M3/multitraits

[mauro3]

Edit: this would be cool but increases the code complexity considerably. It might be easier to provide this functionality with generated functions. So, if demand for this feature materializes, revisit.

I think I'll merge this after a bit of settling. This came out of discussions over in JuliaLang/julia#13222. This adds:

• inheritance
• trait functions dispatching on several traits (although only a fixed set)

This is a breaking change because now trait-functions need to be initialized via:
@traitfn f{X; TR{X}}(x::X)
(sans =...).

I think this is as far as Holy Traits can be taken. Going further would get into the Traits.jl territory of trait-dispatch and trait-definitions via method specifications.

[mauro3]

Part of the addition looks like so (from the README):

Dispatch on several traits

It is also possible to dispatch on several traits:

@traitfn f55{X, Y;  TT1{X},  TT2{Y}}(x::X, y::Y)
@traitfn f55{X, Y;  TT1{X},  TT2{Y}}(x::X, y::Y) = 1
@traitfn f55{X, Y; !TT1{X},  TT2{Y}}(x::X, y::Y) = 2
@traitfn f55{X, Y;  TT1{X}, !TT2{Y}}(x::X, y::Y) = 3
@traitfn f55{X, Y; !TT1{X}, !TT2{Y}}(x::X, y::Y) = 4

Note that all methods need to feature the same traits (possibly
negated) in the same order. Any method violating that will never be
called (without an error being thrown).

[JeffreySarnoff]

This capability is very important to Traits being "trait-like" in the humanistic sense.
Any time I think of a trait-ified way, naturally, there are almost always simultaneous ontologically kinda-orthogonal multiple traits co-dispatching (and with a type or two).

[mauro3]

Thanks @JeffreySarnoff, I take this as a +1 :-)

[vollmersj]

Any update here?

[mauro3]

No. If you're at JuliaCon we could discuss it there.

[InterdisciplinaryPhysicsTeam]

Hi,

Would you have any updates on this?

Thanks

[mauro3]

I don't have any plans with this, nor time.

[MilesCranmer]

Friendly ping :)

I dream of rust-like generics in Julia; this PR would make it so!