cubicweb logo

Table Of Contents

Previous topic

1.1.2. Discovering the web interface

Next topic

1.1.4. What’s next?

This Page

1.1.3. Customizing your application

So far so good. The point is that usually, you won’t get enough by assembling cubes out-of-the-box. You will want to customize them, have a personal look and feel, add your own data model and so on. Or maybe start from scratch?

So let’s get a bit deeper and start coding our own cube. In our case, we want to customize the blog we created to add more features to it. Create your own cube

First, notice that if you’ve installed CubicWeb using Debian packages, you will need the additional cubicweb-dev package to get the commands necessary to CubicWeb development. All cubicweb-ctl commands are described in details in cubicweb-ctl tool.

Once your CubicWeb development environment is set up, you can create a new cube:

cubicweb-ctl newcube myblog

This will create in the cubes directory (/path/to/forest/cubes for source installation, /usr/share/cubicweb/cubes for Debian packages installation) a directory named blog reflecting the structure described in Standard structure for a cube.

For packages installation, you can still create new cubes in your home directory using the following configuration. Let’s say you want to develop your new cubes in ~src/cubes, then set the following environment variables:


and then create your new cube using:

cubicweb-ctl newcube --directory=~/src/cubes myblog


We previously used myblog as the name of our instance. We’re now creating a cube with the same name. Both are different things. We’ll now try to specify when we talk about one or another, but keep in mind this difference. Cube metadata

A simple set of metadata about your cube are stored in the file. In our case, we want to extend the blog cube, so we have to tell that our cube depends on this cube, by modifying the __depends__ dictionary in that file:

__depends__ =  {'cubicweb': '>= 3.10.7',
                'cubicweb-blog': None}

where the None means we do not depends on a particular version of the cube. Extending the data model

The data model or schema is the core of your CubicWeb application. It defines the type of content your application will handle. It is defined in the file of the cube. Defining our model

For the sake of example, let’s say we want a new entity type named Community with a name, a description. A Community will hold several blogs.

from yams.buildobjs import EntityType, RelationDefinition, String, RichString

class Community(EntityType):
    name = String(maxsize=50, required=True)
    description = RichString()

class community_blog(RelationDefinition):
    subject = 'Community'
    object = 'Blog'
    cardinality = '*?'
    composite = 'subject'

The first step is the import from the yams package necessary classes to build the schema.

This file defines the following:

  • a Community has a title and a description as attributes
    • the name is a string that is required and can’t be longer than 50 characters
    • the description is a string that is not constrained and may contains rich content such as HTML or Restructured text.
  • a Community may be linked to a Blog using the community_blog relation
    • * means a community may be linked to 0 to N blog, ? means a blog may be linked to 0 to 1 community. For completeness, remember that you can also use + for 1 to N, and 1 for single, mandatory relation (e.g. one to one);
    • this is a composite relation where Community (e.g. the subject of the relation) is the composite. That means that if you delete a community, its blog will be deleted as well.

Of course, there are a lot of other data types and things such as constraints, permissions, etc, that may be defined in the schema, but those won’t be covered in this tutorial.

Notice that our schema refers to the Blog entity type which is not defined here. But we know this type is available since we depend on the blog cube which is defining it. Applying changes to the model into our instance

Now the problem is that we created an instance using the blog cube, not our myblog cube, so if we don’t do anything there is no way that we’ll see anything changing in the instance.

One easy way, as we’ve no really valuable data in the instance would be to trash and recreated it:

cubicweb-ctl stop myblog # or Ctrl-C in the terminal running the server in debug mode
cubicweb-ctl delete myblog
cubicweb-ctl create myblog
cubicweb-ctl start -D myblog

Another way is to add our cube to the instance using the cubicweb-ctl shell facility. It’s a python shell connected to the instance with some special commands available to manipulate it (the same as you’ll have in migration scripts, which are not covered in this tutorial). In that case, we’re interested in the add_cube command:

$ cubicweb-ctl stop myblog # or Ctrl-C in the terminal running the server in debug mode
$ cubicweb-ctl shell myblog
entering the migration python shell
just type migration commands or arbitrary python code and type ENTER to execute it
type "exit" or Ctrl-D to quit the shell and resume operation
>>> add_cube('myblog')
$ cubicweb-ctl start -D myblog

The add_cube command is enough since it automatically updates our application to the cube’s schema. There are plenty of other migration commands of a more finer grain. They are described in Migration

As explained, leave the shell by typing Ctrl-D. If you restart the instance and take another look at the schema, you’ll see that changes to the data model have actually been applied (meaning database schema updates and all necessary stuff has been done).

the instance schema after adding our cube

If you follow the ‘info’ link in the user pop-up menu, you’ll also see that the instance is using blog and myblog cubes.

the instance schema after adding our cube

You can now add some communities, link them to blog, etc... You’ll see that the framework provides default views for this entity type (we have not yet defined any view for it!), and also that the blog primary view will show the community it’s linked to if any. All this thanks to the model driven interface provided by the framework.

You’ll then be able to redefine each of them according to your needs and preferences. We’ll now see how to do such thing. Defining your views

CubicWeb provides a lot of standard views in directory cubicweb/web/views/. We already talked about ‘primary’ and ‘list’ views, which are views which apply to one ore more entities.

A view is defined by a python class which includes:

  • an identifier: all objects used to build the user interface in CubicWeb are recorded in a registry and this identifier will be used as a key in that registry. There may be multiple views for the same identifier.
  • a selector, which is a kind of filter telling how well a view suit to a particular context. When looking for a particular view (e.g. given an identifier), CubicWeb computes for each available view with that identifier a score which is returned by the selector. Then the view with the highest score is used. The standard library of predicates is in cubicweb.predicates.

A view has a set of methods inherited from the cubicweb.view.View class, though you usually don’t derive directly from this class but from one of its more specific child class.

Last but not least, CubicWeb provides a set of default views accepting any kind of entities.

Want a proof? Create a community as you’ve already done for other entity types through the index page, you’ll then see something like that:

the default primary view for our community entity type

If you notice the weird messages that appear in the page: those are messages generated for the new data model, which have no translation yet. To fix that, we’ll have to use dedicated cubicweb-ctl commands:

You’ll then be able to redefine each of them according to your needs and preferences. So let’s see how to do such thing. Changing the layout of the application

The layout is the general organization of the pages in the site. Views that generate the layout are sometimes referred to as ‘templates’. They are implemented in the framework in the module cubicweb.web.views.basetemplates. By overriding classes in this module, you can customize whatever part you wish of the default layout.

But notice that CubicWeb provides many other ways to customize the interface, thanks to actions and components (which you can individually (de)activate, control their location, customize their look...) as well as “simple” CSS customization. You should first try to achieve your goal using such fine grained parametrization rather then overriding a whole template, which usually embeds customisation access points that you may loose in the process.

But for the sake of example, let’s say we want to change the generic page footer... We can simply add to the module views of our cube, e.g. cubes/myblog/, the code below:

from cubicweb.web.views import basetemplates

class MyHTMLPageFooter(basetemplates.HTMLPageFooter):

    def footer_content(self):
        self.w(u'This website has been created with <a href="">CubicWeb</a>.')

def registration_callback(vreg):
    vreg.register_all(globals().values(), __name__, (MyHTMLPageFooter,))
    vreg.register_and_replace(MyHTMLPageFooter, basetemplates.HTMLPageFooter)
  • Our class inherits from the default page footer to ease getting things right, but this is not mandatory.
  • When we want to write something to the output stream, we simply call self.w, with must be passed an unicode string.
  • The latest function is the most exotic stuff. The point is that without it, you would get an error at display time because the framework wouldn’t be able to choose which footer to use between HTMLPageFooter and MyHTMLPageFooter, since both have the same selector, hence the same score... In this case, we want our footer to replace the default one, so we have to define a registration_callback() function to control object registration: the first instruction tells to register everything in the module but the MyHTMLPageFooter class, then the second to register it instead of HTMLPageFooter. Without this function, everything in the module is registered blindly.


When a view is modified while running in debug mode, it is not required to restart the instance server. Save the Python file and reload the page in your web browser to view the changes.

We will now have this simple footer on every page of the site. Primary view customization

The ‘primary’ view (i.e. any view with the identifier set to ‘primary’) is the one used to display all the information about a single entity. The standard primary view is one of the most sophisticated views of all. It has several customisation points, but its power comes with uicfg, allowing you to control it without having to subclass it.

However this is a bit off-topic for this first tutorial. Let’s say we simply want a custom primary view for my Community entity type, using directly the view interface without trying to benefit from the default implementation (you should do that though if you’re rewriting reusable cubes; everything is described in more details in The Primary View).

So... Some code! That we’ll put again in the module views of our cube.

from cubicweb.predicates import is_instance
from cubicweb.web.views import primary

class CommunityPrimaryView(primary.PrimaryView):
    __select__ = is_instance('Community')

    def cell_call(self, row, col):
        entity = self.cw_rset.get_entity(row, col)
        self.w(u'<h1>Welcome to the "%s" community</h1>' % entity.printable_value('name'))
        if entity.description:
            self.w(u'<p>%s</p>' % entity.printable_value('description'))

What’s going on here?

  • Our class inherits from the default primary view, here mainly to get the correct view identifier, since we don’t use any of its features.
  • We set on it a selector telling that it only applies when trying to display some entity of the Community type. This is enough to get an higher score than the default view for entities of this type.
  • View applying to entities usually have to define cell_call as entry point, and are given row and col arguments tell to which entity in the result set the view is applied. We can then get this entity from the result set (self.cw_rset) by using the get_entity method.
  • To ease thing, we access our entity’s attribute for display using its printable_value method, which will handle formatting and escaping when necessary. As you can see, you can also access attributes by their name on the entity to get the raw value.

You can now reload the page of the community we just created and see the changes.

the custom primary view for our community entity type

We’ve seen here a lot of thing you’ll have to deal with to write views in CubicWeb. The good news is that this is almost everything that is used to build higher level layers.


As things get complicated and the volume of code in your cube increases, you can of course still split your views module into a python package with subpackages.

You can find more details about views and selectors in Principles. Write entities to add logic in your data

CubicWeb provides an ORM to easily programmaticaly manipulate entities (just like the one we have fetched earlier by calling get_entity on a result set). By default, entity types are instances of the AnyEntity class, which holds a set of predefined methods as well as property automatically generated for attributes/relations of the type it represents.

You can redefine each entity to provide additional methods or whatever you want to help you write your application. Customizing an entity requires that your entity:

  • inherits from cubicweb.entities.AnyEntity or any subclass
  • defines a __regid__ linked to the corresponding data type of your schema

You may then want to add your own methods, override default implementation of some method, etc...

from cubicweb.entities import AnyEntity, fetch_config

class Community(AnyEntity):
    """customized class for Community entities"""
    __regid__ = 'Community'

    fetch_attrs, cw_fetch_order = fetch_config(['name'])

    def dc_title(self):

    def display_cw_logo(self):
        return 'CubicWeb' in self.description

In this example:

  • we used convenience fetch_config() function to tell which attributes should be prefetched by the ORM when looking for some related entities of this type, and how they should be ordered
  • we overrode the standard dc_title method, used in various place in the interface to display the entity (though in this case the default implementation would have had the same result)
  • we implemented here a method display_cw_logo() which tests if the blog entry title contains ‘CW’. It can then be used when you’re writing code involving ‘Community’ entities in your views, hooks, etc. For instance, you can modify your previous views as follows:
class CommunityPrimaryView(primary.PrimaryView):
    __select__ = is_instance('Community')

    def cell_call(self, row, col):
        entity = self.cw_rset.get_entity(row, col)
        self.w(u'<h1>Welcome to the "%s" community</h1>' % entity.printable_value('name'))
        if entity.display_cw_logo():
            self.w(u'<img src=""/>')
        if entity.description:
            self.w(u'<p>%s</p>' % entity.printable_value('description'))

Then each community whose description contains ‘CW’ is shown with the CubicWeb logo in front of it.


As for view, you don’t have to restart your instance when modifying some entity classes while your server is running in debug mode, the code will be automatically reloaded. Extending the application by using more cubes!

One of the goal of the CubicWeb framework was to have truly reusable components. To do so, they must both behave nicely when plugged into the application and be easily customisable, from the data model to the user interface. And I think the result is pretty successful, thanks to system such as the selection mechanism and the choice to write views as python code which allows to build our page using true object oriented programming techniques, that no template language provides.

A library of standard cubes is available from CubicWeb Forge, to address a lot of common concerns such has manipulating people, files, things to do, etc. In our community blog case, we could be interested for instance in functionalities provided by the comment and tag cubes. The former provides threaded discussion functionalities, the latter a simple tag mechanism to classify content. Let’s say we want to try those. We will first modify our cube’s file:

__depends__ =  {'cubicweb': '>= 3.10.7',
                'cubicweb-blog': None,
                'cubicweb-comment': None,
                'cubicweb-tag': None}

Now, we’ll simply tell on which entity types we want to activate the ‘comment’ and ‘tag’ facilities by adding respectively the ‘comments’ and ‘tags’ relations on them in our schema (

class comments(RelationDefinition):
    subject = 'Comment'
    object = 'BlogEntry'
    cardinality = '1*'
    composite = 'object'

class tags(RelationDefinition):
    subject = 'Tag'
    object = ('Community', 'BlogEntry')

So in the case above we activated comments on BlogEntry entities and tags on both Community and BlogEntry. Various views from both comment and tag cubes will then be automatically displayed when one of those relations is supported.

Let’s synchronize the data model as we’ve done earlier:

$ cubicweb-ctl stop myblog
$ cubicweb-ctl shell myblog
entering the migration python shell
just type migration commands or arbitrary python code and type ENTER to execute it
type "exit" or Ctrl-D to quit the shell and resume operation
>>> add_cubes(('comment', 'tag'))

Then restart the instance. Let’s look at a blog entry:

the primary view for a blog entry with comments and tags activated

As you can see, we now have a box displaying tags and a section proposing to add a comment and displaying existing one below the post. All this without changing anything in our views, thanks to the design of generic views provided by the framework. Though if we take a look at a community, we won’t see the tags box! That’s because by default this box try to locate itself in the left column within the white frame, and this column is handled by the primary view we hijacked. Let’s change our view to make it more extensible, by keeping both our custom rendering but also extension points provided by the default implementation.

class CommunityPrimaryView(primary.PrimaryView):
    __select__ = is_instance('Community')

    def render_entity_title(self, entity):
        self.w(u'<h1>Welcome to the "%s" community</h1>' % entity.printable_value('name'))

    def render_entity_attributes(self, entity):
        if entity.display_cw_logo():
            self.w(u'<img src=""/>')
        if entity.description:
            self.w(u'<p>%s</p>' % entity.printable_value('description'))

It appears now properly:

the custom primary view for a community entry with tags activated

You can control part of the interface independently from each others, piece by piece. Really.