BlueBream The Web Component Framework

9.8. Configuration

9.8.1. Introduction

Developing components alone does not make a framework. There must be some configuration utility that tells the system how the components work together to create the application server framework. This is done using the Zope Configuration Markup Language (ZCML) for all filesystem-based code. Therefore it is very important that a developer knows how to use ZCML to hook up his/her components to the framework.

As stated above, it became necessary to develop a method to setup and configure the components that make up the application server. While it might seem otherwise, it is not that easy to develop an effective configuration system, since there are several requirements that must be satisfied. Over time the following high-level requirements developed that caused revisions of the implementation and coding styles to be created:

  1. While the developer is certainly the one that writes the initial cut of the configuration, this user is not the real target audience. Once the product is written, you would expect a system administrator to interact much more frequently with the configuration, adding and removing functionality or adjust the configuration of the server setup. System administrators are often not developers, so that it would be unfortunate to write the configuration in the programming language, here Python. But an administrator is familiar with configuration scripts, shell code and XML to some extend. Therefore an easy to read syntax that is similar to other configuration files is of advantage.
  2. Since the configuration is not written in Python, it is very important that the tight integration with Python is given. For example, it must be very simple to refer to the components in the Python modules and to internationalize any human-readable strings.
  3. The configuration mechanism should be declarative and not provide any facilities for logical operations. If the configuration would support logic, it would become very hard to read and the initial state of the entire system would be unclear. This is another reason Python was not suited for this task.
  4. Developing new components sometimes requires to extend the configuration mechanism. So it must be easy for the developer to extend the configuration mechanism without much hassle.

To satisfy the first requirement, we decided to use an XML-based language (as the name already suggests). The advantage of XML is also that it is a “standard format”, which increases the likelihood for people to be able to read it right away. Furthermore, we can use standard Python modules to parse the format and XML namespaces help us to group the configuration by functionality.

A single configuration step is called a directive. Each directive is an XML tag, and therefore the tags are grouped by namespaces. Directives are done either by simple or complex directives. Complex directives can contain other sub-directives. They are usually used to provide a set of common properties, but do not generate an action immediately.

A typical configuration file would be:


      provides="product.interfaces.IY" />


All configuration files are wrapped by the configure tag, which represents the beginning of the configuration. In the opening of this tag, we always list the namespaces we wish to use in this configuration file. Here we only want to use the generic BlueBream namespace, which is used as the default. Then we register an adapter with the system on line 4-7. The interfaces and classes are referred to by a proper Python dotted name. The configure tag might also contain an i18n_domain attribute that contains the domain that is used for all the translatable strings in the configuration.

As everywhere in BlueBream, there are several naming and coding conventions for ZCML inside a package. By default you should name the configuration file, configure.zcml. Inside the file you should only declare namespaces that you are actually going to use. When writing the directives make sure to logically group directives together and use comments as necessary. Comments are written using the common XML syntax: <!--...-->. For more info see Steve’s detailed ZCML Style Guide at for more info.

To satisfy our fourth requirement, it is possible to easily extend ZCML through itself using the meta namespace. A directive can be completely described by four components, its name, the namespace it belongs to, the schema and the directive handler:


These meta-directives are commonly placed in a file called meta.zcml.

The schema of a directive, which commonly lives in a file called, is a simple BlueBream schema whose fields describe the available attributes for the directive. The configuration system uses the fields to convert and validate the values of the configuration for use. For example, dotted names are automatically converted to Python objects. There are several specialized fields specifically for the configuration machinery:

  • PythonIndentifier - This field describes a python identifier, for example a simple variable name.
  • GlobalObject - An object that can be accessed as a module global, such as a class, function or constant.
  • Tokens - A sequence that can be read from a space-separated string. The value_type of the field describes token type.
  • Path - A file path name, which may be input as a relative path. Input paths are converted to absolute paths and normalized.
  • Bool - An extended boolean value. Values may be input (in upper or lower case) as any of: yes, no, y, n, true, false, t, or f.
  • MessageID - Text string that should be translated. Therefore the directive schema is the only place that needs to deal with internationalization. This satisfies part of requirement 2 above.

The handler, which commonly lives in a file called, is a function or another callable object that knows what needs to be done with the given information of the directive. Here is a simple (simplified to the actual code) example:

def adapter(_context, factory, provides, for_, name=''):

        discriminator = ('adapter', for_, provides, name),
        callable = provideAdapter,
        args = (for_, provides, factory, name),

The first argument of the handler is always the _context variable, which has a similar function to self in classes. It provides some common methods necessary for handling directives. The following arguments are the attributes of the directive (and their names must match). If an attribute name equals a Python keyword, like for in the example, then an underscore is appended to the attribute name.

The handler should also not directly execute an action, since the system should first go through all the configuration and detect possible conflicts and overrides. Therefore the _context object has a method called action that registers an action to be executed at the end of the configuration process. The first argument is the discriminator, which uniquely defines a specific directive. The callable is the function that is executed to provoke the action, the args argument is a list of arguments that is passed to the callable and the kw contains the callable’s keywords.

As you can see, there is nothing inheritly difficult about ZCML. Still, people coming to BlueBream often experience ZCML as the most difficult part to understand. This often created huge discussions about the format of ZCML. However, we believe that the problem lies not within ZCML itself, but the task it tries to accomplish. The components themselves always seem so clean in implementation; and then you get to the configuration. There you have to register this adapter and that view, make security assertions, and so on. And this in itself seems overwhelming at first sight. When I look at a configuration file after a long time I often have this feeling too, but reading directive for directive often helps me to get a quick overview of the functionality of the package. In fact, the configuration files can help you understand the processes of the BlueBream framework without reading the code, since all of the interesting interactions are defined right there.

Furthermore, ZCML is well documented at many places, including the BlueBream API documentation tool at . Here is a short list of the most important namespaces:

  • zope - This is the most generic and fundamental namespace of all, since it allows you to register all basic components with the component architecture.
  • browser - This namespace contains all of the directives that deal with HTML output, including managing skins and layer, declare new views (pages) and resources as well as setup auto-generated forms.
  • meta - As discussed above, you can use this namespace to extend available directives.
  • xmlrpc - This is the equivalent to browser, except that allows one to specify methods of components that should be available via XML-RPC.
  • i18n - This namespace contains all internationalization- and localization-specific configuration. Using registerTranslations you can register new message catalogs with a translation domain.
  • help - Using the register directive, you can register new help pages with the help system. This will give you context-sensitive help for the ZMI screens of your products.
  • mail - Using the directives of this namespace you can setup mailing components that your application can use to
blog comments powered by Disqus

Table Of Contents

Previous topic

9.7. Zope Component Architecture

Next topic

9.9. Startup

This Page