# For each row, return n lowest column indexes or column names

**URL:** <https://discourse.julialang.org/t/for-each-row-return-n-lowest-column-indexes-or-column-names/81932>\
**Category:** General Usage\
**Tags:** question, dataframes\
**Created:** [May 30, 2022, 9:01pm UTC](https://discourse.julialang.org/t/for-each-row-return-n-lowest-column-indexes-or-column-names/81932 "2022-05-30T21:01:13Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![Snowy](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/snowy/32/36765_2.png) [@Snowy](https://discourse.julialang.org/u/Snowy)\
**Post date:** [May 30, 2022, 9:01pm UTC](https://discourse.julialang.org/t/for-each-row-return-n-lowest-column-indexes-or-column-names/81932/1 "2022-05-30T21:01:13Z")

</div>

Hi all,  
New to Julia and seeking a little help. I am working with a dataframe and for each row, I want to return either the position of the lowest n values, or the column names of the lowest n value. The package DataStructures ([Heaps · DataStructures.jl](https://juliacollections.github.io/DataStructures.jl/latest/heaps/#Functions-using-heaps-1)) has a function that does this, but I really want to avoid loading a package just for this, as I’m sure the solution is relatively simple.

Say my dataframe was as follows:

```julia
df = DataFrame((name=["S1","S2","S3","S4","S5"],Z1=[41,12,19,17,11],Z2=[13,28,29,99,41],Z3=[7,86,23,12,71],Z4=[4,13,23,11,19],Z5=[41,12,13,19,22],Z6=[11,18,22,46,5]))

```

And if I were looking for the lowest 2 values, the result would be as follows:

```julia
df_sorted = DataFrame((name=["S1","S2","S3","S4","S5"],x1=["Z4","Z1","Z5","Z4","Z6"],x2=["Z3","Z5","Z1","Z3","Z1"]))

```

I thought maybe a combination of eachrow() and PartialQuickSort might do the trick, e.g.:

```julia
sort(eachrow(Matrix(df)),alg=PartialQuickSort(2))

```

But this results in an error.  
If anyone has any suggestions, I would very much appreciate it. Thanks for your help!

---

<div class="post-metadata">

**Author:** ![xinchin](https://avatars.discourse-cdn.com/v4/letter/x/54ee81/32.png) [@xinchin](https://discourse.julialang.org/u/xinchin)\
**Post date:** [May 31, 2022, 11:31pm UTC](https://discourse.julialang.org/t/for-each-row-return-n-lowest-column-indexes-or-column-names/81932/2 "2022-05-31T23:31:57Z")

</div>

You should use `InMemoryDatasets`; transpose each row and sort your data set (sort is very fast in that package)

---

<div class="post-metadata">

**Author:** ![pdeffebach](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/pdeffebach/32/10320_2.png) [@pdeffebach](https://discourse.julialang.org/u/pdeffebach)\
**Post date:** [May 31, 2022, 11:41pm UTC](https://discourse.julialang.org/t/for-each-row-return-n-lowest-column-indexes-or-column-names/81932/3 "2022-05-31T23:41:59Z")

</div>

Here’s a good solution. Its a bit of a complicated function I wrote, though.

```julia
julia> function sort_row_return_top_two(row)
           vec_pairs = collect(pairs(row))
           # Sort by the 2nd element in the pair (the value)
           # and not the name)
           sorted_vec_pairs = sort(vec_pairs, by = last)
           # Get the names of the top two
           string.(first.(sorted_vec_pairs[end-1:end]))
       end;

julia> select(df, "name", AsTable(r"^Z") => ByRow(sort_row_return_top_two) => ["first","second"])
5×3 DataFrame
 Row │ name first second 
     │ String String String 
─────┼────────────────────────
   1 │ S1 Z1 Z5
   2 │ S2 Z2 Z3
   3 │ S3 Z4 Z2
   4 │ S4 Z6 Z2
   5 │ S5 Z2 Z3

```