hcat fails if one column is only 'missing' values

[quatrix]

I'm trying to load a CSV and hcat it with another DataFrame that has the same number of rows and unique columns.

using CSV
using DataFrames

df = DataFrame(a=1:3, b=[missing, missing, missing])
CSV.write("/tmp/a.csv", df)

df_csv = CSV.read("/tmp/a.csv")

hcat(df, DataFrame(c=1:3)) # works
hcat(df_csv, DataFrame(c=1:3)) # throws an exception

The first hcat works, the second throws this exception:

MethodError: no method matching iterate(::Nothing)
Closest candidates are:
  iterate(!Matched::Core.SimpleVector) at essentials.jl:600
  iterate(!Matched::Core.SimpleVector, !Matched::Any) at essentials.jl:600
  iterate(!Matched::ExponentialBackOff) at error.jl:218
  ...

Stacktrace:
 [1] iterate at ./iterators.jl:139 [inlined]
 [2] iterate at ./iterators.jl:138 [inlined]
 [3] foreach(::CSV.var"#43#46", ::Base.Iterators.Enumerate{Nothing}) at ./abstractarray.jl:1920
 [4] copy(::CSV.Column{Missing,Missing}) at /Users/quatrix/.julia/packages/CSV/2VBaR/src/tables.jl:107
 [5] (::DataFrames.var"#DataFrame#99#102")(::Bool, ::Type{DataFrame}, ::Array{AbstractArray{T,1} where T,1}, ::DataFrames.Index) at /Users/quatrix/.julia/packages/DataFrames/uPgZV/src/dataframe/dataframe.jl:126
 [6] Type at ./none:0 [inlined]
 [7] #select#130(::Bool, ::typeof(select), ::DataFrame, ::Base.OneTo{Int64}) at /Users/quatrix/.julia/packages/DataFrames/uPgZV/src/dataframe/dataframe.jl:875
 [8] #select#132 at ./none:0 [inlined]
 [9] #select at ./none:0 [inlined]
 [10] getindex at /Users/quatrix/.julia/packages/DataFrames/uPgZV/src/dataframe/dataframe.jl:395 [inlined]
 [11] #copy#125 at /Users/quatrix/.julia/packages/DataFrames/uPgZV/src/dataframe/dataframe.jl:698 [inlined]
 [12] #copy at ./none:0 [inlined]
 [13] #hcat#143 at /Users/quatrix/.julia/packages/DataFrames/uPgZV/src/dataframe/dataframe.jl:931 [inlined]
 [14] hcat(::DataFrame, ::DataFrame) at /Users/quatrix/.julia/packages/DataFrames/uPgZV/src/dataframe/dataframe.jl:931
 [15] top-level scope at In[6]:10

It seems to happen only when one of the columns is just 'missing' values.

workaround I found is to add a unique id to each row and then join them

another_df = DataFrame(c=1:3)

df_csv._id = 1:size(df_csv,1)
another_df._id = 1:size(another_df, 1)

result = join(df_csv, another_df, on=:_id)

# and then to remove _id column

[quinnj]

Thanks for the report; I have a fix in 1e10c9c