Ergonomic ways of recoding data about dataframesmeta.jl

[nalimilan]

That's really cool. I would say the ideal solution would be to automatically create new columns, choosing the most appropriate type like Julia does for variables. Though I guess it would be tricky to decide whether to use a NullableArray or a standard array.

Given it may be hard to get this to work, an explicit syntax is a good solution (be it temporary or final). I think I'd prefer the @newcol form, but writing @newcol x::Array{Int} to match the standard Julia syntax. Actually, maybe @byrow! could detect type assertions and treat them as declarations of a new column, without even @newcol.

Then there's also the question of how to allow propagating null values. I guess we need to wait for the @lift macro to work fine in NullableArrays.jl before implementing this.

Finally, there's the minor annoyance of writing :x instead of x, which we discussed elsewhere already. But that would really be the icing on the cake: I think we can close this issue even if we don't have it. :-)

PS: Your example above doesn't work when copy/pasting: looks like you forgot one line creating df. See also JuliaData/DataFrames.jl#933