infrae.wsgi provides a support to run Zope 2 as a WSGI application.
It basically does the same work than:
- Zope 2 new builtin WSGI publisher.
- It work with real Zope 2 applications and no monkey-patches,
- It pay specially attention to properly implement streaming. You can use your ZODB connection while streaming, (either with the .write method of request, or returning an IResult or an IStreamIterator iterator). ConflictError generate a final error if they happens during a streaming operation, not re-sending things again on the same HTTP connection,
- All ConflictError are properly managed.
- All those cases are tested.
It does not:
- Provide Zope 2 as a collection of WSGI middlewares, as Zope 2 paradigm / code base is not good for it,
- Do all fancy request body changes that old Zope 2 publisher does, as nobody use it anymore since a very long time.
Error messages are handled a bit differently than in traditional Zope 2, in order to make things simpler.
They are views on errors (called error.html), and wrapped in acquisition around the context where they happened:
from five import grok class CorpError(Exception): """Custom corporate error. """ class CorpErrorMessage(grok.View): grok.context(CorpError) grok.name('error.html') def render(self): # aq_parent(self) is where the error happened return u'I Failed: %s' % str(self.error.args)
Errors are logged with useful information:
- Which is the URL triggering this error,
- The user-agent,
- The referent,
- The logged in username,
- On which object, its physical path and meta_type if possible.
The error log can be accessible online via /errorlog.html on any Zope 2 content.
Errors can also be sent to a Sentry service (see Paste Deploy section).
The error log ignores certain errors by default: NotFound, Redirect, Unauthorized, Forbidden, BadRequest, BrokenReferenceError. The errlog.html page has a form configure some (or all) of these errors to not be ignored. This is not a persistent setting and is forgotten on restart.
infrae.wsgi has been and deployed with Paste Deploy, mod_wsgi and nginx. It correctly respect the WSGI specification and should work with any WSGI server.
The application is available with the entry point infrae.wsgi#zope2.
It expect an option variable called zope_conf that point to the Zope 2 configuration file.
The option debug_mode can as well be specified, to run Zope in debug mode. In debug mode, error pages will not be rendered by Zope and the errors will propagate in the WSGI stack. This behavior will let you debug errors with specialized middlewares.
To disable the error propagation in debug mode, the option debug_exceptions can be set to off.
The option zope_workers can be used to specify the maximum of threads Zope should allow to process requests at the same time (default to 4). This can be usefull if you wish to allow more threads in your wsgi environment, in case you have middlewares or other applications that intercept the requests and support more threads than Zope does.
The option show_errors accepts a comma-separated list of errors which will not be ignored. This overrides the default list of ignored errors (see the Errors section, above)
The option ignore_errors accepts a comma-separated list of errors which will be ignored. This overrides the default list of ignored errors too.
The configuration accepts options for Raven (Sentry's client):
raven.dsn = http://public:email@example.com/1 raven.include_paths = my.package, my.other.package raven.exclude_paths = my.package.crud
Those options requires Raven to be installed.
You can add two headers in your proxy in order to control the virtual hosting:
- X-VHM-URL: That would the complete URL of your site, at which you want to see your Zope application, like http://www.mysite.com/application.
- X-VHM-Path: That would be an optional path to the Zope folder you see at the given URL instead of the Zope root, lile /my/folder.
A test request TestRequest can be imported from infrae.wsgi.testing. It behaves exactly like a request used by Zope. It takes the following parameters when creating it:
- WSGI application to use with the request.
- Zope layers to apply on the request.
- URL used with the request.
- HTTP method used with the request.
- HTTP headers that where sent with the request.
- Should the request ran with the debug mode.
- Form data associated with the request.
A layer inheriting of infrae.testing Zope2Layer layer called ZopeBrowserLayer let you write functional tests.
It provides both an http function and a ZopeBrowser class (like the one provided by zope.testbrowser) that you can use, and that will connect to the tested application using the WSGI support provided by this package.
This will let you do functional testing, and things will work exactly like in your browser, as the requests will be processed the same way than they are in real life (which is not really the case with the Testing module of Zope 2).
You will be actually able to test applications that do use streaming:
import unittest from infrae.wsgi.testing import ZopeBrowser, ZopeBrowserLayer import corp.testpackage class CorpTestCase(unittest.TestCase): layer = BrowserLayer(corp.testpackage) def setUp(self): self.root = self.layer.get_application() # Create some test content def test_feature(self): browser = ZopeBrowser() browser.open('http://localhost/somepage') self.assertEqual(browser.status, 200) ...
|Release version||Release date||Release files|
|infrae.wsgi 2.2.1||2013 10 8 15:20:23||
|infrae.wsgi 2.2||2013 5 22 14:37:47||
|infrae.wsgi 2.1||2012 12 10 17:02:18||
|infrae.wsgi 2.0.1||2012 9 24 12:08:32||
|infrae.wsgi 2.0||2012 9 4 11:42:46||
|infrae.wsgi 1.3||2011 7 27 09:54:12||
|infrae.wsgi 1.2||2011 2 2 15:35:38||
|infrae.wsgi 1.1||2010 10 7 11:40:54||
|infrae.wsgi 1.0||2010 7 15 18:13:16||
License: New BSD