# Index of max values of N columns in a DataFrame without manually entering each column

**URL:** <https://discourse.julialang.org/t/index-of-max-values-of-n-columns-in-a-dataframe-without-manually-entering-each-column/92532>\
**Category:** New to Julia\
**Tags:** question, indexing, dataframes, maxima\
**Created:** [January 5, 2023, 12:50am UTC](https://discourse.julialang.org/t/index-of-max-values-of-n-columns-in-a-dataframe-without-manually-entering-each-column/92532 "2023-01-05T00:50:01Z")\
**Posts on this page:** 1\
**Showing post:** 11

<div class="post-metadata">

**Author:** ![phantom](https://avatars.discourse-cdn.com/v4/letter/p/e0b2c6/32.png) [@phantom](https://discourse.julialang.org/u/phantom)\
**Post date:** [January 6, 2023, 6:51am UTC](https://discourse.julialang.org/t/index-of-max-values-of-n-columns-in-a-dataframe-without-manually-entering-each-column/92532/11 "2023-01-06T06:51:02Z")

</div>

Thank you all again for being so generous with your insight and knowledge. It is incredibly helpful and much appreciated. I marked @Dan 's original answer due to the simplicity in the code and as pointed out by bkamins and nilshg the delta in competition time doesn’t necessarily carry over into larger tables. However, I might be wrong, but is it the case that the DataFramesMeta solution

> [@Dan](#):
>
> ```julia
> julia> using DataFramesMeta
> 
> julia> @orderby df -maximum(AsTable(r"id"))
> 
> ```

should be modified to

```julia
 
@orderby(df7, -max.(AsTable(r"id")...))

```

I am likely missing something here, but in the following example each of these solutions generate the same `DataFrame`

```julia
df7 = DataFrame(Aid = rand(-100:100, 10000000), Bid = rand(-100:100, 10000000), C = rand(-100:100,10000000), Did = rand(-100:100,10000000))

```

```julia
mxbyrow = df7[sortperm(select(df7, r"id" => ByRow(max) => :maxid); rev=true),:]

mxclmn = df7[sortperm(max.(df7.Aid, df7.Bid, df7.Did),rev = true), :]

mxeachclmn = df7[sortperm(max.(eachcol(df7[!, r"id"])...),rev = true), :]

mxfltr = df7[sortperm(max.([df7[!,c] for c in filter(n-> endswith(n,"id"), names(df7))]...), rev=true),:]

using DataFrameMacros

mxmacro = @sort(df7, -maximum({{r"id"}}))

mxfltr == mxclmn == mxeachclmn == mxmacro == mxbyrow
true

```

However with the `DataFramesMeta` approach

```julia
using DataFrames
mxmeta = @orderby df7 -maximum(AsTable(r"id"))

mxclmn == mxmeta
false

```

but if I use the syntax from @bkamins previous post here [Sort DataFrame by the greater of multiple columns](https://discourse.julialang.org/t/sort-dataframe-by-the-greater-of-multiple-columns/92481)

```julia
mxmeta2 = @orderby(df7, -max.(AsTable(r"id")...))

```

then

```julia
mxclmn == mxmeta2
true

```

so

```julia
@orderby(df7, -max.(AsTable(r"id")...)) == @orderby df7 -maximum(AsTable(r"id"))
false

```

Is this correct? I am probably missing something but I can’t quite figure out how `@orderby df7 -maximum(AsTable(r"id"))` is ranking the rows.

Just as an aside in terms of speed, if the rows are ranked in ascending rather than descending order, it seems that the non macro approaches are still a bit faster.

```julia
 @benchmark @orderby($df7, max.(AsTable(r"id")...))
BenchmarkTools.Trial: 5 samples with 1 evaluation.
 Range (min … max): 967.029 ms … 1.118 s ┊ GC (min … max): 0.00% … 13.52%
 Time (median): 1.105 s ┊ GC (median): 12.17%
 Time (mean ± σ): 1.078 s ± 62.659 ms ┊ GC (mean ± σ): 10.06% ± 5.54%

  █ █ ██ █  
  █▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█▁▁▁▁██▁▁▁█ ▁
  967 ms Histogram: frequency by time 1.12 s <

 Memory estimate: 2.24 GiB, allocs estimate: 2957653.

julia> @benchmark $df7[sortperm(max.(eachcol($df7[!, r"id"])...)), :]
BenchmarkTools.Trial: 25 samples with 1 evaluation.
 Range (min … max): 193.540 ms … 233.784 ms ┊ GC (min … max): 0.00% … 16.63%
 Time (median): 196.598 ms ┊ GC (median): 0.00%
 Time (mean ± σ): 200.730 ms ± 10.667 ms ┊ GC (mean ± σ): 1.98% ± 4.84%

  ▁▄▁█▁ ▁▁                                                    
  █████▆▆▁▆██▆▆▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▆▁▁▁▁▁▁▁▁▁▁▁▁▁▁▆▁▁▁▁▁▆ ▁
  194 ms Histogram: frequency by time 234 ms <

 Memory estimate: 457.77 MiB, allocs estimate: 59.

```

---

_[View the full topic](https://discourse.julialang.org/t/index-of-max-values-of-n-columns-in-a-dataframe-without-manually-entering-each-column/92532)._