django-recommend 0.1.dev7

Generate recommendations in Django.

Use pyrecommend in Django projects.

Warning: not yet suitable for production.

django_recommend/ is the actual Django app intended for use in projects.

simplerec/ is a Django project used for testing.

Important note: Model classes/database schemas are only supported if they have unique integer primary keys.

Quickstart

• Add django_recommend to your INSTALLED_APPS, and run python manage.py migrate.

• Set SESSION_SAVE_EVERY_REQUEST=True in your settings, to ensure anonymous users can be tracked.

• In your views, call django_recommend.set_score(request, object, score) to start recording user scores. (Currently this is assumed to be implicit feedback.) Note: This will use session keys to store scores for users who aren’t authenticated.

• In your templates, use {% load django_recommend %} and {{ obj|similar_objects }} to show similar objects to visitors. This filter returns a list, so you may also do, for example:

  {% load django_recommend %}
  {% with similar_products as product|similar_objects %}
      {% if similiar_products %}
          <h2>Other users also liked:</h2>
          <ul>
          {% for product in similar_products %}
              <li><a href="{{ product.get_absolute_url }}">{{ product }}</a></li>
          {% endfor %}
          </ul>
      {% endif %}
  {% endwith %}

Notes

There can be some potentially confusing behavior from this library if your application involves objects that get deleted/deactivated. Currently I am not sure how feasible it is to filter by a flag like ‘deactivated’; tests for that kind of functionality will be put into a future version.

If your data is ever deleted, by default, django_recommend will attempt to delete all recommendation information about it at the next available chance. This most notably can occur in ObjectSimilarityQueryset.get_instances_for, where you may, for example, have a queryset like:

suggested_books = similar_books.order_by(
    '-score')[:5].get_instances_for(curr_book)

If one of the books that would be suggested for this book has since been deleted, the default behavior will be to in fact give you a list of four instances. This is because retrieving the deleted instances raises an ObjectDoesNotExist error in the Django ORM; the default behavior is to simply delete that object’s information from the recommendation data and skip the object for now.

This can be controlled by the RECOMMEND_PURGE_MISSING_DATA boolean setting. If this is False, the exception will be propagated, so you may handle it in a different way.