Integer to category

[MarkoMekapses]

Not sure that it's an issue, but as_categorical throws an error in converting integers to categories, like so
cars = DataFrame(cylinders=[2, 4, 6, 8]) @mutate cars cylinders = as_categorical(cylinders)
throws
MethodError: Cannot convertan object of type Int32 to an object of type String The functionconvert exists, but no method is defined for this combination of argument types.
I understand why this is the case... we could want to depict Int32 as octal, or hex, or whatever. But it seems a common enough problem (cf dataframe column names telling you to cast to Symbol) that there might already be a proper way to deal with this in TidierCats. If not, I propose that there should be.

[drizk1]

there is not a method for that right now.

would something as simple as this solve the problem

function as_categorical(arr::AbstractArray)
    T = eltype(arr)
    if T <: Number
        # keep numeric data as numeric categories
        return CategoricalArray{Union{Missing, T}}(arr)
    else
        # fallback: treat them as strings
        return CategoricalArray(map(x -> ismissing(x) ? missing : string(x), arr))
    end
end
julia> @mutate cars cylinders = as_categorical(cylinders)
4×1 DataFrame
 Row │ cylinders 
     │ Cat…?     
─────┼───────────
   1 │ 2
   2 │ 4
   3 │ 6
   4 │ 8

[drizk1]

@MarkoMekapses This should be fixed in v.2.0 now. Give it a shot and let me know if you run into any issues.

Closing this issue for now but reopen or open a new as things come up.