peterc-rsmaz 0.0.2 → 0.0.3

data/History.txt

@@ -1,3 +1,10 @@
+== 0.0.3 2009-04-02
+
+* Now produces same output as reference smaz implementation
+* Better compression
+* Performance tweaks
+* Initial steps of Rubyizing the algorithm
+
 == 0.0.2 2009-04-02
 
 * Ruby 1.9 support added

data/README.rdoc

@@ -12,7 +12,8 @@
   http://github.com/antirez/smaz/tree/master
   
   I've done some initial cleanup of a pure Ruby->C port, but this
-  is not yet complete. It does pass the specs, however!
+  is not yet complete. It does pass the specs, however! Feel free
+  to clean it up as it's a bit memory inefficient right now... :)
 
 == REQUIREMENTS:
 
@@ -20,13 +21,13 @@
 * Some strings to compress
 * A sense of humor
 
-== INSTALL:
+== USAGE:
 
-* require 'rsmaz'
+  require 'rsmaz'
   r = RSmaz.compress("whatever")
   puts RSmaz.decompress(r)
 
-== LICENSE:
+== RSMAZ LICENSE:
 
 Copyright (c) 2009 Peter Cooper, Salvatore Sanfilippo
 

data/lib/rsmaz.rb

@@ -4,11 +4,15 @@ $:.unshift(File.dirname(__FILE__)) unless
 require 'strscan'
 
 # Silly hack to allow usage of String#ord in Ruby 1.9 without breaking Ruby 1.8
-class Fixnum; def ord; self; end; end
+if RUBY_VERSION < '1.9.0'
+  class String
+    def ord; self[0]; end;
+  end
+end  
 
 # RSmaz is too small to bother splitting into separate files, so I'll be lazy..
 module RSmaz
-  VERSION = '0.0.2'
+  VERSION = '0.0.3'
   
   # From http://github.com/antirez/smaz/blob/4b913924e15b7663ee0240af19cedfd266052aab/smaz.c
   CODEBOOK = ["\002s,\266", "\003had\232\002leW", "\003on \216", "", "\001yS",
@@ -84,50 +88,62 @@ module RSmaz
   
   # Compress a string to Smaz encoding
   def self.compress(input)
-    h1, h2, h3 = 0
     verb = ""
     out = ""
+    input = input.dup
     
     # This algorithm has been ported to Ruby from C and only
     # slightly Rubyized.. still a lonnnng way to go. Wanna give it a crack?
     while (input && input.length > 0)
-      h1 = h2 = input[0].ord << 3
-      h2 += input[1].ord if (input.length > 1)
-      h3 = h2 ^ input[2].ord if (input.length > 2)
+      h1 = h2 = input.ord << 3
+      h2 += input[1,1].ord if (input.length > 1)
+      h3 = (input.length > 2) ? h2 ^ input[2,1].ord : 0
       q = []
 
-      [input.length, 7].min.downto(1) do |j2|
-        slot = if j2 == 1
-          CODEBOOK[h1 % 241]
-        elsif j2 == 2
-          CODEBOOK[h2 % 241] 
+      [input.length, 7].min.downto(1) do |j|
+        slot = if j == 1
+          CODEBOOK[h1 % 241].dup
+        elsif j == 2
+          CODEBOOK[h2 % 241].dup
         else
-          CODEBOOK[h3 % 241] 
+          CODEBOOK[h3 % 241].dup
         end
 
         while (slot && slot[0]) do
-          if (slot[0].ord == j2 && (slot[1,j2] == input[0,j2]))
+          if (slot.ord == j && (slot[1,j] == input[0,j]))
             # Match found in hash table
-            q << verb
-            verb = ""
-            q << slot[slot[0].ord+1].ord
-            input = input[j2..-1]
+            
+            # Add verbatim data, if any (yes, it's quicker with the check)
+            unless verb.empty?
+              q << verb
+              verb = ""
+            end
+            
+            # Add encoded data and ditch unnecessary part of input string
+            q << slot[slot.ord+1,1].ord
+            input.slice!(0..j-1)
+            break
           else
-            slot = slot[2..-1]
+            # This in-place hack is quicker than slot = slot[1..-1]
+            
+            slot.reverse!.chop!.reverse!
+            #slot.slice!(0)
+            #slot[0] = ''
           end
         end
       end
       
       # No queue? It means we matched nothing, so add the current byte to the verbatim buffer
       if q.empty?
-        verb << input[0].ord if input[0]
-        input = input[1..-1]
+        verb << input[0,1] if input[0]
+        input.slice!(0)
       end
 
       # If the verbatim buffer is getting too long or we're at the end of the doc
       # throw the verbatim buffer to the output queue
       q << verb if verb.length == 256 || (verb.length > 0 && input.length == 0)
  
+      # Turn the queue into correctly encoded data
       out << q.collect do |item|
         if item.class == String && item.length == 1
           "\376" + item
@@ -148,11 +164,11 @@ module RSmaz
     out = ""
     s = StringScanner.new(input)
     until s.eos?
-      bv = s.get_byte[0].ord
+      bv = s.get_byte.ord
       if (bv == 254) 
         out << s.get_byte
       elsif (bv == 255)
-        len = s.get_byte[0].ord + 1
+        len = s.get_byte.ord + 1
         len.times do 
           out << s.get_byte
         end

data/spec/rsmaz_spec.rb

@@ -2,12 +2,21 @@ require File.dirname(__FILE__) + '/spec_helper.rb'
 
 describe RSmaz do
   
+  before(:each) do
+    # Do some memory leak checking
+    puts "\nMemory used: #{memory_usage}K"
+  end
+  
+  after(:each) do
+    puts "\nMemory used: #{memory_usage}K"
+  end
+  
   it "should compress 'the' to one byte" do
     RSmaz.compress("the").length.should == 1
   end
   
-  it "should compress 'thex' to three bytes" do
-    RSmaz.compress("thex").length.should == 3
+  it "should compress 'thex' to two bytes" do
+    RSmaz.compress("thex").length.should == 2
   end
   
   it "should compress and decompress strings back to the same string" do
@@ -23,4 +32,25 @@ describe RSmaz do
   it "should properly decode a reference compression (so the internal coding doesn't change)" do
     RSmaz.decompress("\020\230`A\376o\f\026\030").should == "hello world"
   end
+  
+  it "should compress to the same extent as the reference smaz implementation" do
+    RSmaz.compress("foobar").length.should == 4
+    RSmaz.compress("Nel mezzo del cammin di nostra vita, mi ritrovai in una selva oscura").length.should == 46
+  end
+  
+  it "should compress and decompress lots of random strings without issues" do
+    100.times do
+      str = (1..100).map { |a| (rand(26)+97).chr }.join
+      RSmaz.decompress(RSmaz.compress(str)).length.should == str.length
+    end
+  end
+
+  it "should compress and decompress lots of random strings without issues (again)" do
+    100.times do
+      str = (1..100).map { |a| (rand(26)+97).chr }.join
+      RSmaz.decompress(RSmaz.compress(str)).length.should == str.length
+    end
+  end
+  
+  
 end

data/spec/spec_helper.rb

@@ -8,3 +8,7 @@ end
 
 $:.unshift(File.dirname(__FILE__) + '/../lib')
 require 'rsmaz'
+
+def memory_usage
+  `ps -Orss #{Process.pid} | tail -1`.scan(/\d+/)[1].to_i rescue 0
+end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification 
 name: peterc-rsmaz
 version: !ruby/object:Gem::Version 
-  version: 0.0.2
+  version: 0.0.3
 platform: ruby
 authors: 
 - Peter Cooper
@@ -32,7 +32,7 @@ dependencies:
       - !ruby/object:Gem::Version 
         version: 1.8.0
     version: 
-description: "Short String Compression for Ruby.  RSmaz is a pure-Ruby port of the Smaz short string compression algorithm by Salvatore Sanfilippo and released as a C library at: http://github.com/antirez/smaz/tree/master  I've done some initial cleanup of a pure Ruby->C port, but this is not yet complete. It does pass the specs, however!"
+description: "Short String Compression for Ruby.  RSmaz is a pure-Ruby port of the Smaz short string compression algorithm by Salvatore Sanfilippo and released as a C library at: http://github.com/antirez/smaz/tree/master  I've done some initial cleanup of a pure Ruby->C port, but this is not yet complete. It does pass the specs, however! Feel free to clean it up as it's a bit memory inefficient right now... :)"
 email: 
 - pcooper@petercooper.co.uk
 executables: []