Dataframe map function drops empty results

General Usage
1

Which is the result i wanted, but not the result i expected.

The result I expect in my example below is a group with an empty dataframe for each of the odd values from column a.

p.s. documentation does not indicate which will happen.

using DataFrames

function test()
    m = 3
    n = 4
    df = DataFrame([repeat(collect(1:n), inner=m),
                    [rand(1:10) for _ in 1:n*m],
                    [rand(1:10) for _ in 1:n*m]], [:a, :b, :c])

    g0 = groupby(df, :a)
    println(g0)

    g2 = map(sdf->filter(r->r[:a] % 2 == 0, sdf), g0)

    println(g2)
    
end
test()

GroupedDataFrame with 4 groups based on key: a
Group 1 (3 rows): a = 1
│ Row │ a     │ b     │ c     │
│     │ Int64 │ Int64 │ Int64 │
├─────┼───────┼───────┼───────┤
│ 1   │ 1     │ 2     │ 2     │
│ 2   │ 1     │ 3     │ 6     │
│ 3   │ 1     │ 9     │ 7     │
Group 2 (3 rows): a = 2
│ Row │ a     │ b     │ c     │
│     │ Int64 │ Int64 │ Int64 │
├─────┼───────┼───────┼───────┤
│ 1   │ 2     │ 10    │ 9     │
│ 2   │ 2     │ 8     │ 8     │
│ 3   │ 2     │ 8     │ 2     │
Group 3 (3 rows): a = 3
│ Row │ a     │ b     │ c     │
│     │ Int64 │ Int64 │ Int64 │
├─────┼───────┼───────┼───────┤
│ 1   │ 3     │ 10    │ 4     │
│ 2   │ 3     │ 8     │ 7     │
│ 3   │ 3     │ 3     │ 3     │
Group 4 (3 rows): a = 4
│ Row │ a     │ b     │ c     │
│     │ Int64 │ Int64 │ Int64 │
├─────┼───────┼───────┼───────┤
│ 1   │ 4     │ 1     │ 7     │
│ 2   │ 4     │ 1     │ 4     │
│ 3   │ 4     │ 10    │ 2     │
GroupedDataFrame with 2 groups based on key: a
Group 1 (3 rows): a = 2
│ Row │ a     │ a_1   │ b     │ c     │
│     │ Int64 │ Int64 │ Int64 │ Int64 │
├─────┼───────┼───────┼───────┼───────┤
│ 1   │ 2     │ 2     │ 10    │ 9     │
│ 2   │ 2     │ 2     │ 8     │ 8     │
│ 3   │ 2     │ 2     │ 8     │ 2     │
Group 2 (3 rows): a = 4
│ Row │ a     │ a_1   │ b     │ c     │
│     │ Int64 │ Int64 │ Int64 │ Int64 │
├─────┼───────┼───────┼───────┼───────┤
│ 1   │ 4     │ 4     │ 1     │ 7     │
│ 2   │ 4     │ 4     │ 1     │ 4     │
│ 3   │ 4     │ 4     │ 10    │ 2     │
2

I have not designed it (@nalimilan probably can comment more), but looking at the source code this is intended.

The underlying reason is that GroupedDataFrame does not store values of grouping variables for each group but only the groups.

3

Yes, GroupedDataFrame never stores empty groups. We would need to change the internal design to allow that. I guess the docs could be more explicit about that.