Make in_use? 200x faster

[aasmith]

It appears that the current implementation of BenString.in_use? is slow, per the following benchmark results:

[andy@hermit ~/ben_string]$ ruby -Ilib perf.rb 
Rehearsal --------------------------------------------------------
original each_object   0.453125  44.117188  44.570312 ( 45.130141)
proposed each_object   0.000000   0.109375   0.109375 (  0.112148)
---------------------------------------------- total: 44.679688sec

                           user     system      total        real
original each_object   0.078125  20.171875  20.250000 ( 20.553962)
proposed each_object   0.000000   0.101562   0.101562 (  0.102011)

The code for the benchmark is:

require 'benchmark'
require 'ben_string'

# Instantiate some objects.
100.times { [String, Array, Hash, BenString].each { |c| c.new } }

N = 1000

Benchmark.bmbm do |bm|
  bm.report 'original each_object' do
    N.times do
      counts = Hash.new(0)
      ObjectSpace.each_object { |o| counts[o.class] += 1 }
    end
  end

  bm.report 'proposed each_object' do
    N.times do
      ObjectSpace.each_object(BenString).any? { |o| BenString === o }
    end
  end
end

Proposed patch is in this pull request.

[mculp]

whoooosh

[zerowidth]

+1, good work man!

[bleything]

I'm not sure what you're getting at, mculp :)

[dkubb]

Wouldn't ObjectSpace.each_object(self).any? be even faster than this? ;)