Migrate events into a new format

This commit migrates events data in such a way that push events are
stored much more efficiently. This is done by creating a shadow table
called "events_for_migration", and a table called "push_event_payloads"
which is used for storing push data of push events. The background
migration in this commit will copy events from the "events" table into
the "events_for_migration" table, push events in will also have a row
created in "push_event_payloads".

This approach allows us to reclaim space in the next release by simply
swapping the "events" and "events_for_migration" tables, then dropping
the old events (now "events_for_migration") table.

The new table structure is also optimised for storage space, and does
not include the unused "title" column nor the "data" column (since this
data is moved to "push_event_payloads").

== Newly Created Events

Newly created events are inserted into both "events" and
"events_for_migration", both using the exact same primary key value. The
table "push_event_payloads" in turn has a foreign key to the _shadow_
table. This removes the need for recreating and validating the foreign
key after swapping the tables. Since the shadow table also has a foreign
key to "projects.id" we also don't have to worry about orphaned rows.

This approach however does require some additional storage as we're
duplicating a portion of the events data for at least 1 release. The
exact amount is hard to estimate, but for GitLab.com this is expected to
be between 10 and 20 GB at most. The background migration in this commit
deliberately does _not_ update the "events" table as doing so would put
a lot of pressure on PostgreSQL's auto vacuuming system.

== Supporting Both Old And New Events

Application code has also been adjusted to support push events using
both the old and new data formats. This is done by creating a PushEvent
class which extends the regular Event class. Using Rails' Single Table
Inheritance system we can ensure the right class is used for the right
data, which in this case is based on the value of `events.action`. To
support displaying old and new data at the same time the PushEvent class
re-defines a few methods of the Event class, falling back to their
original implementations for push events in the old format.

Once all existing events have been migrated the various push event
related methods can be removed from the Event model, and the calls to
`super` can be removed from the methods in the PushEvent model.

The UI and event atom feed have also been slightly changed to better
handle this new setup, fortunately only a few changes were necessary to
make this work.

== API Changes

The API only displays push data of events in the new format. Supporting
both formats in the API is a bit more difficult compared to the UI.
Since the old push data was not really well documented (apart from one
example that used an incorrect "action" nmae) I decided that supporting
both was not worth the effort, especially since events will be migrated
in a few days _and_ new events are created in the correct format.

9 files changed, 345 insertions, 23 deletions

diff --git a/app/models/event.rb b/app/models/event.rb
index 8d93a228494..a598eb08e82 100644
--- a/app/models/event.rb
+++ b/app/models/event.rb
@@ -48,6 +48,7 @@ class Event < ActiveRecord::Base
   belongs_to :author, class_name: "User"
   belongs_to :project
   belongs_to :target, polymorphic: true # rubocop:disable Cop/PolymorphicAssociations
+  has_one :push_event_payload, foreign_key: :event_id
 
   # For Hash only
   serialize :data # rubocop:disable Cop/ActiveRecordSerialize
@@ -55,6 +56,7 @@ class Event < ActiveRecord::Base
   # Callbacks
   after_create :reset_project_activity
   after_create :set_last_repository_updated_at, if: :push?
+  after_create :replicate_event_for_push_events_migration
 
   # Scopes
   scope :recent, -> { reorder(id: :desc) }
@@ -64,10 +66,36 @@ class Event < ActiveRecord::Base
     where(project_id: projects.pluck(:id)).recent
   end
 
-  scope :with_associations, -> { includes(:author, :project, project: :namespace).preload(:target) }
+  scope :with_associations, -> do
+    # We're using preload for "push_event_payload" as otherwise the association
+    # is not always available (depending on the query being built).
+    includes(:author, :project, project: :namespace)
+      .preload(:target, :push_event_payload)
+  end
+
   scope :for_milestone_id, ->(milestone_id) { where(target_type: "Milestone", target_id: milestone_id) }
 
+  self.inheritance_column = 'action'
+
   class << self
+    def find_sti_class(action)
+      if action.to_i == PUSHED
+        PushEvent
+      else
+        Event
+      end
+    end
+
+    def subclass_from_attributes(attrs)
+      # Without this Rails will keep calling this method on the returned class,
+      # resulting in an infinite loop.
+      return unless self == Event
+
+      action = attrs.with_indifferent_access[inheritance_column].to_i
+
+      PushEvent if action == PUSHED
+    end
+
     # Update Gitlab::ContributionsCalendar#activity_dates if this changes
     def contributions
       where("action = ? OR (target_type IN (?) AND action IN (?)) OR (target_type = ? AND action = ?)",
@@ -290,6 +318,16 @@ class Event < ActiveRecord::Base
     @commits ||= (data[:commits] || []).reverse
   end
 
+  def commit_title
+    commit = commits.last
+
+    commit[:message] if commit
+  end
+
+  def commit_id
+    commit_to || commit_from
+  end
+
   def commits_count
     data[:total_commits_count] || commits.count || 0
   end
@@ -385,6 +423,16 @@ class Event < ActiveRecord::Base
     user ? author_id == user.id : false
   end
 
+  # We're manually replicating data into the new table since database triggers
+  # are not dumped to db/schema.rb. This could mean that a new installation
+  # would not have the triggers in place, thus losing events data in GitLab
+  # 10.0.
+  def replicate_event_for_push_events_migration
+    new_attributes = attributes.with_indifferent_access.except(:title, :data)
+
+    EventForMigration.create!(new_attributes)
+  end
+
   private
 
   def recent_update?
diff --git a/app/models/event_for_migration.rb b/app/models/event_for_migration.rb
new file mode 100644
index 00000000000..a1672da5eec
--- /dev/null
+++ b/app/models/event_for_migration.rb
@@ -0,0 +1,5 @@
+# This model is used to replicate events between the old "events" table and the
+# new "events_for_migration" table that will replace "events" in GitLab 10.0.
+class EventForMigration < ActiveRecord::Base
+  self.table_name = 'events_for_migration'
+end
diff --git a/app/models/push_event.rb b/app/models/push_event.rb
new file mode 100644
index 00000000000..3f1ff979de6
--- /dev/null
+++ b/app/models/push_event.rb
@@ -0,0 +1,126 @@
+class PushEvent < Event
+  # This validation exists so we can't accidentally use PushEvent with a
+  # different "action" value.
+  validate :validate_push_action
+
+  # Authors are required as they're used to display who pushed data.
+  #
+  # We're just validating the presence of the ID here as foreign key constraints
+  # should ensure the ID points to a valid user.
+  validates :author_id, presence: true
+
+  # The project is required to build links to commits, commit ranges, etc.
+  #
+  # We're just validating the presence of the ID here as foreign key constraints
+  # should ensure the ID points to a valid project.
+  validates :project_id, presence: true
+
+  # The "data" field must not be set for push events since it's not used and a
+  # waste of space.
+  validates :data, absence: true
+
+  # These fields are also not used for push events, thus storing them would be a
+  # waste.
+  validates :target_id, absence: true
+  validates :target_type, absence: true
+
+  def self.sti_name
+    PUSHED
+  end
+
+  def push?
+    true
+  end
+
+  def push_with_commits?
+    !!(commit_from && commit_to)
+  end
+
+  def tag?
+    return super unless push_event_payload
+
+    push_event_payload.tag?
+  end
+
+  def branch?
+    return super unless push_event_payload
+
+    push_event_payload.branch?
+  end
+
+  def valid_push?
+    return super unless push_event_payload
+
+    push_event_payload.ref.present?
+  end
+
+  def new_ref?
+    return super unless push_event_payload
+
+    push_event_payload.created?
+  end
+
+  def rm_ref?
+    return super unless push_event_payload
+
+    push_event_payload.removed?
+  end
+
+  def commit_from
+    return super unless push_event_payload
+
+    push_event_payload.commit_from
+  end
+
+  def commit_to
+    return super unless push_event_payload
+
+    push_event_payload.commit_to
+  end
+
+  def ref_name
+    return super unless push_event_payload
+
+    push_event_payload.ref
+  end
+
+  def ref_type
+    return super unless push_event_payload
+
+    push_event_payload.ref_type
+  end
+
+  def branch_name
+    return super unless push_event_payload
+
+    ref_name
+  end
+
+  def tag_name
+    return super unless push_event_payload
+
+    ref_name
+  end
+
+  def commit_title
+    return super unless push_event_payload
+
+    push_event_payload.commit_title
+  end
+
+  def commit_id
+    commit_to || commit_from
+  end
+
+  def commits_count
+    return super unless push_event_payload
+
+    push_event_payload.commit_count
+  end
+
+  def validate_push_action
+    return if action == PUSHED
+
+    errors.add(:action, "the action #{action.inspect} is not valid")
+  end
+end
diff --git a/app/models/push_event_payload.rb b/app/models/push_event_payload.rb
new file mode 100644
index 00000000000..6cdb1cd4fe9
--- /dev/null
+++ b/app/models/push_event_payload.rb
@@ -0,0 +1,22 @@
+class PushEventPayload < ActiveRecord::Base
+  include ShaAttribute
+
+  belongs_to :event, inverse_of: :push_event_payload
+
+  validates :event_id, :commit_count, :action, :ref_type, presence: true
+  validates :commit_title, length: { maximum: 70 }
+
+  sha_attribute :commit_from
+  sha_attribute :commit_to
+
+  enum action: {
+    created: 0,
+    removed: 1,
+    pushed: 2
+  }
+
+  enum ref_type: {
+    branch: 0,
+    tag: 1
+  }
+end
diff --git a/app/services/event_create_service.rb b/app/services/event_create_service.rb
index 0f3a485a3fd..0b7e4f187f7 100644
--- a/app/services/event_create_service.rb
+++ b/app/services/event_create_service.rb
@@ -71,7 +71,14 @@ class EventCreateService
   end
 
   def push(project, current_user, push_data)
-    create_event(project, current_user, Event::PUSHED, data: push_data)
+    # We're using an explicit transaction here so that any errors that may occur
+    # when creating push payload data will result in the event creation being
+    # rolled back as well.
+    Event.transaction do
+      event = create_event(project, current_user, Event::PUSHED)
+
+      PushEventPayloadService.new(event, push_data).execute
+    end
 
     Users::ActivityService.new(current_user, 'push').execute
   end
diff --git a/app/services/push_event_payload_service.rb b/app/services/push_event_payload_service.rb
new file mode 100644
index 00000000000..b0a389c85f9
--- /dev/null
+++ b/app/services/push_event_payload_service.rb
@@ -0,0 +1,120 @@
+# Service class for creating push event payloads as stored in the
+# "push_event_payloads" table.
+#
+# Example:
+#
+#     data = Gitlab::DataBuilder::Push.build(...)
+#     event = Event.create(...)
+#
+#     PushEventPayloadService.new(event, data).execute
+class PushEventPayloadService
+  # event - The event this push payload belongs to.
+  # push_data - A Hash produced by `Gitlab::DataBuilder::Push.build` to use for
+  #             building the push payload.
+  def initialize(event, push_data)
+    @event = event
+    @push_data = push_data
+  end
+
+  # Creates and returns a new PushEventPayload row.
+  #
+  # This method will raise upon encountering validation errors.
+  #
+  # Returns an instance of PushEventPayload.
+  def execute
+    @event.build_push_event_payload(
+      commit_count: commit_count,
+      action: action,
+      ref_type: ref_type,
+      commit_from: commit_from_id,
+      commit_to: commit_to_id,
+      ref: trimmed_ref,
+      commit_title: commit_title,
+      event_id: @event.id
+    )
+
+    @event.push_event_payload.save!
+    @event.push_event_payload
+  end
+
+  # Returns the commit title to use.
+  #
+  # The commit title is limited to the first line and a maximum of 70
+  # characters.
+  def commit_title
+    commit = @push_data.fetch(:commits).last
+
+    return nil unless commit && commit[:message]
+
+    raw_msg = commit[:message]
+
+    # Find where the first line ends, without turning the entire message into an
+    # Array of lines (this is a waste of memory for large commit messages).
+    index = raw_msg.index("\n")
+    message = index ? raw_msg[0..index] : raw_msg
+
+    message.strip.truncate(70)
+  end
+
+  def commit_from_id
+    if create?
+      nil
+    else
+      revision_before
+    end
+  end
+
+  def commit_to_id
+    if remove?
+      nil
+    else
+      revision_after
+    end
+  end
+
+  def commit_count
+    @push_data.fetch(:total_commits_count)
+  end
+
+  def ref
+    @push_data.fetch(:ref)
+  end
+
+  def revision_before
+    @push_data.fetch(:before)
+  end
+
+  def revision_after
+    @push_data.fetch(:after)
+  end
+
+  def trimmed_ref
+    Gitlab::Git.ref_name(ref)
+  end
+
+  def create?
+    Gitlab::Git.blank_ref?(revision_before)
+  end
+
+  def remove?
+    Gitlab::Git.blank_ref?(revision_after)
+  end
+
+  def action
+    if create?
+      :created
+    elsif remove?
+      :removed
+    else
+      :pushed
+    end
+  end
+
+  def ref_type
+    if Gitlab::Git.tag_ref?(ref)
+      :tag
+    else
+      :branch
+    end
+  end
+end
diff --git a/app/views/events/_commit.html.haml b/app/views/events/_commit.html.haml
index ad434a64556..98cdcca3ecc 100644
--- a/app/views/events/_commit.html.haml
+++ b/app/views/events/_commit.html.haml
@@ -1,5 +1,5 @@
 %li.commit
   .commit-row-title
-    = link_to truncate_sha(commit[:id]), project_commit_path(project, commit[:id]), class: "commit-sha", alt: '', title: truncate_sha(commit[:id])
+    = link_to truncate_sha(event.commit_id), project_commit_path(project, event.commit_id), class: "commit-sha", alt: '', title: truncate_sha(event.commit_id)
     &middot;
-    = markdown event_commit_title(commit[:message]), project: project, pipeline: :single_line, author: event.author
+    = markdown event_commit_title(event.commit_title), project: project, pipeline: :single_line, author: event.author
diff --git a/app/views/events/_event_push.atom.haml b/app/views/events/_event_push.atom.haml
index 9fcacfbbf36..bf655f9d21a 100644
--- a/app/views/events/_event_push.atom.haml
+++ b/app/views/events/_event_push.atom.haml
@@ -1,14 +1,13 @@
 %div{ xmlns: "http://www.w3.org/1999/xhtml" }
-  - event.commits.first(15).each do |commit|
-    %p
-      %strong= commit[:author][:name]
-      = link_to "(##{truncate_sha(commit[:id])})", project_commit_path(event.project, id: commit[:id])
-      %i
-        at
-        = commit[:timestamp].to_time.to_s(:short)
-    %blockquote= markdown(escape_once(commit[:message]), pipeline: :atom, project: event.project, author: event.author)
-  - if event.commits_count > 15
+  %p
+    %strong= event.author_name
+    = link_to "(#{truncate_sha(event.commit_id)})", project_commit_path(event.project, event.commit_id)
+    %i
+      at
+      = event.created_at.to_s(:short)
+  %blockquote= markdown(escape_once(event.commit_title), pipeline: :atom, project: event.project, author: event.author)
+  - if event.commits_count > 1
     %p
       %i
         \... and
-        = pluralize(event.commits_count - 15, "more commit")
+        = pluralize(event.commits_count - 1, "more commit")
diff --git a/app/views/events/event/_push.html.haml b/app/views/events/event/_push.html.haml
index 54b414cc62a..973c652ad88 100644
--- a/app/views/events/event/_push.html.haml
+++ b/app/views/events/event/_push.html.haml
@@ -14,9 +14,7 @@
 - if event.push_with_commits?
   .event-body
     %ul.well-list.event_commits
-      - few_commits = event.commits[0...2]
-      - few_commits.each do |commit|
-        = render "events/commit", commit: commit, project: project, event: event
+      = render "events/commit", project: project, event: event
 
       - create_mr = event.new_ref? && create_mr_button?(project.default_branch, event.ref_name, project) && event.authored_by?(current_user)
       - if event.commits_count > 1
@@ -44,9 +42,6 @@
           = link_to create_mr_path(project.default_branch, event.ref_name, project) do
             Create Merge Request
 - elsif event.rm_ref?
-  - repository = project.repository
-  - last_commit = repository.commit(event.commit_from)
-  - if last_commit
-    .event-body
-      %ul.well-list.event_commits
-        = render "events/commit", commit: last_commit, project: project, event: event
+  .event-body
+    %ul.well-list.event_commits
+      = render "events/commit", project: project, event: event