linux
163 system, your best approach is to minimize your use of these features, and watch
164 the Linux build bots to find out if your usage triggered a bug. For example, if
165 you hit a type trait which doesn't work we can then add support to LLVM's
166 traits header to emulate it.
168 .. _the libstdc++ manual:
169 http://gcc.gnu.org/onlinedocs/gcc-4.7.3/libstdc++/manual/manual/status.html#status.iso.2011
171 Mechanical Source Issues
172 ========================
174 Source Code Formatting
175 ----------------------
180 Comments are one critical part of readability and maintainability. Everyone
181 knows they should comment their code, and so should you. When writing comments,
182 write them as English prose, which means they should use proper capitalization,
183 punctuation, etc. Aim to describe what the code is trying to do and why, not
184 *how* it does it at a micro level. Here are a few critical things to document:
186 .. _header file comment:
191 Every source file should have a header on it that describes the basic purpose of
192 the file. If a file does not have a header, it should not be checked into the
193 tree. The standard header looks like this:
197 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
199 // The LLVM Compiler Infrastructure
201 // This file is distributed under the University of Illinois Open Source
202 // License. See LICENSE.TXT for details.
204 //===----------------------------------------------------------------------===//
207 /// \brief This file contains the declaration of the Instruction class, which is
208 /// the base class for all of the VM instructions.
210 //===----------------------------------------------------------------------===//
212 A few things to note about this particular format: The "``-*- C++ -*-``" string
213 on the first line is there to tell Emacs that the source file is a C++ file, not
214 a C file (Emacs assumes ``.h`` files are C files by default).
218 This tag is not necessary in ``.cpp`` files. The name of the file is also
219 on the first line, along with a very short description of the purpose of the
220 file. This is important when printing out code and flipping though lots of
223 The next section in the file is a concise note that defines the license that the
224 file is released under. This makes it perfectly clear what terms the source
225 code can be distributed under and should not be modified in any way.
227 The main body is a ``doxygen`` comment describing the purpose of the file. It
228 should have a ``\brief`` command that describes the file in one or two
229 sentences. Any additional information should be separated by a blank line. If
230 an algorithm is being implemented or something tricky is going on, a reference
231 to the paper where it is published should be included, as well as any notes or
232 *gotchas* in the code to watch out for.
237 Classes are one fundamental part of a good object oriented design. As such, a
238 class definition should have a comment block that explains what the class is
239 used for and how it works. Every non-trivial class is expected to have a
240 ``doxygen`` comment block.
245 Methods defined in a class (as well as any global functions) should also be
246 documented properly. A quick note about what it does and a description of the
247 borderline behaviour is all that is necessary here (unless something
248 particularly tricky or insidious is going on). The hope is that people can
249 figure out how to use your interfaces without reading the code itself.
251 Good things to talk about here are what happens when something unexpected
252 happens: does the method return null? Abort? Format your hard disk?
257 In general, prefer C++ style (``//``) comments. They take less space, require
258 less typing, don't have nesting problems, etc. There are a few cases when it is
259 useful to use C style (``/* */``) comments however:
261 #. When writing C code: Obviously if you are writing C code, use C style
264 #. When writing a header file that may be ``#include``\d by a C source file.
266 #. When writing a source file that is used by a tool that only accepts C style
269 To comment out a large block of code, use ``#if 0`` and ``#endif``. These nest
270 properly and are better behaved in general than C style comments.
272 Doxygen Use in Documentation Comments
273 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
275 Use the ``\file`` command to turn the standard file header into a file-level
278 Include descriptive ``\brief`` paragraphs for all public interfaces (public
279 classes, member and non-member functions). Explain API use and purpose in
280 ``\brief`` paragraphs, don't just restate the information that can be inferred
281 from the API name. Put detailed discussion into separate paragraphs.
283 To refer to parameter names inside a paragraph, use the ``\p name`` command.
284 Don't use the ``\arg name`` command since it starts a new paragraph that
285 contains documentation for the parameter.
287 Wrap non-inline code examples in ``\code ... \endcode``.
289 To document a function parameter, start a new paragraph with the
290 ``\param name`` command. If the parameter is used as an out or an in/out
291 parameter, use the ``\param [out] name`` or ``\param [in,out] name`` command,
294 To describe function return value, start a new paragraph with the ``\returns``
297 A minimal documentation comment:
301 /// \brief Does foo and bar.
302 void fooBar(bool Baz);
304 A documentation comment that uses all Doxygen features in a preferred way:
308 /// \brief Does foo and bar.
310 /// Does not do foo the usual way if \p Baz is true.
314 /// fooBar(false, "quux", Res);
317 /// \param Quux kind of foo to do.
318 /// \param [out] Result filled with bar sequence on foo success.
320 /// \returns true on success.
321 bool fooBar(bool Baz, StringRef Quux, std::vector<int> &Result);
323 Don't duplicate the documentation comment in the header file and in the
324 implementation file. Put the documentation comments for public APIs into the
325 header file. Documentation comments for private APIs can go to the
326 implementation file. In any case, implementation files can include additional
327 comments (not necessarily in Doxygen markup) to explain implementation details
330 Don't duplicate function or class name at the beginning of the comment.
331 For humans it is obvious which function or class is being documented;
332 automatic documentation processing tools are smart enough to bind the comment
333 to the correct declaration.
341 /// Something - An abstraction for some complicated thing.
344 /// fooBar - Does foo and bar.
350 /// fooBar - Does foo and bar.
351 void Something::fooBar() { ... }
359 /// \brief An abstraction for some complicated thing.
362 /// \brief Does foo and bar.
368 // Builds a B-tree in order to do foo. See paper by...
369 void Something::fooBar() { ... }
371 It is not required to use additional Doxygen features, but sometimes it might
372 be a good idea to do so.
376 * adding comments to any narrow namespace containing a collection of
377 related functions or types;
379 * using top-level groups to organize a collection of related functions at
380 namespace scope where the grouping is smaller than the namespace;
382 * using member groups and additional comments attached to member
383 groups to organize within a class.
390 /// \name Functions that do Foo.
401 Immediately after the `header file comment`_ (and include guards if working on a
402 header file), the `minimal list of #includes`_ required by the file should be
403 listed. We prefer these ``#include``\s to be listed in this order:
405 .. _Main Module Header:
406 .. _Local/Private Headers:
408 #. Main Module Header
409 #. Local/Private Headers
411 #. System ``#include``\s
413 and each category should be sorted lexicographically by the full path.
415 The `Main Module Header`_ file applies to ``.cpp`` files which implement an
416 interface defined by a ``.h`` file. This ``#include`` should always be included
417 **first** regardless of where it lives on the file system. By including a
418 header file first in the ``.cpp`` files that implement the interfaces, we ensure
419 that the header does not have any hidden dependencies which are not explicitly
420 ``#include``\d in the header, but should be. It is also a form of documentation
421 in the ``.cpp`` file to indicate where the interfaces it implements are defined.
423 .. _fit into 80 columns:
428 Write your code to fit within 80 columns of text. This helps those of us who
429 like to print out code and look at your code in an ``xterm`` without resizing
432 The longer answer is that there must be some limit to the width of the code in
433 order to reasonably allow developers to have multiple files side-by-side in
434 windows on a modest display. If you are going to pick a width limit, it is
435 somewhat arbitrary but you might as well pick something standard. Going with 90
436 columns (for example) instead of 80 columns wouldn't add any significant value
437 and would be detrimental to printing out code. Also many other projects have
438 standardized on 80 columns, so some people have already configured their editors
439 for it (vs something else, like 90 columns).
441 This is one of many contentious issues in coding standards, but it is not up for
444 Use Spaces Instead of Tabs
445 ^^^^^^^^^^^^^^^^^^^^^^^^^^
447 In all cases, prefer spaces to tabs in source files. People have different
448 preferred indentation levels, and different styles of indentation that they
449 like; this is fine. What isn't fine is that different editors/viewers expand
450 tabs out to different tab stops. This can cause your code to look completely
451 unreadable, and it is not worth dealing with.
453 As always, follow the `Golden Rule`_ above: follow the style of
454 existing code if you are modifying and extending it. If you like four spaces of
455 indentation, **DO NOT** do that in the middle of a chunk of code with two spaces
456 of indentation. Also, do not reindent a whole source file: it makes for
457 incredible diffs that are absolutely worthless.
459 Indent Code Consistently
460 ^^^^^^^^^^^^^^^^^^^^^^^^
462 Okay, in your first year of programming you were told that indentation is
463 important. If you didn't believe and internalize this then, now is the time.
464 Just do it. With the introduction of C++11, there are some new formatting
465 challenges that merit some suggestions to help have consistent, maintainable,
466 and tool-friendly formatting and indentation.
468 Format Lambdas Like Blocks Of Code
469 """"""""""""""""""""""""""""""""""
471 When formatting a multi-line lambda, format it like a block of code, that's
472 what it is. If there is only one multi-line lambda in a statement, and there
473 are no expressions lexically after it in the statement, drop the indent to the
474 standard two space indent for a block of code, as if it were an if-block opened
475 by the preceding part of the statement:
479 std::sort(foo.begin(), foo.end(), [&](Foo a, Foo b) -> bool {
484 return a.bam < b.bam;
487 To take best advantage of this formatting, if you are designing an API which
488 accepts a continuation or single callable argument (be it a functor, or
489 a ``std::function``), it should be the last argument if at all possible.
491 If there are multiple multi-line lambdas in a statement, or there is anything
492 interesting after the lambda in the statement, indent the block two spaces from
493 the indent of the ``[]``:
497 dyn_switch(V->stripPointerCasts(),
501 [] (SelectInst *SI) {
502 // process selects...
507 [] (AllocaInst *AI) {
508 // process allocas...
511 Braced Initializer Lists
512 """"""""""""""""""""""""
514 With C++11, there are significantly more uses of braced lists to perform
515 initialization. These allow you to easily construct aggregate temporaries in
516 expressions among other niceness. They now have a natural way of ending up
517 nested within each other and within function calls in order to build up
518 aggregates (such as option structs) from local variables. To make matters
519 worse, we also have many more uses of braces in an expression context that are
520 *not* performing initialization.
522 The historically common formatting of braced initialization of aggregate
523 variables does not mix cleanly with deep nesting, general expression contexts,
524 function arguments, and lambdas. We suggest new code use a simple rule for
525 formatting braced initialization lists: act as-if the braces were parentheses
526 in a function call. The formatting rules exactly match those already well
527 understood for formatting nested function calls. Examples:
531 foo({a, b, c}, {1, 2, 3});
533 llvm::Constant *Mask[] = {
534 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 0),
535 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 1),
536 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 2)};
538 This formatting scheme also makes it particularly easy to get predictable,
539 consistent, and automatic formatting with tools like `Clang Format`_.
541 .. _Clang Format: http://clang.llvm.org/docs/ClangFormat.html
543 Language and Compiler Issues
544 ----------------------------
546 Treat Compiler Warnings Like Errors
547 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
549 If your code has compiler warnings in it, something is wrong --- you aren't
550 casting values correctly, you have "questionable" constructs in your code, or
551 you are doing something legitimately wrong. Compiler warnings can cover up
552 legitimate errors in output and make dealing with a translation unit difficult.
554 It is not possible to prevent all warnings from all compilers, nor is it
555 desirable. Instead, pick a standard compiler (like ``gcc``) that provides a
556 good thorough set of warnings, and stick to it. At least in the case of
557 ``gcc``, it is possible to work around any spurious errors by changing the
558 syntax of the code slightly. For example, a warning that annoys me occurs when
559 I write code like this:
563 if (V = getValue()) {
567 ``gcc`` will warn me that I probably want to use the ``==`` operator, and that I
568 probably mistyped it. In most cases, I haven't, and I really don't want the
569 spurious errors. To fix this particular problem, I rewrite the code like
574 if ((V = getValue())) {
578 which shuts ``gcc`` up. Any ``gcc`` warning that annoys you can be fixed by
579 massaging the code appropriately.
584 In almost all cases, it is possible and within reason to write completely
585 portable code. If there are cases where it isn't possible to write portable
586 code, isolate it behind a well defined (and well documented) interface.
588 In practice, this means that you shouldn't assume much about the host compiler
589 (and Visual Studio tends to be the lowest common denominator). If advanced
590 features are used, they should only be an implementation detail of a library
591 which has a simple exposed API, and preferably be buried in ``libSystem``.
593 Do not use RTTI or Exceptions
594 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
596 In an effort to reduce code and executable size, LLVM does not use RTTI
597 (e.g. ``dynamic_cast<>;``) or exceptions. These two language features violate
598 the general C++ principle of *"you only pay for what you use"*, causing
599 executable bloat even if exceptions are never used in the code base, or if RTTI
600 is never used for a class. Because of this, we turn them off globally in the
603 That said, LLVM does make extensive use of a hand-rolled form of RTTI that use
604 templates like `isa<>, cast<>, and dyn_cast<> <ProgrammersManual.html#isa>`_.
605 This form of RTTI is opt-in and can be
606 :doc:`added to any class <HowToSetUpLLVMStyleRTTI>`. It is also
607 substantially more efficient than ``dynamic_cast<>``.
609 .. _static constructor:
611 Do not use Static Constructors
612 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
614 Static constructors and destructors (e.g. global variables whose types have a
615 constructor or destructor) should not be added to the code base, and should be
616 removed wherever possible. Besides `well known problems
617 <http://yosefk.com/c++fqa/ctors.html#fqa-10.12>`_ where the order of
618 initialization is undefined between globals in different source files, the
619 entire concept of static constructors is at odds with the common use case of
620 LLVM as a library linked into a larger application.
622 Consider the use of LLVM as a JIT linked into another application (perhaps for
623 `OpenGL, custom languages <http://llvm.org/Users.html>`_, `shaders in movies
624 <http://llvm.org/devmtg/2010-11/Gritz-OpenShadingLang.pdf>`_, etc). Due to the
625 design of static constructors, they must be executed at startup time of the
626 entire application, regardless of whether or how LLVM is used in that larger
627 application. There are two problems with this:
629 * The time to run the static constructors impacts startup time of applications
630 --- a critical time for GUI apps, among others.
632 * The static constructors cause the app to pull many extra pages of memory off
633 the disk: both the code for the constructor in each ``.o`` file and the small
634 amount of data that gets touched. In addition, touched/dirty pages put more
635 pressure on the VM system on low-memory machines.
637 We would really like for there to be zero cost for linking in an additional LLVM
638 target or other library into an application, but static constructors violate
641 That said, LLVM unfortunately does contain static constructors. It would be a
642 `great project <http://llvm.org/PR11944>`_ for someone to purge all static
643 constructors from LLVM, and then enable the ``-Wglobal-constructors`` warning
644 flag (when building with Clang) to ensure we do not regress in the future.
646 Use of ``class`` and ``struct`` Keywords
647 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
649 In C++, the ``class`` and ``struct`` keywords can be used almost
650 interchangeably. The only difference is when they are used to declare a class:
651 ``class`` makes all members private by default while ``struct`` makes all
652 members public by default.
654 Unfortunately, not all compilers follow the rules and some will generate
655 different symbols based on whether ``class`` or ``struct`` was used to declare
656 the symbol (e.g., MSVC). This can lead to problems at link time.
658 * All declarations and definitions of a given ``class`` or ``struct`` must use
659 the same keyword. For example:
665 // Breaks mangling in MSVC.
666 struct Foo { int Data; };
668 * As a rule of thumb, ``struct`` should be kept to structures where *all*
669 members are declared public.
673 // Foo feels like a class... this is strange.
679 int getData() const { return Data; }
680 void setData(int D) { Data = D; }
683 // Bar isn't POD, but it does look like a struct.
689 Do not use Braced Initializer Lists to Call a Constructor
690 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
692 In C++11 there is a "generalized initialization syntax" which allows calling
693 constructors using braced initializer lists. Do not use these to call
694 constructors with any interesting logic or if you care that you're calling some
695 *particular* constructor. Those should look like function calls using
696 parentheses rather than like aggregate initialization. Similarly, if you need
697 to explicitly name the type and call its constructor to create a temporary,
698 don't use a braced initializer list. Instead, use a braced initializer list
699 (without any type for temporaries) when doing aggregate initialization or
700 something notionally equivalent. Examples:
706 // Construct a Foo by reading data from the disk in the whizbang format, ...
707 Foo(std::string filename);
709 // Construct a Foo by looking up the Nth element of some global data ...
715 // The Foo constructor call is very deliberate, no braces.
716 std::fill(foo.begin(), foo.end(), Foo("name"));
718 // The pair is just being constructed like an aggregate, use braces.
719 bar_map.insert({my_key, my_value});
721 If you use a braced initializer list when initializing a variable, use an equals before the open curly brace:
725 int data[] = {0, 1, 2, 3};
727 Use ``auto`` Type Deduction to Make Code More Readable
728 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
730 Some are advocating a policy of "almost always ``auto``" in C++11, however LLVM
731 uses a more moderate stance. Use ``auto`` if and only if it makes the code more
732 readable or easier to maintain. Don't "almost always" use ``auto``, but do use
733 ``auto`` with initializers like ``cast<Foo>(...)`` or other places where the
734 type is already obvious from the context. Another time when ``auto`` works well
735 for these purposes is when the type would have been abstracted away anyways,
736 often behind a container's typedef such as ``std::vector<T>::iterator``.
738 Beware unnecessary copies with ``auto``
739 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
741 The convenience of ``auto`` makes it easy to forget that its default behavior
742 is a copy. Particularly in range-based ``for`` loops, careless copies are
745 As a rule of thumb, use ``const auto &`` unless you need to mutate or copy the
750 // Typically there's no reason to mutate or modify Val.
751 for (const auto &Val : Container) { observe(Val); }
753 // Remove the const if you need to modify Val.
754 for (auto &Val : Container) { Val.change(); }
756 // Remove the reference if you really want a new copy.
757 for (auto Val : Container) { Val.change(); saveSomewhere(Val); }
762 The High-Level Issues
763 ---------------------
765 A Public Header File **is** a Module
766 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
768 C++ doesn't do too well in the modularity department. There is no real
769 encapsulation or data hiding (unless you use expensive protocol classes), but it
770 is what we have to work with. When you write a public header file (in the LLVM
771 source tree, they live in the top level "``include``" directory), you are
772 defining a module of functionality.
774 Ideally, modules should be completely independent of each other, and their
775 header files should only ``#include`` the absolute minimum number of headers
776 possible. A module is not just a class, a function, or a namespace: it's a
777 collection of these that defines an interface. This interface may be several
778 functions, classes, or data structures, but the important issue is how they work
781 In general, a module should be implemented by one or more ``.cpp`` files. Each
782 of these ``.cpp`` files should include the header that defines their interface
783 first. This ensures that all of the dependences of the module header have been
784 properly added to the module header itself, and are not implicit. System
785 headers should be included after user headers for a translation unit.
787 .. _minimal list of #includes:
789 ``#include`` as Little as Possible
790 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
792 ``#include`` hurts compile time performance. Don't do it unless you have to,
793 especially in header files.
795 But wait! Sometimes you need to have the definition of a class to use it, or to
796 inherit from it. In these cases go ahead and ``#include`` that header file. Be
797 aware however that there are many cases where you don't need to have the full
798 definition of a class. If you are using a pointer or reference to a class, you
799 don't need the header file. If you are simply returning a class instance from a
800 prototyped function or method, you don't need it. In fact, for most cases, you
801 simply don't need the definition of a class. And not ``#include``\ing speeds up
804 It is easy to try to go too overboard on this recommendation, however. You
805 **must** include all of the header files that you are using --- you can include
806 them either directly or indirectly through another header file. To make sure
807 that you don't accidentally forget to include a header file in your module
808 header, make sure to include your module header **first** in the implementation
809 file (as mentioned above). This way there won't be any hidden dependencies that
810 you'll find out about later.
812 Keep "Internal" Headers Private
813 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
815 Many modules have a complex implementation that causes them to use more than one
816 implementation (``.cpp``) file. It is often tempting to put the internal
817 communication interface (helper classes, extra functions, etc) in the public
818 module header file. Don't do this!
820 If you really need to do something like this, put a private header file in the
821 same directory as the source files, and include it locally. This ensures that
822 your private interface remains private and undisturbed by outsiders.
826 It's okay to put extra implementation methods in a public class itself. Just
827 make them private (or protected) and all is well.
831 Use Early Exits and ``continue`` to Simplify Code
832 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
834 When reading code, keep in mind how much state and how many previous decisions
835 have to be remembered by the reader to understand a block of code. Aim to
836 reduce indentation where possible when it doesn't make it more difficult to
837 understand the code. One great way to do this is by making use of early exits
838 and the ``continue`` keyword in long loops. As an example of using an early
839 exit from a function, consider this "bad" code:
843 Value *doSomething(Instruction *I) {
844 if (!isa<TerminatorInst>(I) &&
845 I->hasOneUse() && doOtherThing(I)) {
846 ... some long code ....
852 This code has several problems if the body of the ``'if'`` is large. When
853 you're looking at the top of the function, it isn't immediately clear that this
854 *only* does interesting things with non-terminator instructions, and only
855 applies to things with the other predicates. Second, it is relatively difficult
856 to describe (in comments) why these predicates are important because the ``if``
857 statement makes it difficult to lay out the comments. Third, when you're deep
858 within the body of the code, it is indented an extra level. Finally, when
859 reading the top of the function, it isn't clear what the result is if the
860 predicate isn't true; you have to read to the end of the function to know that
863 It is much preferred to format the code like this:
867 Value *doSomething(Instruction *I) {
868 // Terminators never need 'something' done to them because ...
869 if (isa<TerminatorInst>(I))
872 // We conservatively avoid transforming instructions with multiple uses
873 // because goats like cheese.
877 // This is really just here for example.
878 if (!doOtherThing(I))
881 ... some long code ....
884 This fixes these problems. A similar problem frequently happens in ``for``
885 loops. A silly example is something like this:
889 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
890 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
891 Value *LHS = BO->getOperand(0);
892 Value *RHS = BO->getOperand(1);
899 When you have very, very small loops, this sort of structure is fine. But if it
900 exceeds more than 10-15 lines, it becomes difficult for people to read and
901 understand at a glance. The problem with this sort of code is that it gets very
902 nested very quickly. Meaning that the reader of the code has to keep a lot of
903 context in their brain to remember what is going immediately on in the loop,
904 because they don't know if/when the ``if`` conditions will have ``else``\s etc.
905 It is strongly preferred to structure the loop like this:
909 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
910 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
913 Value *LHS = BO->getOperand(0);
914 Value *RHS = BO->getOperand(1);
915 if (LHS == RHS) continue;
920 This has all the benefits of using early exits for functions: it reduces nesting
921 of the loop, it makes it easier to describe why the conditions are true, and it
922 makes it obvious to the reader that there is no ``else`` coming up that they
923 have to push context into their brain for. If a loop is large, this can be a
924 big understandability win.
926 Don't use ``else`` after a ``return``
927 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
929 For similar reasons above (reduction of indentation and easier reading), please
930 do not use ``'else'`` or ``'else if'`` after something that interrupts control
931 flow --- like ``return``, ``break``, ``continue``, ``goto``, etc. For
932 example, this is *bad*:
938 Type = Context.getsigjmp_bufType();
940 Error = ASTContext::GE_Missing_sigjmp_buf;
946 Type = Context.getjmp_bufType();
948 Error = ASTContext::GE_Missing_jmp_buf;
956 It is better to write it like this:
962 Type = Context.getsigjmp_bufType();
964 Error = ASTContext::GE_Missing_sigjmp_buf;
968 Type = Context.getjmp_bufType();
970 Error = ASTContext::GE_Missing_jmp_buf;
976 Or better yet (in this case) as:
982 Type = Context.getsigjmp_bufType();
984 Type = Context.getjmp_bufType();
987 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
988 ASTContext::GE_Missing_jmp_buf;
993 The idea is to reduce indentation and the amount of code you have to keep track
994 of when reading the code.
996 Turn Predicate Loops into Predicate Functions
997 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
999 It is very common to write small loops that just compute a boolean value. There
1000 are a number of ways that people commonly write these, but an example of this
1005 bool FoundFoo = false;
1006 for (unsigned I = 0, E = BarList.size(); I != E; ++I)
1007 if (BarList[I]->isFoo()) {
1016 This sort of code is awkward to write, and is almost always a bad sign. Instead
1017 of this sort of loop, we strongly prefer to use a predicate function (which may
1018 be `static`_) that uses `early exits`_ to compute the predicate. We prefer the
1019 code to be structured like this:
1023 /// \returns true if the specified list has an element that is a foo.
1024 static bool containsFoo(const std::vector<Bar*> &List) {
1025 for (unsigned I = 0, E = List.size(); I != E; ++I)
1026 if (List[I]->isFoo())
1032 if (containsFoo(BarList)) {
1036 There are many reasons for doing this: it reduces indentation and factors out
1037 code which can often be shared by other code that checks for the same predicate.
1038 More importantly, it *forces you to pick a name* for the function, and forces
1039 you to write a comment for it. In this silly example, this doesn't add much
1040 value. However, if the condition is complex, this can make it a lot easier for
1041 the reader to understand the code that queries for this predicate. Instead of
1042 being faced with the in-line details of how we check to see if the BarList
1043 contains a foo, we can trust the function name and continue reading with better
1046 The Low-Level Issues
1047 --------------------
1049 Name Types, Functions, Variables, and Enumerators Properly
1050 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1052 Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
1053 enough how important it is to use *descriptive* names. Pick names that match
1054 the semantics and role of the underlying entities, within reason. Avoid
1055 abbreviations unless they are well known. After picking a good name, make sure
1056 to use consistent capitalization for the name, as inconsistency requires clients
1057 to either memorize the APIs or to look it up to find the exact spelling.
1059 In general, names should be in camel case (e.g. ``TextFileReader`` and
1060 ``isLValue()``). Different kinds of declarations have different rules:
1062 * **Type names** (including classes, structs, enums, typedefs, etc) should be
1063 nouns and start with an upper-case letter (e.g. ``TextFileReader``).
1065 * **Variable names** should be nouns (as they represent state). The name should
1066 be camel case, and start with an upper case letter (e.g. ``Leader`` or
1069 * **Function names** should be verb phrases (as they represent actions), and
1070 command-like function should be imperative. The name should be camel case,
1071 and start with a lower case letter (e.g. ``openFile()`` or ``isFoo()``).
1073 * **Enum declarations** (e.g. ``enum Foo {...}``) are types, so they should
1074 follow the naming conventions for types. A common use for enums is as a
1075 discriminator for a union, or an indicator of a subclass. When an enum is
1076 used for something like this, it should have a ``Kind`` suffix
1077 (e.g. ``ValueKind``).
1079 * **Enumerators** (e.g. ``enum { Foo, Bar }``) and **public member variables**
1080 should start with an upper-case letter, just like types. Unless the
1081 enumerators are defined in their own small namespace or inside a class,
1082 enumerators should have a prefix corresponding to the enum declaration name.
1083 For example, ``enum ValueKind { ... };`` may contain enumerators like
1084 ``VK_Argument``, ``VK_BasicBlock``, etc. Enumerators that are just
1085 convenience constants are exempt from the requirement for a prefix. For
1095 As an exception, classes that mimic STL classes can have member names in STL's
1096 style of lower-case words separated by underscores (e.g. ``begin()``,
1097 ``push_back()``, and ``empty()``). Classes that provide multiple
1098 iterators should add a singular prefix to ``begin()`` and ``end()``
1099 (e.g. ``global_begin()`` and ``use_begin()``).
1101 Here are some examples of good and bad names:
1105 class VehicleMaker {
1107 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
1108 Factory<Tire> Factory; // Better.
1109 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
1110 // kind of factories.
1113 Vehicle MakeVehicle(VehicleType Type) {
1114 VehicleMaker M; // Might be OK if having a short life-span.
1115 Tire Tmp1 = M.makeTire(); // Bad -- 'Tmp1' provides no information.
1116 Light Headlight = M.makeLight("head"); // Good -- descriptive.
1123 Use the "``assert``" macro to its fullest. Check all of your preconditions and
1124 assumptions, you never know when a bug (not necessarily even yours) might be
1125 caught early by an assertion, which reduces debugging time dramatically. The
1126 "``<cassert>``" header file is probably already included by the header files you
1127 are using, so it doesn't cost anything to use it.
1129 To further assist with debugging, make sure to put some kind of error message in
1130 the assertion statement, which is printed if the assertion is tripped. This
1131 helps the poor debugger make sense of why an assertion is being made and
1132 enforced, and hopefully what to do about it. Here is one complete example:
1136 inline Value *getOperand(unsigned I) {
1137 assert(I < Operands.size() && "getOperand() out of range!");
1141 Here are more examples:
1145 assert(Ty->isPointerType() && "Can't allocate a non-pointer type!");
1147 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
1149 assert(idx < getNumSuccessors() && "Successor # out of range!");
1151 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
1153 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
1157 In the past, asserts were used to indicate a piece of code that should not be
1158 reached. These were typically of the form:
1162 assert(0 && "Invalid radix for integer literal");
1164 This has a few issues, the main one being that some compilers might not
1165 understand the assertion, or warn about a missing return in builds where
1166 assertions are compiled out.
1168 Today, we have something much better: ``llvm_unreachable``:
1172 llvm_unreachable("Invalid radix for integer literal");
1174 When assertions are enabled, this will print the message if it's ever reached
1175 and then exit the program. When assertions are disabled (i.e. in release
1176 builds), ``llvm_unreachable`` becomes a hint to compilers to skip generating
1177 code for this branch. If the compiler does not support this, it will fall back
1178 to the "abort" implementation.
1180 Another issue is that values used only by assertions will produce an "unused
1181 value" warning when assertions are disabled. For example, this code will warn:
1185 unsigned Size = V.size();
1186 assert(Size > 42 && "Vector smaller than it should be");
1188 bool NewToSet = Myset.insert(Value);
1189 assert(NewToSet && "The value shouldn't be in the set yet");
1191 These are two interesting different cases. In the first case, the call to
1192 ``V.size()`` is only useful for the assert, and we don't want it executed when
1193 assertions are disabled. Code like this should move the call into the assert
1194 itself. In the second case, the side effects of the call must happen whether
1195 the assert is enabled or not. In this case, the value should be cast to void to
1196 disable the warning. To be specific, it is preferred to write the code like
1201 assert(V.size() > 42 && "Vector smaller than it should be");
1203 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1204 assert(NewToSet && "The value shouldn't be in the set yet");
1206 Do Not Use ``using namespace std``
1207 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1209 In LLVM, we prefer to explicitly prefix all identifiers from the standard
1210 namespace with an "``std::``" prefix, rather than rely on "``using namespace
1213 In header files, adding a ``'using namespace XXX'`` directive pollutes the
1214 namespace of any source file that ``#include``\s the header. This is clearly a
1217 In implementation files (e.g. ``.cpp`` files), the rule is more of a stylistic
1218 rule, but is still important. Basically, using explicit namespace prefixes
1219 makes the code **clearer**, because it is immediately obvious what facilities
1220 are being used and where they are coming from. And **more portable**, because
1221 namespace clashes cannot occur between LLVM code and other namespaces. The
1222 portability rule is important because different standard library implementations
1223 expose different symbols (potentially ones they shouldn't), and future revisions
1224 to the C++ standard will add more symbols to the ``std`` namespace. As such, we
1225 never use ``'using namespace std;'`` in LLVM.
1227 The exception to the general rule (i.e. it's not an exception for the ``std``
1228 namespace) is for implementation files. For example, all of the code in the
1229 LLVM project implements code that lives in the 'llvm' namespace. As such, it is
1230 ok, and actually clearer, for the ``.cpp`` files to have a ``'using namespace
1231 llvm;'`` directive at the top, after the ``#include``\s. This reduces
1232 indentation in the body of the file for source editors that indent based on
1233 braces, and keeps the conceptual context cleaner. The general form of this rule
1234 is that any ``.cpp`` file that implements code in any namespace may use that
1235 namespace (and its parents'), but should not use any others.
1237 Provide a Virtual Method Anchor for Classes in Headers
1238 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1240 If a class is defined in a header file and has a vtable (either it has virtual
1241 methods or it derives from classes with virtual methods), it must always have at
1242 least one out-of-line virtual method in the class. Without this, the compiler
1243 will copy the vtable and RTTI into every ``.o`` file that ``#include``\s the
1244 header, bloating ``.o`` file sizes and increasing link times.
1246 Don't use default labels in fully covered switches over enumerations
1247 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1249 ``-Wswitch`` warns if a switch, without a default label, over an enumeration
1250 does not cover every enumeration value. If you write a default label on a fully
1251 covered switch over an enumeration then the ``-Wswitch`` warning won't fire
1252 when new elements are added to that enumeration. To help avoid adding these
1253 kinds of defaults, Clang has the warning ``-Wcovered-switch-default`` which is
1254 off by default but turned on when building LLVM with a version of Clang that
1255 supports the warning.
1257 A knock-on effect of this stylistic requirement is that when building LLVM with
1258 GCC you may get warnings related to "control may reach end of non-void function"
1259 if you return from each case of a covered switch-over-enum because GCC assumes
1260 that the enum expression may take any representable value, not just those of
1261 individual enumerators. To suppress this warning, use ``llvm_unreachable`` after
1264 Use ``LLVM_DELETED_FUNCTION`` to mark uncallable methods
1265 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1267 Prior to C++11, a common pattern to make a class uncopyable was to declare an
1268 unimplemented copy constructor and copy assignment operator and make them
1269 private. This would give a compiler error for accessing a private method or a
1270 linker error because it wasn't implemented.
1272 With C++11, we can mark methods that won't be implemented with ``= delete``.
1273 This will trigger a much better error message and tell the compiler that the
1274 method will never be implemented. This enables other checks like
1275 ``-Wunused-private-field`` to run correctly on classes that contain these
1278 To maintain compatibility with C++03, ``LLVM_DELETED_FUNCTION`` should be used
1279 which will expand to ``= delete`` if the compiler supports it. These methods
1280 should still be declared private. Example of the uncopyable pattern:
1286 DontCopy(const DontCopy&) LLVM_DELETED_FUNCTION;
1287 DontCopy &operator =(const DontCopy&) LLVM_DELETED_FUNCTION;
1292 Don't evaluate ``end()`` every time through a loop
1293 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1295 Because C++ doesn't have a standard "``foreach``" loop (though it can be
1296 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1297 loops that manually iterate from begin to end on a variety of containers or
1298 through other data structures. One common mistake is to write a loop in this
1303 BasicBlock *BB = ...
1304 for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I)
1307 The problem with this construct is that it evaluates "``BB->end()``" every time
1308 through the loop. Instead of writing the loop like this, we strongly prefer
1309 loops to be written so that they evaluate it once before the loop starts. A
1310 convenient way to do this is like so:
1314 BasicBlock *BB = ...
1315 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
1318 The observant may quickly point out that these two loops may have different
1319 semantics: if the container (a basic block in this case) is being mutated, then
1320 "``BB->end()``" may change its value every time through the loop and the second
1321 loop may not in fact be correct. If you actually do depend on this behavior,
1322 please write the loop in the first form and add a comment indicating that you
1323 did it intentionally.
1325 Why do we prefer the second form (when correct)? Writing the loop in the first
1326 form has two problems. First it may be less efficient than evaluating it at the
1327 start of the loop. In this case, the cost is probably minor --- a few extra
1328 loads every time through the loop. However, if the base expression is more
1329 complex, then the cost can rise quickly. I've seen loops where the end
1330 expression was actually something like: "``SomeMap[X]->end()``" and map lookups
1331 really aren't cheap. By writing it in the second form consistently, you
1332 eliminate the issue entirely and don't even have to think about it.
1334 The second (even bigger) issue is that writing the loop in the first form hints
1335 to the reader that the loop is mutating the container (a fact that a comment
1336 would handily confirm!). If you write the loop in the second form, it is
1337 immediately obvious without even looking at the body of the loop that the
1338 container isn't being modified, which makes it easier to read the code and
1339 understand what it does.
1341 While the second form of the loop is a few extra keystrokes, we do strongly
1344 ``#include <iostream>`` is Forbidden
1345 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1347 The use of ``#include <iostream>`` in library files is hereby **forbidden**,
1348 because many common implementations transparently inject a `static constructor`_
1349 into every translation unit that includes it.
1351 Note that using the other stream headers (``<sstream>`` for example) is not
1352 problematic in this regard --- just ``<iostream>``. However, ``raw_ostream``
1353 provides various APIs that are better performing for almost every use than
1354 ``std::ostream`` style APIs.
1358 New code should always use `raw_ostream`_ for writing, or the
1359 ``llvm::MemoryBuffer`` API for reading files.
1366 LLVM includes a lightweight, simple, and efficient stream implementation in
1367 ``llvm/Support/raw_ostream.h``, which provides all of the common features of
1368 ``std::ostream``. All new code should use ``raw_ostream`` instead of
1371 Unlike ``std::ostream``, ``raw_ostream`` is not a template and can be forward
1372 declared as ``class raw_ostream``. Public headers should generally not include
1373 the ``raw_ostream`` header, but use forward declarations and constant references
1374 to ``raw_ostream`` instances.
1379 The ``std::endl`` modifier, when used with ``iostreams`` outputs a newline to
1380 the output stream specified. In addition to doing this, however, it also
1381 flushes the output stream. In other words, these are equivalent:
1385 std::cout << std::endl;
1386 std::cout << '\n' << std::flush;
1388 Most of the time, you probably have no reason to flush the output stream, so
1389 it's better to use a literal ``'\n'``.
1391 Don't use ``inline`` when defining a function in a class definition
1392 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1394 A member function defined in a class definition is implicitly inline, so don't
1395 put the ``inline`` keyword in this case.
1422 This section describes preferred low-level formatting guidelines along with
1423 reasoning on why we prefer them.
1425 Spaces Before Parentheses
1426 ^^^^^^^^^^^^^^^^^^^^^^^^^
1428 We prefer to put a space before an open parenthesis only in control flow
1429 statements, but not in normal function call expressions and function-like
1430 macros. For example, this is good:
1435 for (I = 0; I != 100; ++I) ...
1436 while (LLVMRocks) ...
1439 assert(3 != 4 && "laws of math are failing me");
1441 A = foo(42, 92) + bar(X);
1448 for(I = 0; I != 100; ++I) ...
1449 while(LLVMRocks) ...
1452 assert (3 != 4 && "laws of math are failing me");
1454 A = foo (42, 92) + bar (X);
1456 The reason for doing this is not completely arbitrary. This style makes control
1457 flow operators stand out more, and makes expressions flow better. The function
1458 call operator binds very tightly as a postfix operator. Putting a space after a
1459 function name (as in the last example) makes it appear that the code might bind
1460 the arguments of the left-hand-side of a binary operator with the argument list
1461 of a function and the name of the right side. More specifically, it is easy to
1462 misread the "``A``" example as:
1466 A = foo ((42, 92) + bar) (X);
1468 when skimming through the code. By avoiding a space in a function, we avoid
1469 this misinterpretation.
1474 Hard fast rule: Preincrement (``++X``) may be no slower than postincrement
1475 (``X++``) and could very well be a lot faster than it. Use preincrementation
1478 The semantics of postincrement include making a copy of the value being
1479 incremented, returning it, and then preincrementing the "work value". For
1480 primitive types, this isn't a big deal. But for iterators, it can be a huge
1481 issue (for example, some iterators contains stack and set objects in them...
1482 copying an iterator could invoke the copy ctor's of these as well). In general,
1483 get in the habit of always using preincrement, and you won't have a problem.
1486 Namespace Indentation
1487 ^^^^^^^^^^^^^^^^^^^^^
1489 In general, we strive to reduce indentation wherever possible. This is useful
1490 because we want code to `fit into 80 columns`_ without wrapping horribly, but
1491 also because it makes it easier to understand the code. To facilitate this and
1492 avoid some insanely deep nesting on occasion, don't indent namespaces. If it
1493 helps readability, feel free to add a comment indicating what namespace is
1494 being closed by a ``}``. For example:
1499 namespace knowledge {
1501 /// This class represents things that Smith can have an intimate
1502 /// understanding of and contains the data associated with it.
1506 explicit Grokable() { ... }
1507 virtual ~Grokable() = 0;
1513 } // end namespace knowledge
1514 } // end namespace llvm
1517 Feel free to skip the closing comment when the namespace being closed is
1518 obvious for any reason. For example, the outer-most namespace in a header file
1519 is rarely a source of confusion. But namespaces both anonymous and named in
1520 source files that are being closed half way through the file probably could use
1525 Anonymous Namespaces
1526 ^^^^^^^^^^^^^^^^^^^^
1528 After talking about namespaces in general, you may be wondering about anonymous
1529 namespaces in particular. Anonymous namespaces are a great language feature
1530 that tells the C++ compiler that the contents of the namespace are only visible
1531 within the current translation unit, allowing more aggressive optimization and
1532 eliminating the possibility of symbol name collisions. Anonymous namespaces are
1533 to C++ as "static" is to C functions and global variables. While "``static``"
1534 is available in C++, anonymous namespaces are more general: they can make entire
1535 classes private to a file.
1537 The problem with anonymous namespaces is that they naturally want to encourage
1538 indentation of their body, and they reduce locality of reference: if you see a
1539 random function definition in a C++ file, it is easy to see if it is marked
1540 static, but seeing if it is in an anonymous namespace requires scanning a big
1543 Because of this, we have a simple guideline: make anonymous namespaces as small
1544 as possible, and only use them for class declarations. For example, this is
1554 bool operator<(const char *RHS) const;
1556 } // end anonymous namespace
1558 static void runHelper() {
1562 bool StringSort::operator<(const char *RHS) const {
1576 bool operator<(const char *RHS) const;
1583 bool StringSort::operator<(const char *RHS) const {
1587 } // end anonymous namespace
1589 This is bad specifically because if you're looking at "``runHelper``" in the middle
1590 of a large C++ file, that you have no immediate way to tell if it is local to
1591 the file. When it is marked static explicitly, this is immediately obvious.
1592 Also, there is no reason to enclose the definition of "``operator<``" in the
1593 namespace just because it was declared there.
1598 A lot of these comments and recommendations have been culled from other sources.
1599 Two particularly important books for our work are:
1602 <http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876>`_
1603 by Scott Meyers. Also interesting and useful are "More Effective C++" and
1604 "Effective STL" by the same author.
1606 #. `Large-Scale C++ Software Design
1607 <http://www.amazon.com/Large-Scale-Software-Design-John-Lakos/dp/0201633620/ref=sr_1_1>`_
1610 If you get some free time, and you haven't read them: do so, you might learn