Traits
A trait describes an abstract interface that types can implement. This interface consists of associated items, which come in three varieties:
All traits define an implicit type parameter Self
that refers to "the type
that is implementing this interface". Traits may also contain additional type
parameters. These type parameters (including Self
) may be constrained by
other traits and so forth as usual.
Traits are implemented for specific types through separate implementations.
Associated functions and methods
Associated functions whose first parameter is named self
are called methods
and may be invoked using .
notation (e.g., x.foo()
) as well as the usual
function call notation (foo(x)
).
Consider the following trait:
# #![allow(unused_variables)] #fn main() { # type Surface = i32; # type BoundingBox = i32; trait Shape { fn draw(&self, Surface); fn bounding_box(&self) -> BoundingBox; } #}
This defines a trait with two methods. All values that have implementations
of this trait in scope can have their draw
and bounding_box
methods called,
using value.bounding_box()
syntax. Note that &self
is short for self: &Self
, and similarly, self
is short for self: Self
and &mut self
is
short for self: &mut Self
.
Traits can include default implementations of methods, as in:
# #![allow(unused_variables)] #fn main() { trait Foo { fn bar(&self); fn baz(&self) { println!("We called baz."); } } #}
Here the baz
method has a default implementation, so types that implement
Foo
need only implement bar
. It is also possible for implementing types to
override a method that has a default implementation.
Type parameters can be specified for a trait to make it generic. These appear after the trait name, using the same syntax used in generic functions.
# #![allow(unused_variables)] #fn main() { trait Seq<T> { fn len(&self) -> u32; fn elt_at(&self, n: u32) -> T; fn iter<F>(&self, F) where F: Fn(T); } #}
Associated functions may lack a self
argument, sometimes called 'static
methods'. This means that they can only be called with function call syntax
(f(x)
) and not method call syntax (obj.f()
). The way to refer to the name
of a static method is to qualify it with the trait name or type name, treating
the trait name like a module. For example:
# #![allow(unused_variables)] #fn main() { trait Num { fn from_i32(n: i32) -> Self; } impl Num for f64 { fn from_i32(n: i32) -> f64 { n as f64 } } let x: f64 = Num::from_i32(42); let x: f64 = f64::from_i32(42); #}
Associated Types
It is also possible to define associated types for a trait. Consider the
following example of a Container
trait. Notice how the type is available for
use in the method signatures:
# #![allow(unused_variables)] #fn main() { trait Container { type E; fn empty() -> Self; fn insert(&mut self, Self::E); } #}
In order for a type to implement this trait, it must not only provide
implementations for every method, but it must specify the type E
. Here's an
implementation of Container
for the standard library type Vec
:
# #![allow(unused_variables)] #fn main() { # trait Container { # type E; # fn empty() -> Self; # fn insert(&mut self, Self::E); # } impl<T> Container for Vec<T> { type E = T; fn empty() -> Vec<T> { Vec::new() } fn insert(&mut self, x: T) { self.push(x); } } #}
Associated Constants
A trait can define constants like this:
trait Foo { const ID: i32; } impl Foo for i32 { const ID: i32 = 1; } fn main() { assert_eq!(1, i32::ID); }
Any implementor of Foo
will have to define ID
. Without the definition:
# #![allow(unused_variables)] #fn main() { trait Foo { const ID: i32; } impl Foo for i32 { } #}
gives
error: not all trait items implemented, missing: `ID` [E0046]
impl Foo for i32 {
}
A default value can be implemented as well:
trait Foo { const ID: i32 = 1; } impl Foo for i32 { } impl Foo for i64 { const ID: i32 = 5; } fn main() { assert_eq!(1, i32::ID); assert_eq!(5, i64::ID); }
As you can see, when implementing Foo
, you can leave it unimplemented, as
with i32
. It will then use the default value. But, as in i64
, we can also
add our own definition.
Associated constants don’t have to be associated with a trait. An impl
block
for a struct
or an enum
works fine too:
# #![allow(unused_variables)] #fn main() { struct Foo; impl Foo { const FOO: u32 = 3; } #}
Trait bounds
Generic functions may use traits as bounds on their type parameters. This will have three effects:
- Only types that have the trait may instantiate the parameter.
- Within the generic function, the methods of the trait can be called on values that have the parameter's type. Associated types can be used in the function's signature, and associated constants can be used in expressions within the function body.
- Generic functions and types with the same or weaker bounds can use the generic type in the function body or signature.
For example:
# #![allow(unused_variables)] #fn main() { # type Surface = i32; # trait Shape { fn draw(&self, Surface); } struct Figure<S: Shape>(S, S); fn draw_twice<T: Shape>(surface: Surface, sh: T) { sh.draw(surface); sh.draw(surface); } fn draw_figure<U: Shape>(surface: Surface, Figure(sh1, sh2): Figure<U>) { sh1.draw(surface); draw_twice(surface, sh2); // Can call this since U: Shape } #}
Object Safety
Object safe traits can be the base trait of a trait object. A trait is object safe if it has the following qualities (defined in RFC 255):
- It must not require
Self: Sized
- All associated functions must either have a
where Self: Sized
bound or- Not have any type parameters (although lifetime parameters are allowed)
- Must be a method: its first parameter must be called self, with type
Self
,&Self
,&mut Self
,Box<Self>
. Self
may only be used in the type of the receiver.
- It must not have any associated constants.
Supertraits
Trait bounds on Self
are considered "supertraits". These are required to be
acyclic. Supertraits are somewhat different from other constraints in that
they affect what methods are available in the vtable when the trait is used as
a trait object. Consider the following example:
# #![allow(unused_variables)] #fn main() { trait Shape { fn area(&self) -> f64; } trait Circle : Shape { fn radius(&self) -> f64; } #}
The syntax Circle : Shape
means that types that implement Circle
must also
have an implementation for Shape
. Multiple supertraits are separated by +
,
trait Circle : Shape + PartialEq { }
. In an implementation of Circle
for a
given type T
, methods can refer to Shape
methods, since the typechecker
checks that any type with an implementation of Circle
also has an
implementation of Shape
:
# #![allow(unused_variables)] #fn main() { struct Foo; trait Shape { fn area(&self) -> f64; } trait Circle : Shape { fn radius(&self) -> f64; } impl Shape for Foo { fn area(&self) -> f64 { 0.0 } } impl Circle for Foo { fn radius(&self) -> f64 { println!("calling area: {}", self.area()); 0.0 } } let c = Foo; c.radius(); #}
In type-parameterized functions, methods of the supertrait may be called on
values of subtrait-bound type parameters. Referring to the previous example of
trait Circle : Shape
:
# #![allow(unused_variables)] #fn main() { # trait Shape { fn area(&self) -> f64; } # trait Circle : Shape { fn radius(&self) -> f64; } fn radius_times_area<T: Circle>(c: T) -> f64 { // `c` is both a Circle and a Shape c.radius() * c.area() } #}
Likewise, supertrait methods may also be called on trait objects.
# #![allow(unused_variables)] #fn main() { # trait Shape { fn area(&self) -> f64; } # trait Circle : Shape { fn radius(&self) -> f64; } # impl Shape for i32 { fn area(&self) -> f64 { 0.0 } } # impl Circle for i32 { fn radius(&self) -> f64 { 0.0 } } # let mycircle = 0i32; let mycircle = Box::new(mycircle) as Box<Circle>; let nonsense = mycircle.radius() * mycircle.area(); #}