make sure getindex on DataFrameRows does not alias passed selector (#…

…3192)

NEWS.md

@@ -1,3 +1,11 @@
+# DataFrames.jl v1.4.1 Patch Release Notes
+
+## Bug fixes
+
+* Make sure we always copy the indexing value when calling `getindex` on
+ `DataFrameRows` object
+ ([#3192](https://github.com/JuliaData/DataFrames.jl/issues/3192))
+
 # DataFrames.jl v1.4 Release Notes
 
 ## Julia compatibility change

Project.toml

@@ -1,6 +1,6 @@
 name = "DataFrames"
 uuid = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
-version = "1.4.0"
+version = "1.4.1"
 
 [deps]
 Compat = "34da2185-b29b-5c13-b0c7-acf172513d20"

src/abstractdataframe/iteration.jl

@@ -79,7 +79,7 @@ Base.size(itr::DataFrameRows) = (size(parent(itr), 1), )
 
 Base.@propagate_inbounds Base.getindex(itr::DataFrameRows, i::Int) = parent(itr)[i, :]
 Base.@propagate_inbounds Base.getindex(itr::DataFrameRows, idx) =
- eachrow(@view parent(itr)[idx, :])
+ eachrow(@view parent(itr)[idx isa AbstractVector && !(eltype(idx) <: Bool) ? copy(idx) : idx, :])
 
 # separate methods are needed due to dispatch ambiguity
 Base.getproperty(itr::DataFrameRows, col_ind::Symbol) =

test/iteration.jl

@@ -223,4 +223,30 @@ end
  @test er2 == er[1:2]
 end
 
+@testset "test unaliasing of index" begin
+ for idx in ([2, 3], [0x2, 0x3], 2:3, Not([1, 4]),
+ [false, true, true, false], big.([2, 3]), big.(2:3), :)
+ df = DataFrame(a=1:4)
+ er = eachrow(df)
+ er2 = er[idx]
+ len = length(er2)
+ @test len == (idx === Colon() ? 4 : 2)
+ p = parent(er2)
+ @test p isa SubDataFrame
+ @test parentindices(p)[1] == (1:4)[idx]
+ if !(idx isa UnitRange)
+ @test parentindices(p)[1] !== idx
+ end
+ if idx isa Vector
+ empty!(idx)
+ end
+ @test length(er2) == len
+
+ er3 = filter(x -> x.a <= 2, er)
+ @test length(er3) == 2
+ @test parent(er3) isa SubDataFrame
+ @test parentindices(parent(er3))[1] == 1:2
+ end
+end
+
 end # module