use strict;
use warnings;

use File::Find;
use List::Util qw( reduce );

# these partition functions treat the input data as having
# already been partitioned; they partition the data further.

sub partition_by_calculated_key(&@)
{
	my $cr = shift;
	map
	{
		my %h;
		push @{ $h{ &$cr } }, $_ for @$_;
		values %h
	}
	@_
}

sub partition_by_comparison_function(&@)
{
	my $cr = shift;
	map
	{
		my %h;
		my $added;
		for ( @$_ )
		{
			for my $seen ( keys %h )
			{
				local( $a, $b ) = ( $_, $seen );
				if ( $cr->( $_, $seen ) ) # returns true if "equal"
				{
					push @{ $h{$seen} }, $_;
					$added = 1;
					last;
				}
			}
			$added or push @{ $h{$_} }, $_;
		}
		values %h
	}
	@_
}

#####################################

my @root = @ARGV;
@root or @root = ('.');

my @files;
find( sub { push @files, $File::Find::name }, $_ ) for @root;

# eliminate directory names:
@files = grep { ! -d $_ } @files;
# eliminate existing symlinks:
@files = grep { ! -l $_ } @files;

my @a = ( \@files );

# first do the cheap one:
@a = partition_by_calculated_key { -s $_ } @a;
@a = grep { @$_ > 1 } @a; # filter out singleton lists

# then the expensive one:
@a = partition_by_calculated_key { qx( sum "$_" ) } @a;
@a = grep { @$_ > 1 } @a; # filter out singleton lists

# now group those which are "equal":
@a = partition_by_comparison_function { ! system qq( cmp -s "$a" "$b" ) } @a;
@a = grep { @$_ > 1 } @a; # filter out singleton lists

# finally, hardlink the files in each group:
reduce { system qq( ln "$a" "$b" ); $b } @$_ for @a;