Merge pull request #1 from annikoff/tenth-task

Tenth task

Author: annikoff

Diff for: .gitignore

@@ -0,0 +1,6 @@
+/.idea/
+/.ruby-version
+/.ruby-gemset
+/tmp
+/bin
+/Gemfile.lock

Diff for: .travis.yml

@@ -0,0 +1,8 @@
+dist: trusty
+language: ruby
+rvm:
+  - 2.3.3
+  - 2.2.2
+before_script:
+  - rake compile
+script: rake test

Diff for: Gemfile

@@ -0,0 +1,9 @@
+source 'https://rubygems.org'
+
+gemspec
+
+group :development, :test do
+  gem 'facets'
+  gem 'rake-compiler'
+  gem 'jeweler'
+end

Diff for: README.rdoc

@@ -1,5 +1,7 @@
 = array_stats
 
+{<img src="https://travis-ci.org/annikoff/array_stats.svg?branch=experiment" alt="Build Status" />}[https://travis-ci.org/annikoff/array_stats]
+
 * Docs[http://benkoski.com/array_stats/]
 * Source[http://github.com/bkoski/array_stats]
 
@@ -38,6 +40,16 @@ Percentile calculations are based on the NIST formula[http://www.itl.nist.gov/di
 
 * <tt>sudo gem install bkoski-array_stats --source http://gemcutter.org</tt>
 
+== BENCHMARK
+
+A comparison of percentile and fast_percentile methods.
+
+```
+       user     system      total        real
+Ruby:  23.120000   0.080000  23.200000 ( 23.197993)
+Golang:  11.280000   0.040000  11.320000 ( 11.292624)
+```
+
 == LICENSE:
 
 (The MIT License)

Diff for: Rakefile

@@ -1,5 +1,11 @@
-require 'rubygems'
 require 'rake'
+require 'rake/extensiontask'
+require 'rubygems'
+
+spec = Gem::Specification.load('array_stats.gemspec')
+Rake::ExtensionTask.new('array_stats', spec) do |ext|
+  ext.lib_dir = 'bin'
+end
 
 begin
   require 'jeweler'
@@ -38,11 +44,9 @@ rescue LoadError
   end
 end
 
-task :test => :check_dependencies
-
-task :default => :test
+task :default => [:compile, :test]
 
-require 'rake/rdoctask'
+require 'rdoc/task'
 Rake::RDocTask.new do |rdoc|
   version = File.exist?('VERSION') ? File.read('VERSION') : ""
 

Diff for: array_stats.gemspec

@@ -6,7 +6,7 @@
 Gem::Specification.new do |s|
   s.name = %q{array_stats}
   s.version = "0.6.0"
-
+  s.platform = Gem::Platform::RUBY
   s.required_rubygems_version = Gem::Requirement.new(">= 0") if s.respond_to? :required_rubygems_version=
   s.authors = ["Ben Koski"]
   s.date = %q{2010-04-05}
@@ -26,7 +26,9 @@ Gem::Specification.new do |s|
      "lib/array_stats/float.rb",
      "test/test_array_stats.rb",
      "test/test_float_extensions.rb",
-     "test/test_helper.rb"
+     "test/test_helper.rb",
+     "ext/array_stats/extconf.rb",
+     "ext/array_stats/array_stats.go"
   ]
   s.homepage = %q{http://github.com/bkoski/array_stats}
   s.rdoc_options = ["--charset=UTF-8"]
@@ -38,6 +40,7 @@ Gem::Specification.new do |s|
      "test/test_float_extensions.rb",
      "test/test_helper.rb"
   ]
+  s.extensions = ['ext/array_stats/extconf.rb']
 
   if s.respond_to? :specification_version then
     current_version = Gem::Specification::CURRENT_SPECIFICATION_VERSION
@@ -51,4 +54,5 @@ Gem::Specification.new do |s|
   else
     s.add_dependency(%q<thoughtbot-shoulda>, [">= 0"])
   end
+  s.add_dependency 'ffi'
 end

Diff for: benchmark

@@ -0,0 +1,15 @@
+#!/usr/bin/env ruby
+require 'benchmark'
+require_relative 'lib/array_stats'
+
+n = 100
+array = (1..1000000).map { rand(10) * 10 }
+Benchmark.bm do |x|
+  x.report('Ruby: ') do
+    n.times { array.percentile(65) }
+  end
+
+  x.report('Golang: ') do
+    n.times { array.fast_percentile(65) }
+  end
+end

Diff for: ext/array_stats/array_stats.go

@@ -0,0 +1,32 @@
+package main
+
+import (
+	"C"
+	"unsafe"
+	"sort"
+	"math"
+)
+
+//export fast_percentile
+func fast_percentile(array unsafe.Pointer, size int, percent float64) float64 {
+	if size == 0 {
+		return 0
+	}
+	elements := (*[1<<30]float64)(array)[:size]
+	sort.Float64s(elements)
+
+	rank := (percent / 100) * (float64(size) + 1)
+	_, rank_frac := math.Modf(rank)
+	rank_frac = math.Abs(rank_frac)
+	rank_truncated := int(math.Trunc(rank))
+
+	if rank_frac > 0 {
+		sample_0 := elements[rank_truncated - 1]
+		sample_1 := elements[rank_truncated]
+		return (rank_frac * (sample_1 - sample_0)) + sample_0
+	}else {
+		return elements[int(rank) - 1]
+	}
+}
+
+func main() {}

Diff for: ext/array_stats/extconf.rb

@@ -0,0 +1,4 @@
+require 'mkmf'
+find_executable('go')
+create_makefile('array_stats/array_stats')
+exec 'go build -buildmode=c-shared -o array_stats.so ../../../../ext/array_stats/array_stats.go'

Diff for: lib/array_stats/array_stats.rb

@@ -1,5 +1,10 @@
+require 'ffi'
+
 module ArrayStats
-  
+  extend FFI::Library
+  ffi_lib File.expand_path('../../../bin/array_stats.so', __FILE__)
+  attach_function :fast_percentile, [:pointer, :int, :double], :double
+
   # Returns the sum of all elements in the array; 0 if array is empty
   def total_sum
     self.inject(0) {|sum, sample| sum += sample}
@@ -18,19 +23,23 @@ def mean
   def median
     percentile(50)
   end
+
+  # Returns the median for the array; nil if array is empty
+  def fast_median
+    fast_percentile(50)
+  end
 
   # Returns the percentile value for percentile _p_; nil if array is empty.
   #
   # _p_ should be expressed as an integer; <tt>percentile(90)</tt> returns the 90th percentile of the array.
   #
   # Algorithm from NIST[http://www.itl.nist.gov/div898/handbook/prc/section2/prc252.htm]
   def percentile p
+    return nil if length == 0
     sorted_array = self.sort
     rank = (p.to_f / 100) * (self.length + 1)
 
-    if self.length == 0
-      return nil
-    elsif rank.fractional_part?
+    if rank.fractional_part?
       sample_0 = sorted_array[rank.truncate - 1]
       sample_1 = sorted_array[rank.truncate]
 
@@ -39,5 +48,11 @@ def percentile p
       return sorted_array[rank - 1]
     end    
   end
-  
-end
+
+  def fast_percentile p
+    return nil if size == 0
+    pointer = FFI::MemoryPointer.new :double, size
+    pointer.put_array_of_double 0, self
+    ArrayStats.fast_percentile pointer, size, p
+  end
+end

Diff for: test/test_array_stats.rb

@@ -1,71 +1,73 @@
 require File.dirname(__FILE__) + '/test_helper.rb'
 
 class TestArrayStats < Test::Unit::TestCase
-
   context "median" do
     should "return the middle of the set if array length is odd" do
       assert_equal 15, [1,2,15,22,38].median
+      assert_equal 15, [1,2,15,22,38].fast_median
     end
-    
+
     should "return the average of the middle of the set if array length is even" do
       assert_equal 10, [1,6,14,22].median
+      assert_equal 10, [1,6,14,22].fast_median
     end
-    
+
     should "return nil if the array is empty" do
       assert_nil [].median
+      assert_nil [].fast_median
     end
-    
+
     should "sort an array before deriving the median" do
       assert_equal 20, [1,20,50,60,10].median
+      assert_equal 20.0, [1,20,50,60,10].fast_median
     end
   end
-  
-  context "percentile" do 
+
+  context "percentile" do
     should "choose the number at a particular rank if array divides cleanly" do
       assert_equal 36, [12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48].percentile(65)
+      assert_equal 36, [12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48].fast_percentile(65)
     end
-    
+
     should "interpolate according to algorithm if array does not divide cleanly" do
       assert_equal 5.5, [3,5,7,8,9,11,13,15].percentile(25)
+      assert_equal 5.5, [3,5,7,8,9,11,13,15].fast_percentile(25)
       assert_equal 95.1981, [95.1772,95.1567,95.1937,95.1959,95.1442,95.0610,95.1591,95.1195,95.1065,95.0925,95.1990,95.1682].percentile(90).round_to(0.0001)
+      assert_equal 95.1981, [95.1772,95.1567,95.1937,95.1959,95.1442,95.0610,95.1591,95.1195,95.1065,95.0925,95.1990,95.1682].fast_percentile(90).round_to(0.0001)
     end
-    
+
     should "return nil if the array is empty" do
       assert_nil [].percentile(30)
+      assert_nil [].fast_percentile(30)
     end
   end
-  
+
   context "total_sum" do
     should "return the sum of all elements in the array" do
       assert_equal 12, [2,4,6].total_sum
     end
-    
+
     should "return an integer if all elements are ints" do
       assert_kind_of Integer, [2,4,6].total_sum
     end
-    
+
     should "return a float if at least some of the elements are floats" do
       assert_kind_of Float, [2.5,3.5,6].total_sum
     end
-    
+
     should "return 0 if the array is empty" do
       assert_equal 0, [].total_sum
     end
   end
-  
+
   context "mean" do
     should "return the mean for the array " do
       assert_equal 7, [2,4,6,8,10,12].mean
       assert_equal 25, [48,29,26,19,3].mean
     end
-    
+
     should "return nil if the array is empty" do
       assert_nil [].mean
     end
   end
-  
-  
-  
 end
-
-

Diff for: test/test_float_extensions.rb

@@ -7,24 +7,24 @@ class TestFloatExtensions < Test::Unit::TestCase
       n = 1.249
       assert n.fractional_part?
     end
-    
+
     should "return true for a number without a fractional part" do
        n = 1.0
        assert !n.fractional_part?
      end
   end
-  
+
   context "fractional_part" do
     should "return the decimal parts for a number" do
       n = 12.2456
       assert_in_delta 0.2456, n.fractional_part, 0.00000001
     end
-    
+
     should "retrun the decimal parts, even if they are 0" do
       n = 12.0
       assert_equal 0.0, n.fractional_part
     end
-    
+
     should "retrun the decimal as a positive number, even if the original float is negative" do
       n = -12.2399
       assert_in_delta 0.2399, n.fractional_part, 0.00000001

Diff for: test/test_helper.rb

@@ -2,4 +2,4 @@
 require 'test/unit'
 require File.dirname(__FILE__) + '/../lib/array_stats'
 require 'shoulda'
-require 'facets'
+require 'facets'