1.. _coding_standards: 2 3===================== 4LLVM Coding Standards 5===================== 6 7.. contents:: 8 :local: 9 10Introduction 11============ 12 13This document attempts to describe a few coding standards that are being used in 14the LLVM source tree. Although no coding standards should be regarded as 15absolute requirements to be followed in all instances, coding standards are 16particularly important for large-scale code bases that follow a library-based 17design (like LLVM). 18 19This document intentionally does not prescribe fixed standards for religious 20issues such as brace placement and space usage. For issues like this, follow 21the golden rule: 22 23.. _Golden Rule: 24 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 27 easy to follow.** 28 29Note that some code bases (e.g. ``libc++``) have really good reasons to deviate 30from the coding standards. In the case of ``libc++``, this is because the 31naming and other conventions are dictated by the C++ standard. If you think 32there is a specific good reason to deviate from the standards here, please bring 33it up on the LLVMdev mailing list. 34 35There 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 37lot of code was written before they were put in place. Our long term goal is 38for the entire codebase to follow the convention, but we explicitly *do not* 39want patches that do large-scale reformating of existing code. On the other 40hand, it is reasonable to rename the methods of a class if you're about to 41change it in some other way. Just do the reformating as a separate commit from 42the functionality change. 43 44The ultimate goal of these guidelines is the increase readability and 45maintainability of our common source base. If you have suggestions for topics to 46be included, please mail them to `Chris <mailto:sabre@nondot.org>`_. 47 48Mechanical Source Issues 49======================== 50 51Source Code Formatting 52---------------------- 53 54Commenting 55^^^^^^^^^^ 56 57Comments are one critical part of readability and maintainability. Everyone 58knows they should comment their code, and so should you. When writing comments, 59write them as English prose, which means they should use proper capitalization, 60punctuation, 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: 62 63.. _header file comment: 64 65File Headers 66"""""""""""" 67 68Every source file should have a header on it that describes the basic purpose of 69the file. If a file does not have a header, it should not be checked into the 70tree. The standard header looks like this: 71 72.. code-block:: c++ 73 74 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===// 75 // 76 // The LLVM Compiler Infrastructure 77 // 78 // This file is distributed under the University of Illinois Open Source 79 // License. See LICENSE.TXT for details. 80 // 81 //===----------------------------------------------------------------------===// 82 /// 83 /// \file 84 /// \brief This file contains the declaration of the Instruction class, which is 85 /// the base class for all of the VM instructions. 86 /// 87 //===----------------------------------------------------------------------===// 88 89A few things to note about this particular format: The "``-*- C++ -*-``" string 90on the first line is there to tell Emacs that the source file is a C++ file, not 91a C file (Emacs assumes ``.h`` files are C files by default). 92 93.. note:: 94 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 98 pages. 99 100The next section in the file is a concise note that defines the license that the 101file is released under. This makes it perfectly clear what terms the source 102code can be distributed under and should not be modified in any way. 103 104The main body is a ``doxygen`` comment describing the purpose of the file. It 105should have a ``\brief`` command that describes the file in one or two 106sentences. Any additional information should be separated by a blank line. If 107an algorithm is being implemented or something tricky is going on, a reference 108to the paper where it is published should be included, as well as any notes or 109*gotchas* in the code to watch out for. 110 111Class overviews 112""""""""""""""" 113 114Classes are one fundamental part of a good object oriented design. As such, a 115class definition should have a comment block that explains what the class is 116used for and how it works. Every non-trivial class is expected to have a 117``doxygen`` comment block. 118 119Method information 120"""""""""""""""""" 121 122Methods defined in a class (as well as any global functions) should also be 123documented properly. A quick note about what it does and a description of the 124borderline behaviour is all that is necessary here (unless something 125particularly tricky or insidious is going on). The hope is that people can 126figure out how to use your interfaces without reading the code itself. 127 128Good things to talk about here are what happens when something unexpected 129happens: does the method return null? Abort? Format your hard disk? 130 131Comment Formatting 132^^^^^^^^^^^^^^^^^^ 133 134In general, prefer C++ style (``//``) comments. They take less space, require 135less typing, don't have nesting problems, etc. There are a few cases when it is 136useful to use C style (``/* */``) comments however: 137 138#. When writing C code: Obviously if you are writing C code, use C style 139 comments. 140 141#. When writing a header file that may be ``#include``\d by a C source file. 142 143#. When writing a source file that is used by a tool that only accepts C style 144 comments. 145 146To comment out a large block of code, use ``#if 0`` and ``#endif``. These nest 147properly and are better behaved in general than C style comments. 148 149``#include`` Style 150^^^^^^^^^^^^^^^^^^ 151 152Immediately after the `header file comment`_ (and include guards if working on a 153header file), the `minimal list of #includes`_ required by the file should be 154listed. We prefer these ``#include``\s to be listed in this order: 155 156.. _Main Module Header: 157.. _Local/Private Headers: 158 159#. Main Module Header 160#. Local/Private Headers 161#. ``llvm/*`` 162#. ``llvm/Analysis/*`` 163#. ``llvm/Assembly/*`` 164#. ``llvm/Bitcode/*`` 165#. ``llvm/CodeGen/*`` 166#. ... 167#. ``llvm/Support/*`` 168#. ``llvm/Config/*`` 169#. System ``#include``\s 170 171and each category should be sorted by name. 172 173The `Main Module Header`_ file applies to ``.cpp`` files which implement an 174interface 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 176header file first in the ``.cpp`` files that implement the interfaces, we ensure 177that 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 179in the ``.cpp`` file to indicate where the interfaces it implements are defined. 180 181.. _fit into 80 columns: 182 183Source Code Width 184^^^^^^^^^^^^^^^^^ 185 186Write your code to fit within 80 columns of text. This helps those of us who 187like to print out code and look at your code in an ``xterm`` without resizing 188it. 189 190The longer answer is that there must be some limit to the width of the code in 191order to reasonably allow developers to have multiple files side-by-side in 192windows on a modest display. If you are going to pick a width limit, it is 193somewhat arbitrary but you might as well pick something standard. Going with 90 194columns (for example) instead of 80 columns wouldn't add any significant value 195and would be detrimental to printing out code. Also many other projects have 196standardized on 80 columns, so some people have already configured their editors 197for it (vs something else, like 90 columns). 198 199This is one of many contentious issues in coding standards, but it is not up for 200debate. 201 202Use Spaces Instead of Tabs 203^^^^^^^^^^^^^^^^^^^^^^^^^^ 204 205In all cases, prefer spaces to tabs in source files. People have different 206preferred indentation levels, and different styles of indentation that they 207like; this is fine. What isn't fine is that different editors/viewers expand 208tabs out to different tab stops. This can cause your code to look completely 209unreadable, and it is not worth dealing with. 210 211As always, follow the `Golden Rule`_ above: follow the style of 212existing code if you are modifying and extending it. If you like four spaces of 213indentation, **DO NOT** do that in the middle of a chunk of code with two spaces 214of indentation. Also, do not reindent a whole source file: it makes for 215incredible diffs that are absolutely worthless. 216 217Indent Code Consistently 218^^^^^^^^^^^^^^^^^^^^^^^^ 219 220Okay, in your first year of programming you were told that indentation is 221important. If you didn't believe and internalize this then, now is the time. 222Just do it. 223 224Compiler Issues 225--------------- 226 227Treat Compiler Warnings Like Errors 228^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 229 230If your code has compiler warnings in it, something is wrong --- you aren't 231casting values correctly, you have "questionable" constructs in your code, or 232you are doing something legitimately wrong. Compiler warnings can cover up 233legitimate errors in output and make dealing with a translation unit difficult. 234 235It is not possible to prevent all warnings from all compilers, nor is it 236desirable. Instead, pick a standard compiler (like ``gcc``) that provides a 237good 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 239syntax of the code slightly. For example, a warning that annoys me occurs when 240I write code like this: 241 242.. code-block:: c++ 243 244 if (V = getValue()) { 245 ... 246 } 247 248``gcc`` will warn me that I probably want to use the ``==`` operator, and that I 249probably mistyped it. In most cases, I haven't, and I really don't want the 250spurious errors. To fix this particular problem, I rewrite the code like 251this: 252 253.. code-block:: c++ 254 255 if ((V = getValue())) { 256 ... 257 } 258 259which shuts ``gcc`` up. Any ``gcc`` warning that annoys you can be fixed by 260massaging the code appropriately. 261 262Write Portable Code 263^^^^^^^^^^^^^^^^^^^ 264 265In almost all cases, it is possible and within reason to write completely 266portable code. If there are cases where it isn't possible to write portable 267code, isolate it behind a well defined (and well documented) interface. 268 269In 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 271features are used, they should only be an implementation detail of a library 272which has a simple exposed API, and preferably be buried in ``libSystem``. 273 274Do not use RTTI or Exceptions 275^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 276 277In 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 279the general C++ principle of *"you only pay for what you use"*, causing 280executable bloat even if exceptions are never used in the code base, or if RTTI 281is never used for a class. Because of this, we turn them off globally in the 282code. 283 284That said, LLVM does make extensive use of a hand-rolled form of RTTI that use 285templates like `isa<>, cast<>, and dyn_cast<> <ProgrammersManual.html#isa>`_. 286This form of RTTI is opt-in and can be added to any class. It is also 287substantially more efficient than ``dynamic_cast<>``. 288 289.. _static constructor: 290 291Do not use Static Constructors 292^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 293 294Static constructors and destructors (e.g. global variables whose types have a 295constructor or destructor) should not be added to the code base, and should be 296removed wherever possible. Besides `well known problems 297<http://yosefk.com/c++fqa/ctors.html#fqa-10.12>`_ where the order of 298initialization is undefined between globals in different source files, the 299entire concept of static constructors is at odds with the common use case of 300LLVM as a library linked into a larger application. 301 302Consider 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 305design of static constructors, they must be executed at startup time of the 306entire application, regardless of whether or how LLVM is used in that larger 307application. There are two problems with this: 308 309* The time to run the static constructors impacts startup time of applications 310 --- a critical time for GUI apps, among others. 311 312* 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. 316 317We would really like for there to be zero cost for linking in an additional LLVM 318target or other library into an application, but static constructors violate 319this goal. 320 321That said, LLVM unfortunately does contain static constructors. It would be a 322`great project <http://llvm.org/PR11944>`_ for someone to purge all static 323constructors from LLVM, and then enable the ``-Wglobal-constructors`` warning 324flag (when building with Clang) to ensure we do not regress in the future. 325 326Use of ``class`` and ``struct`` Keywords 327^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 328 329In C++, the ``class`` and ``struct`` keywords can be used almost 330interchangeably. The only difference is when they are used to declare a class: 331``class`` makes all members private by default while ``struct`` makes all 332members public by default. 333 334Unfortunately, not all compilers follow the rules and some will generate 335different symbols based on whether ``class`` or ``struct`` was used to declare 336the symbol. This can lead to problems at link time. 337 338So, the rule for LLVM is to always use the ``class`` keyword, unless **all** 339members 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. 342 343Style Issues 344============ 345 346The High-Level Issues 347--------------------- 348 349A Public Header File **is** a Module 350^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 351 352C++ doesn't do too well in the modularity department. There is no real 353encapsulation or data hiding (unless you use expensive protocol classes), but it 354is what we have to work with. When you write a public header file (in the LLVM 355source tree, they live in the top level "``include``" directory), you are 356defining a module of functionality. 357 358Ideally, modules should be completely independent of each other, and their 359header files should only ``#include`` the absolute minimum number of headers 360possible. A module is not just a class, a function, or a namespace: it's a 361collection of these that defines an interface. This interface may be several 362functions, classes, or data structures, but the important issue is how they work 363together. 364 365In general, a module should be implemented by one or more ``.cpp`` files. Each 366of these ``.cpp`` files should include the header that defines their interface 367first. This ensures that all of the dependences of the module header have been 368properly added to the module header itself, and are not implicit. System 369headers should be included after user headers for a translation unit. 370 371.. _minimal list of #includes: 372 373``#include`` as Little as Possible 374^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 375 376``#include`` hurts compile time performance. Don't do it unless you have to, 377especially in header files. 378 379But wait! Sometimes you need to have the definition of a class to use it, or to 380inherit from it. In these cases go ahead and ``#include`` that header file. Be 381aware however that there are many cases where you don't need to have the full 382definition of a class. If you are using a pointer or reference to a class, you 383don't need the header file. If you are simply returning a class instance from a 384prototyped function or method, you don't need it. In fact, for most cases, you 385simply don't need the definition of a class. And not ``#include``\ing speeds up 386compilation. 387 388It 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 390them either directly or indirectly through another header file. To make sure 391that you don't accidentally forget to include a header file in your module 392header, make sure to include your module header **first** in the implementation 393file (as mentioned above). This way there won't be any hidden dependencies that 394you'll find out about later. 395 396Keep "Internal" Headers Private 397^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 398 399Many modules have a complex implementation that causes them to use more than one 400implementation (``.cpp``) file. It is often tempting to put the internal 401communication interface (helper classes, extra functions, etc) in the public 402module header file. Don't do this! 403 404If you really need to do something like this, put a private header file in the 405same directory as the source files, and include it locally. This ensures that 406your private interface remains private and undisturbed by outsiders. 407 408.. note:: 409 410 It's okay to put extra implementation methods in a public class itself. Just 411 make them private (or protected) and all is well. 412 413.. _early exits: 414 415Use Early Exits and ``continue`` to Simplify Code 416^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 417 418When reading code, keep in mind how much state and how many previous decisions 419have to be remembered by the reader to understand a block of code. Aim to 420reduce indentation where possible when it doesn't make it more difficult to 421understand the code. One great way to do this is by making use of early exits 422and the ``continue`` keyword in long loops. As an example of using an early 423exit from a function, consider this "bad" code: 424 425.. code-block:: c++ 426 427 Value *doSomething(Instruction *I) { 428 if (!isa<TerminatorInst>(I) && 429 I->hasOneUse() && doOtherThing(I)) { 430 ... some long code .... 431 } 432 433 return 0; 434 } 435 436This code has several problems if the body of the ``'if'`` is large. When 437you'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 439applies to things with the other predicates. Second, it is relatively difficult 440to describe (in comments) why these predicates are important because the ``if`` 441statement makes it difficult to lay out the comments. Third, when you're deep 442within the body of the code, it is indented an extra level. Finally, when 443reading the top of the function, it isn't clear what the result is if the 444predicate isn't true; you have to read to the end of the function to know that 445it returns null. 446 447It is much preferred to format the code like this: 448 449.. code-block:: c++ 450 451 Value *doSomething(Instruction *I) { 452 // Terminators never need 'something' done to them because ... 453 if (isa<TerminatorInst>(I)) 454 return 0; 455 456 // We conservatively avoid transforming instructions with multiple uses 457 // because goats like cheese. 458 if (!I->hasOneUse()) 459 return 0; 460 461 // This is really just here for example. 462 if (!doOtherThing(I)) 463 return 0; 464 465 ... some long code .... 466 } 467 468This fixes these problems. A similar problem frequently happens in ``for`` 469loops. A silly example is something like this: 470 471.. code-block:: c++ 472 473 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); 477 if (LHS != RHS) { 478 ... 479 } 480 } 481 } 482 483When you have very, very small loops, this sort of structure is fine. But if it 484exceeds more than 10-15 lines, it becomes difficult for people to read and 485understand at a glance. The problem with this sort of code is that it gets very 486nested very quickly. Meaning that the reader of the code has to keep a lot of 487context in their brain to remember what is going immediately on in the loop, 488because they don't know if/when the ``if`` conditions will have ``else``\s etc. 489It is strongly preferred to structure the loop like this: 490 491.. code-block:: c++ 492 493 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) { 494 BinaryOperator *BO = dyn_cast<BinaryOperator>(II); 495 if (!BO) continue; 496 497 Value *LHS = BO->getOperand(0); 498 Value *RHS = BO->getOperand(1); 499 if (LHS == RHS) continue; 500 501 ... 502 } 503 504This has all the benefits of using early exits for functions: it reduces nesting 505of the loop, it makes it easier to describe why the conditions are true, and it 506makes it obvious to the reader that there is no ``else`` coming up that they 507have to push context into their brain for. If a loop is large, this can be a 508big understandability win. 509 510Don't use ``else`` after a ``return`` 511^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 512 513For similar reasons above (reduction of indentation and easier reading), please 514do not use ``'else'`` or ``'else if'`` after something that interrupts control 515flow --- like ``return``, ``break``, ``continue``, ``goto``, etc. For 516example, this is *bad*: 517 518.. code-block:: c++ 519 520 case 'J': { 521 if (Signed) { 522 Type = Context.getsigjmp_bufType(); 523 if (Type.isNull()) { 524 Error = ASTContext::GE_Missing_sigjmp_buf; 525 return QualType(); 526 } else { 527 break; 528 } 529 } else { 530 Type = Context.getjmp_bufType(); 531 if (Type.isNull()) { 532 Error = ASTContext::GE_Missing_jmp_buf; 533 return QualType(); 534 } else { 535 break; 536 } 537 } 538 } 539 540It is better to write it like this: 541 542.. code-block:: c++ 543 544 case 'J': 545 if (Signed) { 546 Type = Context.getsigjmp_bufType(); 547 if (Type.isNull()) { 548 Error = ASTContext::GE_Missing_sigjmp_buf; 549 return QualType(); 550 } 551 } else { 552 Type = Context.getjmp_bufType(); 553 if (Type.isNull()) { 554 Error = ASTContext::GE_Missing_jmp_buf; 555 return QualType(); 556 } 557 } 558 break; 559 560Or better yet (in this case) as: 561 562.. code-block:: c++ 563 564 case 'J': 565 if (Signed) 566 Type = Context.getsigjmp_bufType(); 567 else 568 Type = Context.getjmp_bufType(); 569 570 if (Type.isNull()) { 571 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf : 572 ASTContext::GE_Missing_jmp_buf; 573 return QualType(); 574 } 575 break; 576 577The idea is to reduce indentation and the amount of code you have to keep track 578of when reading the code. 579 580Turn Predicate Loops into Predicate Functions 581^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 582 583It is very common to write small loops that just compute a boolean value. There 584are a number of ways that people commonly write these, but an example of this 585sort of thing is: 586 587.. code-block:: c++ 588 589 bool FoundFoo = false; 590 for (unsigned i = 0, e = BarList.size(); i != e; ++i) 591 if (BarList[i]->isFoo()) { 592 FoundFoo = true; 593 break; 594 } 595 596 if (FoundFoo) { 597 ... 598 } 599 600This sort of code is awkward to write, and is almost always a bad sign. Instead 601of this sort of loop, we strongly prefer to use a predicate function (which may 602be `static`_) that uses `early exits`_ to compute the predicate. We prefer the 603code to be structured like this: 604 605.. code-block:: c++ 606 607 /// containsFoo - Return true if the specified list has an element that is 608 /// a foo. 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()) 612 return true; 613 return false; 614 } 615 ... 616 617 if (containsFoo(BarList)) { 618 ... 619 } 620 621There are many reasons for doing this: it reduces indentation and factors out 622code which can often be shared by other code that checks for the same predicate. 623More importantly, it *forces you to pick a name* for the function, and forces 624you to write a comment for it. In this silly example, this doesn't add much 625value. However, if the condition is complex, this can make it a lot easier for 626the reader to understand the code that queries for this predicate. Instead of 627being faced with the in-line details of how we check to see if the BarList 628contains a foo, we can trust the function name and continue reading with better 629locality. 630 631The Low-Level Issues 632-------------------- 633 634Name Types, Functions, Variables, and Enumerators Properly 635^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 636 637Poorly-chosen names can mislead the reader and cause bugs. We cannot stress 638enough how important it is to use *descriptive* names. Pick names that match 639the semantics and role of the underlying entities, within reason. Avoid 640abbreviations unless they are well known. After picking a good name, make sure 641to use consistent capitalization for the name, as inconsistency requires clients 642to either memorize the APIs or to look it up to find the exact spelling. 643 644In general, names should be in camel case (e.g. ``TextFileReader`` and 645``isLValue()``). Different kinds of declarations have different rules: 646 647* **Type names** (including classes, structs, enums, typedefs, etc) should be 648 nouns and start with an upper-case letter (e.g. ``TextFileReader``). 649 650* **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 652 ``Boats``). 653 654* **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()``). 657 658* **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``). 663 664* **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 671 instance: 672 673 .. code-block:: c++ 674 675 enum { 676 MaxSize = 42, 677 Density = 12 678 }; 679 680As an exception, classes that mimic STL classes can have member names in STL's 681style of lower-case words separated by underscores (e.g. ``begin()``, 682``push_back()``, and ``empty()``). 683 684Here are some examples of good and bad names: 685 686.. code-block:: c++ 687 688 class VehicleMaker { 689 ... 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. 694 }; 695 696 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. 700 ... 701 } 702 703Assert Liberally 704^^^^^^^^^^^^^^^^ 705 706Use the "``assert``" macro to its fullest. Check all of your preconditions and 707assumptions, you never know when a bug (not necessarily even yours) might be 708caught early by an assertion, which reduces debugging time dramatically. The 709"``<cassert>``" header file is probably already included by the header files you 710are using, so it doesn't cost anything to use it. 711 712To further assist with debugging, make sure to put some kind of error message in 713the assertion statement, which is printed if the assertion is tripped. This 714helps the poor debugger make sense of why an assertion is being made and 715enforced, and hopefully what to do about it. Here is one complete example: 716 717.. code-block:: c++ 718 719 inline Value *getOperand(unsigned i) { 720 assert(i < Operands.size() && "getOperand() out of range!"); 721 return Operands[i]; 722 } 723 724Here are more examples: 725 726.. code-block:: c++ 727 728 assert(Ty->isPointerType() && "Can't allocate a non pointer type!"); 729 730 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!"); 731 732 assert(idx < getNumSuccessors() && "Successor # out of range!"); 733 734 assert(V1.getType() == V2.getType() && "Constant types must be identical!"); 735 736 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!"); 737 738You get the idea. 739 740Please be aware that, when adding assert statements, not all compilers are aware 741of the semantics of the assert. In some places, asserts are used to indicate a 742piece of code that should not be reached. These are typically of the form: 743 744.. code-block:: c++ 745 746 assert(0 && "Some helpful error message"); 747 748When used in a function that returns a value, they should be followed with a 749return statement and a comment indicating that this line is never reached. This 750will prevent a compiler which is unable to deduce that the assert statement 751never returns from generating a warning. 752 753.. code-block:: c++ 754 755 assert(0 && "Some helpful error message"); 756 return 0; 757 758Another issue is that values used only by assertions will produce an "unused 759value" warning when assertions are disabled. For example, this code will warn: 760 761.. code-block:: c++ 762 763 unsigned Size = V.size(); 764 assert(Size > 42 && "Vector smaller than it should be"); 765 766 bool NewToSet = Myset.insert(Value); 767 assert(NewToSet && "The value shouldn't be in the set yet"); 768 769These 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 771assertions are disabled. Code like this should move the call into the assert 772itself. In the second case, the side effects of the call must happen whether 773the assert is enabled or not. In this case, the value should be cast to void to 774disable the warning. To be specific, it is preferred to write the code like 775this: 776 777.. code-block:: c++ 778 779 assert(V.size() > 42 && "Vector smaller than it should be"); 780 781 bool NewToSet = Myset.insert(Value); (void)NewToSet; 782 assert(NewToSet && "The value shouldn't be in the set yet"); 783 784Do Not Use ``using namespace std`` 785^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 786 787In LLVM, we prefer to explicitly prefix all identifiers from the standard 788namespace with an "``std::``" prefix, rather than rely on "``using namespace 789std;``". 790 791In header files, adding a ``'using namespace XXX'`` directive pollutes the 792namespace of any source file that ``#include``\s the header. This is clearly a 793bad thing. 794 795In implementation files (e.g. ``.cpp`` files), the rule is more of a stylistic 796rule, but is still important. Basically, using explicit namespace prefixes 797makes the code **clearer**, because it is immediately obvious what facilities 798are being used and where they are coming from. And **more portable**, because 799namespace clashes cannot occur between LLVM code and other namespaces. The 800portability rule is important because different standard library implementations 801expose different symbols (potentially ones they shouldn't), and future revisions 802to the C++ standard will add more symbols to the ``std`` namespace. As such, we 803never use ``'using namespace std;'`` in LLVM. 804 805The exception to the general rule (i.e. it's not an exception for the ``std`` 806namespace) is for implementation files. For example, all of the code in the 807LLVM project implements code that lives in the 'llvm' namespace. As such, it is 808ok, and actually clearer, for the ``.cpp`` files to have a ``'using namespace 809llvm;'`` directive at the top, after the ``#include``\s. This reduces 810indentation in the body of the file for source editors that indent based on 811braces, and keeps the conceptual context cleaner. The general form of this rule 812is that any ``.cpp`` file that implements code in any namespace may use that 813namespace (and its parents'), but should not use any others. 814 815Provide a Virtual Method Anchor for Classes in Headers 816^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 817 818If a class is defined in a header file and has a vtable (either it has virtual 819methods or it derives from classes with virtual methods), it must always have at 820least one out-of-line virtual method in the class. Without this, the compiler 821will copy the vtable and RTTI into every ``.o`` file that ``#include``\s the 822header, bloating ``.o`` file sizes and increasing link times. 823 824Don't use default labels in fully covered switches over enumerations 825^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 826 827``-Wswitch`` warns if a switch, without a default label, over an enumeration 828does not cover every enumeration value. If you write a default label on a fully 829covered switch over an enumeration then the ``-Wswitch`` warning won't fire 830when new elements are added to that enumeration. To help avoid adding these 831kinds of defaults, Clang has the warning ``-Wcovered-switch-default`` which is 832off by default but turned on when building LLVM with a version of Clang that 833supports the warning. 834 835A knock-on effect of this stylistic requirement is that when building LLVM with 836GCC you may get warnings related to "control may reach end of non-void function" 837if you return from each case of a covered switch-over-enum because GCC assumes 838that the enum expression may take any representable value, not just those of 839individual enumerators. To suppress this warning, use ``llvm_unreachable`` after 840the switch. 841 842Use ``LLVM_DELETED_FUNCTION`` to mark uncallable methods 843^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 844 845Prior to C++11, a common pattern to make a class uncopyable was to declare an 846unimplemented copy constructor and copy assignment operator and make them 847private. This would give a compiler error for accessing a private method or a 848linker error because it wasn't implemented. 849 850With C++11, we can mark methods that won't be implemented with ``= delete``. 851This will trigger a much better error message and tell the compiler that the 852method will never be implemented. This enables other checks like 853``-Wunused-private-field`` to run correctly on classes that contain these 854methods. 855 856To maintain compatibility with C++03, ``LLVM_DELETED_FUNCTION`` should be used 857which will expand to ``= delete`` if the compiler supports it. These methods 858should still be declared private. Example of the uncopyable pattern: 859 860.. code-block:: c++ 861 862 class DontCopy { 863 private: 864 DontCopy(const DontCopy&) LLVM_DELETED_FUNCTION; 865 DontCopy &operator =(const DontCopy&) LLVM_DELETED_FUNCTION; 866 public: 867 ... 868 }; 869 870Don't evaluate ``end()`` every time through a loop 871^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 872 873Because C++ doesn't have a standard "``foreach``" loop (though it can be 874emulated with macros and may be coming in C++'0x) we end up writing a lot of 875loops that manually iterate from begin to end on a variety of containers or 876through other data structures. One common mistake is to write a loop in this 877style: 878 879.. code-block:: c++ 880 881 BasicBlock *BB = ... 882 for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I) 883 ... use I ... 884 885The problem with this construct is that it evaluates "``BB->end()``" every time 886through the loop. Instead of writing the loop like this, we strongly prefer 887loops to be written so that they evaluate it once before the loop starts. A 888convenient way to do this is like so: 889 890.. code-block:: c++ 891 892 BasicBlock *BB = ... 893 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) 894 ... use I ... 895 896The observant may quickly point out that these two loops may have different 897semantics: 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 899loop may not in fact be correct. If you actually do depend on this behavior, 900please write the loop in the first form and add a comment indicating that you 901did it intentionally. 902 903Why do we prefer the second form (when correct)? Writing the loop in the first 904form has two problems. First it may be less efficient than evaluating it at the 905start of the loop. In this case, the cost is probably minor --- a few extra 906loads every time through the loop. However, if the base expression is more 907complex, then the cost can rise quickly. I've seen loops where the end 908expression was actually something like: "``SomeMap[x]->end()``" and map lookups 909really aren't cheap. By writing it in the second form consistently, you 910eliminate the issue entirely and don't even have to think about it. 911 912The second (even bigger) issue is that writing the loop in the first form hints 913to the reader that the loop is mutating the container (a fact that a comment 914would handily confirm!). If you write the loop in the second form, it is 915immediately obvious without even looking at the body of the loop that the 916container isn't being modified, which makes it easier to read the code and 917understand what it does. 918 919While the second form of the loop is a few extra keystrokes, we do strongly 920prefer it. 921 922``#include <iostream>`` is Forbidden 923^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 924 925The use of ``#include <iostream>`` in library files is hereby **forbidden**, 926because many common implementations transparently inject a `static constructor`_ 927into every translation unit that includes it. 928 929Note that using the other stream headers (``<sstream>`` for example) is not 930problematic in this regard --- just ``<iostream>``. However, ``raw_ostream`` 931provides various APIs that are better performing for almost every use than 932``std::ostream`` style APIs. 933 934.. note:: 935 936 New code should always use `raw_ostream`_ for writing, or the 937 ``llvm::MemoryBuffer`` API for reading files. 938 939.. _raw_ostream: 940 941Use ``raw_ostream`` 942^^^^^^^^^^^^^^^^^^^ 943 944LLVM 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 947``ostream``. 948 949Unlike ``std::ostream``, ``raw_ostream`` is not a template and can be forward 950declared as ``class raw_ostream``. Public headers should generally not include 951the ``raw_ostream`` header, but use forward declarations and constant references 952to ``raw_ostream`` instances. 953 954Avoid ``std::endl`` 955^^^^^^^^^^^^^^^^^^^ 956 957The ``std::endl`` modifier, when used with ``iostreams`` outputs a newline to 958the output stream specified. In addition to doing this, however, it also 959flushes the output stream. In other words, these are equivalent: 960 961.. code-block:: c++ 962 963 std::cout << std::endl; 964 std::cout << '\n' << std::flush; 965 966Most of the time, you probably have no reason to flush the output stream, so 967it's better to use a literal ``'\n'``. 968 969Microscopic Details 970------------------- 971 972This section describes preferred low-level formatting guidelines along with 973reasoning on why we prefer them. 974 975Spaces Before Parentheses 976^^^^^^^^^^^^^^^^^^^^^^^^^ 977 978We prefer to put a space before an open parenthesis only in control flow 979statements, but not in normal function call expressions and function-like 980macros. For example, this is good: 981 982.. code-block:: c++ 983 984 if (x) ... 985 for (i = 0; i != 100; ++i) ... 986 while (llvm_rocks) ... 987 988 somefunc(42); 989 assert(3 != 4 && "laws of math are failing me"); 990 991 a = foo(42, 92) + bar(x); 992 993and this is bad: 994 995.. code-block:: c++ 996 997 if(x) ... 998 for(i = 0; i != 100; ++i) ... 999 while(llvm_rocks) ... 1000 1001 somefunc (42); 1002 assert (3 != 4 && "laws of math are failing me"); 1003 1004 a = foo (42, 92) + bar (x); 1005 1006The reason for doing this is not completely arbitrary. This style makes control 1007flow operators stand out more, and makes expressions flow better. The function 1008call operator binds very tightly as a postfix operator. Putting a space after a 1009function name (as in the last example) makes it appear that the code might bind 1010the arguments of the left-hand-side of a binary operator with the argument list 1011of a function and the name of the right side. More specifically, it is easy to 1012misread the "``a``" example as: 1013 1014.. code-block:: c++ 1015 1016 a = foo ((42, 92) + bar) (x); 1017 1018when skimming through the code. By avoiding a space in a function, we avoid 1019this misinterpretation. 1020 1021Prefer Preincrement 1022^^^^^^^^^^^^^^^^^^^ 1023 1024Hard fast rule: Preincrement (``++X``) may be no slower than postincrement 1025(``X++``) and could very well be a lot faster than it. Use preincrementation 1026whenever possible. 1027 1028The semantics of postincrement include making a copy of the value being 1029incremented, returning it, and then preincrementing the "work value". For 1030primitive types, this isn't a big deal. But for iterators, it can be a huge 1031issue (for example, some iterators contains stack and set objects in them... 1032copying an iterator could invoke the copy ctor's of these as well). In general, 1033get in the habit of always using preincrement, and you won't have a problem. 1034 1035 1036Namespace Indentation 1037^^^^^^^^^^^^^^^^^^^^^ 1038 1039In general, we strive to reduce indentation wherever possible. This is useful 1040because we want code to `fit into 80 columns`_ without wrapping horribly, but 1041also because it makes it easier to understand the code. Namespaces are a funny 1042thing: 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 1044enum or something similar. In order to balance this, we use different 1045approaches for small versus large namespaces. 1046 1047If a namespace definition is small and *easily* fits on a screen (say, less than 104835 lines of code), then you should indent its body. Here's an example: 1049 1050.. code-block:: c++ 1051 1052 namespace llvm { 1053 namespace X86 { 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, 1061 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 1064 /// PIC base is. 1065 reloc_picrel_word = 1, 1066 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 1071 }; 1072 } 1073 } 1074 1075Since the body is small, indenting adds value because it makes it very clear 1076where the namespace starts and ends, and it is easy to take the whole thing in 1077in one "gulp" when reading the code. If the blob of code in the namespace is 1078larger (as it typically is in a header in the ``llvm`` or ``clang`` namespaces), 1079do not indent the code, and add a comment indicating what namespace is being 1080closed. For example: 1081 1082.. code-block:: c++ 1083 1084 namespace llvm { 1085 namespace knowledge { 1086 1087 /// Grokable - This class represents things that Smith can have an intimate 1088 /// understanding of and contains the data associated with it. 1089 class Grokable { 1090 ... 1091 public: 1092 explicit Grokable() { ... } 1093 virtual ~Grokable() = 0; 1094 1095 ... 1096 1097 }; 1098 1099 } // end namespace knowledge 1100 } // end namespace llvm 1101 1102Because the class is large, we don't expect that the reader can easily 1103understand the entire concept in a glance, and the end of the file (where the 1104namespaces end) may be a long ways away from the place they open. As such, 1105indenting the contents of the namespace doesn't add any value, and detracts from 1106the readability of the class. In these cases it is best to *not* indent the 1107contents of the namespace. 1108 1109.. _static: 1110 1111Anonymous Namespaces 1112^^^^^^^^^^^^^^^^^^^^ 1113 1114After talking about namespaces in general, you may be wondering about anonymous 1115namespaces in particular. Anonymous namespaces are a great language feature 1116that tells the C++ compiler that the contents of the namespace are only visible 1117within the current translation unit, allowing more aggressive optimization and 1118eliminating the possibility of symbol name collisions. Anonymous namespaces are 1119to C++ as "static" is to C functions and global variables. While "``static``" 1120is available in C++, anonymous namespaces are more general: they can make entire 1121classes private to a file. 1122 1123The problem with anonymous namespaces is that they naturally want to encourage 1124indentation of their body, and they reduce locality of reference: if you see a 1125random function definition in a C++ file, it is easy to see if it is marked 1126static, but seeing if it is in an anonymous namespace requires scanning a big 1127chunk of the file. 1128 1129Because of this, we have a simple guideline: make anonymous namespaces as small 1130as possible, and only use them for class declarations. For example, this is 1131good: 1132 1133.. code-block:: c++ 1134 1135 namespace { 1136 class StringSort { 1137 ... 1138 public: 1139 StringSort(...) 1140 bool operator<(const char *RHS) const; 1141 }; 1142 } // end anonymous namespace 1143 1144 static void runHelper() { 1145 ... 1146 } 1147 1148 bool StringSort::operator<(const char *RHS) const { 1149 ... 1150 } 1151 1152This is bad: 1153 1154.. code-block:: c++ 1155 1156 namespace { 1157 class StringSort { 1158 ... 1159 public: 1160 StringSort(...) 1161 bool operator<(const char *RHS) const; 1162 }; 1163 1164 void runHelper() { 1165 ... 1166 } 1167 1168 bool StringSort::operator<(const char *RHS) const { 1169 ... 1170 } 1171 1172 } // end anonymous namespace 1173 1174This is bad specifically because if you're looking at "``runHelper``" in the middle 1175of a large C++ file, that you have no immediate way to tell if it is local to 1176the file. When it is marked static explicitly, this is immediately obvious. 1177Also, there is no reason to enclose the definition of "``operator<``" in the 1178namespace just because it was declared there. 1179 1180See Also 1181======== 1182 1183A lot of these comments and recommendations have been culled for other sources. 1184Two particularly important books for our work are: 1185 1186#. `Effective C++ 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. 1190 1191#. `Large-Scale C++ Software Design 1192 <http://www.amazon.com/Large-Scale-Software-Design-John-Lakos/dp/0201633620/ref=sr_1_1>`_ 1193 by John Lakos 1194 1195If you get some free time, and you haven't read them: do so, you might learn 1196something. 1197