[rqlrewrite] fix rqlrewrite unpredictability vs relation sharing. Closes #3036144
The relations index used to determine if relation may be shared only
considered a single relation node per relation type. So when the same
relation type occurs several time, dict order give unpredictable result.
We shall properly consider all relations instead.
Tentative test case included (but bug reproduction is by definition
unpredictable...)
"""misc. unittests for utility functions
"""
from logilab.common.testlib import TestCase, unittest_main
from functools import partial
from cubicweb.devtools.fake import FakeConfig
from cubicweb.web.request import (CubicWebRequestBase, _parse_accept_header,
_mimetype_sort_key, _mimetype_parser, _charset_sort_key)
class AcceptParserTC(TestCase):
def test_parse_accept(self):
parse_accept_header = partial(_parse_accept_header,
value_parser=_mimetype_parser,
value_sort_key=_mimetype_sort_key)
# compare scores
self.assertEqual(parse_accept_header("audio/*;q=0.2, audio/basic"),
[( ('audio/basic', ('audio', 'basic', {}), 1.0 ) ),
( ('audio/*', ('audio', '*', {}), 0.2 ) )])
self.assertEqual(parse_accept_header("text/plain;q=0.5, text/html, text/x-dvi;q=0.8, text/x-c"),
[( ('text/html', ('text', 'html', {}), 1.0 ) ),
( ('text/x-c', ('text', 'x-c', {}), 1.0 ) ),
( ('text/x-dvi', ('text', 'x-dvi', {}), 0.8 ) ),
( ('text/plain', ('text', 'plain', {}), 0.5 ) )])
# compare mimetype precedence for a same given score
self.assertEqual(parse_accept_header("audio/*, audio/basic"),
[( ('audio/basic', ('audio', 'basic', {}), 1.0 ) ),
( ('audio/*', ('audio', '*', {}), 1.0 ) )])
self.assertEqual(parse_accept_header("text/*, text/html, text/html;level=1, */*"),
[( ('text/html', ('text', 'html', {'level': '1'}), 1.0 ) ),
( ('text/html', ('text', 'html', {}), 1.0 ) ),
( ('text/*', ('text', '*', {}), 1.0 ) ),
( ('*/*', ('*', '*', {}), 1.0 ) )])
# free party
self.assertEqual(parse_accept_header("text/*;q=0.3, text/html;q=0.7, text/html;level=1, text/html;level=2;q=0.4, */*;q=0.5"),
[( ('text/html', ('text', 'html', {'level': '1'}), 1.0 ) ),
( ('text/html', ('text', 'html', {}), 0.7 ) ),
( ('*/*', ('*', '*', {}), 0.5 ) ),
( ('text/html', ('text', 'html', {'level': '2'}), 0.4 ) ),
( ('text/*', ('text', '*', {}), 0.3 ) )
])
# chrome sample header
self.assertEqual(parse_accept_header("application/xml,application/xhtml+xml,text/html;q=0.9,text/plain;q=0.8,image/png,*/*;q=0.5"),
[( ('application/xhtml+xml', ('application', 'xhtml+xml', {}), 1.0 ) ),
( ('application/xml', ('application', 'xml', {}), 1.0 ) ),
( ('image/png', ('image', 'png', {}), 1.0 ) ),
( ('text/html', ('text', 'html', {}), 0.9 ) ),
( ('text/plain', ('text', 'plain', {}), 0.8 ) ),
( ('*/*', ('*', '*', {}), 0.5 ) ),
])
def test_parse_accept_language(self):
self.assertEqual(_parse_accept_header('fr,fr-fr;q=0.8,en-us;q=0.5,en;q=0.3'),
[('fr', 'fr', 1.0), ('fr-fr', 'fr-fr', 0.8),
('en-us', 'en-us', 0.5), ('en', 'en', 0.3)])
def test_parse_accept_charset(self):
parse_accept_header = partial(_parse_accept_header,
value_sort_key=_charset_sort_key)
self.assertEqual(parse_accept_header('ISO-8859-1,utf-8;q=0.7,*;q=0.7'),
[('ISO-8859-1', 'ISO-8859-1', 1.0),
('utf-8', 'utf-8', 0.7),
('*', '*', 0.7)])
def test_base_url(self):
dummy_vreg = type('DummyVreg', (object,), {})()
dummy_vreg.config = FakeConfig()
dummy_vreg.config['base-url'] = 'http://babar.com/'
dummy_vreg.config['https-url'] = 'https://toto.com/'
req = CubicWebRequestBase(dummy_vreg, https=False)
self.assertEqual('http://babar.com/', req.base_url())
self.assertEqual('http://babar.com/', req.base_url(False))
self.assertEqual('https://toto.com/', req.base_url(True))
req = CubicWebRequestBase(dummy_vreg, https=True)
self.assertEqual('https://toto.com/', req.base_url())
self.assertEqual('http://babar.com/', req.base_url(False))
self.assertEqual('https://toto.com/', req.base_url(True))
if __name__ == '__main__':
unittest_main()