wwood-rarff 0.2.2 → 0.2.3

data/History.txt

@@ -1,5 +1,9 @@
 == Changes
 
+=== 0.2.3
+
+* Added set_string_attributes_to_nominal for easy conversion of string attributes to nominal ones
+
 === 0.2.2 (unofficial)
 
 * Handles boolean inputs, which are modelled as nominals

data/Rakefile

@@ -3,7 +3,7 @@ require 'hoe'
 #require './lib/rarff.rb'
  
 gem_name = 'rarff'
-hoe = Hoe.new(gem_name,'0.2.2') do |p|
+hoe = Hoe.new(gem_name,'0.2.3') do |p|
   
   p.author = "Andy Payne, Ben J Woodcroft"
   p.email = "apayne .at. gmail.com, b.woodcroft@pgrad.unimelb.edu.au"

data/lib/rarff.rb

@@ -63,7 +63,8 @@ module Rarff
   ################################################################################
 
   class Attribute
-    attr_accessor :name, :type
+    attr_accessor :name
+    attr_reader :type
 
     def initialize(name='', type='')
       @name = name
@@ -88,7 +89,8 @@ module Rarff
         @type_is_nominal = true
         # Example format: "{nom1,nom2, nom3, nom4,nom5 } "
         # Split on '{'  ','  or  '}'
-        @type = @type.gsub(/^\s*\{\s*/, '').gsub(/\s*\}\s*$/, '').split(/\s*\,\s*/)
+        #        @type = @type.gsub(/^\s*\{\s*/, '').gsub(/\s*\}\s*$/, '').split(/\s*\,\s*/)
+        @type = @type.split(/\s*\,\s*/)
       end
     end
 
@@ -104,7 +106,7 @@ module Rarff
 
     def to_arff
       if @type_is_nominal == true
-        ATTRIBUTE_MARKER + " #{@name} {#{@type.join(',').gsub(' ','_')}}"
+        ATTRIBUTE_MARKER + " #{@name} #{@type.join(',').gsub(' ','_')}"
       else
         ATTRIBUTE_MARKER + " #{@name} #{@type}"
       end
@@ -120,7 +122,8 @@ module Rarff
 
 
   class Relation
-    attr_accessor :name, :attributes, :instances
+    attr_accessor :name, :attributes
+    attr_reader :instances
 
 
     def initialize(name='')
@@ -197,10 +200,10 @@ module Rarff
             @attributes[j] = Attribute.new("Attr#{j}", ATTRIBUTE_NUMERIC)
           elsif col.kind_of?(String)
             @attributes[j] = Attribute.new("Attr#{j}", ATTRIBUTE_STRING)
-          elsif col.kind_of?(TrueClass) or col.kind_of?(FalseClass) # How come there is no generic BooleanClass?
+          elsif col == false or col == true #exactly equal to a boolean
             @attributes[j] = Attribute.new("Attr#{j}", ATTRIBUTE_BOOLEAN)
           else
-            raise Exception, "Could not parse attribute: #{col.inspect}"
+            raise Exception, "Could not parse attribute to ARFF data type: #{col.inspect}"
           end
         }
       }
@@ -212,6 +215,26 @@ module Rarff
       end
     end
 
+    # Make all String type attributes into nominal attributes, because
+    # they are more useful in WEKA because more techniques handle them than
+    # strings
+    def set_string_attributes_to_nominal
+      nominals = {}
+      # Frustratingly, we have to traverse this 2D array with the
+      # wrong dimension first. Oh well.
+      @instances.each_with_index do |row, row_index|
+        row.each_with_index do |string, col_index|
+          next unless @attributes[col_index].type == ATTRIBUTE_STRING
+
+          nominals[col_index] ||= {}
+          nominals[col_index][string] ||= true
+        end
+      end
+
+      nominals.each do |index, strings|
+        @attributes[index].type = "{#{strings.keys.join(',')}}"
+      end
+    end
 
     def expand_sparse(str)
       arr = Array.new(@attributes.size, 0)
@@ -247,7 +270,7 @@ module Rarff
           
           # Do the final output
           if sparse
-            if col.nil? or 
+            if col.nil? or
                 (@attributes[i].type =~ /^#{ATTRIBUTE_NUMERIC}$/i and col == 0)
               nil
             else
@@ -267,7 +290,7 @@ module Rarff
         else
           mapped.join(", ")
         end
-      }.join("\n").gsub(/^/, sparse ? '{' : '').gsub(/$/, sparse ? '}' : '')
+      }.join("\n")
     end
 
 

data/test/test_rarff.rb

@@ -101,7 +101,6 @@ class TestArffLib < Test::Unit::TestCase
   #    assert_equal(0, rel.instances[3][12])
   #    #               puts "\n\nARFF: (\n#{rel.to_arff}\n)"
   #  end
-  #  
   def test_output_missing
     arff_file_str = <<-END_OF_ARFF_FILE
 @RELATION MyCoolRelation
@@ -155,12 +154,75 @@ class TestArffLib < Test::Unit::TestCase
     rel.attributes[4].type = 'DATE "yyyy-MM-dd HH:mm:ss"'
 
     #               puts "rel.to_arff:\n(\n#{rel.to_arff}\n)\n"
-    assert_equal(arff_file_str, rel.to_arff, "missing data output failure")
+    assert_equal(arff_file_str, rel.to_arff, "missing data from first line output failure")
   end
   
   def test_boolean
     arff_file_str = <<-END_OF_ARFF_FILE
 @RELATION MyCoolRelation
+@ATTRIBUTE Attr0 {false,true}
+@DATA
+true
+    END_OF_ARFF_FILE
+
+    arff_file_str.gsub!(/\n$/, '')
+
+    instances = [ [true]]
+
+    rel = Rarff::Relation.new('MyCoolRelation')
+    rel.instances = instances
+
+    #               puts "rel.to_arff:\n(\n#{rel.to_arff}\n)\n"
+    assert_equal(arff_file_str, rel.to_arff, "missing data from first line output failure")   
+  end
+  
+  def test_boolean_multipl
+    arff_file_str = <<-END_OF_ARFF_FILE
+@RELATION MyCoolRelation
+@ATTRIBUTE Attr0 {false,true}
+@ATTRIBUTE Attr1 {false,true}
+@ATTRIBUTE Attr2 {false,true}
+@DATA
+true, false, true
+true, true, true
+    END_OF_ARFF_FILE
+
+    arff_file_str.gsub!(/\n$/, '')
+
+    instances = [ [true,false,true],[true,true,true]]
+
+    rel = Rarff::Relation.new('MyCoolRelation')
+    rel.instances = instances
+
+    #               puts "rel.to_arff:\n(\n#{rel.to_arff}\n)\n"
+    assert_equal(arff_file_str, rel.to_arff, "missing data from first line output failure")   
+  end
+
+  def test_strings_as_nominal
+    arff_file_str = <<-END_OF_ARFF_FILE
+@RELATION MyCoolRelation
+@ATTRIBUTE Attr0 {two,one}
+@ATTRIBUTE Attr1 {three,four}
+@DATA
+one, three
+two, four
+    END_OF_ARFF_FILE
+
+    arff_file_str.gsub!(/\n$/, '')
+
+    instances = [ ['one','three'],['two','four']]
+
+    rel = Rarff::Relation.new('MyCoolRelation')
+    rel.instances = instances
+    rel.set_string_attributes_to_nominal
+
+    #               puts "rel.to_arff:\n(\n#{rel.to_arff}\n)\n"
+    assert_equal(arff_file_str, rel.to_arff, "test_strings_as_nominal")
+  end
+  
+  def test_boolean_2
+    arff_file_str = <<-END_OF_ARFF_FILE
+@RELATION MyCoolRelation
 @ATTRIBUTE Attr0 NUMERIC
 @ATTRIBUTE subject STRING
 @ATTRIBUTE Attr2 {false,true}

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification 
 name: wwood-rarff
 version: !ruby/object:Gem::Version 
-  version: 0.2.2
+  version: 0.2.3
 platform: ruby
 authors: 
 - Andy Payne, Ben J Woodcroft
@@ -9,11 +9,12 @@ autorequire:
 bindir: bin
 cert_chain: []
 
-date: 2009-01-30 00:00:00 -08:00
+date: 2009-02-18 00:00:00 -08:00
 default_executable: 
 dependencies: 
 - !ruby/object:Gem::Dependency 
   name: hoe
+  type: :development
   version_requirement: 
   version_requirements: !ruby/object:Gem::Requirement 
     requirements: