Skip to content
/ WebCore Public

WebCore, the super tiny and blazingly fast modular Python web nanoframework.

License

Notifications You must be signed in to change notification settings

marrow/WebCore

Repository files navigation

WebCore

© 2006-2019 Alice Bevan-McGregor and contributors.
https://github.com/marrow/WebCore
Latest released version. Latest Github tagged release. Downloads per week. Release build status. Release test coverage. Status of release dependencies. Subscribe to project activity on Github. Star this project on Github.
  1. What is WebCore?
  2. Installation
    1. Dependencies and Extras
    2. Development Version
  3. Basic Concepts
    1. Application
    2. Context
    3. Controllers, Endpoints, Dispatch, Oh My!
  4. Plugins and Namespaces
  5. Version History
  6. License

What is 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.

Installation

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.

Dependencies and Extras

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 optional DebugExtension), object dispatch, the comprehensive ptipython upgraded REPL, and the waitress development web server.
  • production -- install recommended production-time packages; currently this only installs the flup FastCGI server bridge.

The default choices for dispatch are allowed as extras:

  • object -- install object dispatch
  • route -- install route-based dispatch
  • traversal -- install traversal dispatch

You can also name a supported server bridge as an extra. Currently available bridges with third-party dependencies include:

  • waitress
  • tornado
  • flup

Development Version

Development build status. Development test coverage. Changes since last release. Github Issues Fork this project on Github.

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.

Basic Concepts

Application

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 / requests
  • extensions -- a list of extensions to use with your application
  • logging -- Python logging configuration

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.

Context

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 the Application instance
  • root -- the original object passed when constructing the Application instance
  • extension -- the WebExtensions extension registry
  • dispatch -- the WebDispatchers dispatch protocol bridge and plugin registry
  • view -- the WebViews view 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 handler
  • request -- a webob.Request representing the current HTTP request
  • response -- a webob.Response object corresponding to the response WebCore will return
  • path -- a list of dispatch steps represented by tuples of (handler, script_name)

Additional attributes may be added by other extensions.

Controllers, Endpoints, Dispatch, Oh My!

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 a name field and value of Eve -- 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/baz as an attempt to resolve the Python reference root.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.

Plugins and Namespaces

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 accessories
  • web.app -- a namespace for reusable application components and your own use
  • web.ext -- a namespace for WebCore extensions; your own can be placed here
  • web.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 trees
  • web.extension -- extensions registered by name and "provides" tag
  • web.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.

Version History

Version 2.0

  • 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.

Version 2.0.1

  • Thanks Pypi.

Version 2.0.2

  • Corrected argument specification for transform extension callbacks, fixing AnnotationExtension usage 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_require by the same name which will pull in the required third-party dependency.

Version 2.0.3

  • Argument processing moved out of web.core into extension mutate handlers. 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 (formerly DBExtension) has been moved into its own repository.
  • Content negotiation endpoint return value serialization, with pluggable dumps registry.
  • Complete unit test coverage.

License

WebCore has been released under the MIT Open Source license.

The MIT 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.