vector_embed 0.1.1 → 0.2.0

checksums.yaml

@@ -1,15 +1,15 @@
 ---
 !binary "U0hBMQ==":
   metadata.gz: !binary |-
-    YWRiOGI3YTdkMDIwNDljNTA0NDkzOGNiNGVkNThiZjczNjdlNDU3OQ==
+    YWJmNzczYzU1MjZlODE4NzcxZGMxZTc0MTgzNjdkNjFjYzc5MjRjNA==
   data.tar.gz: !binary |-
-    ZDIxY2Q1NDFjMjkxNDFkOGJkZTk0NTZiMDc4NTgwNjYwZGE5MDI1MQ==
+    NDNhZmQwZDNjOWMwNWJlOGM3NTBiOTVmNjIwZGJjM2EzYjdlZWZiMA==
 !binary "U0hBNTEy":
   metadata.gz: !binary |-
-    YzAwOGQ0NjlmNThiYzNmZmQxNDk0ZTY2ZTIxMTA2M2NjMjZiNGE4MGU0NzE3
-    YTZhNTkzM2ViMWU2M2FhZDE5NTk5YTdkOWU3NGYxYjVjNmJkYmZjNGVhMTU0
-    ODY1OTkzNmFhNDIwODY4MjUwYTJjODU0ZDgzNWYzNmE1ZTljYzg=
+    MDc4NjE0YWQ4ZmYzZWZmNTdhOTM3ZTkxODI0Y2RjNjU4YTAxMjMzMjdlZWQx
+    OTUzMDY4NGE4YjFkOWY1OWJiNTVjNWJmYTFkN2I1ZDdhZTY1OWQ0YzRkYjE3
+    MDkxZWYzYmQ2NDMyOTIxMzEyZDBkNGMwMjdkNDY1Yzc2YTMyNzk=
   data.tar.gz: !binary |-
-    ZjM3OGUyNDk1YzIxMjc5OTgyYmI3ZWY3NGEwYjRjYWY1MTk0MjJhYzhjODU0
-    NTg0OWYxNWU4ZTQ3YTQ0ZDFlOWVjYzMyYmZhOTE0MjNlMDRjMjMyMDkwNGMy
-    NDNhMjhkYjUzNGRhZmRhNjliMGFiNzY1NmNiNzk5MDhhOTUyOTU=
+    ZmI5MzkzYWRmNzY2NjA0ZjBhY2NmNWIzMWMxZmM3OTU3MTVlY2Q5ZTg0YzJi
+    NzUyZDI3MjgzNmJhNDFiYTY0YjRlOTNlNzE4NjYyYjZlMTI5ZDFhMGZhMmY2
+    ZmYyZmU5Nzc1NzBmYjhkMmFmMDkxYzdkZjM0NTY2OTZmY2U0ODU=

data/CHANGELOG

@@ -1,3 +1,13 @@
+0.2.0 / 2013-05-14
+
+* Breaking changes
+
+  * Per the whole point of sparse vectors, don't output numbers features with value 0
+
+* Enhancements
+
+  * More concise number representations per https://github.com/scikit-learn/scikit-learn/pull/1849
+
 0.1.1 / 2013-04-04
 
 * Enhancements

data/lib/vector_embed/maker.rb

@@ -29,11 +29,17 @@ class VectorEmbed
       when Array
         memo = []
         v.each_with_index do |vv, i|
-          memo << [ parent.index([k, i]), value(vv) ]
+          unless (vvv = value(vv)).nil?
+            memo << [ parent.index([k, i]), vvv ]
+          end
         end
         memo
       else
-        [ [ parent.index([k]), value(v) ] ]
+        if (vv = value(v)).nil?
+          []
+        else
+          [ [ parent.index([k]), value(v) ] ]
+        end
       end
     end
   end

data/lib/vector_embed/maker/number.rb

@@ -18,7 +18,9 @@ class VectorEmbed
           else
             v
           end
-          num > 1e10 ? ('%.10e' % num) : num
+          if num.nonzero?
+            '%.16g' % num
+          end
         end
       end
 
@@ -27,7 +29,7 @@ class VectorEmbed
         when Numeric, JUST_A_NUMBER
           Number.numify v
         when NilClass, NULL, SLASH_N
-          0
+          nil
         else
           raise ArgumentError, "Can't embed #{v.inspect} in number feature #{k.inspect}"
         end

data/lib/vector_embed/version.rb

@@ -1,3 +1,3 @@
 class VectorEmbed
-  VERSION = "0.1.1"
+  VERSION = "0.2.0"
 end

data/spec/vector_embed_spec.rb

@@ -122,8 +122,21 @@ describe VectorEmbed do
       v.line(1, 1 => '9').should == "1 #{l_h('1')}:9"
       v.line(1, 1 => 5.4).should == "1 #{l_h('1')}:5.4"
       v.line(1, 1 => '5.4').should == "1 #{l_h('1')}:5.4"
-      v.line(1, 1 => 9e9).should == "1 #{l_h('1')}:9000000000.0"
-      v.line(1, 1 => '9e9').should == "1 #{l_h('1')}:9000000000.0"
+      v.line(1, 1 => 9e9).should == "1 #{l_h('1')}:9000000000"
+      v.line(1, 1 => '9e9').should == "1 #{l_h('1')}:9000000000"
+    end
+
+    it "does not output 0 in number attributes" do
+      v = VectorEmbed.new
+      v.line(3, 1 => 1)
+      v.line(3, 1 => 0).should == "3"
+      v.line(3, 1 => '0').should == "3"
+    end
+
+    it "treats nil like zero in number attributes" do
+      v = VectorEmbed.new
+      v.line(1, 1 => 1)
+      v.line(1, 1 => nil).should == v.line(1, 1 => 0)
     end
 
     it "stores strings as m-category attributes" do
@@ -160,10 +173,10 @@ describe VectorEmbed do
     it "in number mode, treats null as 0" do
       v = VectorEmbed.new
       v.line(1, 1 => 9).should == "1 #{l_h('1')}:9"
-      v.line(1, 1 => nil).should == "1 #{l_h('1')}:0"
-      v.line(1, 1 => 'null').should == "1 #{l_h('1')}:0"
-      v.line(1, 1 => 'NULL').should == "1 #{l_h('1')}:0"
-      v.line(1, 1 => '\N').should == "1 #{l_h('1')}:0"
+      v.line(1, 1 => nil).should == v.line(1, 1 => 0)
+      v.line(1, 1 => 'null').should == v.line(1, 1 => 0)
+      v.line(1, 1 => 'NULL').should == v.line(1, 1 => 0)
+      v.line(1, 1 => '\N').should == v.line(1, 1 => 0)
     end
 
     it "doesn't allow embedding boolean in number mode or vice-versa" do
@@ -186,12 +199,12 @@ describe VectorEmbed do
 
     it "uses scientific notation for large numbers" do
       v = VectorEmbed.new
-      v.line(5, 1 => 8.12e13).should == "5 #{l_h('1')}:8.1200000000e+13"
+      v.line(5, 1 => 8.12e27).should == "5 #{l_h('1')}:8.12e+27"
     end
 
     it "detects numbers in strings" do
       v = VectorEmbed.new
-      v.line(5, 1 => '8.12e13').should == "5 #{l_h('1')}:8.1200000000e+13"
+      v.line(5, 1 => '8.12e13').should == "5 #{l_h('1')}:81200000000000"
     end
 
     it "allows 2 byte n-grams" do

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: vector_embed
 version: !ruby/object:Gem::Version
-  version: 0.1.1
+  version: 0.2.0
 platform: ruby
 authors:
 - Seamus Abshere
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2013-04-09 00:00:00.000000000 Z
+date: 2013-05-15 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: murmurhash3