remove DataFrame!?

[JeffBezanson]

As far as I can tell, this doesn't mutate anything, so the ! is not appropriate. For the cases where it gives errors, DataFrame could instead give the error if copycols=false is passed.

[bkamins]

Yes, DataFrame! does not mutate anything.

For the cases where it gives errors,

Having this method is not related with errors - it is just a convenience.

The reason why it exists is that Base Julia does not provide a convenient syntax for partial function application,
and being able to easily avoid copying data when DataFrame is created was considered important by users for performance reasons. The ! in the name was to warn the user that mutation of the source might happen in the future (as in general DataFrame guarantees to perform a copy), as mutation of the source was a major source of bugs in codes using DataFrames.jl.

Now - the major use case for this method was in the past:

CSV.File(filename) |> DataFrame!

which is relevant for very large source files. However, with the changes in JuliaData/CSV.jl#687 this will not be needed.

Therefore, I think we can deprecate DataFrame!, but let us wait for some feedback about this proposal before making a PR.

[quinnj]

I think we should deprecate it; I don't think anyone really likes the bang operator here because it's just slightly confusing why it's there.