* post keeper is calculated synchronously

* add fallback mechanism in case archive service is not up
* pass along most recently added tags to any keeper calculation

Gemfile

@@ -8,7 +8,7 @@ gem "sprockets-rails", :require => "sprockets/railtie"
 gem "uglifier"
 gem "therubyracer", :platforms => :ruby
 gem "rails", "~> 4.2.0"
-gem "pg"
+gem "pg", "0.21.0"
 gem "dalli", :platforms => :ruby
 gem "memcache-client", :platforms => [:mswin, :mingw, :x64_mingw]
 gem "tzinfo-data", :platforms => [:mswin, :mingw, :x64_mingw]
@@ -67,6 +67,7 @@ group :development, :test do
   gem 'awesome_print'
   gem 'pry-byebug'
   gem 'ruby-prof'
+  gem 'foreman'
 end
 
 group :test do

Gemfile.lock

@@ -130,6 +130,13 @@ GEM
     ffaker (2.8.1)
     ffi (1.9.18)
     ffi (1.9.18-x64-mingw32)
+    foreman (0.63.0)
+      dotenv (>= 0.7)
+      thor (>= 0.13.6)
+    foreman (0.63.0-mingw32)
+      dotenv (>= 0.7)
+      thor (>= 0.13.6)
+      win32console (~> 1.3.0)
     get_process_mem (0.2.1)
     globalid (0.4.1)
       activesupport (>= 4.2.0)
@@ -233,8 +240,8 @@ GEM
       multi_xml (~> 0.5)
       rack (>= 1.2, < 3)
     os (0.9.6)
-    pg (0.19.0)
+    pg (0.21.0)
-    pg (0.19.0-x64-mingw32)
+    pg (0.21.0-x64-mingw32)
     protected_attributes (1.1.4)
       activemodel (>= 4.0.1, < 5.0)
     pry (0.11.3)
@@ -396,6 +403,7 @@ GEM
     webrobots (0.1.2)
     whenever (0.10.0)
       chronic (>= 0.6.3)
+    win32console (1.3.2)
 
 PLATFORMS
   ruby
@@ -423,6 +431,7 @@ DEPENDENCIES
   dtext_rb!
   factory_girl
   ffaker
+  foreman
   google-api-client
   highline
   httparty
@@ -434,7 +443,7 @@ DEPENDENCIES
   net-sftp
   newrelic_rpm
   oauth2
-  pg
+  pg (= 0.21.0)
   protected_attributes
   pry-byebug
   radix62 (~> 1.0.1)

app/helpers/delayed_jobs_helper.rb

@@ -1,6 +1,9 @@
 module DelayedJobsHelper
   def print_name(job)
     case job.name
+    when "PostKeeperManager.check_and_update"
+      "<strong>update post tagger</strong>"
+
     when "Tag.increment_post_counts"
       "<strong>increment post counts</strong>"
 
@@ -68,6 +71,9 @@ module DelayedJobsHelper
 
   def print_handler(job)
     case job.name
+    when "PostKeeperManager.check_and_update"
+      ""
+
     when "Tag.increment_post_counts", "Tag.decrement_post_counts"
       ""
 

app/logical/post_keeper_manager.rb

@@ -3,10 +3,71 @@ class PostKeeperManager
     PostArchive.enabled?
   end
 
-  def self.queue_check(post_id)
+  # these are all class methods to simplify interaction with delayedjob
-    delay(queue: "default").check_and_update(post_id)
+
+  # in general we want to call these methods synchronously because updating
+  # the keeper data with a delay defeats the purpose. but this relies on
+  # archive db being up; we don't want to block updates in case it goes down.
+  # so we need to permit async updates also.
+
+  def self.queue_check(post_id, updater_id, increment_tags)
+    delay(queue: "default").check_and_update(post_id, updater_id, increment_tags, false)
   end
 
+  def self.check_and_update(post, updater_id = nil, increment_tags = nil)
+    post = Post.find(post) unless post.is_a?(Post)
+    keeper_id = check(post, updater_id, increment_tags)
+    post.keeper_data = {uid: keeper_id}
+  end
+
+  # because post archives might get delayed, we need to pass along the most
+  # recently added tags inside the job. downside: this doesn't keep track of
+  # source or rating changes. this method changes no state.
+  def self.check(post, updater_id = nil, increment_tags = nil, enable_async = true)
+    if enable_async && !PostArchive.test_connection
+      # if archive is down, just queue this work and do it later
+      queue_check(post.id, updater_id, increment_tags)
+      return
+    end
+
+    changes = {}
+    final_tags = Set.new(post.tag_array)
+
+    # build a mapping of who added a tag first
+    PostArchive.where(post_id: post.id).order("updated_at").each do |pa|
+      pa.added_tags.each do |at|
+        if pa.updater_id
+          if !changes.has_key?(at) && final_tags.include?(at)
+            changes[at] = pa.updater_id 
+          end
+
+          if pa.source_changed? && pa.source == post.source
+            changes[" source"] = pa.updater_id
+          end
+        end
+      end
+    end
+
+    if updater_id && increment_tags.present?
+      increment_tags.each do |tag|
+        if !changes.has_key?(tag)
+          changes[tag] = updater_id
+        end
+      end
+    end
+
+    # add up how many changes each user has made
+    ranking = changes.values.uniq.inject({}) do |h, user_id|
+      h[user_id] = changes.select {|k, v| v == user_id}.size
+      h
+    end
+
+    ranking.max_by {|k, v| v}.try(:first)
+  end
+
+
+  # these methods are for reporting and are not used
+
   # in general, unweighted changes attribution 5% of the time,
   # weighted changes attribution 12% of the time at w=1000,
   # up to 17% of the time at w=100.
@@ -45,14 +106,6 @@ class PostKeeperManager
     # end
   end
 
-  def self.check_and_update(post_id)
-    post = Post.find(post_id)
-    keeper_id = check(post)
-    CurrentUser.as_system do
-      post.update_column(:keeper_data, {uid: keeper_id})
-    end
-  end
-
   def self.print_weighted(post, w = 1000)
     changes = {}
     final_tags = Set.new(post.tag_array)
@@ -119,50 +172,4 @@ class PostKeeperManager
     ranking.max_by {|k, v| v}.first
   end
 
-  def self.check(post)
-    changes = {}
-    final_tags = Set.new(post.tag_array)
-
-    # build a mapping of who added a tag first
-    PostArchive.where(post_id: post.id).order("updated_at").each do |pa|
-      # Rails.logger.debug "archive #{pa.id}"
-      pa.added_tags.each do |at|
-        # Rails.logger.debug "  checking #{at}"
-        if pa.updater_id
-          if !changes.has_key?(at) && final_tags.include?(at)
-            # Rails.logger.debug "    adding #{at} for #{pa.updater_id}"
-            changes[at] = pa.updater_id 
-          end
-
-          if pa.source_changed? && pa.source == post.source
-            # Rails.logger.debug "    adding source for #{pa.updater_id}"
-            changes[" source"] = pa.updater_id
-          end
-        else
-          # Rails.logger.debug "    no updater"
-        end
-      end
-
-      # easy to double count trivial changes if a user is just fixing mistakes
-      # pa.removed_tags.each do |rt|
-      #   Rails.logger.debug "  checking -#{rt}"
-      #   if pa.updater_id
-      #     if !changes.has_key?("-#{rt}") && !final_tags.include?(rt)
-      #       Rails.logger.debug "    adding -#{rt} for #{pa.updater_id}"
-      #       changes["-#{rt}"] = pa.updater_id
-      #     end
-      #   else
-      #     Rails.logger.debug "    no updater"
-      #   end
-      # end
-    end
-
-    # add up how many changes each user has made
-    ranking = changes.values.uniq.inject({}) do |h, user_id|
-      h[user_id] = changes.select {|k, v| v == user_id}.size
-      h
-    end
-
-    ranking.max_by {|k, v| v}.first
-  end
 end

app/models/application_record.rb

@@ -131,6 +131,13 @@ class ApplicationRecord < ActiveRecord::Base
       def columns(*params)
         super.reject {|x| x.sql_type == "tsvector"}
       end
+
+      def test_connection
+        limit(1).select(:id)
+        return true
+      rescue PG::Error
+        return false
+      end
     end
   end
 

app/models/post.rb

@@ -57,9 +57,7 @@ class Post < ApplicationRecord
   has_many :favorites
   has_many :replacements, class_name: "PostReplacement", :dependent => :destroy
 
-  if PostKeeperManager.enabled?
+  serialize :keeper_data, JSON
-    serialize :keeper_data, JSON
-  end
 
   if PostArchive.enabled?
     has_many :versions, lambda {order("post_versions.updated_at ASC")}, :class_name => "PostArchive", :dependent => :destroy
@@ -76,7 +74,11 @@ class Post < ApplicationRecord
     end
 
     def keeper_id
-      keeper_data ? keeper_data[:uid] : uploader_id
+      if PostKeeperManager.enabled?
+        keeper_data ? keeper_data["uid"] : uploader_id
+      else
+        uploader_id
+      end
     end
 
     def keeper
@@ -664,7 +666,7 @@ class Post < ApplicationRecord
 
       if PostKeeperManager.enabled? && persisted?
         # no need to do this check on the initial create
-        PostKeeperManager.queue_check(id)
+        PostKeeperManager.check_and_update(self, CurrentUser.id, increment_tags)
       end
     end
 

app/presenters/post_set_presenters/wiki_page.rb

@@ -2,7 +2,7 @@ module PostSetPresenters
   class WikiPage < PostSetPresenters::Post
     def posts
       @post_set.posts
-    rescue ActiveRecord::StatementInvalid, PGError
+    rescue ActiveRecord::StatementInvalid, PG::Error
       []
     end
 

test/models/post_keeper_manager_test.rb

@@ -6,8 +6,6 @@ class PostKeeperManagerTest < ActiveSupport::TestCase
   context "#check_and_update" do
     setup do
       Timecop.travel(1.month.ago) do
-        @system = FactoryGirl.create(:user)
-        User.stubs(:system).returns(@system)
         @alice = FactoryGirl.create(:user)
         @bob = FactoryGirl.create(:user)
         @carol = FactoryGirl.create(:user)
@@ -39,8 +37,6 @@ class PostKeeperManagerTest < ActiveSupport::TestCase
   context "#check" do
     setup do
       Timecop.travel(1.month.ago) do
-        @system = FactoryGirl.create(:user)
-        User.stubs(:system).returns(@system)
         @alice = FactoryGirl.create(:user)
         @bob = FactoryGirl.create(:user)
         @carol = FactoryGirl.create(:user)
@@ -62,15 +58,10 @@ class PostKeeperManagerTest < ActiveSupport::TestCase
             @post.update_attributes(tag_string: "aaa bbb ccc")
           end
         end
-        CurrentUser.scoped(@carol) do
-          Timecop.travel(4.hours.from_now) do
-            @post.update_attributes(tag_string: "ccc ddd eee fff ggg")
-          end
-        end
       end
 
       should "find the most frequent tagger for a post" do
-        assert_equal(@carol.id, subject.check(@post))
+        assert_equal(@carol.id, subject.check(@post, @carol.id, %w(ddd eee fff ggg)))
       end
     end
   end