1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104
use crate::prelude::*;
/// Algorithm configurations. A trait for preparing the algorithm.
///
/// The setting type is actually a builder of the [`AlgCfg::Algorithm`] type.
///
/// Please note that the setting should not overlap with the [`SolverBuilder`].
pub trait AlgCfg {
/// Associated algorithm.
type Algorithm<F: ObjFunc>: Algorithm<F> + 'static;
/// Create the algorithm.
fn algorithm<F: ObjFunc>(self) -> Self::Algorithm<F>;
/// Default population number.
fn pop_num() -> usize {
200
}
}
/// The methods of the metaheuristic algorithms.
///
/// 1. Implement [`AlgCfg`] trait then indicate to a "method" type.
/// 1. Implement `Algorithm` trait on the "method" type.
///
/// Usually, the "method" type that implements this trait will not leak from the
/// API. All most common dataset is store in the [`Ctx`] type. So the "method"
/// type is used to store the additional data if any.
///
/// ```
/// use metaheuristics_nature::prelude::*;
///
/// /// A setting with fields.
/// #[derive(Default)]
/// pub struct MySetting1 {
/// my_option: u32,
/// }
///
/// /// The implementation of the structure with fields.
/// impl AlgCfg for MySetting1 {
/// type Algorithm<F: ObjFunc> = Method;
/// fn algorithm<F: ObjFunc>(self) -> Self::Algorithm<F> {
/// Method /* inherit setting */
/// }
/// }
///
/// /// No setting.
/// #[derive(Default)]
/// pub struct MySetting2;
///
/// /// The implementation of a tuple-like structure.
/// impl AlgCfg for MySetting2 {
/// type Algorithm<F: ObjFunc> = Method;
/// fn algorithm<F: ObjFunc>(self) -> Self::Algorithm<F> {
/// Method
/// }
/// }
///
/// /// The type implements our algorithm.
/// pub struct Method;
///
/// impl<F: ObjFunc> Algorithm<F> for Method {
/// fn generation(&mut self, ctx: &mut Ctx<F>, rng: &mut Rng) {
/// /* implement the method */
/// }
/// }
/// ```
///
/// The complete algorithm will be implemented by the [`Solver`](crate::Solver)
/// type automatically. All you have to do is implement the "initialization"
/// method and "generation" method, which are corresponded to the
/// [`Algorithm::init()`] and [`Algorithm::generation()`] respectively.
///
/// The generic type `F: ObjFunc` is the objective function marker, which is
/// used to allow storing the types that are related to the objective function
/// for the implementor `Self`. An actual example is
/// [`crate::methods::pso::Method`].
pub trait Algorithm<F: ObjFunc>: MaybeParallel {
/// Initialization implementation.
///
/// The information of the [`Ctx`] can be obtained or modified at this phase
/// preliminarily.
///
/// The default behavior is do nothing.
#[inline]
#[allow(unused_variables)]
fn init(&mut self, ctx: &mut Ctx<F>, rng: &mut Rng) {}
/// Processing implementation of each generation.
fn generation(&mut self, ctx: &mut Ctx<F>, rng: &mut Rng);
}
/// Implement for `Box<dyn Algorithm<F>>`.
///
/// See also [`SolverBox`].
impl<F: ObjFunc, T: Algorithm<F> + ?Sized> Algorithm<F> for alloc::boxed::Box<T> {
#[inline]
fn init(&mut self, ctx: &mut Ctx<F>, rng: &mut Rng) {
self.as_mut().init(ctx, rng);
}
#[inline]
fn generation(&mut self, ctx: &mut Ctx<F>, rng: &mut Rng) {
self.as_mut().generation(ctx, rng);
}
}