ruby const_get vs ::

benchmark.rb

require 'benchmark'

n = 1_000_000

class Parent
  HANDLER_METHOD = :awesome_method
end

class Children < Parent
  HANDLER_METHOD = :not_very_awesome_method
end

ch = Children.new

Benchmark.bmbm 20 do |results|
  results.report 'cycle' do
    n.times do
    end
  end

  results.report 'class.const_get' do
    n.times do
      ch.class.const_get(:HANDLER_METHOD)
    end
  end

  results.report 'class::' do
    n.times do
      ch.class::HANDLER_METHOD
    end
  end
end

$ uname -a && ruby benchmark.rb 
Linux Inspiron-17R 3.13.0-43-generic #72-Ubuntu SMP Mon Dec 8 19:35:44 UTC 2014 i686 i686 i686 GNU/Linux
Rehearsal --------------------------------------------------------
cycle                  0.050000   0.000000   0.050000 (  0.049341)
class.const_get        0.450000   0.000000   0.450000 (  0.448197)
class::                0.100000   0.000000   0.100000 (  0.098225)
----------------------------------------------- total: 0.600000sec

                           user     system      total        real
cycle                  0.050000   0.000000   0.050000 (  0.046460)
class.const_get        0.440000   0.000000   0.440000 (  0.447367)
class::                0.100000   0.000000   0.100000 (  0.093392)