#! /usr/bin/perl
use strict;
use warnings;
use 5.010;

use Data::Dumper;

my @array = (
    [ 3, 7 ],
    [ 1, 2 ],
    [ 2, 4 ],
);


# Schwartzian Transform; read it from the end
my @sorted =
  # restore original data
  map { $_->[0] }
    # sort numerically by the calculated differemce
    sort { $a->[1] <=> $b->[1] }
      # create temp array ref; first element is the original data
      # second element is the square of the difference between second and first value of original dataset
      map { [ $_, ($_->[1] - $_->[0])**2 ] }
        # parse list of array duples
        @array;

say Dumper \@sorted;