Issue

Tried to implement both non-linear and linear initialization methods, however, they are incompatible. Linear initialization needs to return Vec, so that Genotype can construct an instance of Self. Non-linear (Tree) genotype would need a Vec<Node> to do the same. Current trait based system, does not allow it.

Possible solution

Change trait-based system into an enum-based system:

enum Initialization {
  Linear(Linear),
  Non-linear(NonLinear)
}

with respective implementations, though need to consider grouping mechanism.

struct Linear<T> { field1 ... }
impl<T> Linear<T> { fn initialize(...) -> Vec<T> }

struct NonLinear<T>
impl<T> NonLinear<T> { fn initialize(...) -> Vec<Node<T>> }

Then methods would be used in Genotype via match:

match Initialization {
  Linear(init_scheme) => init_scheme.initialize(...),
  NonLinear(init_scheme) => init_scheme.initialize(...),
  _ => panic!(...) 
  }

README.typ in spec folder at project root should be rewritten and worked on.

Adding logger may be useful in monitoring the execution/state of the process.

Documentation is minimal at the time of creating this issue. Should be properly written and maintained.

README.md file at the project root is minimal and should be properly extended.

Mutation

All basic mutation operators should be implemented and handled via enum-based system.

• Enum to handle execution (c.f. Issue #6).
• Mutation ops to implement: sub-tree, size-fair, point, hoist, shrink, permutation, random, and systematic (c.f. Field Guide to GP page 42).

Crossover

All basic crossover operators should be implemented and handled via enum-based system.

• Enum to handle execution (c.f. Issue #6)
• Crossover ops to implement: one-point, uniform, context-preserving, size-fair, and other (c.f. Field Guide to GP page 45).

How to write tests

Tests for all modules:

• How to test initialization methods?
  • Chi-square tests for distributions? -> Should be a valid method to check if everything is generated properly.
  • Validity tests for individuals? -> Should be a valid method to check if every generated individual is valid and nothing strange happens.
• How to test mutation and crossover methods?
  • Chi-square tests? -> Not really sure if it makes sense, i.e., what distribution should we check for?
  • Validity tests for individuals? -> Should be a valid method to check if every generated individual is valid.