Add indexes methods

[rikhuijzer]

I was trying to use CategoricalArrays with Turing.jl. Surprisingly, it worked after I added some CategoricalArrays specialized type signatures for Base.checkindex and Base.to_index. According to the docstrings of both, these extensions are allowed, see ?Base.checkindex and ?Base.to_index.

With this PR, I suggest to add these specialized type signatures.

By the way, this PR is slightly related to #296, which is about a specialization for InvertedIndices.

[nalimilan]

Thanks, but I don't think that's right. CategoricalValue is not supposed to be used as a numeric index. In particular, CategoricalValue{Int} would be ambiguous as it's not clear whether you'd like to use the reference code corresponding to the level, or the wrapped numeric value.

Can you develop what's the problem you get with Turing.jl?

[rikhuijzer]

Fair enough; good point.

Can you develop what's the problem you get with Turing.jl?

I'm afraid that I don't understand this sentence. You mean to show the exact error here or to see whether the problem can be fixed at Turing.jl?

[nalimilan]

I guess both, but seeing the error would probably help.

[rikhuijzer]

Thanks, but I don't think that's right. CategoricalValue is not supposed to be used as a numeric index. In particular, CategoricalValue{Int} would be ambiguous as it's not clear whether you'd like to use the reference code corresponding to the level, or the wrapped numeric value.

On second thought, isn't this PR a more specific implementation of the existing design? Even though numeric indexing isn't supposed to be used, it is implemented:

julia> using CategoricalArrays

julia> v = categorical([1, 1, 2])
3-element CategoricalArray{Int64,1,UInt32}:
 1
 1
 2

julia> v[3]
CategoricalValue{Int64, UInt32} 2

I guess both, but seeing the error would probably help.

When I use CategoricalArrays for df.culture in this model, the error is as follows:

Sampling 100%|████████████████████████████████████████| Time: 0:00:01
ERROR: LoadError: ArgumentError: unable to check bounds for indices of type CategoricalValue{String, UInt32}
Stacktrace:
  [1] checkindex(#unused#::Type{Bool}, inds::Base.OneTo{Int64}, i::CategoricalValue{String, UInt32})
    @ Base ./abstractarray.jl:665
  [2] checkindex
    @ ./abstractarray.jl:680 [inlined]
  [3] checkbounds
    @ ./abstractarray.jl:595 [inlined]
  [4] checkbounds
    @ ./abstractarray.jl:610 [inlined]
  [5] _getindex
    @ ./multidimensional.jl:831 [inlined]
  [6] getindex
    @ ./abstractarray.jl:1164 [inlined]
  [7] #1
    @ ~/git/TuringModels.jl/scripts/spatial-autocorrelation-oceanic.jl:49 [inlined]
  [8] (::var"#1#2")(_rng::Random._GLOBAL_RNG, _model::DynamicPPL.Model{var"#1#2", (:Dmat, :society, :logpop, :total_tools), (), (), Tuple{Matrix{Float64}, CategoricalVector{String, UInt32, String, CategoricalValue{String, UInt32}, Union{}}, Vector{Float64}, Vector{Int64}}, Tuple{}}, _varinfo::DynamicPPL.ThreadSafeVarInfo{DynamicPPL.UntypedVarInfo{DynamicPPL.Metadata{Dict{DynamicPPL.VarName, Int64}, Vector{Distribution}, Vector{DynamicPPL.VarName}, Vector{Real}, Vector{Set{DynamicPPL.Selector}}}, Float64}, Vector{Base.RefValue{Float64}}}, _sampler::DynamicPPL.SampleFromUniform, _context::DynamicPPL.DefaultContext, Dmat::Matrix{Float64}, society::CategoricalVector{String, UInt32, String, CategoricalValue{String, UInt32}, Union{}}, logpop::Vector{Float64}, total_tools::Vector{Int64})
    @ Main ./none:0
  [9] macro expansion
    @ ~/.julia/packages/DynamicPPL/wf0dU/src/model.jl:0 [inlined]
 [10] _evaluate
    @ ~/.julia/packages/DynamicPPL/wf0dU/src/model.jl:154 [inlined]
...

After a method with a specialized type on CategoricalArrays for Base.checkindex, the error becomes:

ERROR: LoadError: ArgumentError: invalid index: CategoricalValue{String, UInt32} "Malekula" of type CategoricalValue{String, UInt32}
Stacktrace:
  [1] to_index(i::CategoricalValue{String, UInt32})
    @ Base ./indices.jl:297
  [2] to_index(A::Vector{Float64}, i::CategoricalValue{String, UInt32})
    @ Base ./indices.jl:274