Product Generation Process

Basically, product generation is a two step process (focusing on the Python side of things for now):

digraph foo {
    graph [bgcolor="#F8F8F8"];
    node [fontsize=10, shape=box3d];

    "Product Context" [shape=note];
    "Feature Equation" [shape=note];
    "Django web application" [shape=ellipse];

    "Product Context" -> "Context Binding";
    "Context Binding" -> "Feature Composition";
    "Feature Equation" -> "Feature Composition";
    "Feature Composition" -> "Django web application";

Figure: High-Level overview of the product generation process

Product Context
The product context contains the specific configuration of the product e.g. the database configuration and the hostname the product will be serving requests from. The product context is given in JSON format. Features may provide a wishlist of necessary configuration values.
Context Binding
the product context is loaded from file and made available as django_productline.context.PRODUCT_CONTEXT. The product context is considered to be read only — so it may not be written to.
Feature Equation
The list of features selected for the product in the order of composition. The feature equation is given as text file containing one feature per line.
Feature Composition
After the product context has been bound, featuremonkey is used to compose the selected features. This results in a running django web application where introductions and refinements given by the selected features have been patched in.

The product context

the product context captures environment and configuration settings that are specific to each product, e.g. each product requires a different database configuration.

Use the context only for very specific stuff that NEEDS to be configured on a product basis.

The context is loaded from a file in json format.

Composition of application code

featuremonkey is used to compose Python code. It allows introductions of new structures and refinements of existing ones.

For some use cases in the context of django-productline, see Refinements by example.

Also, see the featuremonkey documentation.

Template composition

You can use django-overextends for feature oriented template development. It is automatically installed as a dependency of django-productline.

By default, Django uses the app directories template loader to locate templates. It searches the templates folder of each app in the order that the apps are specified in INSTALLED_APPS. The loader picks the first template with matching name.

Conceptually, this is a form of feature oriented composer with file level replacements:

  • apps represent features
  • INSTALLED_APPS defines the feature selection and their composition order

On top of that, django-overextends provides overextension — template block level refinements.


Consider, we have a template called mytemplate.html in the template directory of a django app called myfeature:


Suppose mytemplate.html looks like this:

        <title>{% block title %}Hello{% endblock %}</title>
        {% block body %}
        {% endblock %}

Django templates already provide blocks, that are used for template inheritance

django-overextends provides template superimposition using the overextends tag: To refine mytemplate.html, all we need to do is to create another template with that name in a django app that is placed before myfeature in INSTALLED_APPS:

{% overextends "mytemplate.html" %}

{% block title %}Replacement{% endblock %}

{% block body %}
{{ block.super }}
Refinements are also possible!
{% endblock %}

Block tags are used to annotate FST-Nodes. Since blocks can be nested, we can build feature structure trees. Nodes with the same name are superimposed, when the template is rendered. {{ block.super }} provides access to the original implementation.

Rendering the above example, would result roughly in the following HTML document:

        Refinements are also possible!

Javascript Composition

If necessary, JavaScript can be composed using featuremonkey.js. Essentially, it works the same way as featuremonkey.

Have a look at the example product line and feel free to snoop around by viewing the source in your browser.

CSS Composition

feature oriented CSS is easy: concatenation is a pretty good composition mechanism for it.

Task Composition

django-productline relies on ape tasks. Features may introduce new tasks and refine existing ones by providing a tasks module.

Please see the ape tasks documentation for details.

Tasks contributed by django-productline are listed in Available tasks.