Support rekeying a KeyedArray

[rofinn]

I decided to go with pairs vs kwargs because:

1. I wanted to support unnamed dims
2. The nested pair syntax of oldname => newname => newvalues seemed more intuitive than mixing pairs and kwargs.

Closes #60

[rofinn]

Nightly error seems unrelated?

[glennmoy]

Nightly error seems unrelated?

nightly failed on my PR due to #57

[mcabbott]

I wonder if this method ought to be restricted to types which have axis keys?

Mostly because it seems a little odd verbally, and redundant to provide a constructor. But, on the other hand, maybe it's useful not to have to care?

Precedent:

julia> rename(rand(3), (:x,))
ERROR: MethodError: no method matching rename(::Vector{Float64}, ::Tuple{Symbol})
Closest candidates are:
  rename(::NamedDimsArray, ::Any) at /Users/me/.julia/dev/NamedDims/src/name_operations.jl:14

[rofinn]

Yeah, I mostly did this to avoid needing the awkward rekey(A::Union{NdaKa{L,T,N}, KaNda{L,T,N}}, k2::Tuple) where {L, T, N} syntax. I assume there isn't a better solution? I suppose if this code is being used more than read then it doesn't really matter, but it's really hard for my brain to parse that Union type.

[mcabbott]

Could be just A::Union{KeyedArray, NdaKa} I think?

[mcabbott]

Unless you wanted to split things so as to preserve the order of wrappers, with a separate A:: NdaKa method which unwraps the names & puts them back on afterwards.

[mcabbott]

Failure in 1.6 is

selectors: Error During Test at /home/runner/work/AxisKeys.jl/AxisKeys.jl/test/_basic.jl:106
128
  Test threw exception
129
  Expression: [V4[Near(x)] for x = xs] == [V5[Near(x)] for x = xs]
130
  BoundsError: attempt to access 10-element StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}} at index [0]
131
  Stacktrace:
132
    [1] throw_boundserror(A::StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}}, I::Tuple{Int64})
133
      @ Base ./abstractarray.jl:651
134

Seems like that should be unrelated... I should take a look at some point.

[mzgubic]

@mcabbott this would be nice to have, any chance you will have time to review at some point?

[oxinabox]

Is rekeying and renaming at some time a common operation?
Mightn't it be clearer to seperate it into two operations?

I would feel more comfortable with this PR if it only rekeyed.
And then we could do a follow up PR with the rekey + rename, that could be reveiwed can considered seperately

[rofinn]

That's a good point. Also, should something like rename be defined for NamedDimsArrays first? I like the syntax for doing both at the same time, but I agree that could be a separate PR.

[oxinabox]

I believe there is a rename in NamedDims.jl

[rofinn]

Are we still okay with the pairs syntax or do we want to use kwargs?

[rofinn]

Okay, I've just dropped the rekey and rename method for now. I think that kind of functionality is only common in the DataFrames ecosystem (combine).

[oxinabox]

Approved subject to splitting the rename+rekey into a seperate PR

[rofinn]

Will likely fail CI with the new CanonicalIndexError on 1.8.