CSV.Rows errors when types for a column is specified to be Symbol

[AayushSabharwal]

Consider the following sample CSV:

id,pos_1,pos_2,pos_3,has_prisoner,capture_time,shape
5,11,86,1,false,0,diamond
4,5,43,1,false,0,diamond
6,16,90,1,false,0,diamond
7,75,35,1,false,0,diamond
2,35,89,1,false,0,diamond
10,81,25,1,false,0,diamond
9,48,98,1,false,0,diamond
8,98,62,1,false,0,diamond
3,50,2,1,false,0,diamond
1,95,24,1,false,0,diamond

The last column (shape) is required to be a Symbol. Specifying this through the types keyword seems to error on 0.9.6:

julia> for r in CSV.Rows(read("test/test.csv"); types = Dict(:id => Int, :pos_1 => Int, :pos_2 => Int, :pos_3 => Int, :has_prisoner => Bool, :capture_time => Int, :shape => Symbol))
           print(r)
       end
ERROR: UndefVarError: values not defined
Stacktrace:
 [1] iterate
   @ ~/.julia/packages/CSV/owrEo/src/rows.jl:266 [inlined]
 [2] iterate(r::CSV.Rows{Vector{UInt8}, Tuple{Symbol}, Any, PosLenString})
   @ CSV ~/.julia/packages/CSV/owrEo/src/rows.jl:274
 [3] top-level scope
   @ ./REPL[4]:1

The same issue does not arise when the type for :shape is specified to be String. It seems to be specific to CSV.Rows. This does not error:

julia> CSV.File(read("test/test.csv"); types = Dict(:id => Int, :pos_1 => Int, :pos_2 => Int, :pos_3 => Int, :has_prisoner => Bool, :capture_time => Int, :shape => Symbol))

Ping JuliaDynamics/Agents.jl#550

[quinnj]

Thanks @AayushSabharwal for the report! A fix (work-around) is up: #931