Attribute macro to implement a trait for tuples
When wanting to implement a trait for combinations of tuples, Rust requires the trait to be implemented
for each combination manually. With this crate you just need to place #[impl_for_tuples(5)]
above
your trait declaration (in full-automatic mode) to implement the trait for the tuple combinations
(), (T0), (T0, T1), (T0, T1, T2), (T0, T1, T2, T3), (T0, T1, T2, T3, T4, T5)
. The number of tuples is the
parameter given to the attribute and can be chosen freely.
This crate provides two modes full-automatic and semi-automatic. The full-automatic mode just requires
the trait definition to implement the trait for the tuple combinations. While being much easier to
use, it also comes with some restrictions like no associated types, no return values or no associated
consts. To support these, the semi-automatic mode is provided. This mode requires a dummy implementation
block of the trait that is expanded to all the tuple combinations implementations. To express the
tuple access in this dummy implementation a special syntax is required for_tuples!( #( Tuple::function(); )* )
.
This would expand to Tuple::function();
for each tuple while Tuple
is chosen by the user and will be
replaced by the corresponding tuple identifier per iteration.
The attribute macro can be called with one #[impl_for_tuples(5)]
or with two #[impl_for_tuples(2, 5)]
parameters. The former instructs the macro to generate up to a tuple of five elements and the later instructs it
to generate from a tuple with two element up to five elements.
trait Trait {
type Ret;
type Arg;
type FixedType;
const VALUE: u32;
fn test(arg: Self::Arg) -> Self::Ret;
fn test_with_self(&self) -> Result<(), ()>;
}
#[impl_for_tuples(1, 5)]
impl Trait for Tuple {
// Here we expand the `Ret` and `Arg` associated types.
for_tuples!( type Ret = ( #( Tuple::Ret ),* ); );
for_tuples!( type Arg = ( #( Tuple::Arg ),* ); );
for_tuples!( const VALUE: u32 = #( Tuple::VALUE )+*; );
// Here we set the `FixedType` to `u32` and add a custom where bound that forces the same
// `FixedType` for all tuple types.
type FixedType = u32;
for_tuples!( where #( Tuple: Trait<FixedType=u32> )* );
fn test(arg: Self::Arg) -> Self::Ret {
for_tuples!( ( #( Tuple::test(arg.Tuple) ),* ) )
}
fn test_with_self(&self) -> Result<(), ()> {
for_tuples!( #( Tuple.test_with_self()?; )* );
Ok(())
}
}
The given example shows all supported combinations of for_tuples!
. When accessing a method from the
self
tuple instance, self.Tuple
is the required syntax and is replaced by self.0
, self.1
, etc.
The placeholder tuple identifer is taken from the self type given to the implementation block. So, it
is up to the user to chose any valid identifier.
The separator given to #( Tuple::something() )SEPARATOR*
can be chosen from ,
, +
, -
,
*
, /
, |
, &
or nothing for no separator.
By adding the #[tuple_types_no_default_trait_bound]
above the impl block, the macro will not add the
automatic bound to the implemented trait for each tuple type.
The trait bound can be customized using #[tuple_types_custom_trait_bound(NewBound)]
.
The new bound will be used instead of the impleted trait for each tuple type.
The macro does not supports for_tuples!
calls in a different macro, so stuff like
vec![ for_tuples!( bla ) ]
will generate invalid code.
#[impl_for_tuples(5)]
trait Notify {
fn notify(&self);
}
trait Notify {
fn notify(&self) -> Result<(), ()>;
}
#[impl_for_tuples(5)]
impl Notify for TupleIdentifier {
fn notify(&self) -> Result<(), ()> {
for_tuples!( #( TupleIdentifier.notify()?; )* );
Ok(())
}
}
Licensed under either of
- Apache License, Version 2.0
- MIT license at your option.
License: Apache-2.0/MIT