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 and Clang are currently written using C++98/03 conforming code, with
60 selective use of C++11 features when they are present in the toolchain.
61 Projects like LLD and LLDB are already heavily using C++11 features.
63 However, LLVM and Clange are also in the process of switching to use C++11 as
64 the base line for standards conformance. Once completed, the same standard
65 baseline will be used for LLVM, Clang, and LLD. LLDB is pushing forward much
66 more aggressively and has their own baseline.
71 Use the C++ standard library facilities whenever they are available for
72 a particular task. LLVM and related projects emphasize and rely on the standard
73 library facilities for as much as possible. Common support libraries providing
74 functionality missing from the standard library for which there are standard
75 interfaces or active work on adding standard interfaces will often be
76 implemented in the LLVM namespace following the expected standard interface.
78 There are some exceptions such as the standard I/O streams library which are
79 avoided. Also, there is much more detailed information on these subjects in the
80 `Programmer's Manual`_.
82 .. _Programmer's Manual:
83 http://llvm.org/docs/ProgrammersManual.html
85 Supported C++11 Language and Library Features
86 -------------------------------------------
89 This section is written to reflect the expected state **AFTER** the
90 transition to C++11 is complete for the LLVM source tree.
92 While LLVM, Clang, and LLD use C++11, not all features are available in all of
93 the toolchains which we support. The set of features supported for use in LLVM
94 is the intersection of those supported in MSVC 2012, GCC 4.7, and Clang 3.1.
95 The ultimate definition of this set is what build bots with those respective
96 toolchains accept. Don't argue with the build bots.
98 Each toolchain provides a good reference for what it accepts:
100 * Clang: http://clang.llvm.org/cxx_status.html
101 * GCC: http://gcc.gnu.org/projects/cxx0x.html
102 * MSVC: http://msdn.microsoft.com/en-us/library/hh567368.aspx
104 In most cases, the MSVC list will be the dominating factor. Here is a summary
105 of the features that are expected to work. Features not on this list are
106 unlikely to be supported by our host compilers.
108 * Rvalue references: N2118_
110 * But *not* Rvalue references for ``*this`` or member qualifiers (N2439_)
112 * Static assert: N1720_
113 * ``auto`` type deduction: N1984_, N1737_
114 * Trailing return types: N2541_
116 * ``decltype``: N2343_
117 * Nested closing right angle brackets: N1757_
118 * Extern templates: N1987_
119 * ``nullptr``: N2431_
120 * Strongly-typed and forward declarable enums: N2347_, N2764_
121 * Local and unnamed types as template arguments: N2657_
122 * Range-based for-loop: N2930_
123 * ``override`` and ``final``: N2928_, N3206_, N3272_
124 * Atomic operations and the C++11 memory model: N2429_
126 .. _N2118: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2118.html
127 .. _N2439: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2439.htm
128 .. _N1720: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1720.html
129 .. _N1984: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1984.pdf
130 .. _N1737: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1737.pdf
131 .. _N2541: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2541.htm
132 .. _N2927: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2927.pdf
133 .. _N2343: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2343.pdf
134 .. _N1757: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1757.html
135 .. _N1987: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1987.htm
136 .. _N2431: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2431.pdf
137 .. _N2347: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2347.pdf
138 .. _N2764: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2764.pdf
139 .. _N2657: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm
140 .. _N2930: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2930.html
141 .. _N2928: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2928.htm
142 .. _N3206: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3206.htm
143 .. _N3272: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3272.htm
144 .. _N2429: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2429.htm
146 The supported features in the C++11 standard libraries are less well tracked,
147 but also much greater. Most of the standard libraries implement most of C++11's
148 library. The most likely lowest common denominator is Linux support. For
149 libc++, the support is just poorly tested and undocumented but expected to be
150 largely complete. YMMV. For libstdc++, the support is documented in detail in
151 `the libstdc++ manual`_. There are some very minor missing facilities that are
152 unlikely to be common problems, and there are a few larger gaps that are worth
155 * Not all of the type traits are implemented
156 * No regular expression library.
157 * While most of the atomics library is well implemented, the fences are
158 missing. Fortunately, they are rarely needed.
159 * The locale support is incomplete.
161 Your best option if you cannot test on a Linux system is to minimize your use
162 of these features, and watch the Linux build bots to find out if your usage
163 triggered a bug. For example if you hit a type trait which doesn't work, we can
164 then add support to LLVM's traits header to emulate it.
166 .. _the libstdc++ manual:
167 http://gcc.gnu.org/onlinedocs/gcc-4.7.3/libstdc++/manual/manual/status.html#status.iso.2011
169 Mechanical Source Issues
170 ========================
172 Source Code Formatting
173 ----------------------
178 Comments are one critical part of readability and maintainability. Everyone
179 knows they should comment their code, and so should you. When writing comments,
180 write them as English prose, which means they should use proper capitalization,
181 punctuation, etc. Aim to describe what the code is trying to do and why, not
182 *how* it does it at a micro level. Here are a few critical things to document:
184 .. _header file comment:
189 Every source file should have a header on it that describes the basic purpose of
190 the file. If a file does not have a header, it should not be checked into the
191 tree. The standard header looks like this:
195 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
197 // The LLVM Compiler Infrastructure
199 // This file is distributed under the University of Illinois Open Source
200 // License. See LICENSE.TXT for details.
202 //===----------------------------------------------------------------------===//
205 /// \brief This file contains the declaration of the Instruction class, which is
206 /// the base class for all of the VM instructions.
208 //===----------------------------------------------------------------------===//
210 A few things to note about this particular format: The "``-*- C++ -*-``" string
211 on the first line is there to tell Emacs that the source file is a C++ file, not
212 a C file (Emacs assumes ``.h`` files are C files by default).
216 This tag is not necessary in ``.cpp`` files. The name of the file is also
217 on the first line, along with a very short description of the purpose of the
218 file. This is important when printing out code and flipping though lots of
221 The next section in the file is a concise note that defines the license that the
222 file is released under. This makes it perfectly clear what terms the source
223 code can be distributed under and should not be modified in any way.
225 The main body is a ``doxygen`` comment describing the purpose of the file. It
226 should have a ``\brief`` command that describes the file in one or two
227 sentences. Any additional information should be separated by a blank line. If
228 an algorithm is being implemented or something tricky is going on, a reference
229 to the paper where it is published should be included, as well as any notes or
230 *gotchas* in the code to watch out for.
235 Classes are one fundamental part of a good object oriented design. As such, a
236 class definition should have a comment block that explains what the class is
237 used for and how it works. Every non-trivial class is expected to have a
238 ``doxygen`` comment block.
243 Methods defined in a class (as well as any global functions) should also be
244 documented properly. A quick note about what it does and a description of the
245 borderline behaviour is all that is necessary here (unless something
246 particularly tricky or insidious is going on). The hope is that people can
247 figure out how to use your interfaces without reading the code itself.
249 Good things to talk about here are what happens when something unexpected
250 happens: does the method return null? Abort? Format your hard disk?
255 In general, prefer C++ style (``//``) comments. They take less space, require
256 less typing, don't have nesting problems, etc. There are a few cases when it is
257 useful to use C style (``/* */``) comments however:
259 #. When writing C code: Obviously if you are writing C code, use C style
262 #. When writing a header file that may be ``#include``\d by a C source file.
264 #. When writing a source file that is used by a tool that only accepts C style
267 To comment out a large block of code, use ``#if 0`` and ``#endif``. These nest
268 properly and are better behaved in general than C style comments.
270 Doxygen Use in Documentation Comments
271 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
273 Use the ``\file`` command to turn the standard file header into a file-level
276 Include descriptive ``\brief`` paragraphs for all public interfaces (public
277 classes, member and non-member functions). Explain API use and purpose in
278 ``\brief`` paragraphs, don't just restate the information that can be inferred
279 from the API name. Put detailed discussion into separate paragraphs.
281 To refer to parameter names inside a paragraph, use the ``\p name`` command.
282 Don't use the ``\arg name`` command since it starts a new paragraph that
283 contains documentation for the parameter.
285 Wrap non-inline code examples in ``\code ... \endcode``.
287 To document a function parameter, start a new paragraph with the
288 ``\param name`` command. If the parameter is used as an out or an in/out
289 parameter, use the ``\param [out] name`` or ``\param [in,out] name`` command,
292 To describe function return value, start a new paragraph with the ``\returns``
295 A minimal documentation comment:
299 /// \brief Does foo and bar.
300 void fooBar(bool Baz);
302 A documentation comment that uses all Doxygen features in a preferred way:
306 /// \brief Does foo and bar.
308 /// Does not do foo the usual way if \p Baz is true.
312 /// fooBar(false, "quux", Res);
315 /// \param Quux kind of foo to do.
316 /// \param [out] Result filled with bar sequence on foo success.
318 /// \returns true on success.
319 bool fooBar(bool Baz, StringRef Quux, std::vector<int> &Result);
321 Don't duplicate the documentation comment in the header file and in the
322 implementation file. Put the documentation comments for public APIs into the
323 header file. Documentation comments for private APIs can go to the
324 implementation file. In any case, implementation files can include additional
325 comments (not necessarily in Doxygen markup) to explain implementation details
328 Don't duplicate function or class name at the beginning of the comment.
329 For humans it is obvious which function or class is being documented;
330 automatic documentation processing tools are smart enough to bind the comment
331 to the correct declaration.
339 /// Something - An abstraction for some complicated thing.
342 /// fooBar - Does foo and bar.
348 /// fooBar - Does foo and bar.
349 void Something::fooBar() { ... }
357 /// \brief An abstraction for some complicated thing.
360 /// \brief Does foo and bar.
366 // Builds a B-tree in order to do foo. See paper by...
367 void Something::fooBar() { ... }
369 It is not required to use additional Doxygen features, but sometimes it might
370 be a good idea to do so.
374 * adding comments to any narrow namespace containing a collection of
375 related functions or types;
377 * using top-level groups to organize a collection of related functions at
378 namespace scope where the grouping is smaller than the namespace;
380 * using member groups and additional comments attached to member
381 groups to organize within a class.
388 /// \name Functions that do Foo.
399 Immediately after the `header file comment`_ (and include guards if working on a
400 header file), the `minimal list of #includes`_ required by the file should be
401 listed. We prefer these ``#include``\s to be listed in this order:
403 .. _Main Module Header:
404 .. _Local/Private Headers:
406 #. Main Module Header
407 #. Local/Private Headers
409 #. System ``#include``\s
411 and each category should be sorted lexicographically by the full path.
413 The `Main Module Header`_ file applies to ``.cpp`` files which implement an
414 interface defined by a ``.h`` file. This ``#include`` should always be included
415 **first** regardless of where it lives on the file system. By including a
416 header file first in the ``.cpp`` files that implement the interfaces, we ensure
417 that the header does not have any hidden dependencies which are not explicitly
418 ``#include``\d in the header, but should be. It is also a form of documentation
419 in the ``.cpp`` file to indicate where the interfaces it implements are defined.
421 .. _fit into 80 columns:
426 Write your code to fit within 80 columns of text. This helps those of us who
427 like to print out code and look at your code in an ``xterm`` without resizing
430 The longer answer is that there must be some limit to the width of the code in
431 order to reasonably allow developers to have multiple files side-by-side in
432 windows on a modest display. If you are going to pick a width limit, it is
433 somewhat arbitrary but you might as well pick something standard. Going with 90
434 columns (for example) instead of 80 columns wouldn't add any significant value
435 and would be detrimental to printing out code. Also many other projects have
436 standardized on 80 columns, so some people have already configured their editors
437 for it (vs something else, like 90 columns).
439 This is one of many contentious issues in coding standards, but it is not up for
442 Use Spaces Instead of Tabs
443 ^^^^^^^^^^^^^^^^^^^^^^^^^^
445 In all cases, prefer spaces to tabs in source files. People have different
446 preferred indentation levels, and different styles of indentation that they
447 like; this is fine. What isn't fine is that different editors/viewers expand
448 tabs out to different tab stops. This can cause your code to look completely
449 unreadable, and it is not worth dealing with.
451 As always, follow the `Golden Rule`_ above: follow the style of
452 existing code if you are modifying and extending it. If you like four spaces of
453 indentation, **DO NOT** do that in the middle of a chunk of code with two spaces
454 of indentation. Also, do not reindent a whole source file: it makes for
455 incredible diffs that are absolutely worthless.
457 Indent Code Consistently
458 ^^^^^^^^^^^^^^^^^^^^^^^^
460 Okay, in your first year of programming you were told that indentation is
461 important. If you didn't believe and internalize this then, now is the time.
467 Treat Compiler Warnings Like Errors
468 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
470 If your code has compiler warnings in it, something is wrong --- you aren't
471 casting values correctly, you have "questionable" constructs in your code, or
472 you are doing something legitimately wrong. Compiler warnings can cover up
473 legitimate errors in output and make dealing with a translation unit difficult.
475 It is not possible to prevent all warnings from all compilers, nor is it
476 desirable. Instead, pick a standard compiler (like ``gcc``) that provides a
477 good thorough set of warnings, and stick to it. At least in the case of
478 ``gcc``, it is possible to work around any spurious errors by changing the
479 syntax of the code slightly. For example, a warning that annoys me occurs when
480 I write code like this:
484 if (V = getValue()) {
488 ``gcc`` will warn me that I probably want to use the ``==`` operator, and that I
489 probably mistyped it. In most cases, I haven't, and I really don't want the
490 spurious errors. To fix this particular problem, I rewrite the code like
495 if ((V = getValue())) {
499 which shuts ``gcc`` up. Any ``gcc`` warning that annoys you can be fixed by
500 massaging the code appropriately.
505 In almost all cases, it is possible and within reason to write completely
506 portable code. If there are cases where it isn't possible to write portable
507 code, isolate it behind a well defined (and well documented) interface.
509 In practice, this means that you shouldn't assume much about the host compiler
510 (and Visual Studio tends to be the lowest common denominator). If advanced
511 features are used, they should only be an implementation detail of a library
512 which has a simple exposed API, and preferably be buried in ``libSystem``.
514 Do not use RTTI or Exceptions
515 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
517 In an effort to reduce code and executable size, LLVM does not use RTTI
518 (e.g. ``dynamic_cast<>;``) or exceptions. These two language features violate
519 the general C++ principle of *"you only pay for what you use"*, causing
520 executable bloat even if exceptions are never used in the code base, or if RTTI
521 is never used for a class. Because of this, we turn them off globally in the
524 That said, LLVM does make extensive use of a hand-rolled form of RTTI that use
525 templates like `isa<>, cast<>, and dyn_cast<> <ProgrammersManual.html#isa>`_.
526 This form of RTTI is opt-in and can be
527 :doc:`added to any class <HowToSetUpLLVMStyleRTTI>`. It is also
528 substantially more efficient than ``dynamic_cast<>``.
530 .. _static constructor:
532 Do not use Static Constructors
533 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
535 Static constructors and destructors (e.g. global variables whose types have a
536 constructor or destructor) should not be added to the code base, and should be
537 removed wherever possible. Besides `well known problems
538 <http://yosefk.com/c++fqa/ctors.html#fqa-10.12>`_ where the order of
539 initialization is undefined between globals in different source files, the
540 entire concept of static constructors is at odds with the common use case of
541 LLVM as a library linked into a larger application.
543 Consider the use of LLVM as a JIT linked into another application (perhaps for
544 `OpenGL, custom languages <http://llvm.org/Users.html>`_, `shaders in movies
545 <http://llvm.org/devmtg/2010-11/Gritz-OpenShadingLang.pdf>`_, etc). Due to the
546 design of static constructors, they must be executed at startup time of the
547 entire application, regardless of whether or how LLVM is used in that larger
548 application. There are two problems with this:
550 * The time to run the static constructors impacts startup time of applications
551 --- a critical time for GUI apps, among others.
553 * The static constructors cause the app to pull many extra pages of memory off
554 the disk: both the code for the constructor in each ``.o`` file and the small
555 amount of data that gets touched. In addition, touched/dirty pages put more
556 pressure on the VM system on low-memory machines.
558 We would really like for there to be zero cost for linking in an additional LLVM
559 target or other library into an application, but static constructors violate
562 That said, LLVM unfortunately does contain static constructors. It would be a
563 `great project <http://llvm.org/PR11944>`_ for someone to purge all static
564 constructors from LLVM, and then enable the ``-Wglobal-constructors`` warning
565 flag (when building with Clang) to ensure we do not regress in the future.
567 Use of ``class`` and ``struct`` Keywords
568 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
570 In C++, the ``class`` and ``struct`` keywords can be used almost
571 interchangeably. The only difference is when they are used to declare a class:
572 ``class`` makes all members private by default while ``struct`` makes all
573 members public by default.
575 Unfortunately, not all compilers follow the rules and some will generate
576 different symbols based on whether ``class`` or ``struct`` was used to declare
577 the symbol. This can lead to problems at link time.
579 So, the rule for LLVM is to always use the ``class`` keyword, unless **all**
580 members are public and the type is a C++ `POD
581 <http://en.wikipedia.org/wiki/Plain_old_data_structure>`_ type, in which case
582 ``struct`` is allowed.
587 The High-Level Issues
588 ---------------------
590 A Public Header File **is** a Module
591 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
593 C++ doesn't do too well in the modularity department. There is no real
594 encapsulation or data hiding (unless you use expensive protocol classes), but it
595 is what we have to work with. When you write a public header file (in the LLVM
596 source tree, they live in the top level "``include``" directory), you are
597 defining a module of functionality.
599 Ideally, modules should be completely independent of each other, and their
600 header files should only ``#include`` the absolute minimum number of headers
601 possible. A module is not just a class, a function, or a namespace: it's a
602 collection of these that defines an interface. This interface may be several
603 functions, classes, or data structures, but the important issue is how they work
606 In general, a module should be implemented by one or more ``.cpp`` files. Each
607 of these ``.cpp`` files should include the header that defines their interface
608 first. This ensures that all of the dependences of the module header have been
609 properly added to the module header itself, and are not implicit. System
610 headers should be included after user headers for a translation unit.
612 .. _minimal list of #includes:
614 ``#include`` as Little as Possible
615 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
617 ``#include`` hurts compile time performance. Don't do it unless you have to,
618 especially in header files.
620 But wait! Sometimes you need to have the definition of a class to use it, or to
621 inherit from it. In these cases go ahead and ``#include`` that header file. Be
622 aware however that there are many cases where you don't need to have the full
623 definition of a class. If you are using a pointer or reference to a class, you
624 don't need the header file. If you are simply returning a class instance from a
625 prototyped function or method, you don't need it. In fact, for most cases, you
626 simply don't need the definition of a class. And not ``#include``\ing speeds up
629 It is easy to try to go too overboard on this recommendation, however. You
630 **must** include all of the header files that you are using --- you can include
631 them either directly or indirectly through another header file. To make sure
632 that you don't accidentally forget to include a header file in your module
633 header, make sure to include your module header **first** in the implementation
634 file (as mentioned above). This way there won't be any hidden dependencies that
635 you'll find out about later.
637 Keep "Internal" Headers Private
638 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
640 Many modules have a complex implementation that causes them to use more than one
641 implementation (``.cpp``) file. It is often tempting to put the internal
642 communication interface (helper classes, extra functions, etc) in the public
643 module header file. Don't do this!
645 If you really need to do something like this, put a private header file in the
646 same directory as the source files, and include it locally. This ensures that
647 your private interface remains private and undisturbed by outsiders.
651 It's okay to put extra implementation methods in a public class itself. Just
652 make them private (or protected) and all is well.
656 Use Early Exits and ``continue`` to Simplify Code
657 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
659 When reading code, keep in mind how much state and how many previous decisions
660 have to be remembered by the reader to understand a block of code. Aim to
661 reduce indentation where possible when it doesn't make it more difficult to
662 understand the code. One great way to do this is by making use of early exits
663 and the ``continue`` keyword in long loops. As an example of using an early
664 exit from a function, consider this "bad" code:
668 Value *doSomething(Instruction *I) {
669 if (!isa<TerminatorInst>(I) &&
670 I->hasOneUse() && doOtherThing(I)) {
671 ... some long code ....
677 This code has several problems if the body of the ``'if'`` is large. When
678 you're looking at the top of the function, it isn't immediately clear that this
679 *only* does interesting things with non-terminator instructions, and only
680 applies to things with the other predicates. Second, it is relatively difficult
681 to describe (in comments) why these predicates are important because the ``if``
682 statement makes it difficult to lay out the comments. Third, when you're deep
683 within the body of the code, it is indented an extra level. Finally, when
684 reading the top of the function, it isn't clear what the result is if the
685 predicate isn't true; you have to read to the end of the function to know that
688 It is much preferred to format the code like this:
692 Value *doSomething(Instruction *I) {
693 // Terminators never need 'something' done to them because ...
694 if (isa<TerminatorInst>(I))
697 // We conservatively avoid transforming instructions with multiple uses
698 // because goats like cheese.
702 // This is really just here for example.
703 if (!doOtherThing(I))
706 ... some long code ....
709 This fixes these problems. A similar problem frequently happens in ``for``
710 loops. A silly example is something like this:
714 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
715 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
716 Value *LHS = BO->getOperand(0);
717 Value *RHS = BO->getOperand(1);
724 When you have very, very small loops, this sort of structure is fine. But if it
725 exceeds more than 10-15 lines, it becomes difficult for people to read and
726 understand at a glance. The problem with this sort of code is that it gets very
727 nested very quickly. Meaning that the reader of the code has to keep a lot of
728 context in their brain to remember what is going immediately on in the loop,
729 because they don't know if/when the ``if`` conditions will have ``else``\s etc.
730 It is strongly preferred to structure the loop like this:
734 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
735 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
738 Value *LHS = BO->getOperand(0);
739 Value *RHS = BO->getOperand(1);
740 if (LHS == RHS) continue;
745 This has all the benefits of using early exits for functions: it reduces nesting
746 of the loop, it makes it easier to describe why the conditions are true, and it
747 makes it obvious to the reader that there is no ``else`` coming up that they
748 have to push context into their brain for. If a loop is large, this can be a
749 big understandability win.
751 Don't use ``else`` after a ``return``
752 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
754 For similar reasons above (reduction of indentation and easier reading), please
755 do not use ``'else'`` or ``'else if'`` after something that interrupts control
756 flow --- like ``return``, ``break``, ``continue``, ``goto``, etc. For
757 example, this is *bad*:
763 Type = Context.getsigjmp_bufType();
765 Error = ASTContext::GE_Missing_sigjmp_buf;
771 Type = Context.getjmp_bufType();
773 Error = ASTContext::GE_Missing_jmp_buf;
781 It is better to write it like this:
787 Type = Context.getsigjmp_bufType();
789 Error = ASTContext::GE_Missing_sigjmp_buf;
793 Type = Context.getjmp_bufType();
795 Error = ASTContext::GE_Missing_jmp_buf;
801 Or better yet (in this case) as:
807 Type = Context.getsigjmp_bufType();
809 Type = Context.getjmp_bufType();
812 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
813 ASTContext::GE_Missing_jmp_buf;
818 The idea is to reduce indentation and the amount of code you have to keep track
819 of when reading the code.
821 Turn Predicate Loops into Predicate Functions
822 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
824 It is very common to write small loops that just compute a boolean value. There
825 are a number of ways that people commonly write these, but an example of this
830 bool FoundFoo = false;
831 for (unsigned I = 0, E = BarList.size(); I != E; ++I)
832 if (BarList[I]->isFoo()) {
841 This sort of code is awkward to write, and is almost always a bad sign. Instead
842 of this sort of loop, we strongly prefer to use a predicate function (which may
843 be `static`_) that uses `early exits`_ to compute the predicate. We prefer the
844 code to be structured like this:
848 /// \returns true if the specified list has an element that is a foo.
849 static bool containsFoo(const std::vector<Bar*> &List) {
850 for (unsigned I = 0, E = List.size(); I != E; ++I)
851 if (List[I]->isFoo())
857 if (containsFoo(BarList)) {
861 There are many reasons for doing this: it reduces indentation and factors out
862 code which can often be shared by other code that checks for the same predicate.
863 More importantly, it *forces you to pick a name* for the function, and forces
864 you to write a comment for it. In this silly example, this doesn't add much
865 value. However, if the condition is complex, this can make it a lot easier for
866 the reader to understand the code that queries for this predicate. Instead of
867 being faced with the in-line details of how we check to see if the BarList
868 contains a foo, we can trust the function name and continue reading with better
874 Name Types, Functions, Variables, and Enumerators Properly
875 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
877 Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
878 enough how important it is to use *descriptive* names. Pick names that match
879 the semantics and role of the underlying entities, within reason. Avoid
880 abbreviations unless they are well known. After picking a good name, make sure
881 to use consistent capitalization for the name, as inconsistency requires clients
882 to either memorize the APIs or to look it up to find the exact spelling.
884 In general, names should be in camel case (e.g. ``TextFileReader`` and
885 ``isLValue()``). Different kinds of declarations have different rules:
887 * **Type names** (including classes, structs, enums, typedefs, etc) should be
888 nouns and start with an upper-case letter (e.g. ``TextFileReader``).
890 * **Variable names** should be nouns (as they represent state). The name should
891 be camel case, and start with an upper case letter (e.g. ``Leader`` or
894 * **Function names** should be verb phrases (as they represent actions), and
895 command-like function should be imperative. The name should be camel case,
896 and start with a lower case letter (e.g. ``openFile()`` or ``isFoo()``).
898 * **Enum declarations** (e.g. ``enum Foo {...}``) are types, so they should
899 follow the naming conventions for types. A common use for enums is as a
900 discriminator for a union, or an indicator of a subclass. When an enum is
901 used for something like this, it should have a ``Kind`` suffix
902 (e.g. ``ValueKind``).
904 * **Enumerators** (e.g. ``enum { Foo, Bar }``) and **public member variables**
905 should start with an upper-case letter, just like types. Unless the
906 enumerators are defined in their own small namespace or inside a class,
907 enumerators should have a prefix corresponding to the enum declaration name.
908 For example, ``enum ValueKind { ... };`` may contain enumerators like
909 ``VK_Argument``, ``VK_BasicBlock``, etc. Enumerators that are just
910 convenience constants are exempt from the requirement for a prefix. For
920 As an exception, classes that mimic STL classes can have member names in STL's
921 style of lower-case words separated by underscores (e.g. ``begin()``,
922 ``push_back()``, and ``empty()``). Classes that provide multiple
923 iterators should add a singular prefix to ``begin()`` and ``end()``
924 (e.g. ``global_begin()`` and ``use_begin()``).
926 Here are some examples of good and bad names:
932 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
933 Factory<Tire> Factory; // Better.
934 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
935 // kind of factories.
938 Vehicle MakeVehicle(VehicleType Type) {
939 VehicleMaker M; // Might be OK if having a short life-span.
940 Tire Tmp1 = M.makeTire(); // Bad -- 'Tmp1' provides no information.
941 Light Headlight = M.makeLight("head"); // Good -- descriptive.
948 Use the "``assert``" macro to its fullest. Check all of your preconditions and
949 assumptions, you never know when a bug (not necessarily even yours) might be
950 caught early by an assertion, which reduces debugging time dramatically. The
951 "``<cassert>``" header file is probably already included by the header files you
952 are using, so it doesn't cost anything to use it.
954 To further assist with debugging, make sure to put some kind of error message in
955 the assertion statement, which is printed if the assertion is tripped. This
956 helps the poor debugger make sense of why an assertion is being made and
957 enforced, and hopefully what to do about it. Here is one complete example:
961 inline Value *getOperand(unsigned I) {
962 assert(I < Operands.size() && "getOperand() out of range!");
966 Here are more examples:
970 assert(Ty->isPointerType() && "Can't allocate a non-pointer type!");
972 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
974 assert(idx < getNumSuccessors() && "Successor # out of range!");
976 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
978 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
982 In the past, asserts were used to indicate a piece of code that should not be
983 reached. These were typically of the form:
987 assert(0 && "Invalid radix for integer literal");
989 This has a few issues, the main one being that some compilers might not
990 understand the assertion, or warn about a missing return in builds where
991 assertions are compiled out.
993 Today, we have something much better: ``llvm_unreachable``:
997 llvm_unreachable("Invalid radix for integer literal");
999 When assertions are enabled, this will print the message if it's ever reached
1000 and then exit the program. When assertions are disabled (i.e. in release
1001 builds), ``llvm_unreachable`` becomes a hint to compilers to skip generating
1002 code for this branch. If the compiler does not support this, it will fall back
1003 to the "abort" implementation.
1005 Another issue is that values used only by assertions will produce an "unused
1006 value" warning when assertions are disabled. For example, this code will warn:
1010 unsigned Size = V.size();
1011 assert(Size > 42 && "Vector smaller than it should be");
1013 bool NewToSet = Myset.insert(Value);
1014 assert(NewToSet && "The value shouldn't be in the set yet");
1016 These are two interesting different cases. In the first case, the call to
1017 ``V.size()`` is only useful for the assert, and we don't want it executed when
1018 assertions are disabled. Code like this should move the call into the assert
1019 itself. In the second case, the side effects of the call must happen whether
1020 the assert is enabled or not. In this case, the value should be cast to void to
1021 disable the warning. To be specific, it is preferred to write the code like
1026 assert(V.size() > 42 && "Vector smaller than it should be");
1028 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1029 assert(NewToSet && "The value shouldn't be in the set yet");
1031 Do Not Use ``using namespace std``
1032 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1034 In LLVM, we prefer to explicitly prefix all identifiers from the standard
1035 namespace with an "``std::``" prefix, rather than rely on "``using namespace
1038 In header files, adding a ``'using namespace XXX'`` directive pollutes the
1039 namespace of any source file that ``#include``\s the header. This is clearly a
1042 In implementation files (e.g. ``.cpp`` files), the rule is more of a stylistic
1043 rule, but is still important. Basically, using explicit namespace prefixes
1044 makes the code **clearer**, because it is immediately obvious what facilities
1045 are being used and where they are coming from. And **more portable**, because
1046 namespace clashes cannot occur between LLVM code and other namespaces. The
1047 portability rule is important because different standard library implementations
1048 expose different symbols (potentially ones they shouldn't), and future revisions
1049 to the C++ standard will add more symbols to the ``std`` namespace. As such, we
1050 never use ``'using namespace std;'`` in LLVM.
1052 The exception to the general rule (i.e. it's not an exception for the ``std``
1053 namespace) is for implementation files. For example, all of the code in the
1054 LLVM project implements code that lives in the 'llvm' namespace. As such, it is
1055 ok, and actually clearer, for the ``.cpp`` files to have a ``'using namespace
1056 llvm;'`` directive at the top, after the ``#include``\s. This reduces
1057 indentation in the body of the file for source editors that indent based on
1058 braces, and keeps the conceptual context cleaner. The general form of this rule
1059 is that any ``.cpp`` file that implements code in any namespace may use that
1060 namespace (and its parents'), but should not use any others.
1062 Provide a Virtual Method Anchor for Classes in Headers
1063 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1065 If a class is defined in a header file and has a vtable (either it has virtual
1066 methods or it derives from classes with virtual methods), it must always have at
1067 least one out-of-line virtual method in the class. Without this, the compiler
1068 will copy the vtable and RTTI into every ``.o`` file that ``#include``\s the
1069 header, bloating ``.o`` file sizes and increasing link times.
1071 Don't use default labels in fully covered switches over enumerations
1072 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1074 ``-Wswitch`` warns if a switch, without a default label, over an enumeration
1075 does not cover every enumeration value. If you write a default label on a fully
1076 covered switch over an enumeration then the ``-Wswitch`` warning won't fire
1077 when new elements are added to that enumeration. To help avoid adding these
1078 kinds of defaults, Clang has the warning ``-Wcovered-switch-default`` which is
1079 off by default but turned on when building LLVM with a version of Clang that
1080 supports the warning.
1082 A knock-on effect of this stylistic requirement is that when building LLVM with
1083 GCC you may get warnings related to "control may reach end of non-void function"
1084 if you return from each case of a covered switch-over-enum because GCC assumes
1085 that the enum expression may take any representable value, not just those of
1086 individual enumerators. To suppress this warning, use ``llvm_unreachable`` after
1089 Use ``LLVM_DELETED_FUNCTION`` to mark uncallable methods
1090 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1092 Prior to C++11, a common pattern to make a class uncopyable was to declare an
1093 unimplemented copy constructor and copy assignment operator and make them
1094 private. This would give a compiler error for accessing a private method or a
1095 linker error because it wasn't implemented.
1097 With C++11, we can mark methods that won't be implemented with ``= delete``.
1098 This will trigger a much better error message and tell the compiler that the
1099 method will never be implemented. This enables other checks like
1100 ``-Wunused-private-field`` to run correctly on classes that contain these
1103 To maintain compatibility with C++03, ``LLVM_DELETED_FUNCTION`` should be used
1104 which will expand to ``= delete`` if the compiler supports it. These methods
1105 should still be declared private. Example of the uncopyable pattern:
1111 DontCopy(const DontCopy&) LLVM_DELETED_FUNCTION;
1112 DontCopy &operator =(const DontCopy&) LLVM_DELETED_FUNCTION;
1117 Don't evaluate ``end()`` every time through a loop
1118 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1120 Because C++ doesn't have a standard "``foreach``" loop (though it can be
1121 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1122 loops that manually iterate from begin to end on a variety of containers or
1123 through other data structures. One common mistake is to write a loop in this
1128 BasicBlock *BB = ...
1129 for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I)
1132 The problem with this construct is that it evaluates "``BB->end()``" every time
1133 through the loop. Instead of writing the loop like this, we strongly prefer
1134 loops to be written so that they evaluate it once before the loop starts. A
1135 convenient way to do this is like so:
1139 BasicBlock *BB = ...
1140 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
1143 The observant may quickly point out that these two loops may have different
1144 semantics: if the container (a basic block in this case) is being mutated, then
1145 "``BB->end()``" may change its value every time through the loop and the second
1146 loop may not in fact be correct. If you actually do depend on this behavior,
1147 please write the loop in the first form and add a comment indicating that you
1148 did it intentionally.
1150 Why do we prefer the second form (when correct)? Writing the loop in the first
1151 form has two problems. First it may be less efficient than evaluating it at the
1152 start of the loop. In this case, the cost is probably minor --- a few extra
1153 loads every time through the loop. However, if the base expression is more
1154 complex, then the cost can rise quickly. I've seen loops where the end
1155 expression was actually something like: "``SomeMap[X]->end()``" and map lookups
1156 really aren't cheap. By writing it in the second form consistently, you
1157 eliminate the issue entirely and don't even have to think about it.
1159 The second (even bigger) issue is that writing the loop in the first form hints
1160 to the reader that the loop is mutating the container (a fact that a comment
1161 would handily confirm!). If you write the loop in the second form, it is
1162 immediately obvious without even looking at the body of the loop that the
1163 container isn't being modified, which makes it easier to read the code and
1164 understand what it does.
1166 While the second form of the loop is a few extra keystrokes, we do strongly
1169 ``#include <iostream>`` is Forbidden
1170 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1172 The use of ``#include <iostream>`` in library files is hereby **forbidden**,
1173 because many common implementations transparently inject a `static constructor`_
1174 into every translation unit that includes it.
1176 Note that using the other stream headers (``<sstream>`` for example) is not
1177 problematic in this regard --- just ``<iostream>``. However, ``raw_ostream``
1178 provides various APIs that are better performing for almost every use than
1179 ``std::ostream`` style APIs.
1183 New code should always use `raw_ostream`_ for writing, or the
1184 ``llvm::MemoryBuffer`` API for reading files.
1191 LLVM includes a lightweight, simple, and efficient stream implementation in
1192 ``llvm/Support/raw_ostream.h``, which provides all of the common features of
1193 ``std::ostream``. All new code should use ``raw_ostream`` instead of
1196 Unlike ``std::ostream``, ``raw_ostream`` is not a template and can be forward
1197 declared as ``class raw_ostream``. Public headers should generally not include
1198 the ``raw_ostream`` header, but use forward declarations and constant references
1199 to ``raw_ostream`` instances.
1204 The ``std::endl`` modifier, when used with ``iostreams`` outputs a newline to
1205 the output stream specified. In addition to doing this, however, it also
1206 flushes the output stream. In other words, these are equivalent:
1210 std::cout << std::endl;
1211 std::cout << '\n' << std::flush;
1213 Most of the time, you probably have no reason to flush the output stream, so
1214 it's better to use a literal ``'\n'``.
1216 Don't use ``inline`` when defining a function in a class definition
1217 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1219 A member function defined in a class definition is implicitly inline, so don't
1220 put the ``inline`` keyword in this case.
1247 This section describes preferred low-level formatting guidelines along with
1248 reasoning on why we prefer them.
1250 Spaces Before Parentheses
1251 ^^^^^^^^^^^^^^^^^^^^^^^^^
1253 We prefer to put a space before an open parenthesis only in control flow
1254 statements, but not in normal function call expressions and function-like
1255 macros. For example, this is good:
1260 for (I = 0; I != 100; ++I) ...
1261 while (LLVMRocks) ...
1264 assert(3 != 4 && "laws of math are failing me");
1266 A = foo(42, 92) + bar(X);
1273 for(I = 0; I != 100; ++I) ...
1274 while(LLVMRocks) ...
1277 assert (3 != 4 && "laws of math are failing me");
1279 A = foo (42, 92) + bar (X);
1281 The reason for doing this is not completely arbitrary. This style makes control
1282 flow operators stand out more, and makes expressions flow better. The function
1283 call operator binds very tightly as a postfix operator. Putting a space after a
1284 function name (as in the last example) makes it appear that the code might bind
1285 the arguments of the left-hand-side of a binary operator with the argument list
1286 of a function and the name of the right side. More specifically, it is easy to
1287 misread the "``A``" example as:
1291 A = foo ((42, 92) + bar) (X);
1293 when skimming through the code. By avoiding a space in a function, we avoid
1294 this misinterpretation.
1299 Hard fast rule: Preincrement (``++X``) may be no slower than postincrement
1300 (``X++``) and could very well be a lot faster than it. Use preincrementation
1303 The semantics of postincrement include making a copy of the value being
1304 incremented, returning it, and then preincrementing the "work value". For
1305 primitive types, this isn't a big deal. But for iterators, it can be a huge
1306 issue (for example, some iterators contains stack and set objects in them...
1307 copying an iterator could invoke the copy ctor's of these as well). In general,
1308 get in the habit of always using preincrement, and you won't have a problem.
1311 Namespace Indentation
1312 ^^^^^^^^^^^^^^^^^^^^^
1314 In general, we strive to reduce indentation wherever possible. This is useful
1315 because we want code to `fit into 80 columns`_ without wrapping horribly, but
1316 also because it makes it easier to understand the code. To facilitate this and
1317 avoid some insanely deep nesting on occasion, don't indent namespaces. If it
1318 helps readability, feel free to add a comment indicating what namespace is
1319 being closed by a ``}``. For example:
1324 namespace knowledge {
1326 /// This class represents things that Smith can have an intimate
1327 /// understanding of and contains the data associated with it.
1331 explicit Grokable() { ... }
1332 virtual ~Grokable() = 0;
1338 } // end namespace knowledge
1339 } // end namespace llvm
1342 Feel free to skip the closing comment when the namespace being closed is
1343 obvious for any reason. For example, the outer-most namespace in a header file
1344 is rarely a source of confusion. But namespaces both anonymous and named in
1345 source files that are being closed half way through the file probably could use
1350 Anonymous Namespaces
1351 ^^^^^^^^^^^^^^^^^^^^
1353 After talking about namespaces in general, you may be wondering about anonymous
1354 namespaces in particular. Anonymous namespaces are a great language feature
1355 that tells the C++ compiler that the contents of the namespace are only visible
1356 within the current translation unit, allowing more aggressive optimization and
1357 eliminating the possibility of symbol name collisions. Anonymous namespaces are
1358 to C++ as "static" is to C functions and global variables. While "``static``"
1359 is available in C++, anonymous namespaces are more general: they can make entire
1360 classes private to a file.
1362 The problem with anonymous namespaces is that they naturally want to encourage
1363 indentation of their body, and they reduce locality of reference: if you see a
1364 random function definition in a C++ file, it is easy to see if it is marked
1365 static, but seeing if it is in an anonymous namespace requires scanning a big
1368 Because of this, we have a simple guideline: make anonymous namespaces as small
1369 as possible, and only use them for class declarations. For example, this is
1379 bool operator<(const char *RHS) const;
1381 } // end anonymous namespace
1383 static void runHelper() {
1387 bool StringSort::operator<(const char *RHS) const {
1401 bool operator<(const char *RHS) const;
1408 bool StringSort::operator<(const char *RHS) const {
1412 } // end anonymous namespace
1414 This is bad specifically because if you're looking at "``runHelper``" in the middle
1415 of a large C++ file, that you have no immediate way to tell if it is local to
1416 the file. When it is marked static explicitly, this is immediately obvious.
1417 Also, there is no reason to enclose the definition of "``operator<``" in the
1418 namespace just because it was declared there.
1423 A lot of these comments and recommendations have been culled from other sources.
1424 Two particularly important books for our work are:
1427 <http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876>`_
1428 by Scott Meyers. Also interesting and useful are "More Effective C++" and
1429 "Effective STL" by the same author.
1431 #. `Large-Scale C++ Software Design
1432 <http://www.amazon.com/Large-Scale-Software-Design-John-Lakos/dp/0201633620/ref=sr_1_1>`_
1435 If you get some free time, and you haven't read them: do so, you might learn