liverpole,
I haven't read your code but in thinking about the problem, you should never need to scan the array. The following is completely untested but it explains why I believe it is unnecessary.
my @array = ((0) x 100);
my $inc_array = gen_min_tracker(0, 100, \@array);
my $floor = 27;
my $new_min = inc_array(42, 7); # increment the 43rd element by 7
if ($new_min >= $floor) {
print "The floor has been reached\n";
}
sub gen_min_tracker {
my ($curr_min, $curr_cnt, $aref) = @_;
my ($next_min, $next_cnt);
return sub {
my ($idx, $inc) = @_;
my $val = $aref->[$idx];
my $add = $val + $inc;
if ($val == $curr_min) {
$curr_cnt--;
if (! defined $next_min || $add <= $next_min) {
$next_min = $add;
$next_cnt++;
}
if (! $curr_cnt) {
($curr_val, $curr_cnt, $next_val, $next_cnt) = ($next_
+val, $next_cnt, undef, undef);
}
}
$aref->[$idx] += $inc; # let's not forget to actually increme
+nt
return $curr_min;
};
}
Update: Due to some unknown copy/paste error, the code above has been altered slightly from its original form to hopefully be correct despite still not being tested.
Warning: In the CB, ambrus points out that this approach is flawed. It is not safe to reset $next_val and $next_cnt to undef. The assumption was that anytime one of the $curr_min values was incremented, it would be incremented to the $next_min and that no other value in the array could possibly be smaller. The solution is simple - put in the O(N) low water mark algorithm keeping track of the min value and count.
Consideration: While you can't keep track of the minimum value without periodically scanning, you can still achieve the OP's goal as suaveant points out.
|