Update scitypes.jl

[OkonSamuel]

Addresses ScientificTypes.jl#179

[juliohm]

So we should only depend on ScientificTypesBase.jl now @OkonSamuel ?

[OkonSamuel]

So we should only depend on ScientificTypesBase.jl now @OkonSamuel ?

Yes. Only extend methods from ScientificTypesBase. But you would still need to import ScientificTypes to use scitype(c)

julia> using CoDa, ScientificTypes

julia> c = Composition(1,2,3)
                  3-part composition
      ┌                                        ┐
   w1 ┤■■■■■■■■■■■■ 1.0
   w2 ┤■■■■■■■■■■■■■■■■■■■■■■■ 2.0
   w3 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 3.0
      └                                        ┘

julia> scitype(c)
Compositional{3}

[juliohm]

But it seems that we still need DefaultConvention from ScientificTypes.jl? It is a bit confusing. Can you please clarify?

[codecov-commenter]

Codecov Report

Merging #29 (7b5d14d) into master (cae693f) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master      #29   +/-   ##
=======================================
  Coverage   84.50%   84.50%           
=======================================
  Files          13       13           
  Lines         355      355           
=======================================
  Hits          300      300           
  Misses         55       55           

Impacted Files	Coverage Δ
src/scitypes.jl	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cae693f...7b5d14d. Read the comment docs.

[OkonSamuel]

But it seems that we still need DefaultConvention from ScientificTypes.jl? It is a bit confusing. Can you please clarify?

Yes, You would still need ScientificTypes, DefaultConvention.

So that could be done by

ST.ScientificTypesBase.scitype(::YourType, ::ST.DefaultConvention) = ...

I hope this is clear enough.

[juliohm]

I find it strange that ScientificTypes.jl doesn't reexport scitype from ScientificTypesBase.jl. Someone depending on the default convention would like to implement methods for functions defined in the same module. Any chance these functions can be reexported in ScientificTypes.jl? I really feel that we should try another round of refactoring in these two packages.

[OkonSamuel]

I find it strange that ScientificTypes.jl doesn't reexport scitype from ScientificTypesBase.jl. Someone depending on the default convention would like to implement methods for functions defined in the same module. Any chance these functions can be reexported in ScientificTypes.jl? I really feel that we should try another round of refactoring in these two packages.

I'll discuss with @ablaom about this possibility.

[OkonSamuel]

@juliohm what if I renamed the scitype method from ScientificTypesBase as scitype_def and exported that?
So you could do

using CoDa, ScientificTypes

ScientificTypes.scitype_def(::YourType, ::ST.DefaultConvention) = ...

c = Composition(1,2,3)

scitype(c)

[juliohm]

Why do we need two different function names @OkonSamuel ? Isn't just a single function with multiple methods we are discussing? Something changed recently? Why things are split between two packages the way they are?

[OkonSamuel]

Why do we need two different function names @OkonSamuel ? Isn't just a single function with multiple methods we are discussing? Something changed recently? Why things are split between two packages the way they are?

Yes, there was a recent change as regards performance, we wanted to avoid setting convention at run time.

We can't just move the DefaultConvention implemted in ScientificTypes to ScientificTypesBase.jl because we wish that ScientificTypesBase.jl be as light as possible.