CSVFiles: make delim a keyword argument in load

[zweiglimmergneis]

See issue #16.

load accepts delim as a keyword argument for CSV files and/or TSV files. The delimiting character may also be specified as the second positional argument (as before).

Tests:

load(File(format"CSV", "../../data/Iris/iris.data"))
# ok

load(File(format"CSV", "../../data/Iris/iris.data"), ',')
#ok

load(File(format"CSV", "../../data/Iris/iris.data"), delim=',')
#ok

load(File(format"CSV", "../../data/Iris/iris.data"), skiplines_begin=3, header_exists=false)
# ok

load(File(format"CSV", "../../data/Iris/iris.data"), delim=',', skiplines_begin=3, header_exists=false)
# ok

load(File(format"CSV", "../../data/Iris/iris.data"), skiplines_begin=3, header_exists=false, delim=',')
# ok
# similar tests with a file containing ';' as a separator are successful, too

# tsv
load("../../data/Iris/iris.tsv")
# ok

load("../../data/Iris/iris.tsv", '\t')
# ok

load("../../data/Iris/iris.tsv", delim='\t')
# ok

When load is called with a stream as first argument, delim is accepted as a keyword argument. However, I could not extract data from a stream obtained with open:

     julia> csio = open("../../data/Iris/iris.csv")
     IOStream(<file ../../data/Iris/iris.csv>)

     julia> csvs = load(Stream(format"CSV", csio), skiplines_begin=3, header_exists=false);

     julia> df = DataFrame(csvs)
     ERROR: MethodError: no method matching String(::Ptr{UInt8}, ::Int64)
     Stacktrace:
      [1] #csvread#17(::Base.Iterators.Pairs{Symbol,Integer,Tuple{Symbol,Symbol},NamedTuple{(:skiplines_begin, :header_exists),Tuple{Int64,Bool}}}, ::Function, ::IOStream, ::Char) at /home/nepomuk/.julia/packages/TextParse/WFgcL/src/csv.jl:80
      [2] (::getfield(TextParse, Symbol("#kw##csvread")))(::NamedTuple{(:skiplines_begin, :header_exists),Tuple{Int64,Bool}}, ::typeof(TextParse.csvread), ::IOStream, ::Char) at ./none:0
      [3] getiterator(::CSVFiles.CSVStream) at /home/nepomuk/projekte/run/julia/CSVFiles/src/CSVFiles.jl:93
      [4] DataFrame(::CSVFiles.CSVStream) at /home/nepomuk/.julia/packages/DataFrames/VC5qi/src/other/tables.jl:22
      [5] top-level scope at none:0

[codecov-io]

Codecov Report

Merging #35 into master will decrease coverage by 32.18%.
The diff coverage is 16.66%.

@@             Coverage Diff             @@
##           master      #35       +/-   ##
===========================================
- Coverage     100%   67.81%   -32.19%     
===========================================
  Files           2        2               
  Lines          59       87       +28     
===========================================
  Hits           59       59               
- Misses          0       28       +28

Impacted Files	Coverage Δ
src/CSVFiles.jl	60% <16.66%> (-40%)	⬇️
src/csv_writer.jl	76.19% <0%> (-23.81%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 153a21a...3a87378. Read the comment docs.

[davidanthoff]

Great!

Maybe we could add some tests? And then I'll merge.

The stream issue looks more like a generic problem, do you also get that with the released version?

Oh, and then I'm wondering whether we should actually deprecate the positional version? I'm not sure...

[davidanthoff]

I just merged another PR that implements this, should be in the tagged version now.