Mercurial > p > roundup > code
view test/wsgi_liveserver.py @ 8177:2967f37e73e4
refactor: issue2551289. invalid REST Accept header stops request
Sending a POST, PUT (maybe PATCH) with an accept header that is not
application/json or xml (if enabled) used to complete the request
before throwing a 406 error. This was wrong.
Now it reports an error without dispatching/processing the requested
transaction. This is the first of a series of refactors of the
dispatch method to make it faster and more readable by using return
early pattern and extracting methods from the code.
changes:
The following now return 406 errors not 400 errors
invalid version specified with @apiver in URL.
invalid version specified with @apiver in payload body
invalid version specified in accept headers as
application/vnd.roundup.test-vz+json or version property
Parsing the accept header returns a 400 when presented with a
parameter without an = sign or other parse error. They used to
return a 406 which is wrong since the header is malformed rather
than having a value I can't respond to.
Some error messages were made clearer.
Results in the case of an error are proper json error object rather
than text/plain strings.
New test added for testdetermine_output_formatBadAccept that test the
new method using the same test cases as for
testDispatchBadAccept. I intend to extend the test coverage for
determine_output_format to cover more cases. This should be a faster
unit test than for dispatch.
Removed .lower() calls for accept_mime_type as the input values are
taken from the values in the __accepted_content_type dict which
only has lower case values.
| author | John Rouillard <rouilj@ieee.org> |
|---|---|
| date | Sun, 08 Dec 2024 01:09:34 -0500 |
| parents | e9760702bf0c |
| children | f6923d2ba9a5 |
line wrap: on
line source
# -*- coding: utf-8 -*- """ wsgi-liveserver provides a simple LiverServerTestCase class that can be used to help start a web server in the background to serve a WSGI compliant application for use with testing. Generally it will be used in conjuction with something like Selenium to perform a series of functional tests using a browser. Licensed under the GNU GPL v3 Copyright (c) 2013 John Kristensen (unless explicitly stated otherwise). """ import threading import socket import unittest from wsgiref.simple_server import make_server, WSGIRequestHandler __author__ = 'John Kristensen' __version__ = '0.3.1' __license__ = 'GPLv3' class QuietHandler(WSGIRequestHandler): def log_request(*args, **kwargs): pass class LiveServerTestCase(unittest.TestCase): port_range = (8080, 8090) def create_app(self): """Create your wsgi app and return it.""" raise NotImplementedError def __call__(self, result=None): """ Do some custom setup stuff and then hand off to TestCase to do its thing. """ try: self._pre_setup() super(LiveServerTestCase, self).__call__(result) finally: self._post_teardown() def url_base(self): """Return the url of the test server.""" return 'http://{0}:{1}'.format(self.host, self.port) def _pre_setup(self): """Setup and start the test server in the background.""" self._server = None self.host = 'localhost' self.port = self.port_range[0] self._thread = None # Get the app self.app = self.create_app() # Cycle through the port range to find a free port while self._server is None and self.port <= self.port_range[1]: try: self._server = make_server(self.host, self.port, self.app, handler_class=QuietHandler) except socket.error: self.port += 1 # No free port, raise an exception if self._server is None: raise socket.error('Ports {0}-{1} are all already in use'.format( *self.port_range)) # Start the test server in the background self._thread = threading.Thread(target=self._server.serve_forever) self._thread.start() def _post_teardown(self): """Stop the test server.""" if self._thread is not None: self._server.shutdown() self._server.server_close() self._thread.join() del self._server
