If a Pyramid developer has obeyed certain constraints while building an application, a third party should be able to change the application’s behavior without needing to modify its source code. The behavior of a Pyramid application that obeys certain constraints can be overridden or extended without modification.
We’ll define some jargon here for the benefit of identifying the parties involved in such an effort.
Other web frameworks, such as Django, advertise that they allow developers to create “pluggable applications”. They claim that if you create an application in a certain way, it will be integratable in a sensible, structured way into another arbitrarily-written application or project created by a third-party developer.
Pyramid, as a platform, does not claim to provide such a feature. The platform provides no guarantee that you can create an application and package it up such that an arbitrary integrator can use it as a subcomponent in a larger Pyramid application or project. Pyramid does not mandate the constraints necessary for such a pattern to work satisfactorily. Because Pyramid is not very “opinionated”, developers are able to use wildly different patterns and technologies to build an application. A given Pyramid application may happen to be reusable by a particular third party integrator, because the integrator and the original developer may share similar base technology choices (such as the use of a particular relational database or ORM). But the same application may not be reusable by a different developer, because he has made different technology choices which are incompatible with the original developer’s.
As a result, the concept of a “pluggable application” is left to layers built above Pyramid, such as a “CMS” layer or “application server” layer. Such layers are apt to provide the necessary “opinions” (such as mandating a storage layer, a templating system, and a structured, well-documented pattern of registering that certain URLs map to certain bits of code) which makes the concept of a “pluggable application” possible. “Pluggable applications”, thus, should not plug in to Pyramid itself but should instead plug into a system written atop Pyramid.
Although it does not provide for “pluggable applications”, Pyramid does provide a rich set of mechanisms which allows for the extension of a single existing application. Such features can be used by frameworks built using Pyramid as a base. All Pyramid applications may not be pluggable, but all Pyramid applications are extensible.
There is only one rule you need to obey if you want to build a maximally extensible Pyramid application: as a developer, you should factor any overrideable imperative configuration you’ve created into functions which can be used via pyramid.config.Configurator.include() rather than inlined as calls to methods of a Configurator within the main function in your application’s __init__.py. For example, rather than:
1 2 3 4 5 6
from pyramid.config import Configurator if __name__ == '__main__': config = Configurator() config.add_view('myapp.views.view1', name='view1') config.add_view('myapp.views.view2', name='view2')
You should do move the calls to add_view outside of the (non-reusable) if __name__ == '__main__' block, and into a reusable function:
1 2 3 4 5 6 7 8 9
from pyramid.config import Configurator if __name__ == '__main__': config = Configurator() config.include(add_views) def add_views(config): config.add_view('myapp.views.view1', name='view1') config.add_view('myapp.views.view2', name='view2')
Doing this allows an integrator to maximally reuse the configuration statements that relate to your application by allowing him to selectively include or disinclude the configuration functions you’ve created from an “override package”.
Alternately, you can use ZCML for the purpose of making configuration extensible and overrideable. ZCML declarations that belong to an application can be overridden and extended by integrators as necessary in a similar fashion. If you use only ZCML to configure your application, it will automatically be maximally extensible without any manual effort. See pyramid_zcml for information about using ZCML.
The fundamental “plug points” of an application developed using Pyramid are routes, views, and assets. Routes are declarations made using the pyramid.config.Configurator.add_route() method. Views are declarations made using the pyramid.config.Configurator.add_view() method. Assets are files that are accessed by Pyramid using the pkg_resources API such as static files and templates via a asset specification. Other directives and configurator methods also deal in routes, views, and assets. For example, the add_handler directive of the pyramid_handlers package adds a single route, and some number of views.
The steps for extending an existing application depend largely on whether the application does or does not use configuration decorators and/or imperative code.
If you just want to extend the application, you can run a scan against the application’s package, then add additional configuration that registers more views or routes.
1 2 3
if __name__ == '__main__': config.scan('someotherpackage') config.add_view('mypackage.views.myview', name='myview')
If you want to override configuration in the application, you may need to run pyramid.config.Configurator.commit() after performing the scan of the original package, then add additional configuration that registers more views or routes which performs overrides.
1 2 3 4
if __name__ == '__main__': config.scan('someotherpackage') config.commit() config.add_view('mypackage.views.myview', name='myview')
Once this is done, you should be able to extend or override the application like any other (see Extending the Application).
You can alternately just prevent a scan from happening (by omitting any call to the pyramid.config.Configurator.scan() method). This will cause the decorators attached to objects in the target application to do nothing. At this point, you will need to convert all the configuration done in decorators into equivalent imperative configuration or ZCML and add that configuration or ZCML to a separate Python package as described in Extending the Application.
To extend or override the behavior of an existing application, you will need to create a new package which includes the configuration of the old package, and you’ll perhaps need to create implementations of the types of things you’d like to override (such as views), which are referred to within the original package.
The general pattern for extending an existing application looks something like this:
The view configuration declarations you make which override application behavior will usually have the same view predicate attributes as the original you wish to override. These <view> declarations will point at “new” view code, in the override package you’ve created. The new view code itself will usually be cut-n-paste copies of view callables from the original application with slight tweaks.
For example, if the original application has the following configure_views configuration method:
def configure_views(config): config.add_view('theoriginalapp.views.theview', name='theview')
You can override the first view configuration statement made by configure_views within the override package, after loading the original configuration function:
1 2 3 4 5 6 7
from pyramid.config import Configurator from originalapp import configure_views if __name == '__main__': config = Configurator() config.include(configure_views) config.add_view('theoverrideapp.views.theview', name='theview')
In this case, the theoriginalapp.views.theview view will never be executed. Instead, a new view, theoverrideapp.views.theview will be executed instead, when request circumstances dictate.
A similar pattern can be used to extend the application with add_view declarations. Just register a new view against some other set of predicates to make sure the URLs it implies are available on some other page rendering.
Route setup is currently typically performed in a sequence of ordered calls to add_route(). Because these calls are ordered relative to each other, and because this ordering is typically important, you should retain their relative ordering when performing an override. Typically, this means copying all the add_route statements into the override package’s file and changing them as necessary. Then disinclude any add_route statements from the original application.
Assets are files on the filesystem that are accessible within a Python package. An entire chapter is devoted to assets: Static Assets. Within this chapter is a section named Overriding Assets. This section of that chapter describes in detail how to override package assets with other assets by using the pyramid.config.Configurator.override_asset() method. Add such override_asset calls to your override package’s __init__.py to perform overrides.