avalanche 0.3.0

Testability

Avalanche was designed in a way that it makes it possible (easier for you) to write good unit-tests for your code. That is not only making it easy to write tests. A unit-test should:

• give a clear error message when it fails

• fail only when the feature under test is broken, not on every code change

• be fast

Reusability

Mostly every framework claims that some kind of reusability is one of their design goals. Avalanche “reusability” goals means source-code reusability.

Many frameworks provide some mechanisms for reusable/plugable sub-applications however it is not always easy to re-use these applications source code in case you need to configure/modify it. Plugable applications is also a very important feature but as of now Avalanche has no support for that.

It should not only be possible to write reusable code, the code should be reusable on the first time you write it. You should not be advised to write the code in one way, and than later have to modify it to make it reusable. I.e. it is opposed to saying “Use view (handler) functions”. And than… “if you want your views to be re-usable convert them to class-based views!”.

Of course Avalanche does not do miracles. Testability and reusability will ultimately depends on the application code. But the framework have a big role on setting up the right path.

The goals are presented below, hopefully after reading the design and tutorial it will be clear to you how these goals are achieved.

beyond MVC (model-view-controller)

MVC is a software architectural pattern created with the goal to isolate “domain logic” from the user interface. This separation of concern enables the creation of better application code. This pattern was very successful for many desktop frameworks and so served as a reference to the creation of web-frameworks. The problem is that this architecture can not be mapped directly to the way web-applications work.

Even the so-called MVC frameworks are not really MVC. So let’s just keep the MVC’s goal. That is to write clean, re-usable and testable code.

web applications

Essentially all a web-application do is to receive a HTTP request, process it and generate a HTTP response.

                    +------------------+
HTTP Request ------>| web application  + ----->  HTTP Response
                    +------------------+

Sending and receiving HTTP is handled by a web-server, so we don’t need to worry about that. Let’s take a closer look into how the web application generates a HTTP response:

                  +------+      +-------+
HTTP request ---->|router|----->|handler|----> HTTP response
                  +------+      +-------+

The router will analyze the request and dispatch it to a request handler that will create the response.

request handlers styles

There are mainly 3 styles of request handlers.

• a single function

• a class method

• a class

Avalanche uses the third style, a class. Using a class as request handler suits better our goals because it provides a greater flexibility, easier to modify/extend and re-use parts of the handler.

request handler processing

The request handler processing can be divided in 3 stages:

             +-----------------+                        +------------+                  +----------+
request ---->| param converter |---- param objects ---->| processing |--- context ----->| renderer |----> response
             +-----------------+                        +------------+                  +----------+

1. param converter - get parameters from HTTP request

HTTP is a text protocol, the application will typically get some parameters from the request and convert string values into some native data types. These parameters are taken from the URI path, URI query, post-data, cookies, etc.

This is where you will write some glue code to map HTTP into your application code.

2. processing / context builder

Processing is the application logic. It will often access a persistence layer (sometimes called Model, but this is entirely up to the application code and the web-framework has no role on that). It might also trigger other operations like doing some batch processing or accessing third-party applications. Note that typically this application logic is independent of HTTP.

This processing may also consist of simple data retrieval. In this case the processing may generate a context. Context is a term used to represent the data that will be used by a renderer to create a HTTP response.

A web page is often composed of several elements so sometimes it makes sense to divide the work into more than one “context builder”.

3. renderer - generate output

The renderer will convert the result of the application processing into text to create a HTTP response. The renderer will typically use a template system to generate HTML code or convert the data to JSON.

On avalanche you should write code for the 3 stages of the handler separately and let the framework put the different parts together.

Move on to the tutorial to see how it looks like.