13 This document attempts to describe a few coding standards that are being used in
14 the LLVM source tree. Although no coding standards should be regarded as
15 absolute requirements to be followed in all instances, coding standards are
16 particularly important for large-scale code bases that follow a library-based
19 This document intentionally does not prescribe fixed standards for religious
20 issues such as brace placement and space usage. For issues like this, follow
25 **If you are extending, enhancing, or bug fixing already implemented code,
26 use the style that is already being used so that the source is uniform and
29 Note that some code bases (e.g. ``libc++``) have really good reasons to deviate
30 from the coding standards. In the case of ``libc++``, this is because the
31 naming and other conventions are dictated by the C++ standard. If you think
32 there is a specific good reason to deviate from the standards here, please bring
33 it up on the LLVMdev mailing list.
35 There are some conventions that are not uniformly followed in the code base
36 (e.g. the naming convention). This is because they are relatively new, and a
37 lot of code was written before they were put in place. Our long term goal is
38 for the entire codebase to follow the convention, but we explicitly *do not*
39 want patches that do large-scale reformating of existing code. On the other
40 hand, it is reasonable to rename the methods of a class if you're about to
41 change it in some other way. Just do the reformating as a separate commit from
42 the functionality change.
44 The ultimate goal of these guidelines is the increase readability and
45 maintainability of our common source base. If you have suggestions for topics to
46 be included, please mail them to `Chris <mailto:sabre@nondot.org>`_.
48 Mechanical Source Issues
49 ========================
51 Source Code Formatting
52 ----------------------
57 Comments are one critical part of readability and maintainability. Everyone
58 knows they should comment their code, and so should you. When writing comments,
59 write them as English prose, which means they should use proper capitalization,
60 punctuation, etc. Aim to describe what the code is trying to do and why, not
61 *how* it does it at a micro level. Here are a few critical things to document:
63 .. _header file comment:
68 Every source file should have a header on it that describes the basic purpose of
69 the file. If a file does not have a header, it should not be checked into the
70 tree. The standard header looks like this:
74 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
76 // The LLVM Compiler Infrastructure
78 // This file is distributed under the University of Illinois Open Source
79 // License. See LICENSE.TXT for details.
81 //===----------------------------------------------------------------------===//
84 /// \brief This file contains the declaration of the Instruction class, which is
85 /// the base class for all of the VM instructions.
87 //===----------------------------------------------------------------------===//
89 A few things to note about this particular format: The "``-*- C++ -*-``" string
90 on the first line is there to tell Emacs that the source file is a C++ file, not
91 a C file (Emacs assumes ``.h`` files are C files by default).
95 This tag is not necessary in ``.cpp`` files. The name of the file is also
96 on the first line, along with a very short description of the purpose of the
97 file. This is important when printing out code and flipping though lots of
100 The next section in the file is a concise note that defines the license that the
101 file is released under. This makes it perfectly clear what terms the source
102 code can be distributed under and should not be modified in any way.
104 The main body is a ``doxygen`` comment describing the purpose of the file. It
105 should have a ``\brief`` command that describes the file in one or two
106 sentences. Any additional information should be separated by a blank line. If
107 an algorithm is being implemented or something tricky is going on, a reference
108 to the paper where it is published should be included, as well as any notes or
109 *gotchas* in the code to watch out for.
114 Classes are one fundamental part of a good object oriented design. As such, a
115 class definition should have a comment block that explains what the class is
116 used for and how it works. Every non-trivial class is expected to have a
117 ``doxygen`` comment block.
122 Methods defined in a class (as well as any global functions) should also be
123 documented properly. A quick note about what it does and a description of the
124 borderline behaviour is all that is necessary here (unless something
125 particularly tricky or insidious is going on). The hope is that people can
126 figure out how to use your interfaces without reading the code itself.
128 Good things to talk about here are what happens when something unexpected
129 happens: does the method return null? Abort? Format your hard disk?
134 In general, prefer C++ style (``//``) comments. They take less space, require
135 less typing, don't have nesting problems, etc. There are a few cases when it is
136 useful to use C style (``/* */``) comments however:
138 #. When writing C code: Obviously if you are writing C code, use C style
141 #. When writing a header file that may be ``#include``\d by a C source file.
143 #. When writing a source file that is used by a tool that only accepts C style
146 To comment out a large block of code, use ``#if 0`` and ``#endif``. These nest
147 properly and are better behaved in general than C style comments.
152 Immediately after the `header file comment`_ (and include guards if working on a
153 header file), the `minimal list of #includes`_ required by the file should be
154 listed. We prefer these ``#include``\s to be listed in this order:
156 .. _Main Module Header:
157 .. _Local/Private Headers:
159 #. Main Module Header
160 #. Local/Private Headers
162 #. ``llvm/Analysis/*``
163 #. ``llvm/Assembly/*``
164 #. ``llvm/Bitcode/*``
165 #. ``llvm/CodeGen/*``
167 #. ``llvm/Support/*``
169 #. System ``#include``\s
171 and each category should be sorted by name.
173 The `Main Module Header`_ file applies to ``.cpp`` files which implement an
174 interface defined by a ``.h`` file. This ``#include`` should always be included
175 **first** regardless of where it lives on the file system. By including a
176 header file first in the ``.cpp`` files that implement the interfaces, we ensure
177 that the header does not have any hidden dependencies which are not explicitly
178 ``#include``\d in the header, but should be. It is also a form of documentation
179 in the ``.cpp`` file to indicate where the interfaces it implements are defined.
181 .. _fit into 80 columns:
186 Write your code to fit within 80 columns of text. This helps those of us who
187 like to print out code and look at your code in an ``xterm`` without resizing
190 The longer answer is that there must be some limit to the width of the code in
191 order to reasonably allow developers to have multiple files side-by-side in
192 windows on a modest display. If you are going to pick a width limit, it is
193 somewhat arbitrary but you might as well pick something standard. Going with 90
194 columns (for example) instead of 80 columns wouldn't add any significant value
195 and would be detrimental to printing out code. Also many other projects have
196 standardized on 80 columns, so some people have already configured their editors
197 for it (vs something else, like 90 columns).
199 This is one of many contentious issues in coding standards, but it is not up for
202 Use Spaces Instead of Tabs
203 ^^^^^^^^^^^^^^^^^^^^^^^^^^
205 In all cases, prefer spaces to tabs in source files. People have different
206 preferred indentation levels, and different styles of indentation that they
207 like; this is fine. What isn't fine is that different editors/viewers expand
208 tabs out to different tab stops. This can cause your code to look completely
209 unreadable, and it is not worth dealing with.
211 As always, follow the `Golden Rule`_ above: follow the style of
212 existing code if you are modifying and extending it. If you like four spaces of
213 indentation, **DO NOT** do that in the middle of a chunk of code with two spaces
214 of indentation. Also, do not reindent a whole source file: it makes for
215 incredible diffs that are absolutely worthless.
217 Indent Code Consistently
218 ^^^^^^^^^^^^^^^^^^^^^^^^
220 Okay, in your first year of programming you were told that indentation is
221 important. If you didn't believe and internalize this then, now is the time.
227 Treat Compiler Warnings Like Errors
228 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
230 If your code has compiler warnings in it, something is wrong --- you aren't
231 casting values correctly, you have "questionable" constructs in your code, or
232 you are doing something legitimately wrong. Compiler warnings can cover up
233 legitimate errors in output and make dealing with a translation unit difficult.
235 It is not possible to prevent all warnings from all compilers, nor is it
236 desirable. Instead, pick a standard compiler (like ``gcc``) that provides a
237 good thorough set of warnings, and stick to it. At least in the case of
238 ``gcc``, it is possible to work around any spurious errors by changing the
239 syntax of the code slightly. For example, a warning that annoys me occurs when
240 I write code like this:
244 if (V = getValue()) {
248 ``gcc`` will warn me that I probably want to use the ``==`` operator, and that I
249 probably mistyped it. In most cases, I haven't, and I really don't want the
250 spurious errors. To fix this particular problem, I rewrite the code like
255 if ((V = getValue())) {
259 which shuts ``gcc`` up. Any ``gcc`` warning that annoys you can be fixed by
260 massaging the code appropriately.
265 In almost all cases, it is possible and within reason to write completely
266 portable code. If there are cases where it isn't possible to write portable
267 code, isolate it behind a well defined (and well documented) interface.
269 In practice, this means that you shouldn't assume much about the host compiler
270 (and Visual Studio tends to be the lowest common denominator). If advanced
271 features are used, they should only be an implementation detail of a library
272 which has a simple exposed API, and preferably be buried in ``libSystem``.
274 Do not use RTTI or Exceptions
275 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
277 In an effort to reduce code and executable size, LLVM does not use RTTI
278 (e.g. ``dynamic_cast<>;``) or exceptions. These two language features violate
279 the general C++ principle of *"you only pay for what you use"*, causing
280 executable bloat even if exceptions are never used in the code base, or if RTTI
281 is never used for a class. Because of this, we turn them off globally in the
284 That said, LLVM does make extensive use of a hand-rolled form of RTTI that use
285 templates like `isa<>, cast<>, and dyn_cast<> <ProgrammersManual.html#isa>`_.
286 This form of RTTI is opt-in and can be added to any class. It is also
287 substantially more efficient than ``dynamic_cast<>``.
289 .. _static constructor:
291 Do not use Static Constructors
292 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
294 Static constructors and destructors (e.g. global variables whose types have a
295 constructor or destructor) should not be added to the code base, and should be
296 removed wherever possible. Besides `well known problems
297 <http://yosefk.com/c++fqa/ctors.html#fqa-10.12>`_ where the order of
298 initialization is undefined between globals in different source files, the
299 entire concept of static constructors is at odds with the common use case of
300 LLVM as a library linked into a larger application.
302 Consider the use of LLVM as a JIT linked into another application (perhaps for
303 `OpenGL, custom languages <http://llvm.org/Users.html>`_, `shaders in movies
304 <http://llvm.org/devmtg/2010-11/Gritz-OpenShadingLang.pdf>`_, etc). Due to the
305 design of static constructors, they must be executed at startup time of the
306 entire application, regardless of whether or how LLVM is used in that larger
307 application. There are two problems with this:
309 * The time to run the static constructors impacts startup time of applications
310 --- a critical time for GUI apps, among others.
312 * The static constructors cause the app to pull many extra pages of memory off
313 the disk: both the code for the constructor in each ``.o`` file and the small
314 amount of data that gets touched. In addition, touched/dirty pages put more
315 pressure on the VM system on low-memory machines.
317 We would really like for there to be zero cost for linking in an additional LLVM
318 target or other library into an application, but static constructors violate
321 That said, LLVM unfortunately does contain static constructors. It would be a
322 `great project <http://llvm.org/PR11944>`_ for someone to purge all static
323 constructors from LLVM, and then enable the ``-Wglobal-constructors`` warning
324 flag (when building with Clang) to ensure we do not regress in the future.
326 Use of ``class`` and ``struct`` Keywords
327 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
329 In C++, the ``class`` and ``struct`` keywords can be used almost
330 interchangeably. The only difference is when they are used to declare a class:
331 ``class`` makes all members private by default while ``struct`` makes all
332 members public by default.
334 Unfortunately, not all compilers follow the rules and some will generate
335 different symbols based on whether ``class`` or ``struct`` was used to declare
336 the symbol. This can lead to problems at link time.
338 So, the rule for LLVM is to always use the ``class`` keyword, unless **all**
339 members are public and the type is a C++ `POD
340 <http://en.wikipedia.org/wiki/Plain_old_data_structure>`_ type, in which case
341 ``struct`` is allowed.
346 The High-Level Issues
347 ---------------------
349 A Public Header File **is** a Module
350 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
352 C++ doesn't do too well in the modularity department. There is no real
353 encapsulation or data hiding (unless you use expensive protocol classes), but it
354 is what we have to work with. When you write a public header file (in the LLVM
355 source tree, they live in the top level "``include``" directory), you are
356 defining a module of functionality.
358 Ideally, modules should be completely independent of each other, and their
359 header files should only ``#include`` the absolute minimum number of headers
360 possible. A module is not just a class, a function, or a namespace: it's a
361 collection of these that defines an interface. This interface may be several
362 functions, classes, or data structures, but the important issue is how they work
365 In general, a module should be implemented by one or more ``.cpp`` files. Each
366 of these ``.cpp`` files should include the header that defines their interface
367 first. This ensures that all of the dependences of the module header have been
368 properly added to the module header itself, and are not implicit. System
369 headers should be included after user headers for a translation unit.
371 .. _minimal list of #includes:
373 ``#include`` as Little as Possible
374 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
376 ``#include`` hurts compile time performance. Don't do it unless you have to,
377 especially in header files.
379 But wait! Sometimes you need to have the definition of a class to use it, or to
380 inherit from it. In these cases go ahead and ``#include`` that header file. Be
381 aware however that there are many cases where you don't need to have the full
382 definition of a class. If you are using a pointer or reference to a class, you
383 don't need the header file. If you are simply returning a class instance from a
384 prototyped function or method, you don't need it. In fact, for most cases, you
385 simply don't need the definition of a class. And not ``#include``\ing speeds up
388 It is easy to try to go too overboard on this recommendation, however. You
389 **must** include all of the header files that you are using --- you can include
390 them either directly or indirectly through another header file. To make sure
391 that you don't accidentally forget to include a header file in your module
392 header, make sure to include your module header **first** in the implementation
393 file (as mentioned above). This way there won't be any hidden dependencies that
394 you'll find out about later.
396 Keep "Internal" Headers Private
397 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
399 Many modules have a complex implementation that causes them to use more than one
400 implementation (``.cpp``) file. It is often tempting to put the internal
401 communication interface (helper classes, extra functions, etc) in the public
402 module header file. Don't do this!
404 If you really need to do something like this, put a private header file in the
405 same directory as the source files, and include it locally. This ensures that
406 your private interface remains private and undisturbed by outsiders.
410 It's okay to put extra implementation methods in a public class itself. Just
411 make them private (or protected) and all is well.
415 Use Early Exits and ``continue`` to Simplify Code
416 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
418 When reading code, keep in mind how much state and how many previous decisions
419 have to be remembered by the reader to understand a block of code. Aim to
420 reduce indentation where possible when it doesn't make it more difficult to
421 understand the code. One great way to do this is by making use of early exits
422 and the ``continue`` keyword in long loops. As an example of using an early
423 exit from a function, consider this "bad" code:
427 Value *doSomething(Instruction *I) {
428 if (!isa<TerminatorInst>(I) &&
429 I->hasOneUse() && doOtherThing(I)) {
430 ... some long code ....
436 This code has several problems if the body of the ``'if'`` is large. When
437 you're looking at the top of the function, it isn't immediately clear that this
438 *only* does interesting things with non-terminator instructions, and only
439 applies to things with the other predicates. Second, it is relatively difficult
440 to describe (in comments) why these predicates are important because the ``if``
441 statement makes it difficult to lay out the comments. Third, when you're deep
442 within the body of the code, it is indented an extra level. Finally, when
443 reading the top of the function, it isn't clear what the result is if the
444 predicate isn't true; you have to read to the end of the function to know that
447 It is much preferred to format the code like this:
451 Value *doSomething(Instruction *I) {
452 // Terminators never need 'something' done to them because ...
453 if (isa<TerminatorInst>(I))
456 // We conservatively avoid transforming instructions with multiple uses
457 // because goats like cheese.
461 // This is really just here for example.
462 if (!doOtherThing(I))
465 ... some long code ....
468 This fixes these problems. A similar problem frequently happens in ``for``
469 loops. A silly example is something like this:
473 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
474 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
475 Value *LHS = BO->getOperand(0);
476 Value *RHS = BO->getOperand(1);
483 When you have very, very small loops, this sort of structure is fine. But if it
484 exceeds more than 10-15 lines, it becomes difficult for people to read and
485 understand at a glance. The problem with this sort of code is that it gets very
486 nested very quickly. Meaning that the reader of the code has to keep a lot of
487 context in their brain to remember what is going immediately on in the loop,
488 because they don't know if/when the ``if`` conditions will have ``else``\s etc.
489 It is strongly preferred to structure the loop like this:
493 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
494 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
497 Value *LHS = BO->getOperand(0);
498 Value *RHS = BO->getOperand(1);
499 if (LHS == RHS) continue;
504 This has all the benefits of using early exits for functions: it reduces nesting
505 of the loop, it makes it easier to describe why the conditions are true, and it
506 makes it obvious to the reader that there is no ``else`` coming up that they
507 have to push context into their brain for. If a loop is large, this can be a
508 big understandability win.
510 Don't use ``else`` after a ``return``
511 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
513 For similar reasons above (reduction of indentation and easier reading), please
514 do not use ``'else'`` or ``'else if'`` after something that interrupts control
515 flow --- like ``return``, ``break``, ``continue``, ``goto``, etc. For
516 example, this is *bad*:
522 Type = Context.getsigjmp_bufType();
524 Error = ASTContext::GE_Missing_sigjmp_buf;
530 Type = Context.getjmp_bufType();
532 Error = ASTContext::GE_Missing_jmp_buf;
540 It is better to write it like this:
546 Type = Context.getsigjmp_bufType();
548 Error = ASTContext::GE_Missing_sigjmp_buf;
552 Type = Context.getjmp_bufType();
554 Error = ASTContext::GE_Missing_jmp_buf;
560 Or better yet (in this case) as:
566 Type = Context.getsigjmp_bufType();
568 Type = Context.getjmp_bufType();
571 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
572 ASTContext::GE_Missing_jmp_buf;
577 The idea is to reduce indentation and the amount of code you have to keep track
578 of when reading the code.
580 Turn Predicate Loops into Predicate Functions
581 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
583 It is very common to write small loops that just compute a boolean value. There
584 are a number of ways that people commonly write these, but an example of this
589 bool FoundFoo = false;
590 for (unsigned i = 0, e = BarList.size(); i != e; ++i)
591 if (BarList[i]->isFoo()) {
600 This sort of code is awkward to write, and is almost always a bad sign. Instead
601 of this sort of loop, we strongly prefer to use a predicate function (which may
602 be `static`_) that uses `early exits`_ to compute the predicate. We prefer the
603 code to be structured like this:
607 /// containsFoo - Return true if the specified list has an element that is
609 static bool containsFoo(const std::vector<Bar*> &List) {
610 for (unsigned i = 0, e = List.size(); i != e; ++i)
611 if (List[i]->isFoo())
617 if (containsFoo(BarList)) {
621 There are many reasons for doing this: it reduces indentation and factors out
622 code which can often be shared by other code that checks for the same predicate.
623 More importantly, it *forces you to pick a name* for the function, and forces
624 you to write a comment for it. In this silly example, this doesn't add much
625 value. However, if the condition is complex, this can make it a lot easier for
626 the reader to understand the code that queries for this predicate. Instead of
627 being faced with the in-line details of how we check to see if the BarList
628 contains a foo, we can trust the function name and continue reading with better
634 Name Types, Functions, Variables, and Enumerators Properly
635 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
637 Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
638 enough how important it is to use *descriptive* names. Pick names that match
639 the semantics and role of the underlying entities, within reason. Avoid
640 abbreviations unless they are well known. After picking a good name, make sure
641 to use consistent capitalization for the name, as inconsistency requires clients
642 to either memorize the APIs or to look it up to find the exact spelling.
644 In general, names should be in camel case (e.g. ``TextFileReader`` and
645 ``isLValue()``). Different kinds of declarations have different rules:
647 * **Type names** (including classes, structs, enums, typedefs, etc) should be
648 nouns and start with an upper-case letter (e.g. ``TextFileReader``).
650 * **Variable names** should be nouns (as they represent state). The name should
651 be camel case, and start with an upper case letter (e.g. ``Leader`` or
654 * **Function names** should be verb phrases (as they represent actions), and
655 command-like function should be imperative. The name should be camel case,
656 and start with a lower case letter (e.g. ``openFile()`` or ``isFoo()``).
658 * **Enum declarations** (e.g. ``enum Foo {...}``) are types, so they should
659 follow the naming conventions for types. A common use for enums is as a
660 discriminator for a union, or an indicator of a subclass. When an enum is
661 used for something like this, it should have a ``Kind`` suffix
662 (e.g. ``ValueKind``).
664 * **Enumerators** (e.g. ``enum { Foo, Bar }``) and **public member variables**
665 should start with an upper-case letter, just like types. Unless the
666 enumerators are defined in their own small namespace or inside a class,
667 enumerators should have a prefix corresponding to the enum declaration name.
668 For example, ``enum ValueKind { ... };`` may contain enumerators like
669 ``VK_Argument``, ``VK_BasicBlock``, etc. Enumerators that are just
670 convenience constants are exempt from the requirement for a prefix. For
680 As an exception, classes that mimic STL classes can have member names in STL's
681 style of lower-case words separated by underscores (e.g. ``begin()``,
682 ``push_back()``, and ``empty()``).
684 Here are some examples of good and bad names:
690 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
691 Factory<Tire> Factory; // Better.
692 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
693 // kind of factories.
696 Vehicle MakeVehicle(VehicleType Type) {
697 VehicleMaker M; // Might be OK if having a short life-span.
698 Tire tmp1 = M.makeTire(); // Bad -- 'tmp1' provides no information.
699 Light headlight = M.makeLight("head"); // Good -- descriptive.
706 Use the "``assert``" macro to its fullest. Check all of your preconditions and
707 assumptions, you never know when a bug (not necessarily even yours) might be
708 caught early by an assertion, which reduces debugging time dramatically. The
709 "``<cassert>``" header file is probably already included by the header files you
710 are using, so it doesn't cost anything to use it.
712 To further assist with debugging, make sure to put some kind of error message in
713 the assertion statement, which is printed if the assertion is tripped. This
714 helps the poor debugger make sense of why an assertion is being made and
715 enforced, and hopefully what to do about it. Here is one complete example:
719 inline Value *getOperand(unsigned i) {
720 assert(i < Operands.size() && "getOperand() out of range!");
724 Here are more examples:
728 assert(Ty->isPointerType() && "Can't allocate a non pointer type!");
730 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
732 assert(idx < getNumSuccessors() && "Successor # out of range!");
734 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
736 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
740 Please be aware that, when adding assert statements, not all compilers are aware
741 of the semantics of the assert. In some places, asserts are used to indicate a
742 piece of code that should not be reached. These are typically of the form:
746 assert(0 && "Some helpful error message");
748 When used in a function that returns a value, they should be followed with a
749 return statement and a comment indicating that this line is never reached. This
750 will prevent a compiler which is unable to deduce that the assert statement
751 never returns from generating a warning.
755 assert(0 && "Some helpful error message");
758 Another issue is that values used only by assertions will produce an "unused
759 value" warning when assertions are disabled. For example, this code will warn:
763 unsigned Size = V.size();
764 assert(Size > 42 && "Vector smaller than it should be");
766 bool NewToSet = Myset.insert(Value);
767 assert(NewToSet && "The value shouldn't be in the set yet");
769 These are two interesting different cases. In the first case, the call to
770 ``V.size()`` is only useful for the assert, and we don't want it executed when
771 assertions are disabled. Code like this should move the call into the assert
772 itself. In the second case, the side effects of the call must happen whether
773 the assert is enabled or not. In this case, the value should be cast to void to
774 disable the warning. To be specific, it is preferred to write the code like
779 assert(V.size() > 42 && "Vector smaller than it should be");
781 bool NewToSet = Myset.insert(Value); (void)NewToSet;
782 assert(NewToSet && "The value shouldn't be in the set yet");
784 Do Not Use ``using namespace std``
785 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
787 In LLVM, we prefer to explicitly prefix all identifiers from the standard
788 namespace with an "``std::``" prefix, rather than rely on "``using namespace
791 In header files, adding a ``'using namespace XXX'`` directive pollutes the
792 namespace of any source file that ``#include``\s the header. This is clearly a
795 In implementation files (e.g. ``.cpp`` files), the rule is more of a stylistic
796 rule, but is still important. Basically, using explicit namespace prefixes
797 makes the code **clearer**, because it is immediately obvious what facilities
798 are being used and where they are coming from. And **more portable**, because
799 namespace clashes cannot occur between LLVM code and other namespaces. The
800 portability rule is important because different standard library implementations
801 expose different symbols (potentially ones they shouldn't), and future revisions
802 to the C++ standard will add more symbols to the ``std`` namespace. As such, we
803 never use ``'using namespace std;'`` in LLVM.
805 The exception to the general rule (i.e. it's not an exception for the ``std``
806 namespace) is for implementation files. For example, all of the code in the
807 LLVM project implements code that lives in the 'llvm' namespace. As such, it is
808 ok, and actually clearer, for the ``.cpp`` files to have a ``'using namespace
809 llvm;'`` directive at the top, after the ``#include``\s. This reduces
810 indentation in the body of the file for source editors that indent based on
811 braces, and keeps the conceptual context cleaner. The general form of this rule
812 is that any ``.cpp`` file that implements code in any namespace may use that
813 namespace (and its parents'), but should not use any others.
815 Provide a Virtual Method Anchor for Classes in Headers
816 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
818 If a class is defined in a header file and has a vtable (either it has virtual
819 methods or it derives from classes with virtual methods), it must always have at
820 least one out-of-line virtual method in the class. Without this, the compiler
821 will copy the vtable and RTTI into every ``.o`` file that ``#include``\s the
822 header, bloating ``.o`` file sizes and increasing link times.
824 Don't use default labels in fully covered switches over enumerations
825 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
827 ``-Wswitch`` warns if a switch, without a default label, over an enumeration
828 does not cover every enumeration value. If you write a default label on a fully
829 covered switch over an enumeration then the ``-Wswitch`` warning won't fire
830 when new elements are added to that enumeration. To help avoid adding these
831 kinds of defaults, Clang has the warning ``-Wcovered-switch-default`` which is
832 off by default but turned on when building LLVM with a version of Clang that
833 supports the warning.
835 A knock-on effect of this stylistic requirement is that when building LLVM with
836 GCC you may get warnings related to "control may reach end of non-void function"
837 if you return from each case of a covered switch-over-enum because GCC assumes
838 that the enum expression may take any representable value, not just those of
839 individual enumerators. To suppress this warning, use ``llvm_unreachable`` after
842 Use ``LLVM_DELETED_FUNCTION`` to mark uncallable methods
843 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
845 Prior to C++11, a common pattern to make a class uncopyable was to declare an
846 unimplemented copy constructor and copy assignment operator and make them
847 private. This would give a compiler error for accessing a private method or a
848 linker error because it wasn't implemented.
850 With C++11, we can mark methods that won't be implemented with ``= delete``.
851 This will trigger a much better error message and tell the compiler that the
852 method will never be implemented. This enables other checks like
853 ``-Wunused-private-field`` to run correctly on classes that contain these
856 To maintain compatibility with C++03, ``LLVM_DELETED_FUNCTION`` should be used
857 which will expand to ``= delete`` if the compiler supports it. These methods
858 should still be declared private. Example of the uncopyable pattern:
864 DontCopy(const DontCopy&) LLVM_DELETED_FUNCTION;
865 DontCopy &operator =(const DontCopy&) LLVM_DELETED_FUNCTION;
870 Don't evaluate ``end()`` every time through a loop
871 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
873 Because C++ doesn't have a standard "``foreach``" loop (though it can be
874 emulated with macros and may be coming in C++'0x) we end up writing a lot of
875 loops that manually iterate from begin to end on a variety of containers or
876 through other data structures. One common mistake is to write a loop in this
882 for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I)
885 The problem with this construct is that it evaluates "``BB->end()``" every time
886 through the loop. Instead of writing the loop like this, we strongly prefer
887 loops to be written so that they evaluate it once before the loop starts. A
888 convenient way to do this is like so:
893 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
896 The observant may quickly point out that these two loops may have different
897 semantics: if the container (a basic block in this case) is being mutated, then
898 "``BB->end()``" may change its value every time through the loop and the second
899 loop may not in fact be correct. If you actually do depend on this behavior,
900 please write the loop in the first form and add a comment indicating that you
901 did it intentionally.
903 Why do we prefer the second form (when correct)? Writing the loop in the first
904 form has two problems. First it may be less efficient than evaluating it at the
905 start of the loop. In this case, the cost is probably minor --- a few extra
906 loads every time through the loop. However, if the base expression is more
907 complex, then the cost can rise quickly. I've seen loops where the end
908 expression was actually something like: "``SomeMap[x]->end()``" and map lookups
909 really aren't cheap. By writing it in the second form consistently, you
910 eliminate the issue entirely and don't even have to think about it.
912 The second (even bigger) issue is that writing the loop in the first form hints
913 to the reader that the loop is mutating the container (a fact that a comment
914 would handily confirm!). If you write the loop in the second form, it is
915 immediately obvious without even looking at the body of the loop that the
916 container isn't being modified, which makes it easier to read the code and
917 understand what it does.
919 While the second form of the loop is a few extra keystrokes, we do strongly
922 ``#include <iostream>`` is Forbidden
923 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
925 The use of ``#include <iostream>`` in library files is hereby **forbidden**,
926 because many common implementations transparently inject a `static constructor`_
927 into every translation unit that includes it.
929 Note that using the other stream headers (``<sstream>`` for example) is not
930 problematic in this regard --- just ``<iostream>``. However, ``raw_ostream``
931 provides various APIs that are better performing for almost every use than
932 ``std::ostream`` style APIs.
936 New code should always use `raw_ostream`_ for writing, or the
937 ``llvm::MemoryBuffer`` API for reading files.
944 LLVM includes a lightweight, simple, and efficient stream implementation in
945 ``llvm/Support/raw_ostream.h``, which provides all of the common features of
946 ``std::ostream``. All new code should use ``raw_ostream`` instead of
949 Unlike ``std::ostream``, ``raw_ostream`` is not a template and can be forward
950 declared as ``class raw_ostream``. Public headers should generally not include
951 the ``raw_ostream`` header, but use forward declarations and constant references
952 to ``raw_ostream`` instances.
957 The ``std::endl`` modifier, when used with ``iostreams`` outputs a newline to
958 the output stream specified. In addition to doing this, however, it also
959 flushes the output stream. In other words, these are equivalent:
963 std::cout << std::endl;
964 std::cout << '\n' << std::flush;
966 Most of the time, you probably have no reason to flush the output stream, so
967 it's better to use a literal ``'\n'``.
972 This section describes preferred low-level formatting guidelines along with
973 reasoning on why we prefer them.
975 Spaces Before Parentheses
976 ^^^^^^^^^^^^^^^^^^^^^^^^^
978 We prefer to put a space before an open parenthesis only in control flow
979 statements, but not in normal function call expressions and function-like
980 macros. For example, this is good:
985 for (i = 0; i != 100; ++i) ...
986 while (llvm_rocks) ...
989 assert(3 != 4 && "laws of math are failing me");
991 a = foo(42, 92) + bar(x);
998 for(i = 0; i != 100; ++i) ...
999 while(llvm_rocks) ...
1002 assert (3 != 4 && "laws of math are failing me");
1004 a = foo (42, 92) + bar (x);
1006 The reason for doing this is not completely arbitrary. This style makes control
1007 flow operators stand out more, and makes expressions flow better. The function
1008 call operator binds very tightly as a postfix operator. Putting a space after a
1009 function name (as in the last example) makes it appear that the code might bind
1010 the arguments of the left-hand-side of a binary operator with the argument list
1011 of a function and the name of the right side. More specifically, it is easy to
1012 misread the "``a``" example as:
1016 a = foo ((42, 92) + bar) (x);
1018 when skimming through the code. By avoiding a space in a function, we avoid
1019 this misinterpretation.
1024 Hard fast rule: Preincrement (``++X``) may be no slower than postincrement
1025 (``X++``) and could very well be a lot faster than it. Use preincrementation
1028 The semantics of postincrement include making a copy of the value being
1029 incremented, returning it, and then preincrementing the "work value". For
1030 primitive types, this isn't a big deal. But for iterators, it can be a huge
1031 issue (for example, some iterators contains stack and set objects in them...
1032 copying an iterator could invoke the copy ctor's of these as well). In general,
1033 get in the habit of always using preincrement, and you won't have a problem.
1036 Namespace Indentation
1037 ^^^^^^^^^^^^^^^^^^^^^
1039 In general, we strive to reduce indentation wherever possible. This is useful
1040 because we want code to `fit into 80 columns`_ without wrapping horribly, but
1041 also because it makes it easier to understand the code. Namespaces are a funny
1042 thing: they are often large, and we often desire to put lots of stuff into them
1043 (so they can be large). Other times they are tiny, because they just hold an
1044 enum or something similar. In order to balance this, we use different
1045 approaches for small versus large namespaces.
1047 If a namespace definition is small and *easily* fits on a screen (say, less than
1048 35 lines of code), then you should indent its body. Here's an example:
1054 /// RelocationType - An enum for the x86 relocation codes. Note that
1055 /// the terminology here doesn't follow x86 convention - word means
1056 /// 32-bit and dword means 64-bit.
1057 enum RelocationType {
1058 /// reloc_pcrel_word - PC relative relocation, add the relocated value to
1059 /// the value already in memory, after we adjust it for where the PC is.
1060 reloc_pcrel_word = 0,
1062 /// reloc_picrel_word - PIC base relative relocation, add the relocated
1063 /// value to the value already in memory, after we adjust it for where the
1065 reloc_picrel_word = 1,
1067 /// reloc_absolute_word, reloc_absolute_dword - Absolute relocation, just
1068 /// add the relocated value to the value already in memory.
1069 reloc_absolute_word = 2,
1070 reloc_absolute_dword = 3
1075 Since the body is small, indenting adds value because it makes it very clear
1076 where the namespace starts and ends, and it is easy to take the whole thing in
1077 in one "gulp" when reading the code. If the blob of code in the namespace is
1078 larger (as it typically is in a header in the ``llvm`` or ``clang`` namespaces),
1079 do not indent the code, and add a comment indicating what namespace is being
1080 closed. For example:
1085 namespace knowledge {
1087 /// Grokable - This class represents things that Smith can have an intimate
1088 /// understanding of and contains the data associated with it.
1092 explicit Grokable() { ... }
1093 virtual ~Grokable() = 0;
1099 } // end namespace knowledge
1100 } // end namespace llvm
1102 Because the class is large, we don't expect that the reader can easily
1103 understand the entire concept in a glance, and the end of the file (where the
1104 namespaces end) may be a long ways away from the place they open. As such,
1105 indenting the contents of the namespace doesn't add any value, and detracts from
1106 the readability of the class. In these cases it is best to *not* indent the
1107 contents of the namespace.
1111 Anonymous Namespaces
1112 ^^^^^^^^^^^^^^^^^^^^
1114 After talking about namespaces in general, you may be wondering about anonymous
1115 namespaces in particular. Anonymous namespaces are a great language feature
1116 that tells the C++ compiler that the contents of the namespace are only visible
1117 within the current translation unit, allowing more aggressive optimization and
1118 eliminating the possibility of symbol name collisions. Anonymous namespaces are
1119 to C++ as "static" is to C functions and global variables. While "``static``"
1120 is available in C++, anonymous namespaces are more general: they can make entire
1121 classes private to a file.
1123 The problem with anonymous namespaces is that they naturally want to encourage
1124 indentation of their body, and they reduce locality of reference: if you see a
1125 random function definition in a C++ file, it is easy to see if it is marked
1126 static, but seeing if it is in an anonymous namespace requires scanning a big
1129 Because of this, we have a simple guideline: make anonymous namespaces as small
1130 as possible, and only use them for class declarations. For example, this is
1140 bool operator<(const char *RHS) const;
1142 } // end anonymous namespace
1144 static void runHelper() {
1148 bool StringSort::operator<(const char *RHS) const {
1161 bool operator<(const char *RHS) const;
1168 bool StringSort::operator<(const char *RHS) const {
1172 } // end anonymous namespace
1174 This is bad specifically because if you're looking at "``runHelper``" in the middle
1175 of a large C++ file, that you have no immediate way to tell if it is local to
1176 the file. When it is marked static explicitly, this is immediately obvious.
1177 Also, there is no reason to enclose the definition of "``operator<``" in the
1178 namespace just because it was declared there.
1183 A lot of these comments and recommendations have been culled for other sources.
1184 Two particularly important books for our work are:
1187 <http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876>`_
1188 by Scott Meyers. Also interesting and useful are "More Effective C++" and
1189 "Effective STL" by the same author.
1191 #. `Large-Scale C++ Software Design
1192 <http://www.amazon.com/Large-Scale-Software-Design-John-Lakos/dp/0201633620/ref=sr_1_1>`_
1195 If you get some free time, and you haven't read them: do so, you might learn