in reply to du -h, sorted

But then I realized it'd probably be better to use Perl for the sorting, and got this shorter result

Now that sounds like a challenge...

  1. The return is clearly superflous. Off with 7.
  2. [0-9] can be replaced by \d, saving a character.
  3. Hoist the exponentation into the hash values and use a bigger scaling factor (that may be written more concisely).
  4. Replace $n and $u with $1 and $2 thus doing away with their initialisation.
  5. Hang on, moving the exponentiation back into h() means that the hash declaration is shorter still.
  6. If one considers that 'du -h|' adds to the overall count, then we can replace the diamond operator with a backtick of du -h. Overall this saves us the | of the pipe.
  7. We can replace the fat commas with ordinary commas (and fortunately not receive warnings).
  8. In fact, that regexp is way too verbose. /^(...)(.)/ will work just as well, saving a couple of characters.
  9. update: I forgot to add the last step, changing shift to pop (since we only read one arg).

Which gives....

perl -e 'sub h{%h=(K,1,M,4,G,7);pop=~/^(...)(.)/&&$1**$h{$2}}print sort{h($b)<=>h($a)}`du -h`'

ok, ok, I promise I'll leave it alone now, there's just the idea of using an anonymous hash to shave off a couple more characters, and the parens can be dropped from h($a), (but not h($b) because of precedences):

perl -e 'sub h{pop=~/(...)(.)/&&$1**{K,1,M,4,G,7}->{$2}}print sort{h($b)<=>h$a}`du -h`'

... 90 88 characters.

• another intruder with the mooring in the heart of the Perl