Bug in setindex

[aaowens]

This works

using IndexedTables
cs = Columns(x=[1.2, 3.4], y=[3,4])
 t = table(cs, copy=false, pkey=:x)
ts = ndsparse(t)
ts[1.3] = (y = 5,)

However, if you first accidentally try to assign the wrong type, assigning correctly breaks the table.

julia> using IndexedTables

julia> cs = Columns(x=[1.2, 3.4], y=[3,4])
2-element StructArrays.StructArray{NamedTuple{(:x, :y),Tuple{Float64,Int64}},1,NamedTuple{(:x, :y),Tuple{Array{Float64,1},Array{Int64,1}}}}:
 (x = 1.2, y = 3)
 (x = 3.4, y = 4)

julia> t = table(cs, copy=false, pkey=:x)
Table with 2 rows, 2 columns:
x    y
──────
1.2  3
3.4  4

julia> ts = ndsparse(t)
1-d NDSparse with 2 values (1 field named tuples):
x   │ y
────┼──
1.2 │ 3
3.4 │ 4

julia> ts[1.3] = 5
ERROR: type Int64 has no field y
...

julia> ts
1-d NDSparse with 2 values (1 field named tuples):
x   │ y
────┼──
1.2 │ 3
3.4 │ 4
julia> ts[1.3] = (y = 5,)
(y = 5,)

julia> ts
Error showing value of type NDSparse{NamedTuple{(:y,),Tuple{Int64}},Tuple{Float64},StructArrays.StructArray{NamedTuple{(:x,),Tuple{Float64}},1,NamedTuple{(:x,),Tuple{Array{Float64,1}}}},StructArrays.StructArray{NamedTuple{(:y,),Tuple{Int64}},1,NamedTuple{(:y,),Tuple{Array{Int64,1}}}}}:
ERROR: index and data must have the same number of elements

I think this is because of this function

function _setindex_scalar!(t, rhs, idxs)
    push!(t.index_buffer, idxs)
    push!(t.data_buffer, rhs)
    t
end

It doesn't check the types beforehand, so it successfully pushes to the index but fails to push to the data.