# Combining a col from each DF group into a single DF

**URL:** <https://discourse.julialang.org/t/combining-a-col-from-each-df-group-into-a-single-df/86311>\
**Category:** New to Julia\
**Tags:** question, dataframes\
**Created:** [August 25, 2022, 8:19am UTC](https://discourse.julialang.org/t/combining-a-col-from-each-df-group-into-a-single-df/86311 "2022-08-25T08:19:36Z")\
**Posts on this page:** 6\
**Page:** 1

<div class="post-metadata">

**Author:** ![Bob\_Jones](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/bob_jones/32/37809_2.png) [@Bob\_Jones](https://discourse.julialang.org/u/Bob_Jones)\
**Post date:** [August 25, 2022, 8:19am UTC](https://discourse.julialang.org/t/combining-a-col-from-each-df-group-into-a-single-df/86311/1 "2022-08-25T08:19:36Z")

</div>

Hi all,

sorry if this obvious, fairly new to Julia. I think there must be a simple way to combine a single col from all DF groups into a single DF. I have this:

`gd = groupby(DataFrame(dtype=repeat(["AA","BB","CC"], inner=4), val=collect(1:12)),:dtype)`

and want this

```julia
      │ AA BB CC    
   ─┼─────────────────────
   1 │ 1 5 9
   2 │ 2 6 10
   3 │ 3 7 11
   4 │ 4 8 12

```

currently doing it like this, but I think it there should be a simpler way:

```julia
res = DataFrame()
for dt in keys(gd)
    res[!,dt[1]] = gd[(dtype=dt[1],)][!,:val]
end

```

Thanks!

---

<div class="post-metadata">

**Author:** ![Jollywatt](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/jollywatt/32/202198_2.png) [@Jollywatt](https://discourse.julialang.org/u/Jollywatt)\
**Post date:** [August 25, 2022, 9:32am UTC](https://discourse.julialang.org/t/combining-a-col-from-each-df-group-into-a-single-df/86311/2 "2022-08-25T09:32:32Z")

</div>

This looks like ‘stacking’ or ‘pivoting’.

```julia
julia> combine(gd, :dtype, :val, eachindex => :row)
12×3 DataFrame
 Row │ dtype val row
     │ String Int64 Int64
─────┼──────────────────────
   1 │ AA 1 1
   2 │ AA 2 2
   3 │ AA 3 3
   4 │ AA 4 4
   5 │ BB 5 1
   6 │ BB 6 2
   7 │ BB 7 3
   8 │ BB 8 4
   9 │ CC 9 1
  10 │ CC 10 2
  11 │ CC 11 3
  12 │ CC 12 4

julia> unstack(ans, :row, :dtype, :val)
4×4 DataFrame
 Row │ row AA BB CC
     │ Int64 Int64? Int64? Int64?
─────┼───────────────────────────────
   1 │ 1 1 5 9
   2 │ 2 2 6 10
   3 │ 3 3 7 11
   4 │ 4 4 8 12

```

The way unstacking works is you specify row labels, column labels and values, `unstack(df, :rows, :columns, :values)`, and you get a table with the values at the corresponding positions. In our case, we need to add the target rows from the position within each group (which is what `eachindex => :row` does), the target columns are from `:dtype`, and the values from `:val`.

> **Rows and columns are ordered by appearance**
>
> The column names in `:target_column` appear in the table in order of appearance, and similarly for `:target_rows`. I.e., `:target_rows` does not specify the row _index_ — in fact, we can use any value for the row label:
> 
> ```julia
> julia> using DataFramesMeta
> 
> julia> @chain(gd,
> @combine(:val, :row = 'v' .+ eachindex(:val)),
> unstack(:row, :dtype, :val)
> )
> 4×4 DataFrame
> Row │ row AA BB CC
> │ Char Int64? Int64? Int64?
> ─────┼──────────────────────────────
> 1 │ w 1 5 9
> 2 │ x 2 6 10
> 3 │ y 3 7 11
> 4 │ z 4 8 12
> 
> ```

---

<div class="post-metadata">

**Author:** ![rocco\_sprmnt21](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/rocco_sprmnt21/32/20127_2.png) [@rocco\_sprmnt21](https://discourse.julialang.org/u/rocco_sprmnt21)\
**Post date:** [August 25, 2022, 12:59pm UTC](https://discourse.julialang.org/t/combining-a-col-from-each-df-group-into-a-single-df/86311/3 "2022-08-25T12:59:37Z")

</div>

> [@Bob\_Jones](#):
>
> gd = groupby(DataFrame(dtype=repeat([“AA”,“BB”,“CC”], inner=4), val=collect(1:12)),:dtype)

```julia
df=DataFrame(dtype=repeat(["AA","BB","CC"], inner=4), val=collect(1:12))
flatten(unstack(df, :dtype, :val, valuestransform=identity), Cols(:))

```

if the ultimate goal is to apply an aggregation function to the values of the various groups, you can avoid going through groupby and / or flatten in the following way

```julia

unstack(df, :dtype, :val, valuestransform=sum)
unstack(df, :dtype, :val, valuestransform=x->sum(x)/length(x))

```

---

<div class="post-metadata">

**Author:** ![digital\_carver](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/digital_carver/32/33818_2.png) [@digital\_carver](https://discourse.julialang.org/u/digital_carver)\
**Post date:** [August 25, 2022, 1:32pm UTC](https://discourse.julialang.org/t/combining-a-col-from-each-df-group-into-a-single-df/86311/4 "2022-08-25T13:32:42Z")

</div>

`valuestransform` isn’t in the stable release of DataFrames yet though, right? I found it on the `latest/` docs pages but not in the `stable/` ones.

---

<div class="post-metadata">

**Author:** ![rocco\_sprmnt21](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/rocco_sprmnt21/32/20127_2.png) [@rocco\_sprmnt21](https://discourse.julialang.org/u/rocco_sprmnt21)\
**Post date:** [August 25, 2022, 1:37pm UTC](https://discourse.julialang.org/t/combining-a-col-from-each-df-group-into-a-single-df/86311/5 "2022-08-25T13:37:03Z")

</div>

I have this version installed.  
I’m not sure, but I think it’s as you say

```julia
Status `C:\Users\sprmn\.julia\v1.8\Project.toml`
  [a93c6f00] DataFrames v1.4.0 `https://github.com/JuliaData/DataFrames.jl.git#main`

```

---

<div class="post-metadata">

**Author:** ![bkamins](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/bkamins/32/208538_2.png) [@bkamins](https://discourse.julialang.org/u/bkamins)\
**Post date:** [August 25, 2022, 2:08pm UTC](https://discourse.julialang.org/t/combining-a-col-from-each-df-group-into-a-single-df/86311/6 "2022-08-25T14:08:57Z")

</div>

Yes - `valuestransform` is not released yet, but it will be released soon.