Add markdown cache columns to the database, but don't use them yet

This commit adds a number of _html columns and, with the exception of Note,
starts updating them whenever the content of their partner fields changes.

Note has a collision with the note_html attr_accessor; that will be fixed later

A background worker for clearing these cache columns is also introduced - use
`rake cache:clear` to set it off. You can clear the database or Redis caches
separately by running `rake cache:clear:db` or `rake cache:clear:redis`,
respectively.

16 files changed, 236 insertions, 9 deletions

diff --git a/app/helpers/gitlab_markdown_helper.rb b/app/helpers/gitlab_markdown_helper.rb
index 1a259656f31..d24680b8617 100644
--- a/app/helpers/gitlab_markdown_helper.rb
+++ b/app/helpers/gitlab_markdown_helper.rb
@@ -51,17 +51,15 @@ module GitlabMarkdownHelper
     context[:project] ||= @project
 
     html = Banzai.render(text, context)
+    banzai_postprocess(html, context)
+  end
 
-    context.merge!(
-      current_user:   (current_user if defined?(current_user)),
+  def markdown_field(object, field)
+    object = object.for_display if object.respond_to?(:for_display)
+    return "" unless object.present?
 
-      # RelativeLinkFilter
-      requested_path: @path,
-      project_wiki:   @project_wiki,
-      ref:            @ref
-    )
-
-    Banzai.post_process(html, context)
+    html = Banzai.render_field(object, field)
+    banzai_postprocess(html, object.banzai_render_context(field))
   end
 
   def asciidoc(text)
@@ -196,4 +194,18 @@ module GitlabMarkdownHelper
       icon(options[:icon])
     end
   end
+
+  # Calls Banzai.post_process with some common context options
+  def banzai_postprocess(html, context)
+    context.merge!(
+      current_user:   (current_user if defined?(current_user)),
+
+      # RelativeLinkFilter
+      requested_path: @path,
+      project_wiki:   @project_wiki,
+      ref:            @ref
+    )
+
+    Banzai.post_process(html, context)
+  end
 end
diff --git a/app/models/abuse_report.rb b/app/models/abuse_report.rb
index b01a244032d..2340453831e 100644
--- a/app/models/abuse_report.rb
+++ b/app/models/abuse_report.rb
@@ -1,4 +1,8 @@
 class AbuseReport < ActiveRecord::Base
+  include CacheMarkdownField
+
+  cache_markdown_field :message, pipeline: :single_line
+
   belongs_to :reporter, class_name: 'User'
   belongs_to :user
 
@@ -7,6 +11,9 @@ class AbuseReport < ActiveRecord::Base
   validates :message, presence: true
   validates :user_id, uniqueness: { message: 'has already been reported' }
 
+  # For CacheMarkdownField
+  alias_method :author, :reporter
+
   def remove_user(deleted_by:)
     user.block
     DeleteUserWorker.perform_async(deleted_by.id, user.id, delete_solo_owned_groups: true)
diff --git a/app/models/appearance.rb b/app/models/appearance.rb
index 4cf8dd9a8ce..e4106e1c2e9 100644
--- a/app/models/appearance.rb
+++ b/app/models/appearance.rb
@@ -1,4 +1,8 @@
 class Appearance < ActiveRecord::Base
+  include CacheMarkdownField
+
+  cache_markdown_field :description
+
   validates :title,       presence: true
   validates :description, presence: true
   validates :logo,        file_size: { maximum: 1.megabyte }
diff --git a/app/models/application_setting.rb b/app/models/application_setting.rb
index 55d2e07de08..c99aa7772bb 100644
--- a/app/models/application_setting.rb
+++ b/app/models/application_setting.rb
@@ -1,5 +1,7 @@
 class ApplicationSetting < ActiveRecord::Base
+  include CacheMarkdownField
   include TokenAuthenticatable
+
   add_authentication_token_field :runners_registration_token
   add_authentication_token_field :health_check_access_token
 
@@ -17,6 +19,11 @@ class ApplicationSetting < ActiveRecord::Base
   serialize :domain_whitelist, Array
   serialize :domain_blacklist, Array
 
+  cache_markdown_field :sign_in_text
+  cache_markdown_field :help_page_text
+  cache_markdown_field :shared_runners_text, pipeline: :plain_markdown
+  cache_markdown_field :after_sign_up_text
+
   attr_accessor :domain_whitelist_raw, :domain_blacklist_raw
 
   validates :session_expire_delay,
diff --git a/app/models/broadcast_message.rb b/app/models/broadcast_message.rb
index 61498140f27..cb40f33932a 100644
--- a/app/models/broadcast_message.rb
+++ b/app/models/broadcast_message.rb
@@ -1,6 +1,9 @@
 class BroadcastMessage < ActiveRecord::Base
+  include CacheMarkdownField
   include Sortable
 
+  cache_markdown_field :message, pipeline: :broadcast_message
+
   validates :message,   presence: true
   validates :starts_at, presence: true
   validates :ends_at,   presence: true
diff --git a/app/models/concerns/cache_markdown_field.rb b/app/models/concerns/cache_markdown_field.rb
new file mode 100644
index 00000000000..90bd6490a02
--- /dev/null
+++ b/app/models/concerns/cache_markdown_field.rb
@@ -0,0 +1,131 @@
+# This module takes care of updating cache columns for Markdown-containing
+# fields. Use like this in the body of your class:
+#
+#     include CacheMarkdownField
+#     cache_markdown_field :foo
+#     cache_markdown_field :bar
+#     cache_markdown_field :baz, pipeline: :single_line
+#
+# Corresponding foo_html, bar_html and baz_html fields should exist.
+module CacheMarkdownField
+  # Knows about the relationship between markdown and html field names, and
+  # stores the rendering contexts for the latter
+  class FieldData
+    extend Forwardable
+
+    def initialize
+      @data = {}
+    end
+
+    def_delegators :@data, :[], :[]=
+    def_delegator :@data, :keys, :markdown_fields
+
+    def html_field(markdown_field)
+      "#{markdown_field}_html"
+    end
+
+    def html_fields
+      markdown_fields.map {|field| html_field(field) }
+    end
+  end
+
+  # Dynamic registries don't really work in Rails as it's not guaranteed that
+  # every class will be loaded, so hardcode the list.
+  CACHING_CLASSES = %w[
+    AbuseReport
+    Appearance
+    ApplicationSetting
+    BroadcastMessage
+    Issue
+    Label
+    MergeRequest
+    Milestone
+    Namespace
+    Note
+    Project
+    Release
+    Snippet
+  ]
+
+  def self.caching_classes
+    CACHING_CLASSES.map(&:constantize)
+  end
+
+  extend ActiveSupport::Concern
+
+  included do
+    cattr_reader :cached_markdown_fields do
+      FieldData.new
+    end
+
+    # Returns the default Banzai render context for the cached markdown field.
+    def banzai_render_context(field)
+      raise ArgumentError.new("Unknown field: #{field.inspect}") unless
+        cached_markdown_fields.markdown_fields.include?(field)
+
+      # Always include a project key, or Banzai complains
+      project = self.project if self.respond_to?(:project)
+      context = cached_markdown_fields[field].merge(project: project)
+
+      # Banzai is less strict about authors, so don't always have an author key
+      context[:author] = self.author if self.respond_to?(:author)
+
+      context
+    end
+
+    # Allow callers to look up the cache field name, rather than hardcoding it
+    def markdown_cache_field_for(field)
+      raise ArgumentError.new("Unknown field: #{field}") unless
+        cached_markdown_fields.markdown_fields.include?(field)
+
+      cached_markdown_fields.html_field(field)
+    end
+
+    # Always exclude _html fields from attributes (including serialization).
+    # They contain unredacted HTML, which would be a security issue
+    alias_method :attributes_before_markdown_cache, :attributes
+    def attributes
+      attrs = attributes_before_markdown_cache
+
+      cached_markdown_fields.html_fields.each do |field|
+        attrs.delete(field)
+      end
+
+      attrs
+    end
+  end
+
+  class_methods do
+    private
+
+    # Specify that a field is markdown. Its rendered output will be cached in
+    # a corresponding _html field. Any custom rendering options may be provided
+    # as a context.
+    def cache_markdown_field(markdown_field, context = {})
+      raise "Add #{self} to CacheMarkdownField::CACHING_CLASSES" unless
+        CacheMarkdownField::CACHING_CLASSES.include?(self.to_s)
+
+      cached_markdown_fields[markdown_field] = context
+
+      html_field = cached_markdown_fields.html_field(markdown_field)
+      cache_method = "#{markdown_field}_cache_refresh".to_sym
+      invalidation_method = "#{html_field}_invalidated?".to_sym
+
+      define_method(cache_method) do
+        html = Banzai::Renderer.cacheless_render_field(self, markdown_field)
+        __send__("#{html_field}=", html)
+        true
+      end
+
+      # The HTML becomes invalid if any dependent fields change. For now, assume
+      # author and project invalidate the cache in all circumstances.
+      define_method(invalidation_method) do
+        changed_fields = changed_attributes.keys
+        invalidations = changed_fields & [markdown_field.to_s, "author", "project"]
+        !invalidations.empty?
+      end
+
+      before_save cache_method, if: invalidation_method
+    end
+  end
+end
diff --git a/app/models/concerns/issuable.rb b/app/models/concerns/issuable.rb
index ff465d2c745..c4b42ad82c7 100644
--- a/app/models/concerns/issuable.rb
+++ b/app/models/concerns/issuable.rb
@@ -6,6 +6,7 @@
 #
 module Issuable
   extend ActiveSupport::Concern
+  include CacheMarkdownField
   include Participable
   include Mentionable
   include Subscribable
@@ -13,6 +14,9 @@ module Issuable
   include Awardable
 
   included do
+    cache_markdown_field :title, pipeline: :single_line
+    cache_markdown_field :description
+
     belongs_to :author, class_name: "User"
     belongs_to :assignee, class_name: "User"
     belongs_to :updated_by, class_name: "User"
diff --git a/app/models/global_label.rb b/app/models/global_label.rb
index ddd4bad5c21..698a7bbd327 100644
--- a/app/models/global_label.rb
+++ b/app/models/global_label.rb
@@ -4,6 +4,10 @@ class GlobalLabel
 
   delegate :color, :description, to: :@first_label
 
+  def for_display
+    @first_label
+  end
+
   def self.build_collection(labels)
     labels = labels.group_by(&:title)
 
diff --git a/app/models/global_milestone.rb b/app/models/global_milestone.rb
index bda2b5c5d5d..cde4a568577 100644
--- a/app/models/global_milestone.rb
+++ b/app/models/global_milestone.rb
@@ -4,6 +4,10 @@ class GlobalMilestone
   attr_accessor :title, :milestones
   alias_attribute :name, :title
 
+  def for_display
+    @first_milestone
+  end
+
   def self.build_collection(milestones)
     milestones = milestones.group_by(&:title)
 
@@ -17,6 +21,7 @@ class GlobalMilestone
     @title = title
     @name = title
     @milestones = milestones
+    @first_milestone = milestones.find {|m| m.description.present? } || milestones.first
   end
 
   def safe_title
diff --git a/app/models/label.rb b/app/models/label.rb
index a23140b7d64..e8e12e2904e 100644
--- a/app/models/label.rb
+++ b/app/models/label.rb
@@ -1,4 +1,5 @@
 class Label < ActiveRecord::Base
+  include CacheMarkdownField
   include Referable
   include Subscribable
 
@@ -8,6 +9,8 @@ class Label < ActiveRecord::Base
   None = LabelStruct.new('No Label', 'No Label')
   Any = LabelStruct.new('Any Label', '')
 
+  cache_markdown_field :description, pipeline: :single_line
+
   DEFAULT_COLOR = '#428BCA'
 
   default_value_for :color, DEFAULT_COLOR
diff --git a/app/models/milestone.rb b/app/models/milestone.rb
index 44c3cbb2c73..23aecbfa3a6 100644
--- a/app/models/milestone.rb
+++ b/app/models/milestone.rb
@@ -6,12 +6,16 @@ class Milestone < ActiveRecord::Base
   Any = MilestoneStruct.new('Any Milestone', '', -1)
   Upcoming = MilestoneStruct.new('Upcoming', '#upcoming', -2)
 
+  include CacheMarkdownField
   include InternalId
   include Sortable
   include Referable
   include StripAttribute
   include Milestoneish
 
+  cache_markdown_field :title, pipeline: :single_line
+  cache_markdown_field :description
+
   belongs_to :project
   has_many :issues
   has_many :labels, -> { distinct.reorder('labels.title') },  through: :issues
diff --git a/app/models/namespace.rb b/app/models/namespace.rb
index 919b3b1f095..b7f2b2bbe61 100644
--- a/app/models/namespace.rb
+++ b/app/models/namespace.rb
@@ -1,9 +1,12 @@
 class Namespace < ActiveRecord::Base
   acts_as_paranoid
 
+  include CacheMarkdownField
   include Sortable
   include Gitlab::ShellAdapter
 
+  cache_markdown_field :description, pipeline: :description
+
   has_many :projects, dependent: :destroy
   belongs_to :owner, class_name: "User"
 
diff --git a/app/models/project.rb b/app/models/project.rb
index ecd742a17d5..88e4bd14860 100644
--- a/app/models/project.rb
+++ b/app/models/project.rb
@@ -6,6 +6,7 @@ class Project < ActiveRecord::Base
   include Gitlab::VisibilityLevel
   include Gitlab::CurrentSettings
   include AccessRequestable
+  include CacheMarkdownField
   include Referable
   include Sortable
   include AfterCommitQueue
@@ -17,6 +18,8 @@ class Project < ActiveRecord::Base
 
   UNKNOWN_IMPORT_URL = 'http://unknown.git'
 
+  cache_markdown_field :description, pipeline: :description
+
   delegate :feature_available?, :builds_enabled?, :wiki_enabled?, :merge_requests_enabled?, to: :project_feature, allow_nil: true
 
   default_value_for :archived, false
diff --git a/app/models/release.rb b/app/models/release.rb
index e196b84eb18..c936899799e 100644
--- a/app/models/release.rb
+++ b/app/models/release.rb
@@ -1,4 +1,8 @@
 class Release < ActiveRecord::Base
+  include CacheMarkdownField
+
+  cache_markdown_field :description
+
   belongs_to :project
 
   validates :description, :project, :tag, presence: true
diff --git a/app/models/snippet.rb b/app/models/snippet.rb
index 8a1730f3f36..2373b445009 100644
--- a/app/models/snippet.rb
+++ b/app/models/snippet.rb
@@ -1,11 +1,21 @@
 class Snippet < ActiveRecord::Base
   include Gitlab::VisibilityLevel
   include Linguist::BlobHelper
+  include CacheMarkdownField
   include Participable
   include Referable
   include Sortable
   include Awardable
 
+  cache_markdown_field :title, pipeline: :single_line
+  cache_markdown_field :content
+
+  # If file_name changes, it invalidates content
+  alias_method :default_content_html_invalidator, :content_html_invalidated?
+  def content_html_invalidated?
+    default_content_html_invalidator || file_name_changed?
+  end
+
   default_value_for :visibility_level, Snippet::PRIVATE
 
   belongs_to :author, class_name: 'User'
diff --git a/app/workers/clear_database_cache_worker.rb b/app/workers/clear_database_cache_worker.rb
new file mode 100644
index 00000000000..c541daba50e
--- /dev/null
+++ b/app/workers/clear_database_cache_worker.rb
@@ -0,0 +1,23 @@
+# This worker clears all cache fields in the database, working in batches.
+class ClearDatabaseCacheWorker
+  include Sidekiq::Worker
+
+  BATCH_SIZE = 1000
+
+  def perform
+    CacheMarkdownField.caching_classes.each do |kls|
+      fields = kls.cached_markdown_fields.html_fields
+      clear_cache_fields = fields.each_with_object({}) do |field, memo|
+        memo[field] = nil
+      end
+
+      Rails.logger.debug("Clearing Markdown cache for #{kls}: #{fields.inspect}")
+
+      kls.unscoped.in_batches(of: BATCH_SIZE) do |relation|
+        relation.update_all(clear_cache_fields)
+      end
+    end
+
+    nil
+  end
+end