ShiftedArray's lag causes an AssertionError when used with groupby and transform

[robsmith11]

See the example below:

julia> using ShiftedArrays                                                   

julia> df = DataFrame(x=rand(10), y=rand(1:3, 10))
10×2 DataFrame
│ Row │ x        │ y     │
│     │ Float64  │ Int64 │
├─────┼──────────┼───────┤
│ 1   │ 0.606215 │ 3     │
│ 2   │ 0.128396 │ 1     │
│ 3   │ 0.194318 │ 3     │
│ 4   │ 0.139702 │ 1     │
│ 5   │ 0.766211 │ 3     │
│ 6   │ 0.806462 │ 2     │
│ 7   │ 0.162772 │ 3     │
│ 8   │ 0.863007 │ 1     │
│ 9   │ 0.269158 │ 3     │
│ 10  │ 0.105715 │ 3     │

julia> @linq df |> groupby(:y) |> transform(prevx=lag(:x))
ERROR: AssertionError: newtype_first === nothing
Stacktrace:
 [1] _transform!(::Array{Missing,1}, ::ShiftedArray{Float64,Missing,1,SubArray{Float64,1,Array{Float64,1},Tuple{Array{Int64,1}},false}}, ::Int64, ::GroupedDataFrame{DataFrame}, ::var"#251#252", ::Array{Int64,1}, ::Array{Int64,1}) at /home/me/.julia/packages/DataFramesMeta/c88dH/src/DataFramesMeta.jl:427
 [2] transform(::GroupedDataFrame{DataFrame}; kwargs::Base.Iterators.Pairs{Symbol,var"#251#252",Tuple{Symbol},NamedTuple{(:prevx,),Tuple{var"#251#252"}}
}) at /home/me/.julia/packages/DataFramesMeta/c88dH/src/DataFramesMeta.jl:0
 [3] top-level scope at /home/me/.julia/packages/DataFramesMeta/c88dH/src/DataFramesMeta.jl:473

However, lag works with both by and transform when called separately:

julia> @linq df |> by(:y, prevx=lag(:x))
10×2 DataFrame
│ Row │ y     │ prevx    │
│     │ Int64 │ Float64⍰ │
├─────┼───────┼──────────┤
│ 1   │ 3     │ missing  │
│ 2   │ 3     │ 0.606215 │
│ 3   │ 3     │ 0.194318 │
│ 4   │ 3     │ 0.766211 │
│ 5   │ 3     │ 0.162772 │
│ 6   │ 3     │ 0.269158 │
│ 7   │ 1     │ missing  │
│ 8   │ 1     │ 0.128396 │
│ 9   │ 1     │ 0.139702 │
│ 10  │ 2     │ missing  │

julia> @linq df |> transform(prevx=lag(:x))
10×3 DataFrame
│ Row │ x        │ y     │ prevx    │
│     │ Float64  │ Int64 │ Float64⍰ │
├─────┼──────────┼───────┼──────────┤
│ 1   │ 0.606215 │ 3     │ missing  │
│ 2   │ 0.128396 │ 1     │ 0.606215 │
│ 3   │ 0.194318 │ 3     │ 0.128396 │
│ 4   │ 0.139702 │ 1     │ 0.194318 │
│ 5   │ 0.766211 │ 3     │ 0.139702 │
│ 6   │ 0.806462 │ 2     │ 0.766211 │
│ 7   │ 0.162772 │ 3     │ 0.806462 │
│ 8   │ 0.863007 │ 1     │ 0.162772 │
│ 9   │ 0.269158 │ 3     │ 0.863007 │
│ 10  │ 0.105715 │ 3     │ 0.269158 │

[bkamins]

It seems that this line https://github.com/JuliaData/DataFramesMeta.jl/blob/master/src/DataFramesMeta.jl#L390 leads to this bug.

[pdeffebach]

Fixed by #163

[bkamins]

closing - right?