particle
v0.1.4
Published
A library for distributed state synchronization.
Downloads
152
Readme
Particle
This is a library for distributed state synchronization. Clients can register with a server, and their local data models will be kept up to date as the data on the server changes. To begin with we are focused on Mongo as the data source, but in the future Redis and other data sources should be supported.
Currently the client side data model is read only. This is not enforced, but if you change it your data will now be out of sync with the server. Instead you should use an out of band method such as RPC or REST to initiate changes on the server, and let them trickle back down.
In the future, it is possible we will support a more direct strategy for updating the data model. This is a complex problem to solve because you need to handle conflict resolution on shared data. Ultimately the server must have final say in what operations are allowed.
Credit/Inspiration
This was mainly inspired by personal experience of trying to build a chat server based on message passing semantics. We came to the conclusion that the tool we were using did not allow us to describe the problem at the right level of abstraction. We don't really care how the data gets there, and we don't want to miss any data, which is a danger with message passing. We need to always have the current state of the data, and we need to be notified of changes.
For an elaboration on these ideas, check out this excellent article:
http://awelonblue.wordpress.com/2012/07/01/why-not-events/
To understand the role that event driven systems play in a business architecture:
http://martinfowler.com/bliki/CQRS.html
Ultimately I see technologies like Particle enabling us to build real time awareness systems which can give modern businesses a huge competitive advantage. These technologies can also be used to build systems that feel responsive and natural, and take a more active role in serving our needs.
Another good resource on the subject is the book Event Processing: Designing IT Systems for Agile Companies.
Here's an interview with the author.
Install
npm install particle
component install TorchlightSoftware/particle
Collector (Client)
particle = require 'particle'
collector = new particle.Collector
identity:
sessionId: 'foo'
# I should recieve delta events
collector.on '**', (data, event) ->
collector.register (err) ->
Stream (Server)
particle = require 'particle'
MongoWatch = require 'mongo-watch'
watcher = new MongoWatch {format: 'normal'}
stream = new particle.Stream
#onDebug: console.log
adapter: watcher
# Identity Lookup, performed once upon registration.
# identityLookup: (identity, done) ->
# done null, {accountId: 1}
# Cache Config, used to alias many-to-many lookups or otherwise force caching of fields
cacheConfig:
userstuffs:
userId: 'users._id'
stuffId: 'stuffs._id'
# Data Sources, the data each connected client will have access to. Any fields used in
# the criteria will be automatically cached.
dataSources:
myProfile:
collection: 'users' # the source collection (in mongo or other adapter)
criteria: {_id: '@userId'} # limit which records come back
manifest: true # limit which fields come back
myStuff:
collection: 'stuffs'
criteria: {_id: '@userId|users._id>userstuffs._id>stuffs._id'}
manifest: true
visibleUsers:
collection: 'users'
criteria: {accountId: '@accountId'}
manifest:
name: true
_id: true
notFound:
collection: 'users'
criteria: {notFound: true}
manifest: true
allUsers:
collection: 'users'
criteria: undefined
manifest: true
Configuring Your Particle Stream
When you create a Stream, which is the server component to Particle, you pass it a configuration. The Stream and its configuration reside server side for security purposes, so you never send out any data that you don't want the user to have access to.
The fields below are required unless you see a * after the name.
Data Sources
Each data source corresponds to a root property on the client side data model (collector.data). The data source configuration controls where data will come from and how it will be filtered.
- Collection
The MongoDB collection you want to monitor.
- Criteria
The recordset you want to pull back. This generally looks like a Mongo query, except it supports looking up data in the user's identity and Particle's internal cache. For instance:
criteria: {_id: '@userId|users._id>userstuffs._id>stuffs._id'}
This means "Grab the 'userId' field from the identity, and pipe it through a relationship to find any related stuffs._id's."
You can also supply a regular string or number:
criteria: {name: 'Joe'}
Or a lookup by string:
criteria: {colorId: 'red|colors.name>colors._id'}
Or just an identity:
criteria: {colorId: '@myColorId'}
- Manifest
The manifest controls what fields get pushed out to a Collector. It acts like a Mongo 'project' query modifier.
You can assign 'true' at any point in the data structure. For instance, if you put 'true' at the root then no data will be filtered.
Identity Lookup*
This is the first step taken whenever a new Collector registers. You will be passed the identity that the Collector is claiming, and you have the opportunity to look up additional data and associate it with this Collector, or you can return an error in which case the registration will fail and the Collector will be notified. Any additional data that you look up will be stored server side and will not be passed back to the Collector. However, it will be passed to the Payload and Delta functions, so you can use it in filtering the data that a Collector receives.
onDebug*
The 'onDebug' value can be set to a logging function, for instance console.log. Whenever Stream receives data from a data source, or sends data to a Collector, your function will be notified. Other important events are sent as well.
This is really useful if you are trying to troubleshoot a configuration to find out why data is not being transmitted.
disconnect*
Each Particle Stream instance exposes a disconnect method which can be used to shut it down. In addition to its internal behavior, you can specify your own behavior, such as shutting down any watchers you established.
Designing Your Data Model
By combining field limiting from the manifest, and the record limiting ability of the payload and delta functions, you can achieve precise control over what data a Collector has access to.
For instance, if a particular Data Source represents the 'current user', you can pull their userID from the Identity argument and query for only the record that matches. Another Data Source might be coming from the same database collection, but retrieve ALL the users' records with a more restrictive Manifest.
Collector Configuration
The collector's configuration is pretty minimal. It doesn't need to know about what data it will be managing, it just needs your credentials and the location of the Stream it's connecting to. Here are the fields you can use:
Identity
A hash object containing any data you wish to use for authentication and access control. This could be a sessionId which you got from logging in via some out-of-band channel. For instance, RPC or basic auth. Particle does not handle authentication - this allows you to integrate it with your existing strategy.
Listening to Change Events
We support path based listeners with wildcards courtesy of EventEmitter2:
collector.on 'users.**', (data, event) ->
collector.on 'users.address.*', (data, event) ->
A single '*' will match one path section, a '**' will match multiple. The data and event structures look like this:
data:
users: [
id: 5
email: '[email protected]'
address:
state: 'Lake Maylie'
zip: '07542'
]
event:
root: 'users'
timestamp: 'Fri May 24 2013 11:44:25 GMT-0700 (MST)'
operation: 'set'
id: 5
path: 'address.state'
data: 'Lake Maylie'
This should be sufficient for basic UI updates. We are working on an API for proxy models which will make it easier to do data bindings with a library such as rivets.
Register*
This is a low level function which should not normally be used. See the debugging section below for details. In its absence the Collector will try to connect to its Stream using websockets (which is usually what you want).
The Particle Lifecycle
When a Stream is created, it stores the configuration and waits for Collectors to register. For each Collector that registers it does the following:
- Verifies the Collector's identity
- Sends a manifest
- Sends payloads for all data sources
- Wires up listeners for deltas
When a Collector is created, it immediately tries to register with a Stream based on its configuration. It sends the Stream its identity and waits for a success/failure. The operation succeeds if the Identity lookup is successful (or none is defined), and then the Collector waits for its initial data.
The Collector starts out with a 'status' property set to 'waiting'. It expects to receive:
- A manifest
- A payload for each data source listed in the manifest
Once it has received the manifest and data for all sources, its status will change to 'ready'. In addition the collector has a function called 'ready' which takes a callback and will either call it immediately if the status is 'ready', or will call it once the status changes to 'ready'. As of this writing, there are no other statuses defined.
The function you provide to collector.on '**' will receive (data, event). Data is the entire data root, and event is a description of the change which occurred. The event format follows the specification for a Delta (see Message Format in next section).
The Particle Message Format
Three types of messages are passed between the Particle Stream and Collector:
- Manifest
- Payload
- Delta
These all have formats which I feel are general enough to apply to many different data sources. They were modeled around MongoDB however, so some adjustment may need to be made for supporting other types of DBs. I am working on formalizing these message formats using JSON Schema, and will make available the documentation as well as the schemas for use in other third party libraries that interact with Particle.
For now, please refer to the tests and also keep an eye on this document. Let me know if you have questions.
Debugging
I tried to make the interface a balance of clean and robust. If you run into trouble and want to find out what's going on inside, the following features will be helpful.
onDebug
This is supported on both Collector and Stream instances. When you pass a logging function such as console.log, you will be notified of relevant events in the Particle lifecycle. Turning this on for most production and even development environments is not recommended, as the data volume will be huge.
onRegister
This function is present as a configuration option on the Collector, and there is a corresponding fully implemented 'register' function exposed by the Stream.
We'll talk about the Collector first. You could add a custom function here for the purpose of:
- debugging the Collector
- connecting to a custom event source
- mocking out the data model
The last case is particularly useful - if you mock out the client side data model you can allow front end developers to continue work unimpeded by the back end implementation.
The Stream's 'register' function accepts the following arguments:
- identity
- a receiver method (messageName, event) - call this whenever you want to send the Collector data
- a callback (err) - to be called when registration has completed (or failed)
It is responsible for establishing a communication channel with a particular Collector. See Particle Message Format above. You can call this function manually in order to mimic a Collector registration.
LICENSE
(MIT License)
Copyright (c) 2013 Torchlight Software [email protected]
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.