CSV.write Cannot write Nothing

[evveric]

df = DataFrame(d=[nothing])
CSV.write(raw"c:\temp\t.txt", df)

This has error

[quinnj]

Yeah, the question is what we should write/print for nothing. In Base, they made the decision that nothing shouldn't be printable. I think in the previous release, CSV.jl treated nothing like missing, so it just printed an empty field (or a provided null string). I guess we could go back to that behavior.

[quinnj]

@nalimilan, @bkamins, do either of you have an opinion here? Should we just treat nothing like missing when writing out a csv file?

[nalimilan]

I guess it would make sense to provide an option to choose the string to use for nothing, as for missing. A simpler solution would be to print it as nothing, as it seems Base is going to switch to that soon anyway (JuliaLang/julia#32148).

At any rate, nothing and missing should probably not be printed the same way by default, as it would make a round-trip lose information.

[quinnj]

Ok, I think I'd like to go with not supporting writing nothing in CSV files, with reasoning as follows:

• CSV.jl isn't really in the business of supporting custom types, particularly on the parsing side; so in that respect, it makes sense, at least for the moment, to restrict parsing and writing to the known set of basic types (date, datetime, time, int, float, bool, string)
• So this makes CSV.jl on the side of not being a "serialization" type package, and more of a dedicated "data format" package. That might change in the future, but for now, I think it makes sense and limits what we need to test/cover for