Pig of the Sea

Seapig is a master-slave object sharing system, that:

uses websockets for communication
synchronizes remote states instead of passing messages
only sends JSON-diffs over websocket when possible
handles re-synchronization of objects on link loss/regain
handles objects' lifetime:
- orders object creation when listener appears
- caches objects in memory
- orders re-creation of objects when their dependencies change (e.g. immediately on db change)
- destroys objects when last listener disconnects
(re-)creates objects in separate processes
works nicely with Mithril
https://www.youtube.com/watch?v=_y4DbZivHCY

Deployment consists of:

central server - message router, doing diffs, re-sends, etc.
clients - can act as masters or slaves of seapig objects

A seapig object is a json hash, associated with an id and a version.

Related repos:

seapig-client-js - https://github.com/yunta-mb/seapig-client-js
seapig-client-ruby (also available as seapig-client gem) - https://github.com/yunta-mb/seapig-client-ruby
seapig-postgresql-notifier - https://github.com/yunta-mb/seapig-postgresql-notifier
seapig-router - https://github.com/yunta-mb/seapig-router

Basic use

Browser JS client (coffee example)

client = new SeapigClient('ws://'+window.location.host+'/seapig')
obj = client.slave('my-object-xxx')
obj.onchange -> console.log(obj.id, obj.version, obj.object)

Ruby client, in EM context:

client = SeapigClient.new('ws://'+window.location.host+'/seapig')
obj = client.slave('my-object-xxx')
obj.onchange { p obj.id, obj.version, obj }

Server side (with rails and postgres helpers):

For rails-less setup you’ll have to provide arguments to some executables (hint: --help).

add following gems to your bundle:

gem 'seapig-server'
gem 'seapig-postgresql-notifier'

add one db table:

rake seapig_engine:install:migrations
rake db:migrate

start seapig executables:

bundle exec seapig-server -v
bundle exec seapig-rails-notifier

in lib/seapigs create file that knows how to generate your my-object-xxx

require './config/environment.rb'

class ExecutionSingle < Producer

	@patterns = [ 'my-object-xxx' ]

	def self.produce(seapig_object_id)
		version = SeapigDependency.versions('list-of-bicycles') #arbitrary string
		data = { cars: Airplanes.all } # your generated object
		[data, version]
	end

end

start seapig worker (one or more, up to you):

bundle exec seapig-worker

in places where you modify airplanes table include this after transaction:

SeapigDependency.bump("list-of-bicycles") #same string as in producer

Debugging/introspection:

Seapig-client-js repo has a seapig-viewer.html that you can use to observe state of your seapig server (use with addresses like ws://localhost:3001).

There is also console tool called seapig-observer in seapig-client-ruby repo.

Advanced concepts:

Wildcard consumers

Following update callback will be called on updates of my-object-xxx, my-object-yyy and all other objects (1) with names matching the wildcard pattern.

client.slave('my-object-*').onupdate { |object| }

(1) all cached plus all (non-wildcard-)generateable objects

Wildcard producers

Following produce will be called to produce my-object-xxx, my-object-yyy and all other objects (2) with names matching the wildcard pattern.

class ExecutionSingle < Producer

	@patterns = [ 'my-object-*' ]

	def self.produce(seapig_object_id)
		...
	end

end

(2) all objects having non-wildcard consumers

Versions and dependencies

When producing an object, you can give it version as one of:

a number (integer or float)
an array of numbers (integers or floats)
a hash of object_id: version

If you go for hash, seapig server will interpret it as dependencies declaration. An object with version declared as { "object-b": 12 } will be automatically re-produced when version of object-b increases above 12.

Client-side master objects

There are 2 ways to do it (don’t mix them!):

you always keep your object up-to-date, and pig lib handles communication and events

client = SeapigClient.new(URL)
obj = client.master('my-object-xxx', object: { "initial" => "content" })

obj["aaa"] = 1234
obj.bump() # increases object version and notifies server

you register onproduce callback and provide new object state and version every time it’s called

client = SeapigClient.new(URL)
obj = client.master('my-object-xxx')
obj.onproduce {
	obj.set(object: { "zzz" => "yyy" }, version: 10)
}

Current limitations:

there is no rate-limiting on object re-creation
server is a single-process (in future release all json operations and client communication will be done in separate processes)
there is no safety/security (clients can easily crash the server, or get any data they want)
object identifiers are strings, leading to linear complexities (will probably be class(string) + descriptor(hash) in future releases)

seapig-server

Runtime