Issorted for Dataframe rows

General Usage Performance
1

I have a dataframe with a few dozen columns.
For each row I want to ensure that a subset of the columns (let’s say the relevant columns are named x1 to x6) is sorted

I was wondering whether there is a faster approach than what I drafted below.

#Discourse MWE check if dataframe columns are sorted 
using DataFrames
using BenchmarkTools

#generate some data 
n=20_000_000;
#the relevant columns are integers
mat=trunc.(Int,90 .* rand(n,6));
#note : my actual data has additional columns (-> I need to work with a dataframe)
df=DataFrame(mat);

#sort columns
#my data is generally sorted. But I want to ensure this is the case (i.e. identify the dataframes where the data is not sorted).

#I do not care about the performance of this function
function mysort!(df)
    cols=propertynames(df)
    for i=1:size(df,1)
        df[i,:] .= sort([df[i,col] for col in cols])
    end
    return nothing
end

function myIsSorted(m::Array)
    @inbounds for i=1:size(m,1)
        if !(m[i,6]>=m[i,5]>=m[i,4]>=m[i,3]>=m[i,2]>=m[i,1])
            return false 
        end
    end
    return true 
end

function myIsSorted(df)
    m=convert(Array{Int64,2},df)
    return myIsSorted(m)
end

#sort data 
mysort!(df) #I do not care about the performance of this function

@btime myIsSorted($df) #0.37s, 915MB for 20m rows
2

How about just

cols_to_checked = [:x1, :x2, :x3, :x4, :x5, :x6]

all(col->issorted(df[!, col]), cols_to_checked)
3

Thanks. This is fast, but it is not what I want.
In my case I want every row df[i,cols_to_checked] to be sorted.

4

Can u give examples that illustrate Ur idea. Maybe I am not getting what u r doing

5

More precise language needed. Each row to be sorted meaning within each row x1 <=x2<=x3 etc?

Or row1<=row2<= row3 if u restrict row to those 6 columns

6

please consider the above code from my MWE from the first post. This should illustrate what I am trying to check. You should be able to run the MWE.