Variant of Prime Perms, checking mersennes from 1 to 1 million, using block extraction for time comparisons. Surprising results were found when t3 dominated the field.

prime_perms2.rb

# From original: https://gist.github.com/baweaver/11163861
# Using Pipeable for clearer code: https://github.com/baweaver/pipeable
# and Benchpress to measure: https://github.com/baweaver/benchpress

# Make Object Pipeable
class Object; include Pipeable end

# Patch numbers to have a prime? method
class Fixnum
  def prime?
    ([1,2].include?(self) || !(2..Math.sqrt(self).ceil).find { |x| self % x == 0 })
  end
end

# The original second fastest solution, taking ~18 seconds
def t1
  (1..1_000_000).select { |n| 
    n == 2 || 
    n.to_s.chars.pipe { |c| 
      (c & %w(0 2 4 5 6 8)).count == 0 && 
      c.permutation
        .flat_map(&:join)
        .map(&:to_i)
        .all?(&:prime?)
    }
  }.count
end

# Let's take the select block out of there into a lambda. About 17 seconds, not as large as I had wanted
def t2
  prime_perms = -> n {
    n == 2 || 
    n.to_s.chars.pipe { |c| 
      (c & %w(0 2 4 5 6 8)).count == 0 && 
      c.permutation
        .flat_map(&:join)
        .map(&:to_i)
        .all?(&:prime?)
    }
  }

  (1..1_000_000).select(&prime_perms).count
end

# How about we take the pipe out of there as well? About 15 seconds. Interesting...
def t3
  c_pipe = -> c { 
    (c & %w(0 2 4 5 6 8)).count == 0 && 
    c.permutation
      .flat_map(&:join)
      .map(&:to_i)
      .all?(&:prime?)
  }

  prime_perms = -> n { n == 2 || n.to_s.chars.pipe(&c_pipe) }

  (1..1_000_000).select(&prime_perms).count
end

# Now let's chart it, get a bit more data here:
Benchpress::Chart.new(
  entities: { t1: -> { t1 }, t2: -> { t2 }, t3: -> { t3 } },
  min: 1, max: 5, step: 1, cycle: 5
).render

# Wait wait, what!?
# http://i59.tinypic.com/whjuvq.png
# t3 slaughtered! Interesting.

very interesting that t3 has such a big performance gain.