Kupo is a Javascript-based web development framework.

It's not meant as a general purpose framework but to support web applications driven by clientside javascript. It acts primarily as an adapter to an object database that also provides RPC functionality.

The Kupo software package is not meant for productive use. Right now it is just a proof-of-concept implementation supporting my diploma thesis.

Kupo should run on all Unix systems. It was tested with MacOS X 10.5.7 and Debian 5.0. To use Kupo you need to have a running MongoDB Server >= 0.9.6 on your machine. (http://www.mongodb.org). You also should have git installed.

To install Kupo, just place the source directory somewhere and make sure the correct versions of the Jack and Narwhal frameworks are in place under the packages directory. The best way to install Kupo is to clone the repository using git:

1. git clone git://github.com/janv/kupo.git
2. cd kupo
3. git submodule update --init

After having installed Kupo, run it using the ./start script

The Kupo framework lives in an environment created by the Rhino JavaScript interpreter and the Narwhal/Jack framework. This environment influences how the files are laid out and how certain tasks are achieved.

The Kupo repository is laid out as a Narwhal package. The root directory contains the application that is developed (a sample application at the moment), all required frameworks are located as packages in the packages directory, so they're automatically discovered by narwhal. These packages are Jack, Narwhal itself and the Kupo framework.

The file main.js is executed by narwhal to run the package. Main.js runs the jackup executable of the Jack framework, passing it the current directory as a Jack application. Jack applications are initialized using the file jackconfig.js in which a the app (a simple function adhering to the Jack protocol) is exported in the exports.app variable.

Narwhal implements the SecurableModules specification devised in the ServerJS working group. This standard defines a mechanism for requiring JavaScript source files (modules) in a secure way that keeps private identifiers private and prevents accidential pollution of the global namespace.

To load a file, the function require is used. Require provides the exports variable to the module, runs the file and returns the contents of exports. In order to export functionality, modules can simply manipulate this variable. All other identifiers in the file are not exported and unaccessible from outside.

The app defined in jackconfig.js is simply a delegation to Kupo's dispatcher which acts inside the current app by using the global variable $KUPO_HOME to locate files and sources.

The dispatcher's handle method adheres to the Jack protocol and acts as the app for Jack to run in a Kupo application. Everything the dispatcher needs to know to control the application is the $KUPO_HOME global variable pointing to the root of the application.

The handle method is called for every incoming request. Inside, the requested URL is analyzed. Based on the first part of the path, a model or controller is looked up (via the hasController/hasModel methods), instantiated (via the requestInstance method in the controller) and used to handle the request.

First, the Dispatcher looks for a controller with the given name. If such a controller does not exist, a model is looked up and handed to the generic ResourceController. Looking up and loading controllers and models is done in the Fetcher (and encapsulated in the dispatcher's fetch/hasModel/Controller methods)

The fetcher contains methods to locate the files in the current Kupo application, load them and make them available to the dispatcher. Its main two methods are fetch and check, both are encapsulated by four corresponding methods which are exported. The check method checks for the presence of a model/controller, the fetch method loads and returns them.

An application's controllers are located in app/controller/<controllername>.js. A controller is a module in the SecurableModules-sense. To declare a controller, the module has to define it using CustomController.define('<controllername>'). The resulting object has to be exported as a property of the exports object by the name <Controllername>Controller (eg. FooController, the controller name has to be capitalized).

To define actions on the created controller, the developer should create functions in the controller's actions property. Inside these functions, this.request, this.cookies and this.session are available to access details of the request. The action should return a Jack response array ([<status>, <headers-object>, <array of strings for the body>]).

The request object is created by Jack and defined in packages/jack/lib/jack/request.js. The cookies object is created in Kupo's controller.js. The Controller object defined in this file is a common ancestor for both the ResourceController and the CustomControllers. It contains a handle method which is called by the dispatcher to handle the request. In handle, the request, cookies and session objects are set up before the current controller instance's process method is called. To interact with the cookie or the session object, simply add/remove/change its properties. They're serialized into the HTTP response automatically.

Controller.js further exports the JRPCRequest constructor for JSON-RPC Requests. The two methods JRPCRequest.fromGET and JRPCRequest.fromPOST are the ones actually used to contruct JRPCRequests. JRPCRequest objects contain four methods:

• getMethodName() returns the name of the method that should be called
• getParameters() returns the parameters for the call
• getNamedParameters() returns the parameters for the call as a JavaScript object
• call() is used on an object to call the requested method on the object, passing the provided parameters

JRPCRequest provides two additional helper methods:

• buildResponse takes a HTTP status code and a result object and constructs a Jack response array containing a JRPC reponse object.
• buildError works analogous and constructs a JRPC error response.

If the Dispatcher doesn't find a controller to handle a request, it looks for a model with a matching name. If one is found, an instance of the ResourceController is instantiated with it.

The ResourceController has a few standard ways to deal with requests. Its process method analyzes the request to determine wether to treat the request as a JSON-RPC request or as a simple GET request. The GET requests are processed by the controller's index and show actions. First, it is made sure that the all/find method of the model are callable, then they are called and their result is sent to the client as a JRPCResponse. The actual method call is surrounded by before/after filters defined on the model. They are executed in the context of the controller by the model's callBack method, so they have the opportunity to manipulate the request (via this.request) or the response (via this.collection/this.object).

In all other cases, the ResourceController's process method constructs a JRPCRequest object from the request and processes it in the processJRPC method.

The models Kupo operates with are defined in app/model/<modelname>.js. Their declaration is similar to to the controllers. To define a model, Model.define is called, passing the model's name and an object (called the specialization object internally) that describes the model. The specialization object contains four properties:

• instance is an object containing several properties in itself:
  • methods is an object containing instance methods for instances of this model
  • callables is an array of method names that can be called remotely on the instance
• callables is an array of method names that can be called remotely on the model
• callbacks is an object containing several properties having certain names. Each of these properties contains an array of functions that are executed at points in the instances lifecycle that is provided by the property's name. All possible execution points are given in model.js
• validations: an array of functions that are used to validate the instance
• associations: an object defining associations this model has to other models

Only one aspect of a model is defined outside of this specialization object and that's class methods. Those are simply created as properties on the defined model.

The ClassPrototype is defined in model.js and exported as Model. As its name indicates, the ClassPrototype is a prototype for all model classes. It is derived into concrete models via the define method. Define clones the prototype, processes the specialization object and opens a connection to this model's database collection. This connection is stored in a closure and can be accessed by calling the collection() method on the model.

The initSpecialization method processes the specialization object to create the model's instancePrototype. This instancePrototype is derived from the CommonInstancePrototype and acts as a prototype for all instances of a model. Most aspects in the specialization object however aren't actually evaluated at this point but looked up later during the model's lifecycle.

The all, find, makeNew and create method in the ClassPrototype are involved in aspects of object persistence and database connectivity. All of them are described in model.js.

The CommonInstanceProtoype is similar to the ClassPrototype. It contains several initialization methods and persistence methods. All are described in model.js. Here only some aspects will be listed that are missing from the source documentation.

The save method, responsible for saving changes to an instance to the database shows nicely how saving is performed depending on the instance's state. The state property of the instance can take one of four values:

• New objects that haven't been saved to the database are 'new'.
• Objects that represent their (last known, concurrency issues aside) state in the database are 'clean'.
• Objects that have been changed since their last save or load operation are marked 'dirty'.
• Finally if an object is removed from the database, it's marked 'removed' and can't be saved anymore.

Depending on which state the instance was in, the callbacks are called on it.

Although an instance's underlying data is openly accessible in its data property, it should not be changed there. Instead the CommonInstancePrototype provides get and set methods to access the data. Using set ensures that the instance's state does not become corrupted. Likewise, the state property should never be overwritten manually. If an instance has to be marked dirty for saving, the taint method should be used.

Validations are described in the DocComment for CommonInstancePrototype.validate. Whenever an object isn't valid, it can't be saved to the database anymore.

The last method of the CommonInstancePrototype is newInstance. Once an instancePrototype has been derived from the CommonInstancePrototype, newInstance creates actual model instances. Notice that this method should never be called manually, it is only used internally by CommonInstancePrototype.makeNew.

Associations provide automated contruction and use of relationships between models. To declare an association, create a property with the name of the association in the associations object inside the specialization object (examples are available in the test suite in packages/kupo/tests/association-tests.js). The value of that property is an association descriptor as returned by Association.belongs_to/belongs_to_many/has_one/has_many. Each of these associations is declared in a file in /packages/kupo/lib/kupo/model/associations/ and consists of two parts.

The Association Proxy is an object that is accessed under the assication name on the final model instance (eg. task.user) and contains various methods to manipulate the association.

The association generator returns an association descriptor object containing two methods:
registerCallbacks is called when the instancePrototype for a model is derived from the CommonInstancePrototype. It installs a callback in the instancePrototype that in turn executes a callback method in the association proxy of the actual instance.
installProxy is called whenever an instance of a model is created. It initializes an AssociationProxy for the association and installs it in the instance.

The MongoAdapter provides access to a MongoDB database by wrapping MongoDB's Java driver in Javascript using Rhino's Java interoperability.

MongoAdapter.Connection can be used to create a new database connection. A default database connection is provided by MongoAdapter.getConnection(). This method always returns the same connection, which by default connects to the kupo database. To change the default connection call MongoAdapter.setConnection.

Calling getCollection on a Connetion returns a new Mongo collection. These are roughly corresponding to tables in SQL databases, so one collection is used for every model in the app. Collections are never created explicitly, they are simply accessed and the first write operation creates them.

For collection methods that return muliple items, a wrapper around the Mongo Cursor has been written. It provides basic means to iterate over the items. Mongo Cursors can not be rewound.

Whenever a single MongoDB object leaves the MongoAdapter, it is converted to a JavaScript object using the fromDoc method at the end of mongo_adapter.js. Similarly Javascript objects are converted to Mongo BasicDBObjects before being processed by the Java driver's methods using the createDoc/convert functions.

Kupo has tests. To run them you need to:

1. Start mongod using ./mongod --dbpath /path/to/kupo/db run
2. Change to /path/to/kupo/packages/kupo/tests
3. Run ../../narwhal/bin/narwhal .

The tests are covering the persistence and model aspects of Kupo, mainly everything in mongo_adapter.js and model.js.

Kupo is licensed under the MIT License

Copyright (c) 2009 Jan Varwig

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.