Selecting more than 1 column of PTDF matrix

Error when selecting specific columns of the PTDF matrix I would expect that the following commands work:

julia> a = rand(4,5)
4×5 Array{Float64,2}:
 0.0246739  0.483415  0.0234004  0.378329  0.965849
 0.665427   0.575655  0.843898   0.457372  0.492012
 0.441002   0.719667  0.619103   0.252873  0.80892
 0.647832   0.472036  0.784917   0.577866  0.25087

julia> b = a[:,[2,4]]
4×2 Array{Float64,2}:
 0.483415  0.378329
 0.575655  0.457372
 0.719667  0.252873
 0.472036  0.577866

However for a system sys the PTDF matrix

julia> PTDF_matrix = PTDF(sys)
PowerNetworkMatrix
    Dimension 1, ["Line1", "Line2", "Line3", "Line4", "Line5"]
    Dimension 2, [1, 2, 3, 4]
And data, a (5, 4):
 0.0  -0.625  -0.375  -0.5
 0.0  -0.375  -0.625  -0.5
 0.0   0.25   -0.25    1.38778e-17
 0.0   0.125  -0.125  -0.5
 0.0  -0.125   0.125  -0.5

julia> b = PTDF_matrix[:,[2,4]]
ERROR: KeyError: key [2, 4] not found
Stacktrace:
 [1] getindex at ./dict.jl:467 [inlined]
 [2] lookup_index at /Users/abrahamalvarezb/.julia/packages/PowerSystems/r86iN/src/utils/network_calculations/common.jl:38 [inlined]
 [3] _to_index_tuple at /Users/abrahamalvarezb/.julia/packages/PowerSystems/r86iN/src/utils/network_calculations/common.jl:52 [inlined] (repeats 2 times)
 [4] to_index at /Users/abrahamalvarezb/.julia/packages/PowerSystems/r86iN/src/utils/network_calculations/common.jl:71 [inlined]
 [5] getindex(::PTDF{Tuple{Array{String,1},Array{Int64,1}},Tuple{Dict{String,Int64},Dict{Int64,Int64}}}, ::Function, ::Array{Int64,1}) at /Users/abrahamalvarezb/.julia/packages/PowerSystems/r86iN/src/utils/network_calculations/common.jl:89
 [6] top-level scope at REPL[107]:1

Is this an issue or can it be improved?

Reproduce the PTDF(sys) example and the failing PTDF_matrix[:, [2,4]] selection. Start with utils/network_calculations/common.jl, focusing on lookup_index, _to_index_tuple, to_index, and getindex from the stack trace. Done means selecting multiple column indices returns the corresponding 5×2 matrix without a KeyError, with regression coverage.