blob: d85f171dbb5b0870364b0792cffa45448a0082e8 [file] [log] [blame]
# -*- coding: iso-8859-1 -*-
""" Test script for the Unicode implementation.
Written by Marc-Andre Lemburg (mal@lemburg.com).
(c) Copyright CNRI, All Rights Reserved. NO WARRANTY.
"""#"
import unittest, sys, string, codecs, new
from test import test_support, string_tests
class UnicodeTest(
string_tests.CommonTest,
string_tests.MixinStrUnicodeUserStringTest,
string_tests.MixinStrUnicodeTest,
):
type2test = unicode
def checkequalnofix(self, result, object, methodname, *args):
method = getattr(object, methodname)
realresult = method(*args)
self.assertEqual(realresult, result)
self.assert_(type(realresult) is type(result))
# if the original is returned make sure that
# this doesn't happen with subclasses
if realresult is object:
class usub(unicode):
def __repr__(self):
return 'usub(%r)' % unicode.__repr__(self)
object = usub(object)
method = getattr(object, methodname)
realresult = method(*args)
self.assertEqual(realresult, result)
self.assert_(object is not realresult)
def test_literals(self):
self.assertEqual(u'\xff', u'\u00ff')
self.assertEqual(u'\uffff', u'\U0000ffff')
self.assertRaises(UnicodeError, eval, 'u\'\\Ufffffffe\'')
self.assertRaises(UnicodeError, eval, 'u\'\\Uffffffff\'')
self.assertRaises(UnicodeError, eval, 'u\'\\U%08x\'' % 0x110000)
def test_repr(self):
if not sys.platform.startswith('java'):
# Test basic sanity of repr()
self.assertEqual(repr(u'abc'), "u'abc'")
self.assertEqual(repr(u'ab\\c'), "u'ab\\\\c'")
self.assertEqual(repr(u'ab\\'), "u'ab\\\\'")
self.assertEqual(repr(u'\\c'), "u'\\\\c'")
self.assertEqual(repr(u'\\'), "u'\\\\'")
self.assertEqual(repr(u'\n'), "u'\\n'")
self.assertEqual(repr(u'\r'), "u'\\r'")
self.assertEqual(repr(u'\t'), "u'\\t'")
self.assertEqual(repr(u'\b'), "u'\\x08'")
self.assertEqual(repr(u"'\""), """u'\\'"'""")
self.assertEqual(repr(u"'\""), """u'\\'"'""")
self.assertEqual(repr(u"'"), '''u"'"''')
self.assertEqual(repr(u'"'), """u'"'""")
latin1repr = (
"u'\\x00\\x01\\x02\\x03\\x04\\x05\\x06\\x07\\x08\\t\\n\\x0b\\x0c\\r"
"\\x0e\\x0f\\x10\\x11\\x12\\x13\\x14\\x15\\x16\\x17\\x18\\x19\\x1a"
"\\x1b\\x1c\\x1d\\x1e\\x1f !\"#$%&\\'()*+,-./0123456789:;<=>?@ABCDEFGHI"
"JKLMNOPQRSTUVWXYZ[\\\\]^_`abcdefghijklmnopqrstuvwxyz{|}~\\x7f"
"\\x80\\x81\\x82\\x83\\x84\\x85\\x86\\x87\\x88\\x89\\x8a\\x8b\\x8c\\x8d"
"\\x8e\\x8f\\x90\\x91\\x92\\x93\\x94\\x95\\x96\\x97\\x98\\x99\\x9a\\x9b"
"\\x9c\\x9d\\x9e\\x9f\\xa0\\xa1\\xa2\\xa3\\xa4\\xa5\\xa6\\xa7\\xa8\\xa9"
"\\xaa\\xab\\xac\\xad\\xae\\xaf\\xb0\\xb1\\xb2\\xb3\\xb4\\xb5\\xb6\\xb7"
"\\xb8\\xb9\\xba\\xbb\\xbc\\xbd\\xbe\\xbf\\xc0\\xc1\\xc2\\xc3\\xc4\\xc5"
"\\xc6\\xc7\\xc8\\xc9\\xca\\xcb\\xcc\\xcd\\xce\\xcf\\xd0\\xd1\\xd2\\xd3"
"\\xd4\\xd5\\xd6\\xd7\\xd8\\xd9\\xda\\xdb\\xdc\\xdd\\xde\\xdf\\xe0\\xe1"
"\\xe2\\xe3\\xe4\\xe5\\xe6\\xe7\\xe8\\xe9\\xea\\xeb\\xec\\xed\\xee\\xef"
"\\xf0\\xf1\\xf2\\xf3\\xf4\\xf5\\xf6\\xf7\\xf8\\xf9\\xfa\\xfb\\xfc\\xfd"
"\\xfe\\xff'")
testrepr = repr(u''.join(map(unichr, xrange(256))))
self.assertEqual(testrepr, latin1repr)
def test_count(self):
string_tests.CommonTest.test_count(self)
# check mixed argument types
self.checkequalnofix(3, 'aaa', 'count', u'a')
self.checkequalnofix(0, 'aaa', 'count', u'b')
self.checkequalnofix(3, u'aaa', 'count', 'a')
self.checkequalnofix(0, u'aaa', 'count', 'b')
self.checkequalnofix(0, u'aaa', 'count', 'b')
self.checkequalnofix(1, u'aaa', 'count', 'a', -1)
self.checkequalnofix(3, u'aaa', 'count', 'a', -10)
self.checkequalnofix(2, u'aaa', 'count', 'a', 0, -1)
self.checkequalnofix(0, u'aaa', 'count', 'a', 0, -10)
def test_find(self):
self.checkequalnofix(0, u'abcdefghiabc', 'find', u'abc')
self.checkequalnofix(9, u'abcdefghiabc', 'find', u'abc', 1)
self.checkequalnofix(-1, u'abcdefghiabc', 'find', u'def', 4)
self.assertRaises(TypeError, u'hello'.find)
self.assertRaises(TypeError, u'hello'.find, 42)
def test_rfind(self):
string_tests.CommonTest.test_rfind(self)
# check mixed argument types
self.checkequalnofix(9, 'abcdefghiabc', 'rfind', u'abc')
self.checkequalnofix(12, 'abcdefghiabc', 'rfind', u'')
self.checkequalnofix(12, u'abcdefghiabc', 'rfind', '')
def test_index(self):
string_tests.CommonTest.test_index(self)
# check mixed argument types
for (t1, t2) in ((str, unicode), (unicode, str)):
self.checkequalnofix(0, t1('abcdefghiabc'), 'index', t2(''))
self.checkequalnofix(3, t1('abcdefghiabc'), 'index', t2('def'))
self.checkequalnofix(0, t1('abcdefghiabc'), 'index', t2('abc'))
self.checkequalnofix(9, t1('abcdefghiabc'), 'index', t2('abc'), 1)
self.assertRaises(ValueError, t1('abcdefghiabc').index, t2('hib'))
self.assertRaises(ValueError, t1('abcdefghiab').index, t2('abc'), 1)
self.assertRaises(ValueError, t1('abcdefghi').index, t2('ghi'), 8)
self.assertRaises(ValueError, t1('abcdefghi').index, t2('ghi'), -1)
def test_rindex(self):
string_tests.CommonTest.test_rindex(self)
# check mixed argument types
for (t1, t2) in ((str, unicode), (unicode, str)):
self.checkequalnofix(12, t1('abcdefghiabc'), 'rindex', t2(''))
self.checkequalnofix(3, t1('abcdefghiabc'), 'rindex', t2('def'))
self.checkequalnofix(9, t1('abcdefghiabc'), 'rindex', t2('abc'))
self.checkequalnofix(0, t1('abcdefghiabc'), 'rindex', t2('abc'), 0, -1)
self.assertRaises(ValueError, t1('abcdefghiabc').rindex, t2('hib'))
self.assertRaises(ValueError, t1('defghiabc').rindex, t2('def'), 1)
self.assertRaises(ValueError, t1('defghiabc').rindex, t2('abc'), 0, -1)
self.assertRaises(ValueError, t1('abcdefghi').rindex, t2('ghi'), 0, 8)
self.assertRaises(ValueError, t1('abcdefghi').rindex, t2('ghi'), 0, -1)
def test_translate(self):
self.checkequalnofix(u'bbbc', u'abababc', 'translate', {ord('a'):None})
self.checkequalnofix(u'iiic', u'abababc', 'translate', {ord('a'):None, ord('b'):ord('i')})
self.checkequalnofix(u'iiix', u'abababc', 'translate', {ord('a'):None, ord('b'):ord('i'), ord('c'):u'x'})
self.checkequalnofix(u'<i><i><i>c', u'abababc', 'translate', {ord('a'):None, ord('b'):u'<i>'})
self.checkequalnofix(u'c', u'abababc', 'translate', {ord('a'):None, ord('b'):u''})
self.checkequalnofix(u'xyyx', u'xzx', 'translate', {ord('z'):u'yy'})
self.assertRaises(TypeError, u'hello'.translate)
self.assertRaises(TypeError, u'abababc'.translate, {ord('a'):''})
def test_split(self):
string_tests.CommonTest.test_split(self)
# Mixed arguments
self.checkequalnofix([u'a', u'b', u'c', u'd'], u'a//b//c//d', 'split', '//')
self.checkequalnofix([u'a', u'b', u'c', u'd'], 'a//b//c//d', 'split', u'//')
self.checkequalnofix([u'endcase ', u''], u'endcase test', 'split', 'test')
def test_join(self):
string_tests.MixinStrUnicodeUserStringTest.test_join(self)
# mixed arguments
self.checkequalnofix(u'a b c d', u' ', 'join', ['a', 'b', u'c', u'd'])
self.checkequalnofix(u'abcd', u'', 'join', (u'a', u'b', u'c', u'd'))
self.checkequalnofix(u'w x y z', u' ', 'join', string_tests.Sequence('wxyz'))
self.checkequalnofix(u'a b c d', ' ', 'join', [u'a', u'b', u'c', u'd'])
self.checkequalnofix(u'a b c d', ' ', 'join', ['a', 'b', u'c', u'd'])
self.checkequalnofix(u'abcd', '', 'join', (u'a', u'b', u'c', u'd'))
self.checkequalnofix(u'w x y z', ' ', 'join', string_tests.Sequence(u'wxyz'))
def test_strip(self):
string_tests.CommonTest.test_strip(self)
self.assertRaises(UnicodeError, u"hello".strip, "\xff")
def test_replace(self):
string_tests.CommonTest.test_replace(self)
# method call forwarded from str implementation because of unicode argument
self.checkequalnofix(u'one@two!three!', 'one!two!three!', 'replace', u'!', u'@', 1)
self.assertRaises(TypeError, 'replace'.replace, u"r", 42)
def test_comparison(self):
# Comparisons:
self.assertEqual(u'abc', 'abc')
self.assertEqual('abc', u'abc')
self.assertEqual(u'abc', u'abc')
self.assert_(u'abcd' > 'abc')
self.assert_('abcd' > u'abc')
self.assert_(u'abcd' > u'abc')
self.assert_(u'abc' < 'abcd')
self.assert_('abc' < u'abcd')
self.assert_(u'abc' < u'abcd')
if 0:
# Move these tests to a Unicode collation module test...
# Testing UTF-16 code point order comparisons...
# No surrogates, no fixup required.
self.assert_(u'\u0061' < u'\u20ac')
# Non surrogate below surrogate value, no fixup required
self.assert_(u'\u0061' < u'\ud800\udc02')
# Non surrogate above surrogate value, fixup required
def test_lecmp(s, s2):
self.assert_(s < s2)
def test_fixup(s):
s2 = u'\ud800\udc01'
test_lecmp(s, s2)
s2 = u'\ud900\udc01'
test_lecmp(s, s2)
s2 = u'\uda00\udc01'
test_lecmp(s, s2)
s2 = u'\udb00\udc01'
test_lecmp(s, s2)
s2 = u'\ud800\udd01'
test_lecmp(s, s2)
s2 = u'\ud900\udd01'
test_lecmp(s, s2)
s2 = u'\uda00\udd01'
test_lecmp(s, s2)
s2 = u'\udb00\udd01'
test_lecmp(s, s2)
s2 = u'\ud800\ude01'
test_lecmp(s, s2)
s2 = u'\ud900\ude01'
test_lecmp(s, s2)
s2 = u'\uda00\ude01'
test_lecmp(s, s2)
s2 = u'\udb00\ude01'
test_lecmp(s, s2)
s2 = u'\ud800\udfff'
test_lecmp(s, s2)
s2 = u'\ud900\udfff'
test_lecmp(s, s2)
s2 = u'\uda00\udfff'
test_lecmp(s, s2)
s2 = u'\udb00\udfff'
test_lecmp(s, s2)
test_fixup(u'\ue000')
test_fixup(u'\uff61')
# Surrogates on both sides, no fixup required
self.assert_(u'\ud800\udc02' < u'\ud84d\udc56')
def test_islower(self):
string_tests.MixinStrUnicodeUserStringTest.test_islower(self)
self.checkequalnofix(False, u'\u1FFc', 'islower')
def test_isupper(self):
string_tests.MixinStrUnicodeUserStringTest.test_isupper(self)
if not sys.platform.startswith('java'):
self.checkequalnofix(False, u'\u1FFc', 'isupper')
def test_istitle(self):
string_tests.MixinStrUnicodeUserStringTest.test_title(self)
self.checkequalnofix(True, u'\u1FFc', 'istitle')
self.checkequalnofix(True, u'Greek \u1FFcitlecases ...', 'istitle')
def test_isspace(self):
string_tests.MixinStrUnicodeUserStringTest.test_isspace(self)
self.checkequalnofix(True, u'\u2000', 'isspace')
self.checkequalnofix(True, u'\u200a', 'isspace')
self.checkequalnofix(False, u'\u2014', 'isspace')
def test_isalpha(self):
string_tests.MixinStrUnicodeUserStringTest.test_isalpha(self)
self.checkequalnofix(True, u'\u1FFc', 'isalpha')
def test_isdecimal(self):
self.checkequalnofix(False, u'', 'isdecimal')
self.checkequalnofix(False, u'a', 'isdecimal')
self.checkequalnofix(True, u'0', 'isdecimal')
self.checkequalnofix(False, u'\u2460', 'isdecimal') # CIRCLED DIGIT ONE
self.checkequalnofix(False, u'\xbc', 'isdecimal') # VULGAR FRACTION ONE QUARTER
self.checkequalnofix(True, u'\u0660', 'isdecimal') # ARABIC-INDIC DIGIT ZERO
self.checkequalnofix(True, u'0123456789', 'isdecimal')
self.checkequalnofix(False, u'0123456789a', 'isdecimal')
self.checkraises(TypeError, 'abc', 'isdecimal', 42)
def test_isdigit(self):
string_tests.MixinStrUnicodeUserStringTest.test_isdigit(self)
self.checkequalnofix(True, u'\u2460', 'isdigit')
self.checkequalnofix(False, u'\xbc', 'isdigit')
self.checkequalnofix(True, u'\u0660', 'isdigit')
def test_isnumeric(self):
self.checkequalnofix(False, u'', 'isnumeric')
self.checkequalnofix(False, u'a', 'isnumeric')
self.checkequalnofix(True, u'0', 'isnumeric')
self.checkequalnofix(True, u'\u2460', 'isnumeric')
self.checkequalnofix(True, u'\xbc', 'isnumeric')
self.checkequalnofix(True, u'\u0660', 'isnumeric')
self.checkequalnofix(True, u'0123456789', 'isnumeric')
self.checkequalnofix(False, u'0123456789a', 'isnumeric')
self.assertRaises(TypeError, u"abc".isnumeric, 42)
def test_contains(self):
# Testing Unicode contains method
self.assert_('a' in u'abdb')
self.assert_('a' in u'bdab')
self.assert_('a' in u'bdaba')
self.assert_('a' in u'bdba')
self.assert_('a' in u'bdba')
self.assert_(u'a' in u'bdba')
self.assert_(u'a' not in u'bdb')
self.assert_(u'a' not in 'bdb')
self.assert_(u'a' in 'bdba')
self.assert_(u'a' in ('a',1,None))
self.assert_(u'a' in (1,None,'a'))
self.assert_(u'a' in (1,None,u'a'))
self.assert_('a' in ('a',1,None))
self.assert_('a' in (1,None,'a'))
self.assert_('a' in (1,None,u'a'))
self.assert_('a' not in ('x',1,u'y'))
self.assert_('a' not in ('x',1,None))
self.assert_(u'abcd' not in u'abcxxxx')
self.assert_(u'ab' in u'abcd')
self.assert_('ab' in u'abc')
self.assert_(u'ab' in 'abc')
self.assert_(u'ab' in (1,None,u'ab'))
self.assert_(u'' in u'abc')
self.assert_('' in u'abc')
# If the following fails either
# the contains operator does not propagate UnicodeErrors or
# someone has changed the default encoding
self.assertRaises(UnicodeError, 'g\xe2teau'.__contains__, u'\xe2')
self.assert_(u'' in '')
self.assert_('' in u'')
self.assert_(u'' in u'')
self.assert_(u'' in 'abc')
self.assert_('' in u'abc')
self.assert_(u'' in u'abc')
self.assert_(u'\0' not in 'abc')
self.assert_('\0' not in u'abc')
self.assert_(u'\0' not in u'abc')
self.assert_(u'\0' in '\0abc')
self.assert_('\0' in u'\0abc')
self.assert_(u'\0' in u'\0abc')
self.assert_(u'\0' in 'abc\0')
self.assert_('\0' in u'abc\0')
self.assert_(u'\0' in u'abc\0')
self.assert_(u'a' in '\0abc')
self.assert_('a' in u'\0abc')
self.assert_(u'a' in u'\0abc')
self.assert_(u'asdf' in 'asdf')
self.assert_('asdf' in u'asdf')
self.assert_(u'asdf' in u'asdf')
self.assert_(u'asdf' not in 'asd')
self.assert_('asdf' not in u'asd')
self.assert_(u'asdf' not in u'asd')
self.assert_(u'asdf' not in '')
self.assert_('asdf' not in u'')
self.assert_(u'asdf' not in u'')
self.assertRaises(TypeError, u"abc".__contains__)
def test_formatting(self):
string_tests.MixinStrUnicodeUserStringTest.test_formatting(self)
# Testing Unicode formatting strings...
self.assertEqual(u"%s, %s" % (u"abc", "abc"), u'abc, abc')
self.assertEqual(u"%s, %s, %i, %f, %5.2f" % (u"abc", "abc", 1, 2, 3), u'abc, abc, 1, 2.000000, 3.00')
self.assertEqual(u"%s, %s, %i, %f, %5.2f" % (u"abc", "abc", 1, -2, 3), u'abc, abc, 1, -2.000000, 3.00')
self.assertEqual(u"%s, %s, %i, %f, %5.2f" % (u"abc", "abc", -1, -2, 3.5), u'abc, abc, -1, -2.000000, 3.50')
self.assertEqual(u"%s, %s, %i, %f, %5.2f" % (u"abc", "abc", -1, -2, 3.57), u'abc, abc, -1, -2.000000, 3.57')
self.assertEqual(u"%s, %s, %i, %f, %5.2f" % (u"abc", "abc", -1, -2, 1003.57), u'abc, abc, -1, -2.000000, 1003.57')
if not sys.platform.startswith('java'):
self.assertEqual(u"%r, %r" % (u"abc", "abc"), u"u'abc', 'abc'")
self.assertEqual(u"%(x)s, %(y)s" % {'x':u"abc", 'y':"def"}, u'abc, def')
self.assertEqual(u"%(x)s, %(\xfc)s" % {'x':u"abc", u'\xfc':"def"}, u'abc, def')
self.assertEqual(u'%c' % 0x1234, u'\u1234')
self.assertRaises(OverflowError, u"%c".__mod__, (sys.maxunicode+1,))
# formatting jobs delegated from the string implementation:
self.assertEqual('...%(foo)s...' % {'foo':u"abc"}, u'...abc...')
self.assertEqual('...%(foo)s...' % {'foo':"abc"}, '...abc...')
self.assertEqual('...%(foo)s...' % {u'foo':"abc"}, '...abc...')
self.assertEqual('...%(foo)s...' % {u'foo':u"abc"}, u'...abc...')
self.assertEqual('...%(foo)s...' % {u'foo':u"abc",'def':123}, u'...abc...')
self.assertEqual('...%(foo)s...' % {u'foo':u"abc",u'def':123}, u'...abc...')
self.assertEqual('...%s...%s...%s...%s...' % (1,2,3,u"abc"), u'...1...2...3...abc...')
self.assertEqual('...%%...%%s...%s...%s...%s...%s...' % (1,2,3,u"abc"), u'...%...%s...1...2...3...abc...')
self.assertEqual('...%s...' % u"abc", u'...abc...')
self.assertEqual('%*s' % (5,u'abc',), u' abc')
self.assertEqual('%*s' % (-5,u'abc',), u'abc ')
self.assertEqual('%*.*s' % (5,2,u'abc',), u' ab')
self.assertEqual('%*.*s' % (5,3,u'abc',), u' abc')
self.assertEqual('%i %*.*s' % (10, 5,3,u'abc',), u'10 abc')
self.assertEqual('%i%s %*.*s' % (10, 3, 5, 3, u'abc',), u'103 abc')
self.assertEqual('%c' % u'a', u'a')
class Wrapper:
def __str__(self):
return u'\u1234'
self.assertEqual('%s' % Wrapper(), u'\u1234')
def test_constructor(self):
# unicode(obj) tests (this maps to PyObject_Unicode() at C level)
self.assertEqual(
unicode(u'unicode remains unicode'),
u'unicode remains unicode'
)
class UnicodeSubclass(unicode):
pass
self.assertEqual(
unicode(UnicodeSubclass('unicode subclass becomes unicode')),
u'unicode subclass becomes unicode'
)
self.assertEqual(
unicode('strings are converted to unicode'),
u'strings are converted to unicode'
)
class UnicodeCompat:
def __init__(self, x):
self.x = x
def __unicode__(self):
return self.x
self.assertEqual(
unicode(UnicodeCompat('__unicode__ compatible objects are recognized')),
u'__unicode__ compatible objects are recognized')
class StringCompat:
def __init__(self, x):
self.x = x
def __str__(self):
return self.x
self.assertEqual(
unicode(StringCompat('__str__ compatible objects are recognized')),
u'__str__ compatible objects are recognized'
)
# unicode(obj) is compatible to str():
o = StringCompat('unicode(obj) is compatible to str()')
self.assertEqual(unicode(o), u'unicode(obj) is compatible to str()')
self.assertEqual(str(o), 'unicode(obj) is compatible to str()')
# %-formatting and .__unicode__()
self.assertEqual(u'%s' %
UnicodeCompat(u"u'%s' % obj uses obj.__unicode__()"),
u"u'%s' % obj uses obj.__unicode__()")
self.assertEqual(u'%s' %
UnicodeCompat(u"u'%s' % obj falls back to obj.__str__()"),
u"u'%s' % obj falls back to obj.__str__()")
for obj in (123, 123.45, 123L):
self.assertEqual(unicode(obj), unicode(str(obj)))
# unicode(obj, encoding, error) tests (this maps to
# PyUnicode_FromEncodedObject() at C level)
if not sys.platform.startswith('java'):
self.assertRaises(
TypeError,
unicode,
u'decoding unicode is not supported',
'utf-8',
'strict'
)
self.assertEqual(
unicode('strings are decoded to unicode', 'utf-8', 'strict'),
u'strings are decoded to unicode'
)
if not sys.platform.startswith('java'):
self.assertEqual(
unicode(
buffer('character buffers are decoded to unicode'),
'utf-8',
'strict'
),
u'character buffers are decoded to unicode'
)
self.assertRaises(TypeError, unicode, 42, 42, 42)
def test_codecs_utf7(self):
utfTests = [
(u'A\u2262\u0391.', 'A+ImIDkQ.'), # RFC2152 example
(u'Hi Mom -\u263a-!', 'Hi Mom -+Jjo--!'), # RFC2152 example
(u'\u65E5\u672C\u8A9E', '+ZeVnLIqe-'), # RFC2152 example
(u'Item 3 is \u00a31.', 'Item 3 is +AKM-1.'), # RFC2152 example
(u'+', '+-'),
(u'+-', '+--'),
(u'+?', '+-?'),
(u'\?', '+AFw?'),
(u'+?', '+-?'),
(ur'\\?', '+AFwAXA?'),
(ur'\\\?', '+AFwAXABc?'),
(ur'++--', '+-+---')
]
for (x, y) in utfTests:
self.assertEqual(x.encode('utf-7'), y)
# surrogates not supported
self.assertRaises(UnicodeError, unicode, '+3ADYAA-', 'utf-7')
self.assertEqual(unicode('+3ADYAA-', 'utf-7', 'replace'), u'\ufffd')
def test_codecs_utf8(self):
self.assertEqual(u''.encode('utf-8'), '')
self.assertEqual(u'\u20ac'.encode('utf-8'), '\xe2\x82\xac')
self.assertEqual(u'\ud800\udc02'.encode('utf-8'), '\xf0\x90\x80\x82')
self.assertEqual(u'\ud84d\udc56'.encode('utf-8'), '\xf0\xa3\x91\x96')
self.assertEqual(u'\ud800'.encode('utf-8'), '\xed\xa0\x80')
self.assertEqual(u'\udc00'.encode('utf-8'), '\xed\xb0\x80')
self.assertEqual(
(u'\ud800\udc02'*1000).encode('utf-8'),
'\xf0\x90\x80\x82'*1000
)
self.assertEqual(
u'\u6b63\u78ba\u306b\u8a00\u3046\u3068\u7ffb\u8a33\u306f'
u'\u3055\u308c\u3066\u3044\u307e\u305b\u3093\u3002\u4e00'
u'\u90e8\u306f\u30c9\u30a4\u30c4\u8a9e\u3067\u3059\u304c'
u'\u3001\u3042\u3068\u306f\u3067\u305f\u3089\u3081\u3067'
u'\u3059\u3002\u5b9f\u969b\u306b\u306f\u300cWenn ist das'
u' Nunstuck git und'.encode('utf-8'),
'\xe6\xad\xa3\xe7\xa2\xba\xe3\x81\xab\xe8\xa8\x80\xe3\x81'
'\x86\xe3\x81\xa8\xe7\xbf\xbb\xe8\xa8\xb3\xe3\x81\xaf\xe3'
'\x81\x95\xe3\x82\x8c\xe3\x81\xa6\xe3\x81\x84\xe3\x81\xbe'
'\xe3\x81\x9b\xe3\x82\x93\xe3\x80\x82\xe4\xb8\x80\xe9\x83'
'\xa8\xe3\x81\xaf\xe3\x83\x89\xe3\x82\xa4\xe3\x83\x84\xe8'
'\xaa\x9e\xe3\x81\xa7\xe3\x81\x99\xe3\x81\x8c\xe3\x80\x81'
'\xe3\x81\x82\xe3\x81\xa8\xe3\x81\xaf\xe3\x81\xa7\xe3\x81'
'\x9f\xe3\x82\x89\xe3\x82\x81\xe3\x81\xa7\xe3\x81\x99\xe3'
'\x80\x82\xe5\xae\x9f\xe9\x9a\x9b\xe3\x81\xab\xe3\x81\xaf'
'\xe3\x80\x8cWenn ist das Nunstuck git und'
)
# UTF-8 specific decoding tests
self.assertEqual(unicode('\xf0\xa3\x91\x96', 'utf-8'), u'\U00023456' )
self.assertEqual(unicode('\xf0\x90\x80\x82', 'utf-8'), u'\U00010002' )
self.assertEqual(unicode('\xe2\x82\xac', 'utf-8'), u'\u20ac' )
# Other possible utf-8 test cases:
# * strict decoding testing for all of the
# UTF8_ERROR cases in PyUnicode_DecodeUTF8
def test_codecs_idna(self):
# Test whether trailing dot is preserved
self.assertEqual(u"www.python.org.".encode("idna"), "www.python.org.")
def test_codecs_errors(self):
# Error handling (encoding)
self.assertRaises(UnicodeError, u'Andr\202 x'.encode, 'ascii')
self.assertRaises(UnicodeError, u'Andr\202 x'.encode, 'ascii','strict')
self.assertEqual(u'Andr\202 x'.encode('ascii','ignore'), "Andr x")
self.assertEqual(u'Andr\202 x'.encode('ascii','replace'), "Andr? x")
# Error handling (decoding)
self.assertRaises(UnicodeError, unicode, 'Andr\202 x', 'ascii')
self.assertRaises(UnicodeError, unicode, 'Andr\202 x', 'ascii','strict')
self.assertEqual(unicode('Andr\202 x','ascii','ignore'), u"Andr x")
self.assertEqual(unicode('Andr\202 x','ascii','replace'), u'Andr\uFFFD x')
# Error handling (unknown character names)
self.assertEqual("\\N{foo}xx".decode("unicode-escape", "ignore"), u"xx")
# Error handling (truncated escape sequence)
self.assertRaises(UnicodeError, "\\".decode, "unicode-escape")
# Error handling (bad decoder return)
def search_function(encoding):
def decode1(input, errors="strict"):
return 42 # not a tuple
def encode1(input, errors="strict"):
return 42 # not a tuple
def encode2(input, errors="strict"):
return (42, 42) # no unicode
def decode2(input, errors="strict"):
return (42, 42) # no unicode
if encoding=="test.unicode1":
return (encode1, decode1, None, None)
elif encoding=="test.unicode2":
return (encode2, decode2, None, None)
else:
return None
codecs.register(search_function)
self.assertRaises(TypeError, "hello".decode, "test.unicode1")
self.assertRaises(TypeError, unicode, "hello", "test.unicode2")
self.assertRaises(TypeError, u"hello".encode, "test.unicode1")
self.assertRaises(TypeError, u"hello".encode, "test.unicode2")
# executes PyUnicode_Encode()
import imp
self.assertRaises(
ImportError,
imp.find_module,
"non-existing module",
[u"non-existing dir"]
)
# Error handling (wrong arguments)
self.assertRaises(TypeError, u"hello".encode, 42, 42, 42)
# Error handling (PyUnicode_EncodeDecimal())
self.assertRaises(UnicodeError, int, u"\u0200")
def test_codecs(self):
# Encoding
self.assertEqual(u'hello'.encode('ascii'), 'hello')
self.assertEqual(u'hello'.encode('utf-7'), 'hello')
self.assertEqual(u'hello'.encode('utf-8'), 'hello')
self.assertEqual(u'hello'.encode('utf8'), 'hello')
self.assertEqual(u'hello'.encode('utf-16-le'), 'h\000e\000l\000l\000o\000')
self.assertEqual(u'hello'.encode('utf-16-be'), '\000h\000e\000l\000l\000o')
self.assertEqual(u'hello'.encode('latin-1'), 'hello')
# Roundtrip safety for BMP (just the first 1024 chars)
u = u''.join(map(unichr, xrange(1024)))
for encoding in ('utf-7', 'utf-8', 'utf-16', 'utf-16-le', 'utf-16-be',
'raw_unicode_escape', 'unicode_escape', 'unicode_internal'):
self.assertEqual(unicode(u.encode(encoding),encoding), u)
# Roundtrip safety for BMP (just the first 256 chars)
u = u''.join(map(unichr, xrange(256)))
for encoding in ('latin-1',):
self.assertEqual(unicode(u.encode(encoding),encoding), u)
# Roundtrip safety for BMP (just the first 128 chars)
u = u''.join(map(unichr, xrange(128)))
for encoding in ('ascii',):
self.assertEqual(unicode(u.encode(encoding),encoding), u)
# Roundtrip safety for non-BMP (just a few chars)
u = u'\U00010001\U00020002\U00030003\U00040004\U00050005'
for encoding in ('utf-8', 'utf-16', 'utf-16-le', 'utf-16-be',
#'raw_unicode_escape',
'unicode_escape', 'unicode_internal'):
self.assertEqual(unicode(u.encode(encoding),encoding), u)
# UTF-8 must be roundtrip safe for all UCS-2 code points
# This excludes surrogates: in the full range, there would be
# a surrogate pair (\udbff\udc00), which gets converted back
# to a non-BMP character (\U0010fc00)
u = u''.join(map(unichr, range(0,0xd800)+range(0xe000,0x10000)))
for encoding in ('utf-8',):
self.assertEqual(unicode(u.encode(encoding),encoding), u)
def test_codecs_charmap(self):
# 0-127
s = ''.join(map(chr, xrange(128)))
for encoding in (
'cp037', 'cp1026',
'cp437', 'cp500', 'cp737', 'cp775', 'cp850',
'cp852', 'cp855', 'cp860', 'cp861', 'cp862',
'cp863', 'cp865', 'cp866',
'iso8859_10', 'iso8859_13', 'iso8859_14', 'iso8859_15',
'iso8859_2', 'iso8859_3', 'iso8859_4', 'iso8859_5', 'iso8859_6',
'iso8859_7', 'iso8859_9', 'koi8_r', 'latin_1',
'mac_cyrillic', 'mac_latin2',
'cp1250', 'cp1251', 'cp1252', 'cp1253', 'cp1254', 'cp1255',
'cp1256', 'cp1257', 'cp1258',
'cp856', 'cp857', 'cp864', 'cp869', 'cp874',
'mac_greek', 'mac_iceland','mac_roman', 'mac_turkish',
'cp1006', 'iso8859_8',
### These have undefined mappings:
#'cp424',
### These fail the round-trip:
#'cp875'
):
self.assertEqual(unicode(s, encoding).encode(encoding), s)
# 128-255
s = ''.join(map(chr, xrange(128, 256)))
for encoding in (
'cp037', 'cp1026',
'cp437', 'cp500', 'cp737', 'cp775', 'cp850',
'cp852', 'cp855', 'cp860', 'cp861', 'cp862',
'cp863', 'cp865', 'cp866',
'iso8859_10', 'iso8859_13', 'iso8859_14', 'iso8859_15',
'iso8859_2', 'iso8859_4', 'iso8859_5',
'iso8859_9', 'koi8_r', 'latin_1',
'mac_cyrillic', 'mac_latin2',
### These have undefined mappings:
#'cp1250', 'cp1251', 'cp1252', 'cp1253', 'cp1254', 'cp1255',
#'cp1256', 'cp1257', 'cp1258',
#'cp424', 'cp856', 'cp857', 'cp864', 'cp869', 'cp874',
#'iso8859_3', 'iso8859_6', 'iso8859_7',
#'mac_greek', 'mac_iceland','mac_roman', 'mac_turkish',
### These fail the round-trip:
#'cp1006', 'cp875', 'iso8859_8',
):
self.assertEqual(unicode(s, encoding).encode(encoding), s)
def test_concatenation(self):
self.assertEqual((u"abc" u"def"), u"abcdef")
self.assertEqual(("abc" u"def"), u"abcdef")
self.assertEqual((u"abc" "def"), u"abcdef")
self.assertEqual((u"abc" u"def" "ghi"), u"abcdefghi")
self.assertEqual(("abc" "def" u"ghi"), u"abcdefghi")
def test_printing(self):
class BitBucket:
def write(self, text):
pass
out = BitBucket()
print >>out, u'abc'
print >>out, u'abc', u'def'
print >>out, u'abc', 'def'
print >>out, 'abc', u'def'
print >>out, u'abc\n'
print >>out, u'abc\n',
print >>out, u'abc\n',
print >>out, u'def\n'
print >>out, u'def\n'
def test_ucs4(self):
if sys.maxunicode == 0xFFFF:
return
x = u'\U00100000'
y = x.encode("raw-unicode-escape").decode("raw-unicode-escape")
self.assertEqual(x, y)
def test_conversion(self):
# Make sure __unicode__() works properly
class Foo0:
def __str__(self):
return "foo"
class Foo1:
def __unicode__(self):
return u"foo"
class Foo2(object):
def __unicode__(self):
return u"foo"
class Foo3(object):
def __unicode__(self):
return "foo"
class Foo4(str):
def __unicode__(self):
return "foo"
class Foo5(unicode):
def __unicode__(self):
return "foo"
class Foo6(str):
def __str__(self):
return "foos"
def __unicode__(self):
return u"foou"
class Foo7(unicode):
def __str__(self):
return "foos"
def __unicode__(self):
return u"foou"
class Foo8(unicode):
def __new__(cls, content=""):
return unicode.__new__(cls, 2*content)
def __unicode__(self):
return self
class Foo9(unicode):
def __str__(self):
return "string"
def __unicode__(self):
return "not unicode"
self.assertEqual(unicode(Foo0()), u"foo")
self.assertEqual(unicode(Foo1()), u"foo")
self.assertEqual(unicode(Foo2()), u"foo")
self.assertEqual(unicode(Foo3()), u"foo")
self.assertEqual(unicode(Foo4("bar")), u"foo")
self.assertEqual(unicode(Foo5("bar")), u"foo")
self.assertEqual(unicode(Foo6("bar")), u"foou")
self.assertEqual(unicode(Foo7("bar")), u"foou")
self.assertEqual(unicode(Foo8("foo")), u"foofoo")
self.assertEqual(str(Foo9("foo")), "string")
self.assertEqual(unicode(Foo9("foo")), u"not unicode")
def test_main():
test_support.run_unittest(UnicodeTest)
if __name__ == "__main__":
test_main()