Unusual constructor design

[JeffFessler]

See:
#29 (comment)

[johnnychen94]

It seems to me that the Gauss3(...) function is made for two purposes: 1) interpreting convenient input types (cx, cy, cz, wx, wy, wz, Φ, Θ, value) to a more structural one (center, radii, angle, v), and 2) create an AbstractObject type that wraps it (IIUC this is because AbstractObject and AbstractShape shares different type hierarchy).

How about separating these coupled functionalities apart via:

# type annotations are skipped for illustration purpose
struct Gauss3 <: AbstractShape3
    center
    radii
    angle
    v
end

struct Object{S, D, ...} <: AbstractObject
    # no `shape` field
    center
    radii
    angle
    v
    param
end
Object(shape, params) = Object{typeof(shape)}(shape.center, shape.radii, shape.angle, shape.v, param)

I guess the reason you want to make Gauss3 a type with empty fields is because you don't want to store the same information in both ob.shape and ob. This can be done equivalently to only store the type information in type parameters S so that functions can still dispatch on types S.

But another question is, how necessary it is to store information in ob.center and not ob.shape.center?

[JeffFessler]

Thanks for the input! I would prefer to avoid the nesting of ob.shape.center and I'd also prefer to avoid the repetition of many shapes all having the same args (center, radii, angle, v) so after I finish up some other work I will revisit this and probably adopt your initial suggestion of lower case gauss3 for the constructors, perhaps as wrappers around something like your suggestion of
Object{typeof(shape)}(shape.center, shape.radii, shape.angle, shape.v, param).
I'll ping you eventually when I make a PR related to this.

[JeffFessler]

closed by #38