Reply to thread

Message: [QUOTE="sln, post: 4863101"] This is not a very clever idea. First, using numeric context in the hash. Given the first, second treating it as a string. Given the first and second, third is to treat it as a 10 digit string. Example: two => 2e10 in the hash becomes two => 20000000000 , length = 11 when it hits the length() function So as a general rule, sorting by length in numeric context is not a good idea. The fourth is that its not portable. There is two parts to this, one is forcing non-existing elements to the bottom, the other to sort the existing ones and bring to the top. These two parts have to be kept separate for portablity. for (sort { (exists $h{$b} && $h{$b}) <=> (exists $h{$a} && $h{$a}) } @ks) { print "$_\n" } ^^^^ This wont work if you decide to sort by value instead of length. The most portable approach is a scheme like this: length: !($xb & $xa) ? $xb cmp $xa : length $h{$a} <=> length $h{$b} !($xb & $xa) ? $xb cmp $xa : length $h{$b} <=> length $h{$a} value: !($xb & $xa) ? $xb cmp $xa : $h{$a} <=> $h{$b} !($xb & $xa) ? $xb cmp $xa : $h{$b} <=> $h{$a} or more complex: !($xb & $xa) ? $xb cmp $xa : func($h{$a},$h{$b}) Any scheme will work as long as there is true separation between parts. -sln [/QUOTE]