allow :col => AsTable and :col => cols (#2780)

NEWS.md

@@ -8,6 +8,9 @@
   but preserved in right; in `innerjoin` missings are dropped in both data frames;
   in `outerjoin` this value of keyword argument is not supported
   ([#2724](https://github.com/JuliaData/DataFrames.jl/pull/2724))
+* correctly handle selectors of the form `:col => AsTable` and `:col => cols`
+  by expanding a single column into multiple columns
+  ([#2780](https://github.com/JuliaData/DataFrames.jl/pull/2780))
 
 ## Bug fixes
 

docs/src/man/split_apply_combine.md

@@ -53,7 +53,7 @@ each subset of the `DataFrame`. This specification can be of the following forms
 3. a `cols => function => target_cols` form additionally explicitly specifying
    the target column or columns.
 4. a `col => target_cols` pair, which renames the column `col` to `target_cols`, which
-   must be single name (as a `Symbol` or a string).
+   must be single name (as a `Symbol` or a string), a vector of names or `AsTable`.
 5. a `nrow` or `nrow => target_cols` form which efficiently computes the number of rows
    in a group; without `target_cols` the new column is called `:nrow`, otherwise
    it must be single name (as a `Symbol` or a string).

src/abstractdataframe/selection.jl

@@ -47,7 +47,7 @@ const TRANSFORMATION_COMMON_RULES =
     3. a `cols => function => target_cols` form additionally explicitly specifying
        the target column or columns.
     4. a `col => target_cols` pair, which renames the column `col` to `target_cols`, which
-       must be single name (as a `Symbol` or a string).
+       must be single name (as a `Symbol` or a string), a vector of names or `AsTable`.
     5. a `nrow` or `nrow => target_cols` form which efficiently computes the number of rows
        in a group; without `target_cols` the new column is called `:nrow`, otherwise
        it must be single name (as a `Symbol` or a string).
@@ -212,7 +212,13 @@ end
 
 normalize_selection(idx::AbstractIndex, sel::Pair{<:ColumnIndex, <:AbstractString},
                     renamecols::Bool) =
-    normalize_selection(idx, first(sel) => Symbol(last(sel)), renamecols::Bool)
+    normalize_selection(idx, first(sel) => Symbol(last(sel)), renamecols)
+
+normalize_selection(idx::AbstractIndex, sel::Pair{<:ColumnIndex,
+                                                  <:Union{AbstractVector{Symbol},
+                                                          AbstractVector{<:AbstractString}}},
+                    renamecols::Bool) =
+    normalize_selection(idx, first(sel) => identity => last(sel), renamecols)
 
 function normalize_selection(idx::AbstractIndex,
                              @nospecialize(sel::Pair{<:ColumnIndex,
@@ -287,6 +293,11 @@ function normalize_selection(idx::AbstractIndex,
                              @nospecialize(sel::Pair{<:ColumnIndex, <:Base.Callable}), renamecols::Bool)
     c = idx[first(sel)]
     fun = last(sel)
+
+    if fun === AsTable
+        return normalize_selection(idx, first(sel) => identity => AsTable, renamecols)
+    end
+
     if renamecols
         newcol = Symbol(_names(idx)[c], "_", funname(fun))
     else
@@ -320,6 +331,9 @@ function normalize_selection(idx::AbstractIndex,
             end
     end
     fun = last(sel)
+
+    fun === AsTable && throw(ArgumentError("Passing AsTable in $sel is not supported"))
+
     if length(c) > 3
         prefix = join(@views(_names(idx)[c[1:2]]), '_')
         if renamecols

test/select.jl

@@ -1616,11 +1616,11 @@ end
     @test DataFrames.normalize_selection(DataFrames.Index(Dict(:a => 1, :b => 2), [:a, :b]),
                                          [:a] => sum, false) == (1 => (sum => :a))
 
-    @test DataFrames.normalize_selection(DataFrames.Index(Dict(:a => 1, :b => 2), [:a, :b]), 
+    @test DataFrames.normalize_selection(DataFrames.Index(Dict(:a => 1, :b => 2), [:a, :b]),
                                          [:a] => sum => [:new],  false) == (1 => (sum => [:new]))
 
     # Test that target col strings are converted to Symbols
-    @test DataFrames.normalize_selection(DataFrames.Index(Dict(:a => 1, :b => 2), [:a, :b]), 
+    @test DataFrames.normalize_selection(DataFrames.Index(Dict(:a => 1, :b => 2), [:a, :b]),
                                          [:a] => sum => ["new"],  false) == (1 => (sum => [:new]))
 end
 
@@ -1647,5 +1647,22 @@ end
         @test df.a === a
     end
 end
-
+
+@testset ":col => AsTable and :col => cols" begin
+    df = DataFrame(id=1:2, c1=[(a=1, b=2), (a=3, b=4)], c2=[(11, 12), (13, 14)])
+    @test select(df, :c1 => AsTable) == DataFrame(a=[1, 3], b=[2, 4])
+    @test select(df, :c1 => [:p, :q]) == DataFrame(p=[1, 3], q=[2, 4])
+    @test select(df, :c2 => AsTable) == DataFrame(x1=[11, 13], x2=[12, 14])
+    @test select(df, :c2 => [:p, :q]) == DataFrame(p=[11, 13], q=[12, 14])
+    @test_throws ArgumentError select(df, [:c1, :c2] => AsTable)
+    @test_throws ArgumentError select(df, [:c1, :c2] => AsTable)
+    gdf = groupby(df, :id)
+    @test select(gdf, :c1 => AsTable) == DataFrame(id=1:2, a=[1, 3], b=[2, 4])
+    @test select(gdf, :c1 => [:p, :q]) == DataFrame(id=1:2, p=[1, 3], q=[2, 4])
+    @test select(gdf, :c2 => AsTable) == DataFrame(id=1:2, x1=[11, 13], x2=[12, 14])
+    @test select(gdf, :c2 => [:p, :q]) == DataFrame(id=1:2, p=[11, 13], q=[12, 14])
+    @test_throws ArgumentError select(gdf, [:c1, :c2] => AsTable)
+    @test_throws ArgumentError select(gdf, [:c1, :c2] => AsTable)
+end
+
 end # module