From 0557c0e5dc140e7bbf48e3b2ca8c8fdd8bfda031 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bogumi=C5=82=20Kami=C5=84ski?= <bkamins@sgh.waw.pl>
Date: Thu, 19 Apr 2018 10:11:58 +0200
Subject: [PATCH] Always collect AbstractRange (#1392)

---
 src/dataframe/dataframe.jl | 16 +++++++++++-----
 test/dataframe.jl          |  7 +++++++
 2 files changed, 18 insertions(+), 5 deletions(-)

diff --git a/src/dataframe/dataframe.jl b/src/dataframe/dataframe.jl
index 454e82ff54..47c277e572 100644
--- a/src/dataframe/dataframe.jl
+++ b/src/dataframe/dataframe.jl
@@ -40,12 +40,17 @@ Each column in `columns` should be the same length.
 **Notes**
 
 A `DataFrame` is a lightweight object. As long as columns are not
-manipulated, creation of a DataFrame from existing AbstractVectors is
+manipulated, creation of a `DataFrame` from existing AbstractVectors is
 inexpensive. For example, indexing on columns is inexpensive, but
 indexing by rows is expensive because copies are made of each column.
 
-Because column types can vary, a DataFrame is not type stable. For
-performance-critical code, do not index into a DataFrame inside of
+If a column is passed to a `DataFrame` constructor or is assigned as a whole
+using `setindex!` then its reference is stored in the `DataFrame`. An exception
+to this rule is assignment of an `AbstractRange` as a column, in which case the
+range is collected to a `Vector`.
+
+Because column types can vary, a `DataFrame` is not type stable. For
+performance-critical code, do not index into a `DataFrame` inside of
 loops.
 
 **Examples**
@@ -325,12 +330,13 @@ end
 
 # Will automatically add a new column if needed
 function insert_single_column!(df::DataFrame,
-                               dv::AbstractVector,
+                               v::AbstractVector,
                                col_ind::ColumnIndex)
 
-    if ncol(df) != 0 && nrow(df) != length(dv)
+    if ncol(df) != 0 && nrow(df) != length(v)
         throw(ArgumentError("New columns must have the same length as old columns"))
     end
+    dv = isa(v, AbstractRange) ? collect(v) : v
     if haskey(index(df), col_ind)
         j = index(df)[col_ind]
         df.columns[j] = dv
diff --git a/test/dataframe.jl b/test/dataframe.jl
index 9e2b9e2f43..58483c656c 100644
--- a/test/dataframe.jl
+++ b/test/dataframe.jl
@@ -635,6 +635,13 @@ module TestDataFrame
         @test names(df) == [:x3, :x3_1, :x3_2, :x4]
     end
 
+    @testset "passing range to a DataFrame" begin
+        df = DataFrame(a=1:3, b='a':'c')
+        df[:c] = 1:3
+        df[:d] = 'a':'c'
+        @test all(typeof(df[i]) <: Vector for i in 1:ncol(df))
+    end
+
     @testset "handling of end in indexing" begin
         z = DataFrame(rand(4,5))
         for x in [z, view(z, 1:4)]