21 KiB
Serenity C++ coding style
This document describes the coding style used for C++ code in the Serenity Operating System project. All new code should conform to this style.
We'll definitely be tweaking and amending this over time, so let's consider it a living document. :)
Indentation
Use spaces, not tabs. Tabs should only appear in files that require them for semantic meaning, like Makefiles.
Right:
int main()
{
return 0;
}
Wrong:
int main()
{
return 0;
}
The contents of an outermost namespace block should not be indented.
Right:
// Container.h
namespace AK {
class Container {
Container();
...
};
}
// Container.cpp
namespace AK {
Container::Container()
{
...
}
}
Wrong:
// Container.h
namespace AK {
class Container {
Container();
...
};
}
// Container.cpp
namespace AK {
Container::Container()
{
...
}
}
A case label should line up with its switch statement. The case statement is indented.
Right:
switch (condition) {
case FooCondition:
case BarCondition:
++i;
break;
default:
++i;
}
Wrong:
switch (condition) {
case FooCondition:
case BarCondition:
++i;
break;
default:
--i;
}
Boolean expressions at the same nesting level that span multiple lines should have their operators on the left side of the line instead of the right side.
Right:
return type == Type::SomeType
|| type == Type::OtherType
|| (another_condition && !bad_thing);
Wrong:
return type == Type::SomeType ||
type == Type::OtherType ||
(another_condition && !bad_thing);
Spacing
Do not place spaces around unary operators.
Right:
++i;
Wrong:
++ i;
Do place spaces around binary and ternary operators.
Right:
y = m * x + b;
f(a, b);
c = a | b;
return condition ? 1 : 0;
Wrong:
y=m*x+b;
f(a,b);
c = a|b;
return condition ? 1:0;
Place spaces around the colon in a range-based for loop.
Right:
Vector<ModuleInfo> modules;
for (auto& module : modules)
register_module(module);
Wrong:
Vector<ModuleInfo> modules;
for (auto& module: modules)
register_module(module);
Do not place spaces before comma and semicolon.
Right:
for (int i = 0; i < 10; ++i)
do_something();
f(a, b);
Wrong:
for (int i = 0 ; i < 10 ; ++i)
do_something();
f(a , b) ;
Place spaces between control statements and their parentheses.
Right:
if (condition)
do_it();
Wrong:
if(condition)
do_it();
Do not place spaces between a function and its parentheses, or between a parenthesis and its content.
Right:
f(a, b);
Wrong:
f (a, b);
f( a, b );
When initializing an object, place a space before the leading brace as well as between the braces and their content.
Right:
Foo foo { bar };
Wrong:
Foo foo{ bar };
Foo foo {bar};
Line breaking
Each statement should get its own line.
Right:
++x;
++y;
if (condition)
do_it();
Wrong:
x++; y++;
if (condition) do_it();
An else statement should go on the same line as a preceding close brace if one is present, else it should line up with the if statement.
Right:
if (condition) {
...
} else {
...
}
if (condition)
do_something();
else
do_something_else();
if (condition) {
do_something();
} else {
...
}
Wrong:
if (condition) {
...
}
else {
...
}
if (condition) do_something(); else do_something_else();
if (condition) do_something(); else {
...
}
An else if statement should be written as an if statement when the prior if concludes with a return statement.
Right:
if (condition) {
...
return some_value;
}
if (condition) {
...
}
Wrong:
if (condition) {
...
return some_value;
} else if (condition) {
...
}
Braces
Function definitions: place each brace on its own line.
Right:
int main()
{
...
}
Wrong:
int main() {
...
}
Other braces: place the open brace on the line preceding the code block; place the close brace on its own line.
Right:
class MyClass {
...
};
namespace AK {
...
}
for (int i = 0; i < 10; ++i) {
...
}
Wrong:
class MyClass
{
...
};
One-line control clauses should not use braces unless comments are included or a single statement spans multiple lines.
Right:
if (condition)
do_it();
if (condition) {
// Some comment.
do_it();
}
if (condition) {
my_function(really_long_param1, really_long_param2, ...
really_long_param5);
}
Wrong:
if (condition) {
do_it();
}
if (condition)
// Some comment.
do_it();
if (condition)
my_function(really_long_param1, really_long_param2, ...
really_long_param5);
Control clauses without a body should use empty braces:
Right:
for ( ; current; current = current->next) { }
Wrong:
for ( ; current; current = current->next);
Null, false and zero
In C++, the null pointer value should be written as nullptr. In C, it should be written as NULL.
C++ and C bool values should be written as true and false.
Tests for true/false, null/non-null, and zero/non-zero should all be done without equality comparisons.
Right:
if (condition)
do_it();
if (!ptr)
return;
if (!count)
return;
Wrong:
if (condition == true)
do_it();
if (ptr == nullptr)
return;
if (count == 0)
return;
Floating point literals
Unless required in order to force floating point math, do not append .0, .f and .0f to floating point literals.
Right:
const double duration = 60;
void set_diameter(float diameter)
{
radius = diameter / 2;
}
set_diameter(10);
const int frames_per_second = 12;
double frame_duration = 1.0 / frames_per_second;
Wrong:
const double duration = 60.0;
void set_diameter(float diameter)
{
radius = diameter / 2.f;
}
set_diameter(10.f);
const int frames_per_second = 12;
double frame_duration = 1 / frames_per_second; // Integer division.
Names
A combination of CamelCase and snake_case. Use CamelCase (Capitalize the first letter, including all letters in an acronym) in a class, struct, or namespace name. Use snake_case (all lowercase, with underscores separating words) for variable and function names.
Right:
struct Entry;
size_t buffer_size;
class FileDescriptor;
String absolute_path();
Wrong:
struct data;
size_t bufferSize;
class Filedescriptor;
String MIME_Type();
Use full words, except in the rare case where an abbreviation would be more canonical and easier to understand.
Right:
size_t character_size;
size_t length;
short tab_index; // More canonical.
Wrong:
size_t char_size;
size_t len;
short tabulation_index; // Bizarre.
Data members in C++ classes should be private. Static data members should be prefixed by "s_". Other data members should be prefixed by "m_". Global variables should be prefixed by "g_".
Right:
class String {
public:
...
private:
int m_length;
};
Wrong:
class String {
public:
...
int length;
};
Precede boolean values with words like "is" and "did".
Right:
bool is_valid;
bool did_send_data;
Wrong:
bool valid;
bool sent_data;
Precede setters with the word "set". Use bare words for getters. Setter and getter names should match the names of the variables being set/gotten.
Right:
void set_count(int); // Sets m_count.
int count() const; // Returns m_count.
Wrong:
void set_count(int); // Sets m_the_count.
int get_count() const; // Returns m_the_count.
Precede getters that return values through out arguments with the word "get".
Right:
void get_filename_and_inode_id(String&, InodeIdentifier&) const;
Wrong:
void filename_and_inode_id(String&, InodeIdentifier&) const;
Use descriptive verbs in function names.
Right:
bool convert_to_ascii(short*, size_t);
Wrong:
bool to_ascii(short*, size_t);
The getter function for a member variable should not have any suffix or prefix indicating the function can optionally create or initialize the member variable. Prefix the getter function which automatically creates the object with ensure_ if there is a variant which doesn't.
Right:
Inode* inode();
Inode& ensure_inode();
Wrong:
Inode& inode();
Inode* ensure_inode();
Right:
Frame* frame();
Wrong:
Frame* ensure_frame();
Leave meaningless variable names out of function declarations. A good rule of thumb is if the parameter type name contains the parameter name (without trailing numbers or pluralization), then the parameter name isn't needed. Usually, there should be a parameter name for bools, strings, and numerical types.
Right:
void set_count(int);
void do_something(Context*);
Wrong:
void set_count(int count);
void do_something(Context* context);
Prefer enums to bools on function parameters if callers are likely to be passing constants, since named constants are easier to read at the call site. An exception to this rule is a setter function, where the name of the function already makes clear what the boolean is.
Right:
do_something(something, AllowFooBar);
paint_text_with_shadows(context, ..., text_stroke_width > 0, is_horizontal());
set_resizable(false);
Wrong:
do_something(something, false);
set_resizable(NotResizable);
Objective-C method names should follow the Cocoa naming guidelines — they should read like a phrase and each piece of the selector should start with a lowercase letter and use intercaps.
Enum members should use InterCaps with an initial capital letter.
Prefer const to #define. Prefer inline functions to macros.
#defined constants should use all uppercase names with words separated by underscores.
Use #pragma once instead of #define and #ifdef for header guards.
Right:
// MyClass.h
#pragma once
Wrong:
// MyClass.h
#ifndef MyClass_h
#define MyClass_h
Other Punctuation
Constructors for C++ classes should initialize their members using C++ initializer syntax. Each member (and superclass) should be indented on a separate line, with the colon or comma preceding the member on that line. Prefer initialization at member definition whenever possible.
Right:
class MyClass {
...
Document* m_document { nullptr };
int m_my_member { 0 };
};
MyClass::MyClass(Document* document)
: MySuperClass()
, m_document(document)
{
}
MyOtherClass::MyOtherClass()
: MySuperClass()
{
}
Wrong:
MyClass::MyClass(Document* document) : MySuperClass()
{
m_myMember = 0;
m_document = document;
}
MyClass::MyClass(Document* document) : MySuperClass()
: m_my_member(0) // This should be in the header.
{
m_document = document;
}
MyOtherClass::MyOtherClass() : MySuperClass() {}
Prefer index or range-for over iterators in Vector iterations for terse, easier-to-read code.
Right:
for (auto& child : children)
child->do_child_thing();
OK:
for (int i = 0; i < children.size(); ++i)
children[i]->do_child_thing();
Wrong:
for (auto it = children.begin(); it != children.end(); ++it)
(*it)->do_child_thing();
Pointers and References
Pointer and reference types in C++ code
Both pointer types and reference types should be written with no space between the type name and the * or &.
Right:
GraphicsBitmap* Thingy::bitmap_for(Context context)
{
auto* bitmap = static_cast<SVGStyledElement*>(node());
const KCDashArray& dashes = dashArray();
Wrong:
Image *SVGStyledElement::doSomething(PaintInfo &paintInfo)
{
SVGStyledElement *element = static_cast<SVGStyledElement *>(node());
const KCDashArray &dashes = dashArray();
An out argument of a function should be passed by reference except rare cases where it is optional in which case it should be passed by pointer.
Right:
void MyClass::get_some_value(OutArgumentType& out_argument) const
{
out_argument = m_value;
}
void MyClass::do_something(OutArgumentType* out_argument) const
{
do_the_thing();
if (out_argument)
*out_argument = m_value;
}
Wrong:
void MyClass::get_some_value(OutArgumentType* outArgument) const
{
*out_argument = m_value;
}
#include Statements
All implementation files must #include config.h first. Header files should never include config.h.
Right:
// RenderLayer.h
#include "Node.h"
#include "RenderObject.h"
#include "RenderView.h"
Wrong:
// RenderLayer.h
#include "config.h"
#include "RenderObject.h"
#include "RenderView.h"
#include "Node.h"
All implementation files must #include the primary header second, just after config.h. So for example, Node.cpp should include Node.h first, before other files. This guarantees that each header's completeness is tested. This also assures that each header can be compiled without requiring any other header files be included first.
Other #include statements should be in sorted order (case sensitive, as done by the command-line sort tool or the Xcode sort selection command). Don't bother to organize them in a logical order.
Right:
// HTMLDivElement.cpp
#include "config.h"
#include "HTMLDivElement.h"
#include "Attribute.h"
#include "HTMLElement.h"
#include "QualifiedName.h"
Wrong:
// HTMLDivElement.cpp
#include "HTMLElement.h"
#include "HTMLDivElement.h"
#include "QualifiedName.h"
#include "Attribute.h"
"using" Statements
In header files in the AK sub-library, however, it is acceptable to use "using" declarations at the end of the file to import one or more names in the AK namespace into the global scope.
Right:
// AK/Vector.h
namespace AK {
} // namespace AK
using AK::Vector;
Wrong:
// AK/Vector.h
namespace AK {
} // namespace AK
using namespace AK;
Wrong:
// runtime/Object.h
namespace AK {
} // namespace AK
using AK::SomethingOrOther;
In C++ implementation files, do not use "using" declarations of any kind to import names in the standard template library. Directly qualify the names at the point they're used instead.
Right:
// File.cpp
std::swap(a, b);
c = std::numeric_limits<int>::max()
Wrong:
// File.cpp
using std::swap;
swap(a, b);
Wrong:
// File.cpp
using namespace std;
swap(a, b);
Types
Omit "int" when using "unsigned" modifier. Do not use "signed" modifier. Use "int" by itself instead.
Right:
unsigned a;
int b;
Wrong:
unsigned int a; // Doesn't omit "int".
signed b; // Uses "signed" instead of "int".
signed int c; // Doesn't omit "signed".
Classes
Use a constructor to do an implicit conversion when the argument is reasonably thought of as a type conversion and the type conversion is fast. Otherwise, use the explicit keyword or a function returning the type. This only applies to single argument constructors.
Right:
class LargeInt {
public:
LargeInt(int);
...
class Vector {
public:
explicit Vector(int size); // Not a type conversion.
Vector create(Array); // Costly conversion.
...
Wrong:
class Task {
public:
Task(ExecutionContext&); // Not a type conversion.
explicit Task(); // No arguments.
explicit Task(ExecutionContext&, Other); // More than one argument.
...
Singleton pattern
Use a static member function named "the()" to access the instance of the singleton.
Right:
class UniqueObject {
public:
static UniqueObject& the();
...
Wrong:
class UniqueObject {
public:
static UniqueObject& shared();
...
Wrong:
class UniqueObject {
...
};
UniqueObject& my_unique_object(); // Free function.
Comments
Use only one space before end of line comments and in between sentences in comments.
Right:
f(a, b); // This explains why the function call was done. This is another sentence.
Wrong:
int i; // This is a comment with several spaces before it, which is a non-conforming style.
double f; // This is another comment. There are two spaces before this sentence which is a non-conforming style.
Make comments look like sentences by starting with a capital letter and ending with a period (punctation). One exception may be end of line comments like this if (x == y) // false for NaN.
Use FIXME: (without attribution) to denote items that need to be addressed in the future.
Right:
draw_jpg(); // FIXME: Make this code handle jpg in addition to the png support.
Wrong:
draw_jpg(); // FIXME(joe): Make this code handle jpg in addition to the png support.
draw_jpg(); // TODO: Make this code handle jpg in addition to the png support.
Overriding Virtual Methods
The declaration of a virtual method inside a class must be declared with the virtual keyword. All subclasses of that class must either specify the override keyword when overriding the virtual method or the final keyword when overriding the virtual method and requiring that no further subclasses can override it.
Right:
class Person {
public:
virtual String description() { ... };
}
class Student : public Person {
public:
virtual String description() override { ... }; // This is correct because it only contains the "override" keyword to indicate that the method is overridden.
}
class Person {
public:
virtual String description() { ... };
}
class Student : public Person {
public:
virtual String description() final { ... }; // This is correct because it only contains the "final" keyword to indicate that the method is overridden and that no subclasses of "Student" can override "description".
}
Wrong:
class Person {
public:
virtual String description() { ... };
}
class Student : public Person {
public:
String description() override { ... }; // This is incorrect because it uses only the "override" keywords to indicate that the method is virtual. Instead, it should use both the "virtual" and "override" keywords.
}
class Person {
public:
virtual String description() { ... };
}
class Student : public Person {
public:
String description() final { ... }; // This is incorrect because it uses only the "final" keywords to indicate that the method is virtual and final. Instead, it should use both the "virtual" and "final" keywords.
}
class Person {
public:
virtual String description() { ... };
}
class Student : public Person {
public:
virtual String description() { ... }; // This is incorrect because it uses the "virtual" keyword to indicate that the method is overridden.
}