Merge pull request #18 from nashbridges/performance-improvements

Avoid creating intermediate arrays when possible, it hurt performance

Gemfile

@@ -4,3 +4,7 @@ gemspec
 gem "rspec"
 gem "rake"
 gem "bump"
+
+group :benchmark do
+  gem "benchmark-ips"
+end

Gemfile.lock

@@ -7,6 +7,7 @@ PATH
 GEM
   remote: https://rubygems.org/
   specs:
+    benchmark-ips (2.5.0)
     bump (0.5.2)
     diff-lcs (1.2.5)
     rake (10.4.2)
@@ -29,6 +30,7 @@ PLATFORMS
   ruby
 
 DEPENDENCIES
+  benchmark-ips
   bump
   rake
   rspec

benchmarks/less_arrays.rb

@@ -0,0 +1,96 @@
+require "bundler/setup"
+require "benchmark/ips"
+
+require 'unicode_utils/compatibility_decomposition'
+require 'unicode_utils/general_category'
+
+module SortAlphabetical
+  extend self
+
+  def sort(set)
+    set.sort_by do |item|
+      if block_given?
+        item = yield(item).to_s
+      else
+        item = item.to_s
+      end
+      [normalize(item), item] # when both á and a are present, sort them a, á
+    end
+  end
+
+  def normalize(string)
+    UnicodeUtils.compatibility_decomposition(string).split(''.freeze).select do |c|
+      UnicodeUtils.general_category(c) =~ /Letter|Separator|Punctuation|Number/
+    end.join
+  end
+end
+
+module SortAlphabeticalEachChar
+  extend self
+
+  def sort(set)
+    set.sort_by do |item|
+      if block_given?
+        item = yield(item).to_s
+      else
+        item = item.to_s
+      end
+      [normalize(item), item] # when both á and a are present, sort them a, á
+    end
+  end
+
+  def normalize(string)
+    result = String.new
+    UnicodeUtils.compatibility_decomposition(string).each_char do |c|
+      result << c if UnicodeUtils.general_category(c) =~ /Letter|Separator|Punctuation|Number/
+    end
+    result
+  end
+end
+
+module SortAlphabeticalEachWithObject
+  extend self
+
+  def sort(set)
+    set.sort_by do |item|
+      if block_given?
+        item = yield(item).to_s
+      else
+        item = item.to_s
+      end
+      [normalize(item), item] # when both á and a are present, sort them a, á
+    end
+  end
+
+  def normalize(string)
+    UnicodeUtils.compatibility_decomposition(string).each_char.each_with_object(String.new) do |c, result|
+      result << c if UnicodeUtils.general_category(c) =~ /Letter|Separator|Punctuation|Number/
+    end
+  end
+end
+
+sample = %w[b a á ä o ó x ö í i c]
+
+Benchmark.ips do |x|
+  x.report("regular") { SortAlphabetical.sort(sample) }
+  x.report("each_char") { SortAlphabeticalEachChar.sort(sample) }
+  x.report("each_with_object") { SortAlphabeticalEachWithObject.sort(sample) }
+
+  x.compare!
+end
+
+=begin
+Warming up --------------------------------------
+             regular   662.000  i/100ms
+           each_char     1.247k i/100ms
+    each_with_object     1.070k i/100ms
+Calculating -------------------------------------
+             regular      6.816k (± 3.1%) i/s -     34.424k
+           each_char     12.762k (± 2.5%) i/s -     64.844k
+    each_with_object     10.842k (± 2.7%) i/s -     54.570k
+
+Comparison:
+           each_char:    12762.0 i/s
+    each_with_object:    10842.4 i/s - 1.18x slower
+             regular:     6816.2 i/s - 1.87x slower
+=end

lib/sort_alphabetical.rb

@@ -19,8 +19,12 @@ def sort(set)
   end
 
   def normalize(string)
-    UnicodeUtils.compatibility_decomposition(string).split(''.freeze).select do |c|
-      UnicodeUtils.general_category(c) =~ /Letter|Separator|Punctuation|Number/
-    end.join
+    result = String.new
+    UnicodeUtils.compatibility_decomposition(string).each_char do |c|
+      if UnicodeUtils.general_category(c) =~ /Letter|Separator|Punctuation|Number/
+        result << c
+      end
+    end
+    result
   end
 end