From: Boris Schmid <boris@...>
Date: 2008-05-29T03:28:58+09:00
Subject: subclass of hash, that times last read

Hi all. I'm rather unfamiliar with the subclass / superclass principle,
but I just tried to write one. The result is below here, and should be a
hash-like class, with the exception that it stores the last-accessed
time (the purpose of this being that I can more easily delete entries
from the hash that haven't been used in a while - I'm using a hash for
caching a costly function, but over time the hash grows big and slow,
with a lot of old entries in it.. I'm trying out different ways of
pruning the hash, and this seemed to be an interesting option).

The below code seems to work fine, but it would be a comfort to me to
know if I'm introducing several hairy bug-prone things if I add this to
my code, or that this is the relatively safe and correct way to do it.

class Myhash < Hash
   alias :old_write :[]=
   def []=(key,value)
      old_write(key,[value,Time.now])
   end

   alias :old_read :[]
   def [](key)
      value = old_read(key)
      old_write(key,[value[0],Time.now]) if value.respond_to?(:[])
      old_read(key)
   end
end

hash = Myhash.new
1000.times {hash[rand] = rand}
sleep 7
1000.times {hash[rand] = rand}
hash.delete_if {|k,v| Time.now - v[1] > 5}

Which should delete the 1000 entries from the hash that were created in
the first loop.
-- 
Posted via http://www.ruby-forum.com/.