pirate

Argh! What's This?

Pirate provides a simple key-value storage interface with adapters for different storage systems. Pirate currently supports MongoDB, Elasticsearch, Redis and in-memory storage.

Benefits

The benefits of this approach are:

• Simplify your code. The Pareto Principle often applies to storage systems, where you only need 20% of the features 80% of the time. Pirate optimizes that 80% while still allowing you to extend adapters to handle the other 20%, specific to your requirements.

• Eliminate the impedance mismatch between HTTP and storage. Pirate follows a similar interface to that supported by HTTP: get, put, patch, and delete. There's no equivalent to post and there are a few additional methods, but semantically, they're very close.

• Easily switch between storage implementations. Pirate's adapters not only hide the complexity of the underlying storage implementation, they make it much easier to change it. You can prototype using an in-memory solution, then use a database and later partition your data across servers.

• Make use of powerful event-based interfaces. Node-style callbacks provide a reasonable least-common-denominator, but for more sophisticated applications, they can be tedious. Pirate uses a library called Mutual to provide a simple event-based interface. Each method returns an events object to which event handlers can be attached. Events "bubble up" (think DOM) so that error-handling no longer needs to be done local to the call.

Example

Here's a simple program to put and get and object from MongoDB.

{log} = console
{Mongo} = require "pirate"
{EventChannel} = require "mutual"

# Create the top-level events object
events = new EventChannel

# Default error handler just logs the error
events.on "error", (error) -> log error

# Create an adapter, passing in the events object
Mongo.Adapter.make
  events: events
  port: 27018
  host: "127.0.0.1"
  database: "foo"
  options:
    auto_reconnect: true

# When the adapter is ready, we can do stuff
events.on "ready", (adapter) ->

  # First, let's add a second event handler to 
  # close the connection and exit
  events.on "error", ->
    adapter.close
    process.exit -1

  # Okay, let's get the collection we're going to use
  (adapter.collection "bar")

  # Once we have the collection, let's put something
  .on "success", (collection) ->
    (collection.put "baz", baz: "hello")

    # If the put works, try getting the same thing back out
    .on "success", (object) ->
      (collection.get "baz")

      # If the get works, show the result and close the 
      # adapter because we're done!
      .on "success", (object) ->
        log object
        adapter.close()

Adapter API

The elements of the interface are:

• get key Returns the object associated with the key or null.

• put key, object Overwrites the object associated with key with object. Returns the updated object.

• delete key Deletes the object associated with key. Returns nothing.

• patch key, patch Updates the object associated with key by overlaying patch. Returns the updated object.

• all Returns all the objects in the collection.

• count Returns a count of all the objects in the collection.

All API methods actually return an EventChannel object. The success event is how a value is ultimately returned, if necessary.