Shell – Sorting more than one column

sort -k1,1 -nk2 is the same as sort -k1,1 -n -k2, same as sort -n -k1,1 -k2, as in the numerical sorting is turned on globally, for all the keys.

To sort the 2^nd key only numerically, you need to add n to that sort key description as in:

sort -k1,1 -k2n

Or:

sort -k1,1 -k2,2n

With n and with the default field separator 2 is the same as 2,2 though. 2 would be the part of the line starting from the second field, but when interpreted as a number, that's the same as the second field alone (2,2).

Here, you could also sort numerically on the number that is after chr and then alphabetically on the rest of the first field and then numerically on the second field with:

sort -k1.4n -k1,1 -k2n

You can do it in a single GNU awk:

gawk -F ',' '
    {
        for(i=1;i<=NF;i++){matrix[i][NR]=$i}
    }
    END{
        for(i=1;i<=NF;i++){asort(matrix[i])}
        for(j=1;j<=NR;j++){
            for(i=1;i<NF;i++){
                printf "%s,",matrix[i][j]
            }
            print matrix[i][j]
        }
    }
' file

• for(i=1;i<=NF;i++){matrix[i][NR]=$i}

Multidimensional array (GNU extension) matrix gets populated, so that matrix[i][j] contains the number of column i, row j.

• for(i=1;i<=NF;i++){asort(matrix[i])}

Sorts each column (GNU extension).

• Finally

  for(j=1;j<=NR;j++){
      for(i=1;i<NF;i++){
          printf "%s,",matrix[i][j]
      }
      print matrix[i][j]
  }
  

Prints a sequence of a[1],, a[2],, ..., a[NF-1],, a[NF]\n for each line.