Commits on Jul 10, 2020

1. Change deprecation warning for CSV.read …

   Fixes JuliaData/DataFrames.jl#2309. After much
   discussion, people really like the `CSV.read` function call naming, so
   it was decided to keep it around, while still allowing a break from
   DataFrames.jl as a dependency.
   
   I also realized that `CSV.read` does indeed provide one bit of
   value/functionality: we can wrap `CSV.File` in `Tables.CopiedColumns`.
   What this means is that columnar sinks can safely use the columns passed
   without needing to make copies; i.e. they can assume ownership of the
   columns. In `CSV.read`, the user is essentially saying, "I want to make
   a `CSV.File` and pass it directly to `sink`" which also implies that
   `CSV.File` doesn't need to "own" its own columns.
   
   The only question left in my mind is, with the 1.0 release, what to do
   with `CSV.read(file)` when no `sink` argument is passed. Suggestions
   have included just returning a `CSV.File`, since it's a valid table
   anyway. Or allowing DataFrames.jl to define the no-sink-arg version and
   return a `DataFrame`; that one's a bit awkward as a form of "blessed"
   type piracy, but could also be useful for users as convenience (they
   just have to remember to do `using DataFrames` before trying to use it).
   The other awkward part is that we're currently warning users of the
   deprecation and that they should explicitly spell out `DataFrame`
   whereas we might not actually require that.

   

   quinnj committed Jul 10, 2020
   Configuration menu

   • View commit details
   • Copy full SHA for ac3761d
   • Browse repository at this point

   Copy the full SHA

   ac3761d View commit details

   Browse the repository at this point in the history