© 2006-2019 Alice Bevan-McGregor and contributors.
https://github.com/marrow/WebCore
WebCore is a nanoframework, a fraction of the size of competing "microframeworks", and culmination of more than ten years of web development experience. It provides a clean API for standard points of extension while strongly encouraging model, view, controller separation. Being less than 400 source lines of code (SLoC; excluding comments and documentation) and containing more comments and lines of documentation than lines of code, WebCore is built to be insanely easy to test, adapt, and use, allowing any developer familiar with programming (not just Python programming) to be able to read and understand the entirety of the framework in an evening.
It is substantially smaller and more efficient than monolithic frameworks such as Django or Pyramid:
from web.core import Application
Application("Hi.").serve('wsgiref')
Really; that's it. (It can be made into one line if you're willing to make the import ugly using __import__.) The
Application class represents a standard Python WSGI application, the rest is up to you to pick the components that
best fit your own needs.
Installing WebCore is easy, just execute the following in a terminal:
pip install WebCore
Note: We strongly recommend always using a container, virtualization, or sandboxing environment of some kind when developing using Python; installing things system-wide is yucky (for a variety of reasons) nine times out of ten. We prefer light-weight virtualenv, others prefer solutions as robust as Vagrant.
If you add WebCore to the install_requires argument of the call to setup() in your application's
setup.py file, WebCore will be automatically installed and made available when your own application or
library is installed. We recommend using "less than" version numbers to ensure there are no unintentional
side-effects when updating. Use WebCore<2.1 to get all bugfixes for the current release, and
WebCore<3.0 to get bugfixes and feature updates while ensuring that large breaking changes are not installed.
WebCore only depends on the excellent webob package to provide request, response, and HTTP
status code exception helpers and the marrow.package utility package for plugin
management. All other dependencies will be application dependencies; choice of template engine,
database layer, session storage mechanism, and even dispatch method are left entirely up to the developer making use
of the framework. Provided are a number of extras requirements, which you can define using a comma-separated list
appended to the package name during installation from the command-line, or within your own package's
install_requires. For example, to install a typical set of development tools at the same time as WebCore, run:
pip install WebCore[development]
The available extras are:
development-- this installs a recommended set of development-time packages, including pytest and a suite of plugins for it, plus the backlash interactive debugger (needed by the optionalDebugExtension), object dispatch, the comprehensiveptipythonupgraded REPL, and the waitress development web server.production-- install recommended production-time packages; currently this only installs theflupFastCGI server bridge.
The default choices for dispatch are allowed as extras:
object-- install object dispatchroute-- install route-based dispatchtraversal-- install traversal dispatch
You can also name a supported server bridge as an extra. Currently available bridges with third-party dependencies include:
waitresstornadoflup
Development takes place on GitHub in the
WebCore project. Issue tracking, documentation, and downloads
are provided there. Development chat (both development of WebCore and chat for users using WebCore to develop their
own solutions) is graciously provided by Freenode in the #webcore
channel.
Installing the current development version requires Git, a distributed source code management system. If you have Git you can run the following to download and link the development version into your Python runtime:
git clone https://github.com/marrow/WebCore.git pip install -e WebCore
You can then upgrade to the latest version at any time:
(cd WebCore; git pull; pip install -e .)
Extra dependenies can be declared the same as per web-based installation:
pip install -e WebCore[development]
If you would like to make changes and contribute them back to the project, fork the GitHub project, make your changes, and submit a pull request. This process is beyond the scope of this documentation; for more information see GitHub's documentation.
The Application class is the primary entry point for the web framework. Its constructor currently takes up to
three arguments:
root-- the root object to use as the controller for/requestsextensions-- a list of extensions to use with your applicationlogging-- Pythonloggingconfiguration
The "root controller" is used as the starting point for dispatch resolution of the endpoint for a request, see the Controllers section below for details on what can be used here, but it's basically anything.
By default the BaseExtension, providing basic request and response objects and basic views, is always enabled for
your application, has no configuration, and does not need to be instantiated yourself. Other extensions should be
instantiated and passed in the extensions list.
Logging configuration offers two choices: simple "global logging level" by defining logging as a dictionary
only containing a level key naming the level to set, or full logging.config.dictConfig configuration. Passing
only a level is equivalent to running logging.basicConfig.
This configuration can entirely come from YAML, for example:
root: !!python/name:web.app.example.RootController
extensions:
- !!python/object:web.ext.debug.DebugExtension
- !!python/object:web.ext.analytics.AnalyticsExtension
- !!python/object:web.ext.annotation:AnnotationExtension
logging:
level: debug
This would make managing complex extension configuration easier. One way to invoke WebCore with a configuration like this, while allowing for a distinction between production and development environments and use under ModWSGI would be:
import yaml
from web.core import Application
fname = 'development.yaml' if __debug__ else 'production.yaml'
with open(fname, 'r', encoding='utf-8') as fh:
config = yaml.load(fh)
app = Application(**config)
if __name__ == "__main__":
app.serve('wsgiref')
Now, running python run.py (if saved as run.py) would serve the development.yaml configuration, and
running as python -O run.py (optimization enabled) or with PYTHONOPTIMIZE=1 set in the environment will
utilize the production.yaml file.
WebCore is highly aware running with optimizations enabled, eliminating many of the expensive validation checks that
are only really useful in development. For example, calling an endpoint with invalid arguments will 404 with a
friendly warning in development, but 500 in production as the TypeError is not preemptively checked and
caught; this is one of the most expensive validation checks. Feel free to browse the code looking for if __debug__
blocks to see what else changes in "production mode".
The order you define the extensions in does not matter; they declare dependencies and will be automatically
dependency-ordered when collecting callbacks. Please see the extension.py example for additional information on
what you can do with them.
The overall application has an ApplicationContext associated with it. This object is passed around to the various
extension callbacks and acts as an attribute access dictionary. (All of the typical dictionary methods will work,
and the keys can be accessed as attributes instead, saving some typing.) During the processing of a request a subclass
is constructed called RequestContext and in-request extension callbacks, and your controller endpoints, are given
a reference to this instance.
The attributes present in the base ApplicationContext are:
app-- a reference to theApplicationinstanceroot-- the original object passed when constructing theApplicationinstanceextension-- theWebExtensionsextension registrydispatch-- theWebDispatchersdispatch protocol bridge and plugin registryview-- theWebViewsview handler registry
Extensions would access these during start and stop events, for example to register new view handlers.
The attributes present in the RequestContext (added by WebCore itself or the BaseExtension during request
processing) are:
environ-- the WSGI request environment as passed to WebCore's WSGI handlerrequest-- awebob.Requestrepresenting the current HTTP requestresponse-- awebob.Responseobject corresponding to the response WebCore will returnpath-- a list of dispatch steps represented by tuples of(handler, script_name)
Additional attributes may be added by other extensions.
Controllers and, more generally, callable endpoints, are functions or methods called to process a request and return
a value for view or raise an exception. Non-method callables are passed the context as a first argument; methods are
assumed to have access via self as the context will have been passed as the only positional argument to the class
constructor. Callable endpoints are additionally passed any unprocessed path elements as positional parameters, and
a combination of query string arguments (GET values) and form-encoded body elements (POST values) as keyword
arguments, with arguments from the request body taking precedence and duplicated keys being passed as a list of
values. They may return any value there is a view registered for, see the
docstring of the view manager for details.
Static endpoints, on the other hand, are non-callable objects that can be handled by a view. The very first example at the top of this document relies on the fact that there is a view to handle strings, both static, and as returned by a callable endpoint such as:
def hello(context):
return "Hello world!"
To allow for customization of the name, you would write this endpoint as:
def hello(context, name="world"):
return "Hello {}!".format(name)
As noted in the Application section, when Python is run with optimizations enabled (-O or PYTHONOPTIMIZE set)
unknown arguments being passed (unknown query string arguments or form values) will result in a TypeError being
raised and thus a 500 Internal Server Error due to the uncaught exception. In development (without optimizations)
a 404 Not Found error with a message indicating the mismatched values will be the result. You can use *args
and **kwargs to capture any otherwise undefined positional and keyword arguments, or use an extension to mutate
the incoming data and strip invalid arguments prior to the endpoint being called.
That "hello world" endpoint, however, may be called in one of several different ways, as no other restrictions have been put in place:
GET /-- Hello world! (Default used.)GET /Alice-- Hello Alice! (Passed positionally.)GET /?name=Bob-- Hello Bob! (Via query string assignment.)POST /submitting a form with anamefield and value ofEve-- Hello Eve! (Via form-encoded body assignment.)
Other HTTP verbs will work as well, but a form-encoded body is only expected and processed on POST requests.
The process of finding the endpoint to use to process a request is called dispatch. There are a number of forms of dispatch available, some should be immediately familiar.
- Object dispatch. This is the default (providided by the
web.dispatch.object package) form of dispatch for WebCore, and
is also utilized by several other frameworks such as TurboGears. Essentially each path element is looked up as
an attribute of the previously looked up object treating a path such as
/foo/bar/bazas an attempt to resolve the Python referenceroot.foo.bar.baz. This is quite flexible, allowing easy redirection of descent using Python-standard protocols such as__getattr__methods, use of lazy evaluation descriptors, etc., etc. - Registered routes. This will likely be the approach most familiar to developers switching from PHP frameworks or who have used any of the major macro- or micro-frameworks in Python such as Django, Flask, etc. You explicitly map URLs, generally using a regular expression or regular expression short-hand, to specific callable endpoints. Often this is a accomplished using a decorator. WebCore offers this form of dispatch throuhg the web.dispatch.route package.
- Traversal. This is similar to object dispatch, but descending through mapping keys. The previous example then
translates to
root['foo']['bar']['baz'], allowing managed descent through the__getitem__protocol. This is one of the methods (the other being routes) provided by Pyramid. We offer this form of dispatch through the web.dispatch.traversal package.
There may be other dispatchers available and the protocol allows for "dispatch middleware" to offer even more flexible approaches to endpoint lookup. The dispatch protocol itself is framework agnostic (these example dispatchers are in no way WebCore-specific) and has its own documentation.
WebCore recommends registration of extensions and other plugins as Python-standard entry_points references.
Please see the relevant setuptools documentation for details on this
process. Additionally, WebCore marks package namespaces for shared use. The namespaces used, and their purposes, are:
web-- the top level shared namespace for WebCore and WebCore accessoriesweb.app-- a namespace for reusable application components and your own useweb.ext-- a namespace for WebCore extensions; your own can be placed hereweb.server-- light-weight WSGI server adapters; your own WSGI server can define a dependency-free adapter here, for example
The plugin namespaces follow a similar pattern:
web.app-- re-usable components you can attach to your own controller treesweb.extension-- extensions registered by name and "provides" tagweb.server-- similarly, server adapters registered by name
WebCore also makes use of the web.dispatch namespace to look up dispatchers. Other WebCore-related packages and
extensions may make use of other plugin namespaces. Have a gander at WebCore's setup.py file for an example of
how to register plugins this way, and copy the __init__.py file from the web package into the overlay in your
own package (and declare such in your setup.py package metadata as the namespace_packages argument) to
participate in the Python package namespaces.
- A complete modernization rewrite of WebCore, from the ground up.
- Features multiple extension interfaces to extend registered view handlers and provide a uniform callback mechanism.
- Standard usage makes use of no superglobals or "thread locals", instead relying on a context object collaboratively populated by extensions.
- WebCore's former "dialect" system is now dispatch.
- Thanks Pypi.
- Corrected argument specification for
transformextension callbacks, fixingAnnotationExtensionusage as per #163. - Additional source-level documentation and expanded examples.
- An excessively large number of additional WSGI server adapters; now supported are: waitress
tornado, fcgi,
cherrypy, appengine, paste,
eventlet, gevent, diesel,
and bjoern. Each is available as an
extras_requireby the same name which will pull in the required third-party dependency.
- Argument processing moved out of
web.coreinto extensionmutatehandlers. Features improved rich unflattening of query string and form encoded body parameters. Configurable behaviour. For details, see: web/ext/args.py - Extensively documented access control list extension validating endpoint security and return value permission using context-aware predicates.
- The ability for extensions to define additional callbacks for collection.
- The
DatabaseExtension(formerlyDBExtension) has been moved into its own repository. - Content negotiation endpoint return value serialization, with pluggable
dumpsregistry. - Complete unit test coverage.
WebCore has been released under the MIT Open Source license.
Copyright © 2006-2019 Alice Bevan-McGregor and contributors.
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.