Fix vcat in case no data frames are passed (#3081)

JuliaData · Jun 19, 2022 · d645276 · d645276
1 parent 0ce9b0f
commit d645276
Show file tree

Hide file tree

Showing 3 changed files with 51 additions and 7 deletions.
diff --git a/NEWS.md b/NEWS.md
@@ -57,6 +57,9 @@
 * Make sure we avoid aliasing when repeating the same column
   in `select[!]` and `transform[!]` on `GroupedDataFrame`
   ([#3070](https://github.com/JuliaData/DataFrames.jl/pull/3070))
+* Make `vcat` correctly handle `cols` keyword argument if only
+  data frames having no columns are passed
+  ([#3081](https://github.com/JuliaData/DataFrames.jl/pull/3081))
 
 ## Performance
 

diff --git a/src/abstractdataframe/abstractdataframe.jl b/src/abstractdataframe/abstractdataframe.jl
@@ -1945,8 +1945,11 @@ end
 function _vcat(dfs::AbstractVector{AbstractDataFrame};
                cols::Union{Symbol, AbstractVector{Symbol},
                            AbstractVector{<:AbstractString}}=:setequal)
-
-    isempty(dfs) && return DataFrame()
+    # note that empty DataFrame() objects are dropped from dfs before we call _vcat
+    if isempty(dfs)
+        cols isa Symbol && return DataFrame()
+        return DataFrame([col => Missing[] for col in cols])
+    end
     # Array of all headers
     allheaders = map(names, dfs)
     # Array of unique headers across all data frames

diff --git a/test/dataframe.jl b/test/dataframe.jl
@@ -7,6 +7,9 @@ using OffsetArrays: OffsetArray
 const ≅ = isequal
 const ≇ = !isequal
 
+isequal_coltyped(df1::AbstractDataFrame, df2::AbstractDataFrame) =
+    isequal(df1, df2) && typeof.(eachcol(df1)) == typeof.(eachcol(df2))
+
 # randomized test from https://github.com/JuliaData/DataFrames.jl/pull/1974
 @testset "randomized tests for rename!" begin
     n = Symbol.('a':'z')
@@ -2049,7 +2052,7 @@ end
                                     cols=:orderequal)
 end
 
-@testset "vcat with source" begin
+@testset "vcat with source and reduce(vcat, ...)" begin
     df1 = DataFrame(A=1:3, B=1:3)
     df2 = DataFrame(A=4:6, B=4:6)
     df3 = DataFrame(A=7:9, C=7:9)
@@ -2058,20 +2061,55 @@ end
     for col in [:source, "source"]
         @test vcat(df1, df2, df3, df4, cols=:union, source=col) ≅
               vcat(df1, df2, df3, df4, cols=:union, source=col => [1, 2, 3, 4]) ≅
+              reduce(vcat, [df1, df2, df3, df4], cols=:union, source=col) ≅
+              reduce(vcat, [df1, df2, df3, df4], cols=:union, source=col => [1, 2, 3, 4]) ≅
               DataFrame(A=1:9, B=[1:6; fill(missing, 3)],
                         C=[fill(missing, 6); 7:9],
                         source=[1, 1, 1, 2, 2, 2, 3, 3, 3])
         res = vcat(df1, df2, df3, df4, cols=:union, source=col => categorical(-4:-1))
-        @test res ≅ DataFrame(A=1:9, B=[1:6; fill(missing, 3)],
-                              C=[fill(missing, 6); 7:9],
-                              source=[-4, -4, -4, -3, -3, -3, -2, -2, -2])
-        @test res.source isa CategoricalVector
+        @test isequal_coltyped(res, DataFrame(A=1:9, B=[1:6; fill(missing, 3)],
+                                              C=[fill(missing, 6); 7:9],
+                                              source=categorical([-4, -4, -4, -3, -3, -3, -2, -2, -2])))
+
+        res = reduce(vcat, [df1, df2, df3, df4], cols=:union, source=col => categorical(-4:-1))
+        @test isequal_coltyped(res, DataFrame(A=1:9, B=[1:6; fill(missing, 3)],
+                                              C=[fill(missing, 6); 7:9],
+                                              source=categorical([-4, -4, -4, -3, -3, -3, -2, -2, -2])))
+
+        @test reduce(vcat, DataFrame[]) == DataFrame()
+        @test isequal_coltyped(reduce(vcat, DataFrame[], source=:src),
+                               DataFrame(src=Int[]))
+        @test isequal_coltyped(reduce(vcat, DataFrame[], cols=[:a, :b]),
+                               DataFrame(a=Missing[], b=Missing[]))
+        @test isequal_coltyped(reduce(vcat, DataFrame[], cols=[:a, :b], source=:src),
+                               DataFrame(a=Missing[], b=Missing[], src=Int[]))
     end
 
     @test_throws TypeError vcat(df1, df2, df3, df4, cols=:union, source=1)
     @test_throws TypeError vcat(df1, df2, df3, df4, cols=:union, source=:a => 1)
     @test_throws ArgumentError vcat(df1, df2, df3, df4, cols=:union, source=:C)
     @test_throws ArgumentError vcat(df1, df2, df3, df4, cols=:union, source=:a => [1])
+    @test_throws TypeError reduce(vcat, [df1, df2, df3, df4], cols=:union, source=1)
+    @test_throws TypeError reduce(vcat, [df1, df2, df3, df4], cols=:union, source=:a => 1)
+    @test_throws ArgumentError reduce(vcat, [df1, df2, df3, df4], cols=:union, source=:C)
+    @test_throws ArgumentError reduce(vcat, [df1, df2, df3, df4], cols=:union, source=:a => [1])
+
+    @test vcat(DataFrame(), DataFrame()) ==
+          reduce(vcat, [DataFrame(), DataFrame()]) ==
+          DataFrame()
+    @test isequal_coltyped(vcat(DataFrame(), DataFrame(), cols=[:a, :b]),
+                           DataFrame(a=Missing[], b=Missing[]))
+    @test isequal_coltyped(reduce(vcat, (DataFrame(), DataFrame()), cols=[:a, :b]),
+                           DataFrame(a=Missing[], b=Missing[]))
+    @test isequal_coltyped(vcat(DataFrame(a=1:2), DataFrame(), cols=[:a, :b]),
+                           DataFrame(a=1:2, b=missing))
+    @test isequal_coltyped(reduce(vcat, (DataFrame(a=1:2), DataFrame()), cols=[:a, :b]),
+                           DataFrame(a=1:2, b=missing))
+    @test vcat(DataFrame(a=1), DataFrame(b=2), cols=[:a]) ≅ DataFrame(a=[1, missing])
+    @test vcat(DataFrame(a=1), DataFrame(b=2), cols=[:b]) ≅ DataFrame(b=[missing, 2])
+    @test vcat(DataFrame(a=1), DataFrame(b=2), cols=Symbol[]) == DataFrame()
+    @test isequal_coltyped(vcat(DataFrame(a=1), DataFrame(b=2), cols=[:c]),
+                           DataFrame(c=[missing, missing]))
 end
 
 @testset "push! with :subset" begin