1
0
mirror of https://github.com/django/django.git synced 2024-12-27 03:25:58 +00:00
django/tests/regressiontests/utils/tests.py

260 lines
9.4 KiB
Python
Raw Normal View History

"""
Tests for django.utils.
"""
from unittest import TestCase
from django.utils import html, checksums, text
from django.utils.functional import SimpleLazyObject
import timesince
import datastructures
import datetime_safe
import tzinfo
from decorators import DecoratorFromMiddlewareTests
from functional import FunctionalTestCase
# Extra tests
__test__ = {
'timesince': timesince,
'datastructures': datastructures,
'datetime_safe': datetime_safe,
'tzinfo': tzinfo,
}
from dateformat import *
from feedgenerator import *
from module_loading import *
from termcolors import *
class TestUtilsHtml(TestCase):
def check_output(self, function, value, output=None):
"""
Check that function(value) equals output. If output is None,
check that function(value) equals value.
"""
if output is None:
output = value
self.assertEqual(function(value), output)
def test_escape(self):
f = html.escape
items = (
('&','&'),
('<', '&lt;'),
('>', '&gt;'),
('"', '&quot;'),
("'", '&#39;'),
)
# Substitution patterns for testing the above items.
patterns = ("%s", "asdf%sfdsa", "%s1", "1%sb")
for value, output in items:
for pattern in patterns:
self.check_output(f, pattern % value, pattern % output)
# Check repeated values.
self.check_output(f, value * 2, output * 2)
# Verify it doesn't double replace &.
self.check_output(f, '<&', '&lt;&amp;')
def test_linebreaks(self):
f = html.linebreaks
items = (
("para1\n\npara2\r\rpara3", "<p>para1</p>\n\n<p>para2</p>\n\n<p>para3</p>"),
("para1\nsub1\rsub2\n\npara2", "<p>para1<br />sub1<br />sub2</p>\n\n<p>para2</p>"),
("para1\r\n\r\npara2\rsub1\r\rpara4", "<p>para1</p>\n\n<p>para2<br />sub1</p>\n\n<p>para4</p>"),
("para1\tmore\n\npara2", "<p>para1\tmore</p>\n\n<p>para2</p>"),
)
for value, output in items:
self.check_output(f, value, output)
def test_strip_tags(self):
f = html.strip_tags
items = (
('<adf>a', 'a'),
('</adf>a', 'a'),
('<asdf><asdf>e', 'e'),
('<f', '<f'),
('</fe', '</fe'),
('<x>b<y>', 'b'),
)
for value, output in items:
self.check_output(f, value, output)
def test_strip_spaces_between_tags(self):
f = html.strip_spaces_between_tags
# Strings that should come out untouched.
items = (' <adf>', '<adf> ', ' </adf> ', ' <f> x</f>')
for value in items:
self.check_output(f, value)
# Strings that have spaces to strip.
items = (
('<d> </d>', '<d></d>'),
('<p>hello </p>\n<p> world</p>', '<p>hello </p><p> world</p>'),
('\n<p>\t</p>\n<p> </p>\n', '\n<p></p><p></p>\n'),
)
for value, output in items:
self.check_output(f, value, output)
def test_strip_entities(self):
f = html.strip_entities
# Strings that should come out untouched.
values = ("&", "&a", "&a", "a&#a")
for value in values:
self.check_output(f, value)
# Valid entities that should be stripped from the patterns.
entities = ("&#1;", "&#12;", "&a;", "&fdasdfasdfasdf;")
patterns = (
("asdf %(entity)s ", "asdf "),
("%(entity)s%(entity)s", ""),
("&%(entity)s%(entity)s", "&"),
("%(entity)s3", "3"),
)
for entity in entities:
for in_pattern, output in patterns:
self.check_output(f, in_pattern % {'entity': entity}, output)
def test_fix_ampersands(self):
f = html.fix_ampersands
# Strings without ampersands or with ampersands already encoded.
values = ("a&#1;", "b", "&a;", "&amp; &x; ", "asdf")
patterns = (
("%s", "%s"),
("&%s", "&amp;%s"),
("&%s&", "&amp;%s&amp;"),
)
for value in values:
for in_pattern, out_pattern in patterns:
self.check_output(f, in_pattern % value, out_pattern % value)
# Strings with ampersands that need encoding.
items = (
("&#;", "&amp;#;"),
("&#875 ;", "&amp;#875 ;"),
("&#4abc;", "&amp;#4abc;"),
)
for value, output in items:
self.check_output(f, value, output)
class TestUtilsChecksums(TestCase):
def check_output(self, function, value, output=None):
"""
Check that function(value) equals output. If output is None,
check that function(value) equals value.
"""
if output is None:
output = value
self.assertEqual(function(value), output)
def test_luhn(self):
f = checksums.luhn
items = (
(4111111111111111, True), ('4111111111111111', True),
(4222222222222, True), (378734493671000, True),
(5424000000000015, True), (5555555555554444, True),
(1008, True), ('0000001008', True), ('000000001008', True),
(4012888888881881, True), (1234567890123456789012345678909, True),
(4111111111211111, False), (42222222222224, False),
(100, False), ('100', False), ('0000100', False),
('abc', False), (None, False), (object(), False),
)
for value, output in items:
self.check_output(f, value, output)
class _ComplexObject(object):
def __init__(self, name):
self.name = name
def __eq__(self, other):
return self.name == other.name
def __hash__(self):
return hash(self.name)
def __str__(self):
return "I am _ComplexObject(%r)" % self.name
def __unicode__(self):
return unicode(self.name)
def __repr__(self):
return "_ComplexObject(%r)" % self.name
complex_object = lambda: _ComplexObject("joe")
class TestUtilsSimpleLazyObject(TestCase):
"""
Tests for SimpleLazyObject
"""
# Note that concrete use cases for SimpleLazyObject are also found in the
# auth context processor tests (unless the implementation of that function
# is changed).
def test_equality(self):
self.assertEqual(complex_object(), SimpleLazyObject(complex_object))
self.assertEqual(SimpleLazyObject(complex_object), complex_object())
def test_hash(self):
# hash() equality would not be true for many objects, but it should be
# for _ComplexObject
self.assertEqual(hash(complex_object()),
hash(SimpleLazyObject(complex_object)))
def test_repr(self):
# For debugging, it will really confuse things if there is no clue that
# SimpleLazyObject is actually a proxy object. So we don't
# proxy __repr__
self.assert_("SimpleLazyObject" in repr(SimpleLazyObject(complex_object)))
def test_str(self):
self.assertEqual("I am _ComplexObject('joe')", str(SimpleLazyObject(complex_object)))
def test_unicode(self):
self.assertEqual(u"joe", unicode(SimpleLazyObject(complex_object)))
def test_class(self):
# This is important for classes that use __class__ in things like
# equality tests.
self.assertEqual(_ComplexObject, SimpleLazyObject(complex_object).__class__)
def test_deepcopy(self):
import django.utils.copycompat as copy
# Check that we *can* do deep copy, and that it returns the right
# objects.
# First, for an unevaluated SimpleLazyObject
s = SimpleLazyObject(complex_object)
assert s._wrapped is None
s2 = copy.deepcopy(s)
assert s._wrapped is None # something has gone wrong is s is evaluated
self.assertEqual(s2, complex_object())
# Second, for an evaluated SimpleLazyObject
name = s.name # evaluate
assert s._wrapped is not None
s3 = copy.deepcopy(s)
self.assertEqual(s3, complex_object())
class TestUtilsText(TestCase):
def test_truncate_words(self):
self.assertEqual(u'The quick brown fox jumped over the lazy dog.',
text.truncate_words(u'The quick brown fox jumped over the lazy dog.', 10))
self.assertEqual(u'The quick brown fox ...',
text.truncate_words('The quick brown fox jumped over the lazy dog.', 4))
self.assertEqual(u'The quick brown fox ....',
text.truncate_words('The quick brown fox jumped over the lazy dog.', 4, '....'))
self.assertEqual(u'<p><strong><em>The quick brown fox jumped over the lazy dog.</em></strong></p>',
text.truncate_html_words('<p><strong><em>The quick brown fox jumped over the lazy dog.</em></strong></p>', 10))
self.assertEqual(u'<p><strong><em>The quick brown fox ...</em></strong></p>',
text.truncate_html_words('<p><strong><em>The quick brown fox jumped over the lazy dog.</em></strong></p>', 4))
self.assertEqual(u'<p><strong><em>The quick brown fox ....</em></strong></p>',
text.truncate_html_words('<p><strong><em>The quick brown fox jumped over the lazy dog.</em></strong></p>', 4, '....'))
self.assertEqual(u'<p><strong><em>The quick brown fox</em></strong></p>',
text.truncate_html_words('<p><strong><em>The quick brown fox jumped over the lazy dog.</em></strong></p>', 4, None))
if __name__ == "__main__":
import doctest
doctest.testmod()