summaryrefslogtreecommitdiffstats
path: root/Lib/test/test_multibytecodec_support.py
diff options
context:
space:
mode:
Diffstat (limited to 'Lib/test/test_multibytecodec_support.py')
-rw-r--r--Lib/test/test_multibytecodec_support.py40
1 files changed, 20 insertions, 20 deletions
diff --git a/Lib/test/test_multibytecodec_support.py b/Lib/test/test_multibytecodec_support.py
index d94d115..6abcdd6 100644
--- a/Lib/test/test_multibytecodec_support.py
+++ b/Lib/test/test_multibytecodec_support.py
@@ -18,7 +18,7 @@ class TestBase:
roundtriptest = 1 # set if roundtrip is possible with unicode
has_iso10646 = 0 # set if this encoding contains whole iso10646 map
xmlcharnametest = None # string to test xmlcharrefreplace
- unmappedunicode = u'\udeee' # a unicode codepoint that is not mapped.
+ unmappedunicode = '\udeee' # a unicode codepoint that is not mapped.
def setUp(self):
if self.codec is None:
@@ -54,7 +54,7 @@ class TestBase:
if self.has_iso10646:
return
- s = u"\u0b13\u0b23\u0b60 nd eggs"
+ s = "\u0b13\u0b23\u0b60 nd eggs"
self.assertEqual(
self.encode(s, "xmlcharrefreplace")[0],
"ଓଣୠ nd eggs"
@@ -72,17 +72,17 @@ class TestBase:
l = []
for c in exc.object[exc.start:exc.end]:
if ord(c) in codepoint2name:
- l.append(u"&%s;" % codepoint2name[ord(c)])
+ l.append("&%s;" % codepoint2name[ord(c)])
else:
- l.append(u"&#%d;" % ord(c))
- return (u"".join(l), exc.end)
+ l.append("&#%d;" % ord(c))
+ return ("".join(l), exc.end)
codecs.register_error("test.xmlcharnamereplace", xmlcharnamereplace)
if self.xmlcharnametest:
sin, sout = self.xmlcharnametest
else:
- sin = u"\xab\u211c\xbb = \u2329\u1234\u232a"
+ sin = "\xab\u211c\xbb = \u2329\u1234\u232a"
sout = "«ℜ» = ⟨ሴ⟩"
self.assertEqual(self.encode(sin,
"test.xmlcharnamereplace")[0], sout)
@@ -98,20 +98,20 @@ class TestBase:
def test_callback_long_index(self):
def myreplace(exc):
- return (u'x', int(exc.end))
+ return ('x', int(exc.end))
codecs.register_error("test.cjktest", myreplace)
- self.assertEqual(self.encode(u'abcd' + self.unmappedunicode + u'efgh',
+ self.assertEqual(self.encode('abcd' + self.unmappedunicode + 'efgh',
'test.cjktest'), ('abcdxefgh', 9))
def myreplace(exc):
- return (u'x', sys.maxint + 1)
+ return ('x', sys.maxint + 1)
codecs.register_error("test.cjktest", myreplace)
self.assertRaises(IndexError, self.encode, self.unmappedunicode,
'test.cjktest')
def test_callback_None_index(self):
def myreplace(exc):
- return (u'x', None)
+ return ('x', None)
codecs.register_error("test.cjktest", myreplace)
self.assertRaises(TypeError, self.encode, self.unmappedunicode,
'test.cjktest')
@@ -120,25 +120,25 @@ class TestBase:
def myreplace(exc):
if myreplace.limit > 0:
myreplace.limit -= 1
- return (u'REPLACED', 0)
+ return ('REPLACED', 0)
else:
- return (u'TERMINAL', exc.end)
+ return ('TERMINAL', exc.end)
myreplace.limit = 3
codecs.register_error("test.cjktest", myreplace)
- self.assertEqual(self.encode(u'abcd' + self.unmappedunicode + u'efgh',
+ self.assertEqual(self.encode('abcd' + self.unmappedunicode + 'efgh',
'test.cjktest'),
('abcdREPLACEDabcdREPLACEDabcdREPLACEDabcdTERMINALefgh', 9))
def test_callback_forward_index(self):
def myreplace(exc):
- return (u'REPLACED', exc.end + 2)
+ return ('REPLACED', exc.end + 2)
codecs.register_error("test.cjktest", myreplace)
- self.assertEqual(self.encode(u'abcd' + self.unmappedunicode + u'efgh',
+ self.assertEqual(self.encode('abcd' + self.unmappedunicode + 'efgh',
'test.cjktest'), ('abcdREPLACEDgh', 9))
def test_callback_index_outofbound(self):
def myreplace(exc):
- return (u'TERM', 100)
+ return ('TERM', 100)
codecs.register_error("test.cjktest", myreplace)
self.assertRaises(IndexError, self.encode, self.unmappedunicode,
'test.cjktest')
@@ -191,7 +191,7 @@ class TestBase:
e.reset()
def tempreplace(exc):
- return (u'called', exc.end)
+ return ('called', exc.end)
codecs.register_error('test.incremental_error_callback', tempreplace)
e.errors = 'test.incremental_error_callback'
self.assertEqual(e.encode(inv, True), 'called')
@@ -243,7 +243,7 @@ class TestBase:
self.assertEqual(ostream.getvalue(), self.tstring[0])
-if len(u'\U00012345') == 2: # ucs2 build
+if len('\U00012345') == 2: # ucs2 build
_unichr = unichr
def unichr(v):
if v >= 0x10000:
@@ -272,7 +272,7 @@ class TestBase_Mapping(unittest.TestCase):
return test_support.open_urlresource(self.mapfileurl)
def test_mapping_file(self):
- unichrs = lambda s: u''.join(map(unichr, map(eval, s.split('+'))))
+ unichrs = lambda s: ''.join(map(unichr, map(eval, s.split('+'))))
urt_wa = {}
for line in self.open_mapping_file():
@@ -311,7 +311,7 @@ class TestBase_Mapping(unittest.TestCase):
if (csetch, unich) not in self.pass_enctest:
self.assertEqual(unich.encode(self.encoding), csetch)
if (csetch, unich) not in self.pass_dectest:
- self.assertEqual(unicode(csetch, self.encoding), unich)
+ self.assertEqual(str(csetch, self.encoding), unich)
def load_teststring(encoding):
from test import cjkencodings_test