more general solution for modular agents? about loneparentsmodel.jl HOT 6 CLOSED

Hope I am understanding that correctly. Say we are dealing with module health, why the function add_child! would be needed?

from loneparentsmodel.jl.

add_child! is obviously part of module Kinship. The problem is that a separate module Kinship would have to define add_child! without knowing in advance how that module will be incorporated into an agent. And forwarding every function a module defines from the agent type to the module type is annoying and error-prone.

from loneparentsmodel.jl.

I guess with some representative (psuode-)code, I could follow it better. So my understanding

module Kinship 
  ...
  add_child!(::Kinship) = ..
  ..  
end # Ki

module Person 
   .. 
   # you don't want add_child(::Person) here? 
   .. 
end # Person 

from loneparentsmodel.jl.

Having

add_child!(p::Person) = add_child!(p.kinship)

would be the obvious way to handle these modules. However, I think that's not very satisfying since it means every function defined in any module has to be replicated in every definition of an agent type that uses that module.

from loneparentsmodel.jl.

Ok. You may want to propose if this is milestone V0.X or V{1,2}.X

from loneparentsmodel.jl.

Haven't tried it, but this looks like it might do exactly what we need: https://github.com/JeffreySarnoff/TypedDelegation.jl

from loneparentsmodel.jl.