Surprising setindex/setproperty behaviour

[nickrobinson251]

Here's a surprising result: df.col = 1 and df[:col] = 1 broadcast i.e. act like .= 1 - which is inconsistent with df.col returning the underlying vector.

Example:

julia> df = DataFrame(:a => 1:10, :z => 1:10);

julia> df.a
10-element Array{Int64,1}:
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10

julia> b = collect(1:10);

julia> df.a == b
true

julia> df.a = 1;

julia> df.a
10-element Array{Int64,1}:
 1
 1
 1
 1
 1
 1
 1
 1
 1
 1

julia> b = 1;

julia> b
1

julia> df.a == b
false

[bkamins]

This will be fixed soon. See #1806 and #1804 for first steps towards that. The design is not simple though.

[bkamins]

See the new target rules here: https://github.com/JuliaData/DataFrames.jl/blob/master/docs/src/lib/indexing.md

They should be consistent and work as expected (we need another release cycle to get there).

I am closing this PR, but feel free to reopen if you have more comments.