From c280760b44b16ef43f1c4fa35ce8a0f2291fa455 Mon Sep 17 00:00:00 2001
From: r888888888 <r888888888@gmail.com>
Date: Sat, 2 Jan 2016 09:38:24 -0800
Subject: [PATCH] add cosine similarity algo for related tag calculator

---
 Gemfile.lock                          |  3 --
 app/logical/related_tag_calculator.rb | 41 +++++++++++++++++++++++++--
 2 files changed, 38 insertions(+), 6 deletions(-)

diff --git a/Gemfile.lock b/Gemfile.lock
index f61479045..9f360122b 100644
--- a/Gemfile.lock
+++ b/Gemfile.lock
@@ -384,6 +384,3 @@ DEPENDENCIES
   vcr
   webmock
   whenever
-
-BUNDLED WITH
-   1.10.0
diff --git a/app/logical/related_tag_calculator.rb b/app/logical/related_tag_calculator.rb
index 2215a6976..e290f3216 100644
--- a/app/logical/related_tag_calculator.rb
+++ b/app/logical/related_tag_calculator.rb
@@ -2,7 +2,7 @@ class RelatedTagCalculator
   def self.find_tags(tag, limit)
     CurrentUser.without_safe_mode do
       Post.with_timeout(5_000, []) do
-        Post.tag_match(tag).limit(limit).select("posts.tag_string").reorder("posts.md5").map(&:tag_string)
+        Post.tag_match(tag).limit(limit).select("posts.tag_string").reorder("posts.md5").pluck(:tag_string)
       end
     end
   end
@@ -34,6 +34,41 @@ class RelatedTagCalculator
     counts
   end
 
+  def self.calculate_similar_from_sample(tag)
+    # this uses cosine similarity to produce more useful
+    # related tags, but is more db intensive
+    counts = Hash.new {|h, k| h[k] = 0}
+
+    CurrentUser.without_safe_mode do
+      Post.with_timeout(5_000, []) do
+        Post.tag_match(tag).limit(400).select("posts.tag_string").reorder("posts.md5").pluck(:tag_string).each do |tag_string|
+          tag_string.scan(/\S+/).each do |tag|
+            counts[tag] += 1
+          end
+        end
+      end
+    end
+
+    tag_record = Tag.find_by_name(tag)
+    candidates = convert_hash_to_array(counts, 100)
+    similar_counts = Hash.new {|h, k| h[k] = 0}
+    CurrentUser.without_safe_mode do
+      Post.with_timeout(5_000) do
+        candidates.each do |ctag, _|
+          acount = Post.tag_match("#{tag} #{ctag}").count
+          ctag_record = Tag.find_by_name(ctag)
+          div = Math.sqrt(tag_record.post_count * ctag_record.post_count)
+          if div != 0
+            c = acount / div
+            similar_counts[ctag] = c
+          end
+        end
+      end
+    end
+
+    convert_hash_to_array(similar_counts)
+  end
+
   def self.calculate_from_sample(tags, limit, category_constraint = nil)
     counts = Hash.new {|h, k| h[k] = 0}
 
@@ -56,8 +91,8 @@ class RelatedTagCalculator
     counts
   end
 
-  def self.convert_hash_to_array(hash)
-    hash.to_a.sort_by {|x| -x[1]}.slice(0, 25)
+  def self.convert_hash_to_array(hash, limit = 25)
+    hash.to_a.sort_by {|x| -x[1]}.slice(0, limit)
   end
 
   def self.convert_hash_to_string(hash)