add a fix to the latin1prober to use float values in confidence calculat...

lib/rchardet/jpcntx.rb

@@ -170,7 +170,7 @@ def got_enough_data
     def get_confidence
       # This is just one way to calculate confidence. It works well for me.
       if @totalRel > MINIMUM_DATA_THRESHOLD
-        return (@totalRel - @relSample[0]) / @totalRel
+        return (@totalRel - @relSample[0]).to_f / @totalRel
       else
         return DONT_KNOW
       end

lib/rchardet/latin1prober.rb

@@ -133,7 +133,7 @@ def get_confidence
       if total < 0.01
         confidence = 0.0
       else
-        confidence = (@freqCounter[3] / total) - (@freqCounter[1] * 20.0 / total)
+        confidence = (@freqCounter[3].to_f / total) - (@freqCounter[1] * 20.0 / total)
       end
       if confidence < 0.0
         confidence = 0.0

lib/rchardet/sbcharsetprober.rb

@@ -110,7 +110,7 @@ def feed(aBuf)
     def get_confidence
       r = 0.01
       if @totalSeqs > 0
-        r = (1.0 * @seqCounters[POSITIVE_CAT]) / @totalSeqs / @model['mTypicalPositiveRatio']
+        r = @seqCounters[POSITIVE_CAT].to_f / @totalSeqs / @model['mTypicalPositiveRatio']
         r = r * @freqChar / @totalChar
         if r >= 1.0
           r = 0.99

test/complete_test.rb

@@ -31,6 +31,8 @@
 
       assert u.result, "result should not be nil"
       assert u.result['encoding'], "encoding should not be nil, result: #{u.result.inspect}"
+      assert u.result['confidence'], "confidence should not be nil, result: #{u.result.inspect}"
+      assert_equal u.result['confidence'].class, Float
       assert_equal encoding, u.result['encoding'].downcase
     end
   end

test/simple_assets/windows-1252.txt

@@ -0,0 +1 @@
+1234567890,ASDF,JKL,123 WHEREVER AVE,SOMEWHERE TOWN,UBERL�NDIA,USER@EXAMPLE.COM

test/simple_test.rb

@@ -4,7 +4,9 @@
 describe "Simple" do
   def assert_chardet_spec_detect(file, expected)
     content = File.open("test/simple_assets/#{file}.txt", 'rb'){|io| io.read }
-    assert_equal expected, CharDet.detect(content)
+    detected = CharDet.detect(content)
+    assert_equal expected, detected
+    assert_equal detected['confidence'].class, Float
   end
 
   def pending
@@ -23,7 +25,7 @@ def pending
 
   it "detects Shift_JIS" do
     assert_chardet_spec_detect 'Shift_JIS', {
-      "encoding" => 'SHIFT_JIS', "confidence" => (RUBY_VERSION > "1.9.3" ? 0.99 : 1) # TODO the 1.9 value might be wrong but I cannot find any bug
+      "encoding" => 'SHIFT_JIS', "confidence" => (RUBY_VERSION > "1.9.3" ? 0.99 : 1.0) # TODO the 1.9 value might be wrong but I cannot find any bug
     }
   end
 
@@ -41,13 +43,13 @@ def pending
 
   it "detects UTF_16BE" do
     assert_chardet_spec_detect 'UTF-16BE' , {
-      "encoding" => 'UTF-16BE', "confidence" => 1
+      "encoding" => 'UTF-16BE', "confidence" => 1.0
     }
   end
 
   it "detects UTF_16LE" do
     assert_chardet_spec_detect 'UTF-16LE' , {
-      "encoding" => 'UTF-16LE', "confidence" => 1
+      "encoding" => 'UTF-16LE', "confidence" => 1.0
     }
   end  
 
@@ -63,6 +65,13 @@ def pending
     }
   end
 
+  it "detects windows-1252" do
+    assert_chardet_spec_detect  'windows-1252'  , {
+      # not perfect, but better than detecting nil
+      "encoding" => 'windows-1252', "confidence" => 0.36875
+    }
+  end
+
   it "detects russian" do
     # this failed when using $KCODE='u' on 1.8 ... just making sure it stays put
     CharDet.detect("Toto je zpr\xE1va ve form\xE1tu MIME s n\xECkolika \xE8\xE1stmi.\n")["encoding"].must_equal "windows-1251"