Indexing error with Union{Float64,Missing} variables

[randyzwitch]

Given the dataframe described below, freqtable will fail. Switching the type from Union{Float64, Missings.Missing} to just Float64 allows the function to work

julia> showcols(storedf)
2080×10 DataFrames.DataFrame
│ Col # │ Name     │ Eltype                                                               │ Missing │ Values        │
├───────┼──────────┼──────────────────────────────────────────────────────────────────────┼─────────┼───────────────┤
│ 1     │ storeNum │ Union{Int64, Missings.Missing}                                       │ 0       │ 101  …  120   │
│ 2     │ Year     │ Union{Int64, Missings.Missing}                                       │ 0       │ 1  …  2       │
│ 3     │ Week     │ Union{Int64, Missings.Missing}                                       │ 0       │ 1  …  52      │
│ 4     │ p1sales  │ Union{Int64, Missings.Missing}                                       │ 0       │ 127  …  136   │
│ 5     │ p2sales  │ Union{Int64, Missings.Missing}                                       │ 0       │ 106  …  99    │
│ 6     │ p1price  │ Union{Float64, Missings.Missing}                                     │ 0       │ 2.29  …  2.49 │
│ 7     │ p2price  │ Union{Float64, Missings.Missing}                                     │ 0       │ 2.29  …  2.49 │
│ 8     │ p1prom   │ Union{Int64, Missings.Missing}                                       │ 0       │ 0  …  0       │
│ 9     │ p2prom   │ Union{Int64, Missings.Missing}                                       │ 0       │ 0  …  0       │
│ 10    │ country  │ Union{CategoricalArrays.CategoricalString{UInt32}, Missings.Missing} │ 0       │ US  …  CN     │
julia> using FreqTables

julia> freqtable(storedf, :p1price)
ERROR: ArgumentError: invalid index: 2.99
Stacktrace:
 [1] setindex!(::NamedArrays.NamedArray{Int64,1,Array{Int64,1},Tuple{DataStructures.OrderedDict{Union{Float64, Missings.Missing},Int64}}}, ::Int64, ::Float64) at /Users/randyzwitch/.julia/v0.6/NamedArrays/src/index.jl:134
 [2] _freqtable(::Tuple{Array{Union{Float64, Missings.Missing},1}}, ::Bool, ::FreqTables.UnitWeights, ::Void) at /Users/randyzwitch/.julia/v0.6/FreqTables/src/freqtable.jl:79
 [3] freqtable(::Array{Union{Float64, Missings.Missing},1}) at /Users/randyzwitch/.julia/v0.6/FreqTables/src/freqtable.jl:88
 [4] #freqtable#21(::Array{Any,1}, ::Function, ::DataFrames.DataFrame, ::Symbol, ::Vararg{Symbol,N} where N) at /Users/randyzwitch/.julia/v0.6/FreqTables/src/freqtable.jl:148
 [5] freqtable(::DataFrames.DataFrame, ::Symbol, ::Vararg{Symbol,N} where N) at /Users/randyzwitch/.julia/v0.6/FreqTables/src/freqtable.jl:148

[nalimilan]

Good catch, that's a bug in NamedArrays: davidavdav/NamedArrays.jl#62.

[mirestrepo]

I run into the same issue with Union{Int64, Missings.Missing} and works if I convert to Int64