/**************************************************************************** ** ** This file is part of Qt Creator ** ** Copyright (c) 2012 Nokia Corporation and/or its subsidiary(-ies). ** ** Contact: Nokia Corporation (qt-info@nokia.com) ** ** ** GNU Free Documentation License ** ** Alternatively, this file may be used under the terms of the GNU Free ** Documentation License version 1.3 as published by the Free Software ** Foundation and appearing in the file included in the packaging of this ** file. ** ** If you have questions regarding the use of this file, please contact ** Nokia at qt-info@nokia.com. ** ****************************************************************************/ /*! \contentspage{index.html}{Qt Creator} \previouspage qtcreator-ui-text.html \page coding-style.html \nextpage qtcreator-api.html \title Qt Creator Coding Rules \note This document is work in progress. The coding rules aim to guide Qt Creator developers, to help them write understandable and maintainable code, and to minimize confusion and surprises. As usual, rules are not set in stone. If you have a good reason to break one, do so. But first make sure that at least some other developers agree with you. To contribute to the main Qt Creator source, you should comply to the following rules: \list \o The most important rule is: KISS (keep it short and simple). Always choose the simpler implementation option over the more complicated one. This makes maintenance a lot easier. \o Write good C++ code. That is, readable, well commented when necessary, and object-oriented. \o Take advantage of Qt. Do not re-invent the wheel. Think about which parts of your code are generic enough that they might be incorporated into Qt instead of Qt Creator. \o Adapt the code to the existing structures in Qt Creator. If you have improvement ideas, discuss them with other developers before writing the code. \o Follow the guidelines in \l{Code Constructs}, \l{Formatting}, and \l{Patterns and Practices}. \o Document interfaces. Right now we use qdoc, but changing to doxygen is being considered. \endlist \section1 Submitting Code To submit code to Qt Creator, you must understand the tools and mechanics as well as the philosophy behind Qt development. For more information about how to set up the development environment for working on Qt Creator and how to submit code and documentation for inclusion, see \l{http://wiki.qt-project.org/index.php/Main_Page}{Guidelines for Contributions to the Qt Project}. \section1 Binary and Source Compatibility The following list describes how the releases are numbered and defines \e {binary compatibility} and \e {source code compatibility} between releases: \list \o Qt Creator 2.0.0 is a \e {major release}, Qt Creator 2.1.0 is a \e {minor release}, and Qt Creator 2.1.3 is a \e {patch release}. \o \e {Backward binary compatibility} means that code linked to an earlier version of the library still works. \o \e {Forward binary compatibility} means that code linked to a newer version of the library works with an older library. \o \e {Source code compatibility} means that code compiles without modification. \endlist We do not currently guarantee API nor ABI (\l{http://en.wikipedia.org/wiki/Application_binary_interface}{application binary interface}) compatibility between major releases and minor releases. However, we try to preserve backward and forward binary compatibility and forward and backward source code compatibility in patch releases, so: \list \o Do not add or remove any public API (e.g. global functions,x public/protected/private methods). \o Do not reimplement methods (not even inlines, nor protected or private methods). \o Check \l {http://wiki.qt-project.org/index.php/Binary_Compatibility_Workarounds}{Binary Compatibility Workarounds} for ways to preserve binary compatibility. \endlist \note This is not yet mandatory. For more information on binary compatibility, see \l{http://techbase.kde.org/Policies/Binary_Compatibility_Issues_With_C++}{Binary Compatibility Issues With C++}. \section1 Code Constructs Follow the guidelines for code constructs to make the code faster and clearer. In addition, the guidelines allow you to take advantage of the strong type checking in C++. \list \o Prefer preincrement to postincrement whenever possible. Preincrement is potentially faster than postincrement. Just think about the obvious implementations of pre/post-increment. This rule applies to decrement too: \code ++T; --U; -NOT- T++; U--; \endcode \o Try to minimize evaluation of the same code over and over. This is aimed especially at loops: \code Container::iterator end = large.end(); for (Container::iterator it = large.begin(); it != end; ++it) { ...; } -NOT- for (Container::iterator it = large.begin(); it != large.end(); ++it) { ...; } \endcode \o You can use the Qt \c foreach loop in non-time-critical code with a Qt container. It is a nice way to keep line noise down and to give the loop variable a proper name: \code foreach (QWidget *widget, container) doSomething(widget); -NOT- Container::iterator end = container.end(); for (Container::iterator it = container.begin(); it != end; ++it) doSomething(*it); \endcode Make the loop variable const, if possible. This might prevent unnecessary detaching of shared data: \code foreach (const QString &name, someListOfNames) doSomething(name); - NOT - foreach (QString name, someListOfNames) doSomething(name); \endcode \endlist \section1 Formatting \section2 Capitalizing Identifiers Use \l{http://en.wikipedia.org/wiki/CamelCase}{camel case} in identifiers. Capitalize the first word in an identifier as follows: \list \o Class names begin with a capital letter. \o Function names begin with a lower case letter. \o Variable names begin with a lower case letter. \o Enum names begin with a capital letter. Enum values use lower case and contain some part of the name of the enum type. \endlist \section2 Whitespace \list \o Use four spaces for indentation, no tabs. \o Use blank lines to group statements together where suited. \o Always use only one blank line. \endlist \section3 Pointers and References For pointers or references, always use a single space before an asterisk (*) or an ampersand (&), but never after. Avoid C-style casts when possible: \code char *blockOfMemory = (char *)malloc(data.size()); char *blockOfMemory = reinterpret_cast(malloc(data.size())); -NOT- char* blockOfMemory = (char* ) malloc(data.size()); \endcode Of course, in this particulare case, using \c new might be an even better option. \section3 Operator Names and Parentheses Do not use spaces between operator names and function names. The equation marks (==) are a part of the function name, and therefore, spaces make the declaration look like an expression: \code operator==(type) -NOT- operator == (type) \endcode \section3 Function Names and Parentheses Do not use spaces between function names and parentheses: \code void mangle() -NOT- void mangle () \endcode \section3 Keywords Always use a single space after a keyword, and before a curly brace: \code if (foo) { } -NOT- if(foo){ } \endcode \section2 Braces As a base rule, place the left curly brace on the same line as the start of the statement: \code if (codec) { } -NOT- if (codec) { } \endcode Exception: Function implementations and class declarations always have the left brace in the beginning of a line: \code static void foo(int g) { qDebug("foo: %i", g); } class Moo { }; \endcode Use curly braces when the body of a conditional statement contains more than one line, and also if a single line statement is somewhat complex. Otherwise, omit them: \code if (address.isEmpty()) return false; for (int i = 0; i < 10; ++i) qDebug("%i", i); -NOT- if (address.isEmpty()) { return false; } for (int i = 0; i < 10; ++i) { qDebug("%i", i); } \endcode Exception 1: Use braces also if the parent statement covers several lines or if it wraps: \code if (address.isEmpty() || !isValid() || !codec) { return false; } \endcode \note This could be re-written as: \code if (address.isEmpty()) return false; if (!isValid()) return false; if (!codec) return false; \endcode Exception 2: Use braces also in if-then-else blocks where either the if-code or the else-code covers several lines: \code if (address.isEmpty()) { --it; } else { qDebug("%s", qPrintable(address)); ++it; } -NOT- if (address.isEmpty()) --it; else { qDebug("%s", qPrintable(address)); ++it; } \endcode \code if (a) { if (b) ... else ... } -NOT- if (a) if (b) ... else ... \endcode Use curly braces when the body of a conditional statement is empty: \code while (a) {} -NOT- while (a); \endcode \section2 Parentheses Use parentheses to group expressions: \code if ((a && b) || c) -NOT- if (a && b || c) \endcode \code (a + b) & c -NOT- a + b & c \endcode \section2 Line Breaks \list \o Keep lines shorter than 100 characters. \o Insert line breaks if necessary. \o Commas go at the end of a broken line. \o Operators start at the beginning of the new line. \code if (longExpression || otherLongExpression || otherOtherLongExpression) { } -NOT- if (longExpression || otherLongExpression || otherOtherLongExpression) { } \endcode \endlist \section2 Declarations \list \o Use this order for the access sections of your class: public, protected, private. The public section is interesting for every user of the class. The private section is only of interest for the implementors of the class (you). \o Avoid declaring global objects in the declaration file of the class. If the same variable is used for all objects, use a static member. \o Use \c{class} instead of \c{struct}. Some compilers mangle that difference into the symbol names and spit out warnings if a struct declaration is followed by a class definition. To avoid ongoing changes from one to the other we declare \c{class} the prefered way. \endlist \section3 Declaring Variables \list \o Avoid global variables of class type to rule out initialization order problems. Consider using \c Q_GLOBAL_STATIC if they cannot be avoided. \o Declare global string literals as \code const char aString[] = "Hello"; \endcode \o Avoid short names (such as, a, rbarr, nughdeget) whenever possible. Use single-character variable names only for counters and temporaries, where the purpose of the variable is obvious. \o Declare each variable on a separate line: \code QString a = "Joe"; QString b = "Foo"; -NOT- QString a = "Joe", b = "Foo"; \endcode \note \c{QString a = "Joe"} formally calls a copy constructor on a temporary that is constructed from a string literal. Therefore, it is potentially more expensive than direct construction by \c {QString a("Joe")}. However, the compiler is allowed to elide the copy (even if this has side effects), and modern compilers typically do so. Given these equal costs, Qt Creator code favours the '=' idiom as it is in line with the traditional C-style initialization, it cannot be mistaken as function declaration, and it reduces the level of nested parantheses in more initializations. \o Avoid abbreviations: \code int height; int width; char *nameOfThis; char *nameOfThat; -NOT- int a, b; char *c, *d; \endcode \o Wait with declaring a variable until it is needed. This is especially important when initialization is done at the same time. \endlist \section2 Namespaces \list \o Put the left curly brace on the same line as the \namespace keyword. \o Do not indent declarations or definitions inside. \o Optional, but recommended if the namespaces spans more than a few lines: Add a comment after the right curly brace repeating the namespace. \code namespace MyPlugin { void someFunction() { ... } } // namespace MyPlugin \endcode \o As an exception, if there is only a single class declaration inside the namespace, all can go on a single line: \code namespace MyPlugin { class MyClass; } \endcode \endlist \section1 Patterns and Practices \target coding-rules-namespacing \section2 Namespacing Read \l {http://wiki.qt-project.org/index.php/Qt_In_Namespace}{Qt In Namespace} and keep in mind that all of Qt Creator is \e{namespace aware} code. The namespacing policy within Qt Creator is as follows: \list \o Classes/Symbols of a library or plugin that are exported for use of other libraries or plugins are in a namespace specific to that library/plugin, e.g. \c{MyPlugin}. \o Classes/Symbols of a library or plugin that are not exported are in an additional \c{Internal} namespace, e.g. \c{MyPlugin::Internal}. \endlist \section2 Passing File Names Qt Creator API expects file names in portable format, that is, with slashes (/) instead of backslashes (\\) even on Windows. To pass a file name from the user to the API, convert it with QDir::fromNativeSeparators first. To present a file name to the user, convert it back to native format with QDir::toNativeSeparators. When comparing file names, consider using DocumentManager::fixFileName which makes sure that paths are clean and absolute and also takes Windows case-insensitivity into account (even if it is an expensive operation). \section2 Plugin Extension Points A plugin extension point is an interface that is provided by one plugin to be implemented by others. The plugin then retrieves all implementations of the interface and uses them. That is, they \e extend the functionality of the plugin. Typically, the implementations of the interface are put into the global object pool during plugin initialization, and the plugin retrieves them from the object pool at the end of plugin initialization. For example, the Find plugin provides the FindFilter interface for other plugins to implement. With the FindFilter interface, additional search scopes can be added, that appear in the \gui {Advanced Search} dialog. The Find plugin retrieves all FindFilter implementations from the global object pool and presents them in the dialog. The plugin forwards the actual search request to the correct FindFilter implementation, which then performs the search. \section2 Using the Global Object Pool You can add objects to the global object pool via \l{ExtensionSystem::PluginManager::addObject()}, and retrieve objects of a specific type again via \l{ExtensionSystem::PluginManager::getObjects()}. This should mostly be used for implementations of \l{Plugin Extension Points}. \note Do not put a singleton into the pool, and do not retrieve it from there. Use the singleton pattern instead. \section2 C++ Features \list \o Do not use exceptions, unless you know what you do. \o Do not use RTTI (Run-Time Type Information; that is, the typeinfo struct, the dynamic_cast or the typeid operators, including throwing exceptions), unless you know what you do. \o Do not use virtual inheritance, unless you know what you do. \o Use templates wisely, not just because you can. Hint: Use the compile autotest to see whether a C++ feature is supported by all compilers in the test farm. \o All code is ASCII only (7-bit characters only, run \c {man ascii} if unsure) \list \o Rationale: We have too many locales inhouse and an unhealthy mix of UTF-8 and Latin1 systems. Usually, characters > 127 can be broken without you even knowing by clicking Save in your favourite editor. \o For strings: Use \\nnn (where nnn is the octal representation of whatever locale you want your string in) or \xnn (where nn is hexadecimal). For example: QString s = QString::fromUtf8("\\213\\005"); \o For umlauts in documentation, or other non-ASCII characters, either use the qdoc \c {\unicode} command or use the relevant macro. For example: \c{\uuml} for \uuml. \endlist \o Use static keywords instead of anonymous namespaces whenever possible. A name localized to the compilation unit with static is guaranteed to have internal linkage. For names declared in anonymous namespaces, the C++ standard unfortunately mandates external linkage (ISO/IEC 14882, 7.1.1/6, or see various discussions about this on the gcc mailing lists). \endlist \section3 Null Pointers Using a plain zero (0) for null pointer constants is always correct and least effort to type. \code void *p = 0; -NOT- void *p = NULL; -NOT- void *p = '\0'; -NOT- void *p = 42 - 7 * 6; \endcode \note As an exception, imported third party code as well as code interfacing the native APIs (src/support/os_*) can use NULL. \section2 Using QObject \list \o Every QObject subclass must have a Q_OBJECT macro, even if it does not have signals or slots, if it is intended to be used with qobject_cast<>. See also \l{Casting}. \o Normalize the arguments for signals and slots (see \l{http://doc.qt.nokia.com/4.7/qmetaobject.html#normalizedSignature}{QMetaObject::normalizedSignature} inside connect statements to safely make signal and slot lookup a few cycles faster. You can use $QTDIR/util/normalize to normalize existing code. \endlist \section2 File Headers If you create a new file, the top of the file should include a header comment equal to the one found in other source files of Qt Creator. \section2 Including Headers \list \o Use the following format to include Qt headers: \c{#include }. \o Arrange includes in an order that goes from specific to generic to ensure that the headers are self-contained. For example: \list \o \c{#include "myclass.h"} \o \c{#include "otherclassinplugin.h"} \o \c{#include } \o \c{#include } \o \c{#include } \o \c{#include } \endlist \o Enclose headers from other plugins in square brackets (<>) rather than quotation marks ("") to make it easier to spot external dependencies in the sources. \o Add empty lines between long blocks of \e peer headers and try to arrange the headers in alphabetic order within a block. \endlist \section2 Casting \list \o Avoid C casts, prefer C++ casts (\c static_cast, \c const_cast, \c reinterpret_cast) Both \c reinterpret_cast and C-style casts are dangerous, but at least \c reinterpret_cast will not remove the const modifier. \o Do not use \c dynamic_cast, use \c {qobject_cast} for QObjects, or refactor your design, for example by introducing a \c {type()} method (see QListWidgetItem), unless you know what you do. \endlist \section2 Compiler and Platform-specific Issues \list \o Be extremely careful when using the question mark operator. If the returned types are not identical, some compilers generate code that crashes at runtime (you will not even get a compiler warning): \code QString s; // crash at runtime - QString vs. const char * return condition ? s : "nothing"; \endcode \o Be extremely careful about alignment. Whenever a pointer is cast such that the required alignment of the target is increased, the resulting code might crash at runtime on some architectures. For example, if a \c {const char *} is cast to a \c {const int *}, it will crash on machines where integers have to be aligned at two-byte or four-byte boundaries. Use a union to force the compiler to align variables correctly. In the example below, you can be sure that all instances of AlignHelper are aligned at integer-boundaries: \code union AlignHelper { char c; int i; }; \endcode \o Anything that has a constructor or needs to run code to be initialized cannot be used as global object in library code, since it is undefined when that constructor or code will be run (on first usage, on library load, before \c {main()} or not at all). Even if the execution time of the initializer is defined for shared libraries, you will get into trouble when moving that code in a plugin or if the library is compiled statically: \code // global scope -NOT- // Default constructor needs to be run to initialize x: static const QString x; -NOT- // Constructor that takes a const char * has to be run: static const QString y = "Hello"; -NOT- QString z; -NOT- // Call time of foo() undefined, might not be called at all: static const int i = foo(); \endcode Things you can do: \code // global scope // No constructor must be run, x set at compile time: static const char x[] = "someText"; // y will be set at compile time: static int y = 7; // Will be initialized statically, no code being run. static MyStruct s = {1, 2, 3}; // Pointers to objects are OK, no code needed to be run to // initialize ptr: static QString *ptr = 0; // Use Q_GLOBAL_STATIC to create static global objects instead: Q_STATIC_GLOBAL(QString, s) void foo() { s()->append("moo"); } \endcode \note Static objects in function scope are no problem. The constructor will be run the first time the function is entered. The code is not reentrant, though. \o A \c char is signed or unsigned dependent on the architecture. Use signed \c char or \c uchar if you explicitely want a signed or unsigned char. The following code will break on PowerPC, for example: \code // Condition is always true on platforms where the // default is unsigned: if (c >= 0) { ... } \endcode \o Avoid 64-bit enum values. The AAPCS (Procedure Call Standard for the ARM Architecture) embedded ABI hard codes all enum values to a 32-bit integer. \o Do not mix const and non-const iterators. This will silently crash on broken compilers. \code for (Container::const_iterator it = c.constBegin(); it != c.constEnd(); ++it) -NOT- for (Container::const_iterator it = c.begin(); it != c.end(); ++it) \endcode \endlist \section2 Esthetics \list \o Prefer enums to define const over static const int or defines. Enumeration values will be replaced by the compiler at compile time, resulting in faster code. Defines are not namespace safe. \o Prefer verbose argument names in headers. Qt Creator will show the argument names in their completion box. It will look better in the documentation. \endlist \section2 Inheriting from Template or Tool Classes Inheriting from template or tool classes has the following potential pitfalls: \list \o The destructors are not virtual, which can lead to memory leaks. \o The symbols are not exported (and mostly inline), which can lead to symbol clashes. \endlist For example, library A has class \c {Q_EXPORT X: public QList {};} and library B has class \c {Q_EXPORT Y: public QList {};}. Suddenly, QList symbols are exported from two libraries which results in a clash. \section2 Conventions for Public Header Files Our public header files have to survive the strict settings of some of our users. All installed headers have to follow these rules: \list \o No C style casts (\c{-Wold-style-cast}). Use \c static_cast, \c const_cast or \c reinterpret_cast, for basic types, use the constructor form: \c {int(a)} instead of \c {(int)a}. For more information, see \l{Casting}. \o No float comparisons (\c{-Wfloat-equal}). Use \c qFuzzyCompare to compare values with a delta. Use \c qIsNull to check whether a float is binary 0, instead of comparing it to 0.0, or, prefered, move such code into an implementation file. \o Do not hide virtual methods in subclasses (\{-Woverloaded-virtual}). If the baseclass A has a virtual \c {int val()} and subclass B an overload with the same name, \c {int val(int x)}, the A \c val function is hidden. Use the \c using keyword to make it visible again, and add the following silly workaround for broken compilers: \code class B: public A { #ifdef Q_NO_USING_KEYWORD inline int val() { return A::val(); } #else using A::val; #endif }; \endcode \o Do not shadow variables (\c{-Wshadow}). \o Avoid things like \c {this->x = x;} if possible. \o Do not give variables the same name as functions declared in your class. \o To improve code readability, always check whether a preprocessor variable is defined before probing its value (\c{-Wundef}). \code #if defined(Foo) && Foo == 0 -NOT- #if Foo == 0 -NOT- #if Foo - 0 == 0 \endcode \o When checking for a preprocessor define using the \c{defined} operator, always include the variable name in parentheses. \code #if defined(Foo) -NOT- #if defined Foo \endcode \endlist \section1 Class Member Names We use the "m_" prefix convention, except for public struct members (typically in *Private classes and the very rare cases of really public structures). The \c{d} and \c{q} pointers are exempt from the "m_" rule. The \c{d} pointers ("Pimpls") are named "d", not "m_d". The type of the \c{d} pointer in \c{class Foo} is \c{FooPrivate *}, where \c{FooPrivate} is declared in the same namespace as \c{Foo}, or if \c{Foo} is exported, in the corresponding \{Internal} namespace. If needed (for example when the private object needs to emit signals of the proper class), \c{FooPrivate} can be a friend of \c{Foo}. If the private class needs a backreference to the real class, the pointer is named \c{q}, and its type is \c{Foo *}. (Same convention as in Qt: "q" looks like an inverted "d".) Do not use smart pointers to guard the \c{d} pointer as it imposes a compile and link time overhead and creates fatter object code with more symbols, leading, for instance to slowed down debugger startup: \code ############### bar.h #include //#include struct BarPrivate; struct Bar { Bar(); ~Bar(); int value() const; QScopedPointer d; //std::auto_ptr d; }; ############### bar.cpp #include "bar.h" struct BarPrivate { BarPrivate() : i(23) {} int i; }; Bar::Bar() : d(new BarPrivate) {} Bar::~Bar() {} int Bar::value() const { return d->i; } ############### baruser.cpp #include "bar.h" int barUser() { Bar b; return b.value(); } ############### baz.h struct BazPrivate; struct Baz { Baz(); ~Baz(); int value() const; BazPrivate *d; }; ############### baz.cpp #include "baz.h" struct BazPrivate { BazPrivate() : i(23) {} int i; }; Baz::Baz() : d(new BazPrivate) {} Baz::~Baz() { delete d; } int Baz::value() const { return d->i; } ############### bazuser.cpp #include "baz.h" int bazUser() { Baz b; return b.value(); } ############### main.cpp int barUser(); int bazUser(); int main() { return barUser() + bazUser(); } \endcode Results: \code Object file size: 14428 bar.o 4744 baz.o 8508 baruser.o 2952 bazuser.o Symbols in bar.o: 00000000 W _ZN3Foo10BarPrivateC1Ev 00000036 T _ZN3Foo3BarC1Ev 00000000 T _ZN3Foo3BarC2Ev 00000080 T _ZN3Foo3BarD1Ev 0000006c T _ZN3Foo3BarD2Ev 00000000 W _ZN14QScopedPointerIN3Foo10BarPrivateENS_21QScopedPointerDeleterIS2_EEEC1EPS2_ 00000000 W _ZN14QScopedPointerIN3Foo10BarPrivateENS_21QScopedPointerDeleterIS2_EEED1Ev 00000000 W _ZN21QScopedPointerDeleterIN3Foo10BarPrivateEE7cleanupEPS2_ 00000000 W _ZN7qt_noopEv U _ZN9qt_assertEPKcS1_i 00000094 T _ZNK3Foo3Bar5valueEv 00000000 W _ZNK14QScopedPointerIN3Foo10BarPrivateENS_21QScopedPointerDeleterIS2_EEEptEv U _ZdlPv U _Znwj U __gxx_personality_v0 Symbols in baz.o: 00000000 W _ZN3Foo10BazPrivateC1Ev 0000002c T _ZN3Foo3BazC1Ev 00000000 T _ZN3Foo3BazC2Ev 0000006e T _ZN3Foo3BazD1Ev 00000058 T _ZN3Foo3BazD2Ev 00000084 T _ZNK3Foo3Baz5valueEv U _ZdlPv U _Znwj U __gxx_personality_v0 \endcode \section1 Documentation The documentation is generated from source and header files. You document for the other developers, not for yourself. In the header files, document interfaces. That is, what the function does, not the implementation. In the .cpp files, you can document the implementation if the implementation is not obvious. */