Activr
Activr is the Ruby gem created by Fotonauts to manage activity feeds on Fotopedia.
With Activr you can create:
- a Global Activity Feed to display all activities in your website in a single feed
- a User Activity Feed to display all actions performed by a specific user
- a User News Feeds so that each user can get news from friends they follow, from albums they own or follow, etc.
- an Album Activity Feed to display what happens in a specific album
- ...
Activities are stored in a MongoDB database.
Some magic is invoked when running inside a Rails application but Activr can be used without Rails.
If Resque is detected in a Rails application then it is automatically used to run some parts of Activr code asynchronously.
A demo app is available on heroku, feel free to create an account and try it. Demo source code is on github too.
- More information on our tumblr
- Source code on github
- Code documentation on rubydoc
Install
$ [sudo] gem install activr
In Rails, add it to your Gemfile:
gem 'activr'
Quick start
Define an activity
An activity is an event that is (most of the time) performed by a user in your application.
When defining an activity you specify allowed entities and a humanization template.
Let's generate our first activity that will be dispatched when a user adds a picture to an album:
$ rails g activr:activity add_picture actor:User picture album
The file app/activities/add_picture_activity.rb
is created:
class AddPictureActivity < Activr::Activity
entity :actor, :class => User, :humanize => :fullname
entity :picture, :humanize => :title
entity :album, :humanize => :name
humanize "{{{actor}}} add picture {{{picture}}} {{{album}}}"
end
Entities
An entity represents one of your application models that is involved in the activity.
By convention, the entity that corresponds to the user performing the action should be named :actor
.
The entity class is inferred thanks to entity name, so by default the :picture
entity has the Picture
class, but you can still provide the :class
option to specify another class.
Activity humanization
The humanize
method defines a sentence that describes the activity and it is a Mustache template. Let's change the generated sentence by a better one:
humanize "{{{actor}}} added picture {{{picture}}} to the album {{{album}}}"
The :humanize
option on entity corresponds to a method that is called on the entity instance to humanize it. Note that the generator tries to find that method by itself.
Usage
Here is an example of activity instanciation and humanization:
user = User.create!({ :_id => 'john', :first_name => "John", :last_name => "WILLIAMS"})
picture = Picture.create!({ :_id => 'my_face', :title => "My Face"})
album = Album.create!({ :name => "My Selfies"})
activity = AddPictureActivity.new(:actor => user, :picture => picture, :album => album)
activity.humanize
# => John WILLIAMS added picture My Face to the album My Selfies
activity.humanize(:html => true)
# => <a href="/users/john">John WILLIAMS</a> added picture <a href="/pictures/my_face">My Face</a> to the album <a href="/albums/5295bc9261796d649f080000">My Selfies</a>
Dispatch an activity
You can now dispatch this activity in your application when a picture is added to an album:
class Album
include Mongoid::Document
field :name, :type => String
has_and_belongs_to_many :pictures, :class_name => "Picture", :inverse_of => :albums
def add_picture(picture, user)
unless self.pictures.include?(picture)
self.pictures << picture
# dispatch activity
Activr.dispatch!(AddPictureActivity.new(:actor => user, :picture => picture, :album => self))
end
end
end
Once dispatched the activity is stored in the activities
MongoDB collection:
> db.activities.findOne()
{
"_id" : ObjectId("5295bc9f61796d649f140000"),
"at" : ISODate("2013-11-27T09:34:23.850Z"),
"kind" : "add_picture",
"actor" : "john",
"picture" : "my_face",
"album" : ObjectId("5295bc9261796d649f080000")
}
Basic activity feeds
Several basic activity feeds are now available:
- the global feed: all activities in your application
- per entity feed
Global activity feed
Use Activr#activities
to fetch the latest activities in your application:
puts "There are #{Activr.activities_count} activites. Here are the 10 most recent:"
activities = Activr.activities(10)
activities.each do |activity|
puts activity.humanize
end
Note that you can paginate thanks to the :skip
option of the #activities
method.
Entity activity feed
Each entity involved in an activity can have its own activity feed.
To activate entity activity feed, include the mixin Activr::Entity::ModelMixin
into the corresponding model class, and setup the :feed_index
option:
include Activr::Entity::ModelMixin
activr_entity :feed_index => true
Then launch the task that setup indexes on the activities
collection:
$ rake activr:create_indexes
Example: actor activity feed
To fetch actor activities, include the mixin Activr::Entity::ModelMixin
into your actor class:
class User
# inject sugar methods
include Activr::Entity::ModelMixin
activr_entity :feed_index => true
include Mongoid::Document
field :_id, :type => String
field :first_name, :type => String
field :last_name, :type => String
def fullname
"#{self.first_name} #{self.last_name}"
end
end
Now the User
class has two new methods: #activities
and #activities_count
:
user = User.find('john')
puts "#{user.fullname} have #{user.activities_count} activites. Here are the 10 most recent:"
user.activities(10).each do |activity|
puts activity.humanize
end
Example: album activity feed
You can also fetch a per-album activity feed by including the mixin Activr::Entity::ModelMixin
into the Album
class:
class Album
# inject sugar methods
include Activr::Entity::ModelMixin
activr_entity :feed_index => true
# ...
end
Example:
album = Album.find(BSON::ObjectId.from_string('5295bc9261796d649f080000'))
puts "There are #{album.activities_count} activites in the album #{album.name}. Here are the 10 most recent:"
album.activities(10).each do |activity|
puts activity.humanize
end
News Feed
Now we want a User News Feed, so that each user can get news from friends he follows and from albums he owns or follows. That is the goal of a timeline: to create a complex activity feed.
Timeline
Let's generate a timeline class:
$ rails g activr:timeline user_news_feed User
The file app/timelines/user_news_feed_timeline.rb
is created:
class UserNewsFeedTimeline < Activr::Timeline
recipient User
#
# Routes
#
# route FollowBuddyActivity, :to => 'buddy', :humanize => "{{{actor}}} is now following you"
#
# Callbacks
#
# def self.should_route_activity?(activity)
# # return `false` to cancel activity routing
# true
# end
# def should_handle_activity?(activity, route)
# # return `false` to skip routed activity
# true
# end
# def should_store_timeline_entry?(timeline_entry)
# # return `false` to cancel timeline entry storing
# true
# end
# def will_store_timeline_entry(timeline_entry)
# # this is your last chance to modify timeline entry before it is stored
# end
# def did_store_timeline_entry(timeline_entry)
# # the timeline entry was stored, you can now do some post-processing
# end
end
When defining a Timeline
class you specify:
- what model in your application owns that timeline: the
recipient
- which activities are displayed in that timeline: the
routes
Routes
Routes describe which activities must be stored in the timeline and how to resolve recipients for those activities.
When an activity is dispatched, Activr tries to resolve all routes of every timeline with that activity. The result of a route resolving must be either an array of recipient instances/ids or a unique recipient instance/id.
Let's add some routes:
class UserNewsFeedTimeline < Activr::Timeline
recipient User
# this is a predefined routing, to fetch all followers of an activity actor
routing :actor_follower, :to => Proc.new{ |activity| activity.actor.followers }
# define a routing with a class method, to fetch all followers of an activity album
def self.album_follower(activity)
activity.album.followers
end
#
# Routes
#
# activity path: users will see in their news feed when someone adds a picture in one of their albums
route AddPictureActivity, :to => 'album.owner'
# predefined routing: users will see in their news feed when a friend they follow likes a picture
route AddPictureActivity, :using => :actor_follower
# method call: users will see in their news feed when someone adds a picture in an album they follow
route AddPictureActivity, :using => :album_follower
# ...
end
As you can see there are several ways to define a route:
Route with an activity path
# activity path: users will see in their news feed when someone adds a picture in one of their albums
route AddPictureActivity, :to => 'album.owner'
The path is specified with the :to
route setting. It describes a method chaining to call on dispatched activities.
So with our example the route is resolved that way:
album = activity.album
recipient = album.owner
Route with a predefined routing
First, declare a predefined routing
:
# this is a predefined routing, to fetch all followers of an activity actor
routing :actor_follower, :to => Proc.new{ |activity| activity.actor.followers }
Then use it with the :using
route setting:
# predefined routing: users will see in their news feed when a friend they follow likes a picture
route AddPictureActivity, :using => :actor_follower
Note that you can also use a block syntax:
routing :actor_follower do |activity|
activity.actor.followers
end
Route with a call on timeline class method
You can resolve a route with a timeline class method:
# define a routing with a class method, to fetch all followers of an activity album
def self.album_follower(activity)
activity.album.followers
end
Then use it with the :using
route setting:
# method call: users will see in their news feed when someone adds a picture in an album they follow
route AddPictureActivity, :using => :album_follower
Preferred route syntax
For the sake of demonstration you can see the three ways in previous timeline code example, but when a route is simple to resolve it is preferred to use an activity path like that:
class UserNewsFeedTimeline < Activr::Timeline
recipient User
#
# Routes
#
# activity path: users will see in their news feed when someone adds a picture in one of their albums
route AddPictureActivity, :to => 'album.owner'
# predefined routing: users will see in their news feed when a friend they follow likes a picture
route AddPictureActivity, :to => 'actor.followers'
# method call: users will see in their news feed when someone adds a picture in an album they follow
route AddPictureActivity, :to => 'album.followers'
# ...
end
Timeline Entry
When an activity is routed to a timeline, that activity is copied to a Timeline Entry that is then stored into database (so Activr uses a Fanout on write mecanism to dispatch activities to timelines).
A routed timeline entry is stored in the <timeline kind>_timelines
MongoDB collection.
For example, Corinne received the previously generated activity because John added a picture to an album she owns:
> db.user_news_feed_timelines.findOne({"rcpt": "corinne"})
{
"_id" : ObjectId("5295c06b61796d673b010000"),
"rcpt" : "corinne",
"routing" : "album_owner",
"activity" : {
"_id" : ObjectId("5295bc9f61796d649f140000"),
"at" : ISODate("2013-11-27T09:34:23.850Z"),
"kind" : "add_picture",
"actor" : "john",
"picture" : "my_face",
"album" : ObjectId("5295bc9261796d649f080000")
}
}
As you can see, a Timeline Entry contains:
- a copy of the original activity
- the recipient id:
rcpt
- the
routing
kind: here,album_owner
means that Corinne received that activity in her News Feed because she is the owner of the album
You can also add meta data. For example you may add a read
meta data if you want to implement a read/unread mecanism in your News Feed.
When you create a new timeline class don't forget to launch the task that setup indexes in the corresponding timelines
collection:
$ rake activr:create_indexes
Timeline Entry humanization
Specify a :humanize
setting on a route
to specialize humanization of corresponding timeline entries. For example:
# activity path: users will see in their news feed when someone adds a picture in one of their albums
route AddPictureActivity, :to => 'album.owner', :humanize => "{{{actor}}} added a picture to your album {{{album}}}"
If you do not set a :humanize
setting then the humanization of the embedded activity is used instead.
Callbacks
Several callbacks are invoked on timeline instance so you can hook your own code during the activity dispatching workflow:
class UserNewsFeedTimeline < Activr::Timeline
# ...
#
# Callbacks
#
def self.should_route_activity?(activity)
# if you return `false` then nobody will receive that activity for that timeline class
true
end
def should_handle_activity?(activity, route)
# if you return `false` then current recipient won't receive that routed activity
true
end
def should_store_timeline_entry?(timeline_entry)
# if you return `false` then current recipient won't receive that timeline entry
true
end
def will_store_timeline_entry(timeline_entry)
# this is your last chance to modify timeline entry before it is stored
end
def did_store_timeline_entry(timeline_entry)
# the timeline entry was stored, you can now do some post-processing
# for example you can send notifications
end
end
Fetching / Display
Two methods are injected in the timeline recipient class: #<timeline_kind>
and #<timeline_kind>_count
. So in our case: #user_news_feed
and #user_news_feed_count
:
class UsersController < ApplicationController
def news_feed
user = User.find(params[:id])
@news_feed = user.user_news_feed(10)
@news_feed_count = user.user_news_feed_count
end
end
Here is simple view:
<p>
You have <%= @news_feed_count %> entries in your News Feed. Here are the 10 most recent:
</p>
<ul id='news_feed'>
<% @news_feed.each do |timeline_entry| %>
<li><%= raw timeline_entry.humanize(:html => true) %></li>
<% end %>
</ul>
<% end %>
Here is a view taken from Activr Demo:
<div id='news_feed'>
<% @news_feed.each do |timeline_entry| %>
<% activity = timeline_entry.activity %>
<div class="activity <%= activity.kind %>">
<div class="icon">
<%= link_to(image_tag(activity.actor.avatar.thumb.url, :title => activity.actor.fullname), activity.actor) %>
</div>
<div class="content">
<div class="title"><%= timeline_entry.humanize(:html => true).html_safe %></div>
<% if activity.buddy %>
<div class="buddy">
<%= link_to(image_tag(activity.buddy.avatar.url, :title => activity.buddy.fullname), activity.buddy) %>
</div>
<% elsif activity.picture %>
<div class="picture">
<%= link_to(image_tag(activity.picture.image.small.url, :title => activity.picture.title), activity.picture) %>
</div>
<% elsif activity.album %>
<div class="album">
<%= link_to(image_tag(activity.album.cover.image.small.url, :title => activity.album.name), activity.album) %>
</div>
<% end %>
<small class="date text-muted"><%= distance_of_time_in_words_to_now(activity.at, :include_seconds => true) %> ago</small>
</div>
</div>
<% end %>
</div>
Entity model deletion
When one of your entities models instance is deleted you should probably call the delete_activities!
method. This method deletes all activities that refer to the deleted entity from the activities
and timelines
collections.
You should too add activr_entity :deletable => true
to your model class to ensure that a deletion index is correctly setup when running the rake activr:create_indexes
task.
class Picture
include Activr::Entity::ModelMixin
# picture can be deleted
activr_entity :deletable => true
include Mongoid::Document
# ...
# delete all activities
after_destroy :delete_activities!
end
Async
You can plug a job system to run some parts of Activr code asynchronously.
Possible hooks are:
-
:route_activity
- Activity is routed by the dispatcher -
:timeline_handle
- Activity is handled by a timeline
For example, here is the default :route_activity
hook handler that is provided out of the box when Resque is detected in a Rails application:
# config
Activr.configure do |config|
config.async[:route_activity] ||= Activr::Async::Resque::RouteActivity
end
class Activr::Async::Resque::RouteActivity
@queue = 'activr_route_activity'
class << self
def enqueue(activity)
::Resque.enqueue(self, activity.to_hash)
end
def perform(activity_hash)
# unserialize argument
activity_hash = Activr::Activity.unserialize_hash(activity_hash)
activity = Activr::Activity.from_hash(activity_hash)
# call hook
Activr::Async.route_activity(activity)
end
end # class << self
end # class RouteActivity
A hook class:
- implements an
#enqueue
method, used to enqueue the async job - calls
Activr::Async.<hook_name>
method in the async job
Hook classes are specified thanks to the config.async
hash.
If you are writing a Rails application you just need to add the Resque
gem to your Gemfile
to enable async hooks. If you want to use another job system then you have to write your own async hook handlers. If you want to force disabling of async hooks, for example when deploying your app on Heroku with only one dyno, just set the environment variable ACTIVR_FORCE_SYNC
to true
.
Railties
The default mongodb connection uri is mongodb://127.0.0.1/activr
, but if you are using Activr inside a Rails application with mongoid gem loaded then the mongoid database connection will be used instead. If you don't want that behaviour then set the environment variable ACTIVR_SKIP_MONGOID_RAILTIE
to true
, or set the Fwissr key /activr/skip_mongoid_railtie
to true.
Skip duplicates activities
Use the :skip_dup_period
option when dispatching an activity to avoid duplicates.
# User is now following Buddy
activity = FollowBuddyActivity.new(:actor => user, :buddy => followee)
# skip activity if User already followed Buddy during the last hour
Activr.dispatch!(activity, :skip_dup_period => 3600)
Or you can set that option in global activr configuration:
Activr.config.skip_dup_period = 3600
Trim Timelines
Set max_length
on a timeline class to specify the maximum number of timeline entries allowed. When a recipient timeline exceed that number then old timeline entries are automatically deleted.
class UserNewsFeedTimeline < Activr::Timeline
recipient User
max_length 100
# ...
end
Todo
- Activities aggregation in timelines
- Rails generator to setup basic views
- Rails generator to setup admin controllers
- Permits "Fanout on read" for inactive entities, to preserve db size
- Permits "Fanout on write with buckets", for maximum read perfs
References
- http://blog.mongodb.org/post/65612078649/using-mongodb-schema-design-to-create-inboxes
- http://www.slideshare.net/danmckinley/etsy-activity-feeds-architecture
Credits
From Fotonauts:
- Aymerick Jéhanne @aymerick
Copyright (c) 2013-2014 Fotonauts released under the MIT license.