blob: ee1e28a763ca5ce30acfb949b7a8eefa2adbf547 [file] [log] [blame]
import codecs
import html.entities
import sys
import test.support
import unicodedata
import unittest
import warnings
class PosReturn:
# this can be used for configurable callbacks
def __init__(self):
self.pos = 0
def handle(self, exc):
oldpos = self.pos
realpos = oldpos
if realpos<0:
realpos = len(exc.object) + realpos
# if we don't advance this time, terminate on the next call
# otherwise we'd get an endless loop
if realpos <= exc.start:
self.pos = len(exc.object)
return ("<?>", oldpos)
# A UnicodeEncodeError object with a bad start attribute
class BadStartUnicodeEncodeError(UnicodeEncodeError):
def __init__(self):
UnicodeEncodeError.__init__(self, "ascii", "", 0, 1, "bad")
self.start = []
# A UnicodeEncodeError object with a bad object attribute
class BadObjectUnicodeEncodeError(UnicodeEncodeError):
def __init__(self):
UnicodeEncodeError.__init__(self, "ascii", "", 0, 1, "bad")
self.object = []
# A UnicodeDecodeError object without an end attribute
class NoEndUnicodeDecodeError(UnicodeDecodeError):
def __init__(self):
UnicodeDecodeError.__init__(self, "ascii", bytearray(b""), 0, 1, "bad")
del self.end
# A UnicodeDecodeError object with a bad object attribute
class BadObjectUnicodeDecodeError(UnicodeDecodeError):
def __init__(self):
UnicodeDecodeError.__init__(self, "ascii", bytearray(b""), 0, 1, "bad")
self.object = []
# A UnicodeTranslateError object without a start attribute
class NoStartUnicodeTranslateError(UnicodeTranslateError):
def __init__(self):
UnicodeTranslateError.__init__(self, "", 0, 1, "bad")
del self.start
# A UnicodeTranslateError object without an end attribute
class NoEndUnicodeTranslateError(UnicodeTranslateError):
def __init__(self):
UnicodeTranslateError.__init__(self, "", 0, 1, "bad")
del self.end
# A UnicodeTranslateError object without an object attribute
class NoObjectUnicodeTranslateError(UnicodeTranslateError):
def __init__(self):
UnicodeTranslateError.__init__(self, "", 0, 1, "bad")
del self.object
class CodecCallbackTest(unittest.TestCase):
def test_xmlcharrefreplace(self):
# replace unencodable characters which numeric character entities.
# For ascii, latin-1 and charmaps this is completely implemented
# in C and should be reasonably fast.
s = "\u30b9\u30d1\u30e2 \xe4nd eggs"
self.assertEqual(
s.encode("ascii", "xmlcharrefreplace"),
b"&#12473;&#12497;&#12514; &#228;nd eggs"
)
self.assertEqual(
s.encode("latin-1", "xmlcharrefreplace"),
b"&#12473;&#12497;&#12514; \xe4nd eggs"
)
def test_xmlcharnamereplace(self):
# This time use a named character entity for unencodable
# characters, if one is available.
def xmlcharnamereplace(exc):
if not isinstance(exc, UnicodeEncodeError):
raise TypeError("don't know how to handle %r" % exc)
l = []
for c in exc.object[exc.start:exc.end]:
try:
l.append("&%s;" % html.entities.codepoint2name[ord(c)])
except KeyError:
l.append("&#%d;" % ord(c))
return ("".join(l), exc.end)
codecs.register_error(
"test.xmlcharnamereplace", xmlcharnamereplace)
sin = "\xab\u211c\xbb = \u2329\u1234\u20ac\u232a"
sout = b"&laquo;&real;&raquo; = &lang;&#4660;&euro;&rang;"
self.assertEqual(sin.encode("ascii", "test.xmlcharnamereplace"), sout)
sout = b"\xab&real;\xbb = &lang;&#4660;&euro;&rang;"
self.assertEqual(sin.encode("latin-1", "test.xmlcharnamereplace"), sout)
sout = b"\xab&real;\xbb = &lang;&#4660;\xa4&rang;"
self.assertEqual(sin.encode("iso-8859-15", "test.xmlcharnamereplace"), sout)
def test_uninamereplace(self):
# We're using the names from the unicode database this time,
# and we're doing "syntax highlighting" here, i.e. we include
# the replaced text in ANSI escape sequences. For this it is
# useful that the error handler is not called for every single
# unencodable character, but for a complete sequence of
# unencodable characters, otherwise we would output many
# unnecessary escape sequences.
def uninamereplace(exc):
if not isinstance(exc, UnicodeEncodeError):
raise TypeError("don't know how to handle %r" % exc)
l = []
for c in exc.object[exc.start:exc.end]:
l.append(unicodedata.name(c, "0x%x" % ord(c)))
return ("\033[1m%s\033[0m" % ", ".join(l), exc.end)
codecs.register_error(
"test.uninamereplace", uninamereplace)
sin = "\xac\u1234\u20ac\u8000"
sout = b"\033[1mNOT SIGN, ETHIOPIC SYLLABLE SEE, EURO SIGN, CJK UNIFIED IDEOGRAPH-8000\033[0m"
self.assertEqual(sin.encode("ascii", "test.uninamereplace"), sout)
sout = b"\xac\033[1mETHIOPIC SYLLABLE SEE, EURO SIGN, CJK UNIFIED IDEOGRAPH-8000\033[0m"
self.assertEqual(sin.encode("latin-1", "test.uninamereplace"), sout)
sout = b"\xac\033[1mETHIOPIC SYLLABLE SEE\033[0m\xa4\033[1mCJK UNIFIED IDEOGRAPH-8000\033[0m"
self.assertEqual(sin.encode("iso-8859-15", "test.uninamereplace"), sout)
def test_backslashescape(self):
# Does the same as the "unicode-escape" encoding, but with different
# base encodings.
sin = "a\xac\u1234\u20ac\u8000\U0010ffff"
sout = b"a\\xac\\u1234\\u20ac\\u8000\\U0010ffff"
self.assertEqual(sin.encode("ascii", "backslashreplace"), sout)
sout = b"a\xac\\u1234\\u20ac\\u8000\\U0010ffff"
self.assertEqual(sin.encode("latin-1", "backslashreplace"), sout)
sout = b"a\xac\\u1234\xa4\\u8000\\U0010ffff"
self.assertEqual(sin.encode("iso-8859-15", "backslashreplace"), sout)
def test_nameescape(self):
# Does the same as backslashescape, but prefers ``\N{...}`` escape
# sequences.
sin = "a\xac\u1234\u20ac\u8000\U0010ffff"
sout = (b'a\\N{NOT SIGN}\\N{ETHIOPIC SYLLABLE SEE}\\N{EURO SIGN}'
b'\\N{CJK UNIFIED IDEOGRAPH-8000}\\U0010ffff')
self.assertEqual(sin.encode("ascii", "namereplace"), sout)
sout = (b'a\xac\\N{ETHIOPIC SYLLABLE SEE}\\N{EURO SIGN}'
b'\\N{CJK UNIFIED IDEOGRAPH-8000}\\U0010ffff')
self.assertEqual(sin.encode("latin-1", "namereplace"), sout)
sout = (b'a\xac\\N{ETHIOPIC SYLLABLE SEE}\xa4'
b'\\N{CJK UNIFIED IDEOGRAPH-8000}\\U0010ffff')
self.assertEqual(sin.encode("iso-8859-15", "namereplace"), sout)
def test_decoding_callbacks(self):
# This is a test for a decoding callback handler
# that allows the decoding of the invalid sequence
# "\xc0\x80" and returns "\x00" instead of raising an error.
# All other illegal sequences will be handled strictly.
def relaxedutf8(exc):
if not isinstance(exc, UnicodeDecodeError):
raise TypeError("don't know how to handle %r" % exc)
if exc.object[exc.start:exc.start+2] == b"\xc0\x80":
return ("\x00", exc.start+2) # retry after two bytes
else:
raise exc
codecs.register_error("test.relaxedutf8", relaxedutf8)
# all the "\xc0\x80" will be decoded to "\x00"
sin = b"a\x00b\xc0\x80c\xc3\xbc\xc0\x80\xc0\x80"
sout = "a\x00b\x00c\xfc\x00\x00"
self.assertEqual(sin.decode("utf-8", "test.relaxedutf8"), sout)
# "\xc0\x81" is not valid and a UnicodeDecodeError will be raised
sin = b"\xc0\x80\xc0\x81"
self.assertRaises(UnicodeDecodeError, sin.decode,
"utf-8", "test.relaxedutf8")
def test_charmapencode(self):
# For charmap encodings the replacement string will be
# mapped through the encoding again. This means, that
# to be able to use e.g. the "replace" handler, the
# charmap has to have a mapping for "?".
charmap = dict((ord(c), bytes(2*c.upper(), 'ascii')) for c in "abcdefgh")
sin = "abc"
sout = b"AABBCC"
self.assertEqual(codecs.charmap_encode(sin, "strict", charmap)[0], sout)
sin = "abcA"
self.assertRaises(UnicodeError, codecs.charmap_encode, sin, "strict", charmap)
charmap[ord("?")] = b"XYZ"
sin = "abcDEF"
sout = b"AABBCCXYZXYZXYZ"
self.assertEqual(codecs.charmap_encode(sin, "replace", charmap)[0], sout)
charmap[ord("?")] = "XYZ" # wrong type in mapping
self.assertRaises(TypeError, codecs.charmap_encode, sin, "replace", charmap)
def test_decodeunicodeinternal(self):
with test.support.check_warnings(('unicode_internal codec has been '
'deprecated', DeprecationWarning)):
self.assertRaises(
UnicodeDecodeError,
b"\x00\x00\x00\x00\x00".decode,
"unicode-internal",
)
if len('\0'.encode('unicode-internal')) == 4:
def handler_unicodeinternal(exc):
if not isinstance(exc, UnicodeDecodeError):
raise TypeError("don't know how to handle %r" % exc)
return ("\x01", 1)
self.assertEqual(
b"\x00\x00\x00\x00\x00".decode("unicode-internal", "ignore"),
"\u0000"
)
self.assertEqual(
b"\x00\x00\x00\x00\x00".decode("unicode-internal", "replace"),
"\u0000\ufffd"
)
self.assertEqual(
b"\x00\x00\x00\x00\x00".decode("unicode-internal", "backslashreplace"),
"\u0000\\x00"
)
codecs.register_error("test.hui", handler_unicodeinternal)
self.assertEqual(
b"\x00\x00\x00\x00\x00".decode("unicode-internal", "test.hui"),
"\u0000\u0001\u0000"
)
def test_callbacks(self):
def handler1(exc):
r = range(exc.start, exc.end)
if isinstance(exc, UnicodeEncodeError):
l = ["<%d>" % ord(exc.object[pos]) for pos in r]
elif isinstance(exc, UnicodeDecodeError):
l = ["<%d>" % exc.object[pos] for pos in r]
else:
raise TypeError("don't know how to handle %r" % exc)
return ("[%s]" % "".join(l), exc.end)
codecs.register_error("test.handler1", handler1)
def handler2(exc):
if not isinstance(exc, UnicodeDecodeError):
raise TypeError("don't know how to handle %r" % exc)
l = ["<%d>" % exc.object[pos] for pos in range(exc.start, exc.end)]
return ("[%s]" % "".join(l), exc.end+1) # skip one character
codecs.register_error("test.handler2", handler2)
s = b"\x00\x81\x7f\x80\xff"
self.assertEqual(
s.decode("ascii", "test.handler1"),
"\x00[<129>]\x7f[<128>][<255>]"
)
self.assertEqual(
s.decode("ascii", "test.handler2"),
"\x00[<129>][<128>]"
)
self.assertEqual(
b"\\u3042\u3xxx".decode("unicode-escape", "test.handler1"),
"\u3042[<92><117><51>]xxx"
)
self.assertEqual(
b"\\u3042\u3xx".decode("unicode-escape", "test.handler1"),
"\u3042[<92><117><51>]xx"
)
self.assertEqual(
codecs.charmap_decode(b"abc", "test.handler1", {ord("a"): "z"})[0],
"z[<98>][<99>]"
)
self.assertEqual(
"g\xfc\xdfrk".encode("ascii", "test.handler1"),
b"g[<252><223>]rk"
)
self.assertEqual(
"g\xfc\xdf".encode("ascii", "test.handler1"),
b"g[<252><223>]"
)
def test_longstrings(self):
# test long strings to check for memory overflow problems
errors = [ "strict", "ignore", "replace", "xmlcharrefreplace",
"backslashreplace", "namereplace"]
# register the handlers under different names,
# to prevent the codec from recognizing the name
for err in errors:
codecs.register_error("test." + err, codecs.lookup_error(err))
l = 1000
errors += [ "test." + err for err in errors ]
for uni in [ s*l for s in ("x", "\u3042", "a\xe4") ]:
for enc in ("ascii", "latin-1", "iso-8859-1", "iso-8859-15",
"utf-8", "utf-7", "utf-16", "utf-32"):
for err in errors:
try:
uni.encode(enc, err)
except UnicodeError:
pass
def check_exceptionobjectargs(self, exctype, args, msg):
# Test UnicodeError subclasses: construction, attribute assignment and __str__ conversion
# check with one missing argument
self.assertRaises(TypeError, exctype, *args[:-1])
# check with one argument too much
self.assertRaises(TypeError, exctype, *(args + ["too much"]))
# check with one argument of the wrong type
wrongargs = [ "spam", b"eggs", b"spam", 42, 1.0, None ]
for i in range(len(args)):
for wrongarg in wrongargs:
if type(wrongarg) is type(args[i]):
continue
# build argument array
callargs = []
for j in range(len(args)):
if i==j:
callargs.append(wrongarg)
else:
callargs.append(args[i])
self.assertRaises(TypeError, exctype, *callargs)
# check with the correct number and type of arguments
exc = exctype(*args)
self.assertEqual(str(exc), msg)
def test_unicodeencodeerror(self):
self.check_exceptionobjectargs(
UnicodeEncodeError,
["ascii", "g\xfcrk", 1, 2, "ouch"],
"'ascii' codec can't encode character '\\xfc' in position 1: ouch"
)
self.check_exceptionobjectargs(
UnicodeEncodeError,
["ascii", "g\xfcrk", 1, 4, "ouch"],
"'ascii' codec can't encode characters in position 1-3: ouch"
)
self.check_exceptionobjectargs(
UnicodeEncodeError,
["ascii", "\xfcx", 0, 1, "ouch"],
"'ascii' codec can't encode character '\\xfc' in position 0: ouch"
)
self.check_exceptionobjectargs(
UnicodeEncodeError,
["ascii", "\u0100x", 0, 1, "ouch"],
"'ascii' codec can't encode character '\\u0100' in position 0: ouch"
)
self.check_exceptionobjectargs(
UnicodeEncodeError,
["ascii", "\uffffx", 0, 1, "ouch"],
"'ascii' codec can't encode character '\\uffff' in position 0: ouch"
)
self.check_exceptionobjectargs(
UnicodeEncodeError,
["ascii", "\U00010000x", 0, 1, "ouch"],
"'ascii' codec can't encode character '\\U00010000' in position 0: ouch"
)
def test_unicodedecodeerror(self):
self.check_exceptionobjectargs(
UnicodeDecodeError,
["ascii", bytearray(b"g\xfcrk"), 1, 2, "ouch"],
"'ascii' codec can't decode byte 0xfc in position 1: ouch"
)
self.check_exceptionobjectargs(
UnicodeDecodeError,
["ascii", bytearray(b"g\xfcrk"), 1, 3, "ouch"],
"'ascii' codec can't decode bytes in position 1-2: ouch"
)
def test_unicodetranslateerror(self):
self.check_exceptionobjectargs(
UnicodeTranslateError,
["g\xfcrk", 1, 2, "ouch"],
"can't translate character '\\xfc' in position 1: ouch"
)
self.check_exceptionobjectargs(
UnicodeTranslateError,
["g\u0100rk", 1, 2, "ouch"],
"can't translate character '\\u0100' in position 1: ouch"
)
self.check_exceptionobjectargs(
UnicodeTranslateError,
["g\uffffrk", 1, 2, "ouch"],
"can't translate character '\\uffff' in position 1: ouch"
)
self.check_exceptionobjectargs(
UnicodeTranslateError,
["g\U00010000rk", 1, 2, "ouch"],
"can't translate character '\\U00010000' in position 1: ouch"
)
self.check_exceptionobjectargs(
UnicodeTranslateError,
["g\xfcrk", 1, 3, "ouch"],
"can't translate characters in position 1-2: ouch"
)
def test_badandgoodstrictexceptions(self):
# "strict" complains about a non-exception passed in
self.assertRaises(
TypeError,
codecs.strict_errors,
42
)
# "strict" complains about the wrong exception type
self.assertRaises(
Exception,
codecs.strict_errors,
Exception("ouch")
)
# If the correct exception is passed in, "strict" raises it
self.assertRaises(
UnicodeEncodeError,
codecs.strict_errors,
UnicodeEncodeError("ascii", "\u3042", 0, 1, "ouch")
)
self.assertRaises(
UnicodeDecodeError,
codecs.strict_errors,
UnicodeDecodeError("ascii", bytearray(b"\xff"), 0, 1, "ouch")
)
self.assertRaises(
UnicodeTranslateError,
codecs.strict_errors,
UnicodeTranslateError("\u3042", 0, 1, "ouch")
)
def test_badandgoodignoreexceptions(self):
# "ignore" complains about a non-exception passed in
self.assertRaises(
TypeError,
codecs.ignore_errors,
42
)
# "ignore" complains about the wrong exception type
self.assertRaises(
TypeError,
codecs.ignore_errors,
UnicodeError("ouch")
)
# If the correct exception is passed in, "ignore" returns an empty replacement
self.assertEqual(
codecs.ignore_errors(
UnicodeEncodeError("ascii", "a\u3042b", 1, 2, "ouch")),
("", 2)
)
self.assertEqual(
codecs.ignore_errors(
UnicodeDecodeError("ascii", bytearray(b"a\xffb"), 1, 2, "ouch")),
("", 2)
)
self.assertEqual(
codecs.ignore_errors(
UnicodeTranslateError("a\u3042b", 1, 2, "ouch")),
("", 2)
)
def test_badandgoodreplaceexceptions(self):
# "replace" complains about a non-exception passed in
self.assertRaises(
TypeError,
codecs.replace_errors,
42
)
# "replace" complains about the wrong exception type
self.assertRaises(
TypeError,
codecs.replace_errors,
UnicodeError("ouch")
)
self.assertRaises(
TypeError,
codecs.replace_errors,
BadObjectUnicodeEncodeError()
)
self.assertRaises(
TypeError,
codecs.replace_errors,
BadObjectUnicodeDecodeError()
)
# With the correct exception, "replace" returns an "?" or "\ufffd" replacement
self.assertEqual(
codecs.replace_errors(
UnicodeEncodeError("ascii", "a\u3042b", 1, 2, "ouch")),
("?", 2)
)
self.assertEqual(
codecs.replace_errors(
UnicodeDecodeError("ascii", bytearray(b"a\xffb"), 1, 2, "ouch")),
("\ufffd", 2)
)
self.assertEqual(
codecs.replace_errors(
UnicodeTranslateError("a\u3042b", 1, 2, "ouch")),
("\ufffd", 2)
)
def test_badandgoodxmlcharrefreplaceexceptions(self):
# "xmlcharrefreplace" complains about a non-exception passed in
self.assertRaises(
TypeError,
codecs.xmlcharrefreplace_errors,
42
)
# "xmlcharrefreplace" complains about the wrong exception types
self.assertRaises(
TypeError,
codecs.xmlcharrefreplace_errors,
UnicodeError("ouch")
)
# "xmlcharrefreplace" can only be used for encoding
self.assertRaises(
TypeError,
codecs.xmlcharrefreplace_errors,
UnicodeDecodeError("ascii", bytearray(b"\xff"), 0, 1, "ouch")
)
self.assertRaises(
TypeError,
codecs.xmlcharrefreplace_errors,
UnicodeTranslateError("\u3042", 0, 1, "ouch")
)
# Use the correct exception
cs = (0, 1, 9, 10, 99, 100, 999, 1000, 9999, 10000, 99999, 100000,
999999, 1000000)
cs += (0xd800, 0xdfff)
s = "".join(chr(c) for c in cs)
self.assertEqual(
codecs.xmlcharrefreplace_errors(
UnicodeEncodeError("ascii", "a" + s + "b",
1, 1 + len(s), "ouch")
),
("".join("&#%d;" % c for c in cs), 1 + len(s))
)
def test_badandgoodbackslashreplaceexceptions(self):
# "backslashreplace" complains about a non-exception passed in
self.assertRaises(
TypeError,
codecs.backslashreplace_errors,
42
)
# "backslashreplace" complains about the wrong exception types
self.assertRaises(
TypeError,
codecs.backslashreplace_errors,
UnicodeError("ouch")
)
# Use the correct exception
tests = [
("\u3042", "\\u3042"),
("\n", "\\x0a"),
("a", "\\x61"),
("\x00", "\\x00"),
("\xff", "\\xff"),
("\u0100", "\\u0100"),
("\uffff", "\\uffff"),
("\U00010000", "\\U00010000"),
("\U0010ffff", "\\U0010ffff"),
# Lone surrogates
("\ud800", "\\ud800"),
("\udfff", "\\udfff"),
("\ud800\udfff", "\\ud800\\udfff"),
]
for s, r in tests:
with self.subTest(str=s):
self.assertEqual(
codecs.backslashreplace_errors(
UnicodeEncodeError("ascii", "a" + s + "b",
1, 1 + len(s), "ouch")),
(r, 1 + len(s))
)
self.assertEqual(
codecs.backslashreplace_errors(
UnicodeTranslateError("a" + s + "b",
1, 1 + len(s), "ouch")),
(r, 1 + len(s))
)
tests = [
(b"a", "\\x61"),
(b"\n", "\\x0a"),
(b"\x00", "\\x00"),
(b"\xff", "\\xff"),
]
for b, r in tests:
with self.subTest(bytes=b):
self.assertEqual(
codecs.backslashreplace_errors(
UnicodeDecodeError("ascii", bytearray(b"a" + b + b"b"),
1, 2, "ouch")),
(r, 2)
)
def test_badandgoodnamereplaceexceptions(self):
# "namereplace" complains about a non-exception passed in
self.assertRaises(
TypeError,
codecs.namereplace_errors,
42
)
# "namereplace" complains about the wrong exception types
self.assertRaises(
TypeError,
codecs.namereplace_errors,
UnicodeError("ouch")
)
# "namereplace" can only be used for encoding
self.assertRaises(
TypeError,
codecs.namereplace_errors,
UnicodeDecodeError("ascii", bytearray(b"\xff"), 0, 1, "ouch")
)
self.assertRaises(
TypeError,
codecs.namereplace_errors,
UnicodeTranslateError("\u3042", 0, 1, "ouch")
)
# Use the correct exception
tests = [
("\u3042", "\\N{HIRAGANA LETTER A}"),
("\x00", "\\x00"),
("\ufbf9", "\\N{ARABIC LIGATURE UIGHUR KIRGHIZ YEH WITH "
"HAMZA ABOVE WITH ALEF MAKSURA ISOLATED FORM}"),
("\U000e007f", "\\N{CANCEL TAG}"),
("\U0010ffff", "\\U0010ffff"),
# Lone surrogates
("\ud800", "\\ud800"),
("\udfff", "\\udfff"),
("\ud800\udfff", "\\ud800\\udfff"),
]
for s, r in tests:
with self.subTest(str=s):
self.assertEqual(
codecs.namereplace_errors(
UnicodeEncodeError("ascii", "a" + s + "b",
1, 1 + len(s), "ouch")),
(r, 1 + len(s))
)
def test_badandgoodsurrogateescapeexceptions(self):
surrogateescape_errors = codecs.lookup_error('surrogateescape')
# "surrogateescape" complains about a non-exception passed in
self.assertRaises(
TypeError,
surrogateescape_errors,
42
)
# "surrogateescape" complains about the wrong exception types
self.assertRaises(
TypeError,
surrogateescape_errors,
UnicodeError("ouch")
)
# "surrogateescape" can not be used for translating
self.assertRaises(
TypeError,
surrogateescape_errors,
UnicodeTranslateError("\udc80", 0, 1, "ouch")
)
# Use the correct exception
for s in ("a", "\udc7f", "\udd00"):
with self.subTest(str=s):
self.assertRaises(
UnicodeEncodeError,
surrogateescape_errors,
UnicodeEncodeError("ascii", s, 0, 1, "ouch")
)
self.assertEqual(
surrogateescape_errors(
UnicodeEncodeError("ascii", "a\udc80b", 1, 2, "ouch")),
(b"\x80", 2)
)
self.assertRaises(
UnicodeDecodeError,
surrogateescape_errors,
UnicodeDecodeError("ascii", bytearray(b"a"), 0, 1, "ouch")
)
self.assertEqual(
surrogateescape_errors(
UnicodeDecodeError("ascii", bytearray(b"a\x80b"), 1, 2, "ouch")),
("\udc80", 2)
)
def test_badandgoodsurrogatepassexceptions(self):
surrogatepass_errors = codecs.lookup_error('surrogatepass')
# "surrogatepass" complains about a non-exception passed in
self.assertRaises(
TypeError,
surrogatepass_errors,
42
)
# "surrogatepass" complains about the wrong exception types
self.assertRaises(
TypeError,
surrogatepass_errors,
UnicodeError("ouch")
)
# "surrogatepass" can not be used for translating
self.assertRaises(
TypeError,
surrogatepass_errors,
UnicodeTranslateError("\ud800", 0, 1, "ouch")
)
# Use the correct exception
for enc in ("utf-8", "utf-16le", "utf-16be", "utf-32le", "utf-32be"):
with self.subTest(encoding=enc):
self.assertRaises(
UnicodeEncodeError,
surrogatepass_errors,
UnicodeEncodeError(enc, "a", 0, 1, "ouch")
)
self.assertRaises(
UnicodeDecodeError,
surrogatepass_errors,
UnicodeDecodeError(enc, "a".encode(enc), 0, 1, "ouch")
)
for s in ("\ud800", "\udfff", "\ud800\udfff"):
with self.subTest(str=s):
self.assertRaises(
UnicodeEncodeError,
surrogatepass_errors,
UnicodeEncodeError("ascii", s, 0, len(s), "ouch")
)
tests = [
("utf-8", "\ud800", b'\xed\xa0\x80', 3),
("utf-16le", "\ud800", b'\x00\xd8', 2),
("utf-16be", "\ud800", b'\xd8\x00', 2),
("utf-32le", "\ud800", b'\x00\xd8\x00\x00', 4),
("utf-32be", "\ud800", b'\x00\x00\xd8\x00', 4),
("utf-8", "\udfff", b'\xed\xbf\xbf', 3),
("utf-16le", "\udfff", b'\xff\xdf', 2),
("utf-16be", "\udfff", b'\xdf\xff', 2),
("utf-32le", "\udfff", b'\xff\xdf\x00\x00', 4),
("utf-32be", "\udfff", b'\x00\x00\xdf\xff', 4),
("utf-8", "\ud800\udfff", b'\xed\xa0\x80\xed\xbf\xbf', 3),
("utf-16le", "\ud800\udfff", b'\x00\xd8\xff\xdf', 2),
("utf-16be", "\ud800\udfff", b'\xd8\x00\xdf\xff', 2),
("utf-32le", "\ud800\udfff", b'\x00\xd8\x00\x00\xff\xdf\x00\x00', 4),
("utf-32be", "\ud800\udfff", b'\x00\x00\xd8\x00\x00\x00\xdf\xff', 4),
]
for enc, s, b, n in tests:
with self.subTest(encoding=enc, str=s, bytes=b):
self.assertEqual(
surrogatepass_errors(
UnicodeEncodeError(enc, "a" + s + "b",
1, 1 + len(s), "ouch")),
(b, 1 + len(s))
)
self.assertEqual(
surrogatepass_errors(
UnicodeDecodeError(enc, bytearray(b"a" + b[:n] + b"b"),
1, 1 + n, "ouch")),
(s[:1], 1 + n)
)
def test_badhandlerresults(self):
results = ( 42, "foo", (1,2,3), ("foo", 1, 3), ("foo", None), ("foo",), ("foo", 1, 3), ("foo", None), ("foo",) )
encs = ("ascii", "latin-1", "iso-8859-1", "iso-8859-15")
for res in results:
codecs.register_error("test.badhandler", lambda x: res)
for enc in encs:
self.assertRaises(
TypeError,
"\u3042".encode,
enc,
"test.badhandler"
)
for (enc, bytes) in (
("ascii", b"\xff"),
("utf-8", b"\xff"),
("utf-7", b"+x-"),
("unicode-internal", b"\x00"),
):
with test.support.check_warnings():
# unicode-internal has been deprecated
self.assertRaises(
TypeError,
bytes.decode,
enc,
"test.badhandler"
)
def test_lookup(self):
self.assertEqual(codecs.strict_errors, codecs.lookup_error("strict"))
self.assertEqual(codecs.ignore_errors, codecs.lookup_error("ignore"))
self.assertEqual(codecs.strict_errors, codecs.lookup_error("strict"))
self.assertEqual(
codecs.xmlcharrefreplace_errors,
codecs.lookup_error("xmlcharrefreplace")
)
self.assertEqual(
codecs.backslashreplace_errors,
codecs.lookup_error("backslashreplace")
)
self.assertEqual(
codecs.namereplace_errors,
codecs.lookup_error("namereplace")
)
def test_unencodablereplacement(self):
def unencrepl(exc):
if isinstance(exc, UnicodeEncodeError):
return ("\u4242", exc.end)
else:
raise TypeError("don't know how to handle %r" % exc)
codecs.register_error("test.unencreplhandler", unencrepl)
for enc in ("ascii", "iso-8859-1", "iso-8859-15"):
self.assertRaises(
UnicodeEncodeError,
"\u4242".encode,
enc,
"test.unencreplhandler"
)
def test_badregistercall(self):
# enhance coverage of:
# Modules/_codecsmodule.c::register_error()
# Python/codecs.c::PyCodec_RegisterError()
self.assertRaises(TypeError, codecs.register_error, 42)
self.assertRaises(TypeError, codecs.register_error, "test.dummy", 42)
def test_badlookupcall(self):
# enhance coverage of:
# Modules/_codecsmodule.c::lookup_error()
self.assertRaises(TypeError, codecs.lookup_error)
def test_unknownhandler(self):
# enhance coverage of:
# Modules/_codecsmodule.c::lookup_error()
self.assertRaises(LookupError, codecs.lookup_error, "test.unknown")
def test_xmlcharrefvalues(self):
# enhance coverage of:
# Python/codecs.c::PyCodec_XMLCharRefReplaceErrors()
# and inline implementations
v = (1, 5, 10, 50, 100, 500, 1000, 5000, 10000, 50000, 100000,
500000, 1000000)
s = "".join([chr(x) for x in v])
codecs.register_error("test.xmlcharrefreplace", codecs.xmlcharrefreplace_errors)
for enc in ("ascii", "iso-8859-15"):
for err in ("xmlcharrefreplace", "test.xmlcharrefreplace"):
s.encode(enc, err)
def test_decodehelper(self):
# enhance coverage of:
# Objects/unicodeobject.c::unicode_decode_call_errorhandler()
# and callers
self.assertRaises(LookupError, b"\xff".decode, "ascii", "test.unknown")
def baddecodereturn1(exc):
return 42
codecs.register_error("test.baddecodereturn1", baddecodereturn1)
self.assertRaises(TypeError, b"\xff".decode, "ascii", "test.baddecodereturn1")
self.assertRaises(TypeError, b"\\".decode, "unicode-escape", "test.baddecodereturn1")
self.assertRaises(TypeError, b"\\x0".decode, "unicode-escape", "test.baddecodereturn1")
self.assertRaises(TypeError, b"\\x0y".decode, "unicode-escape", "test.baddecodereturn1")
self.assertRaises(TypeError, b"\\Uffffeeee".decode, "unicode-escape", "test.baddecodereturn1")
self.assertRaises(TypeError, b"\\uyyyy".decode, "raw-unicode-escape", "test.baddecodereturn1")
def baddecodereturn2(exc):
return ("?", None)
codecs.register_error("test.baddecodereturn2", baddecodereturn2)
self.assertRaises(TypeError, b"\xff".decode, "ascii", "test.baddecodereturn2")
handler = PosReturn()
codecs.register_error("test.posreturn", handler.handle)
# Valid negative position
handler.pos = -1
self.assertEqual(b"\xff0".decode("ascii", "test.posreturn"), "<?>0")
# Valid negative position
handler.pos = -2
self.assertEqual(b"\xff0".decode("ascii", "test.posreturn"), "<?><?>")
# Negative position out of bounds
handler.pos = -3
self.assertRaises(IndexError, b"\xff0".decode, "ascii", "test.posreturn")
# Valid positive position
handler.pos = 1
self.assertEqual(b"\xff0".decode("ascii", "test.posreturn"), "<?>0")
# Largest valid positive position (one beyond end of input)
handler.pos = 2
self.assertEqual(b"\xff0".decode("ascii", "test.posreturn"), "<?>")
# Invalid positive position
handler.pos = 3
self.assertRaises(IndexError, b"\xff0".decode, "ascii", "test.posreturn")
# Restart at the "0"
handler.pos = 6
self.assertEqual(b"\\uyyyy0".decode("raw-unicode-escape", "test.posreturn"), "<?>0")
class D(dict):
def __getitem__(self, key):
raise ValueError
self.assertRaises(UnicodeError, codecs.charmap_decode, b"\xff", "strict", {0xff: None})
self.assertRaises(ValueError, codecs.charmap_decode, b"\xff", "strict", D())
self.assertRaises(TypeError, codecs.charmap_decode, b"\xff", "strict", {0xff: sys.maxunicode+1})
def test_encodehelper(self):
# enhance coverage of:
# Objects/unicodeobject.c::unicode_encode_call_errorhandler()
# and callers
self.assertRaises(LookupError, "\xff".encode, "ascii", "test.unknown")
def badencodereturn1(exc):
return 42
codecs.register_error("test.badencodereturn1", badencodereturn1)
self.assertRaises(TypeError, "\xff".encode, "ascii", "test.badencodereturn1")
def badencodereturn2(exc):
return ("?", None)
codecs.register_error("test.badencodereturn2", badencodereturn2)
self.assertRaises(TypeError, "\xff".encode, "ascii", "test.badencodereturn2")
handler = PosReturn()
codecs.register_error("test.posreturn", handler.handle)
# Valid negative position
handler.pos = -1
self.assertEqual("\xff0".encode("ascii", "test.posreturn"), b"<?>0")
# Valid negative position
handler.pos = -2
self.assertEqual("\xff0".encode("ascii", "test.posreturn"), b"<?><?>")
# Negative position out of bounds
handler.pos = -3
self.assertRaises(IndexError, "\xff0".encode, "ascii", "test.posreturn")
# Valid positive position
handler.pos = 1
self.assertEqual("\xff0".encode("ascii", "test.posreturn"), b"<?>0")
# Largest valid positive position (one beyond end of input
handler.pos = 2
self.assertEqual("\xff0".encode("ascii", "test.posreturn"), b"<?>")
# Invalid positive position
handler.pos = 3
self.assertRaises(IndexError, "\xff0".encode, "ascii", "test.posreturn")
handler.pos = 0
class D(dict):
def __getitem__(self, key):
raise ValueError
for err in ("strict", "replace", "xmlcharrefreplace",
"backslashreplace", "namereplace", "test.posreturn"):
self.assertRaises(UnicodeError, codecs.charmap_encode, "\xff", err, {0xff: None})
self.assertRaises(ValueError, codecs.charmap_encode, "\xff", err, D())
self.assertRaises(TypeError, codecs.charmap_encode, "\xff", err, {0xff: 300})
def test_translatehelper(self):
# enhance coverage of:
# Objects/unicodeobject.c::unicode_encode_call_errorhandler()
# and callers
# (Unfortunately the errors argument is not directly accessible
# from Python, so we can't test that much)
class D(dict):
def __getitem__(self, key):
raise ValueError
#self.assertRaises(ValueError, "\xff".translate, D())
self.assertRaises(ValueError, "\xff".translate, {0xff: sys.maxunicode+1})
self.assertRaises(TypeError, "\xff".translate, {0xff: ()})
def test_bug828737(self):
charmap = {
ord("&"): "&amp;",
ord("<"): "&lt;",
ord(">"): "&gt;",
ord('"'): "&quot;",
}
for n in (1, 10, 100, 1000):
text = 'abc<def>ghi'*n
text.translate(charmap)
def test_mutatingdecodehandler(self):
baddata = [
("ascii", b"\xff"),
("utf-7", b"++"),
("utf-8", b"\xff"),
("utf-16", b"\xff"),
("utf-32", b"\xff"),
("unicode-escape", b"\\u123g"),
("raw-unicode-escape", b"\\u123g"),
("unicode-internal", b"\xff"),
]
def replacing(exc):
if isinstance(exc, UnicodeDecodeError):
exc.object = 42
return ("\u4242", 0)
else:
raise TypeError("don't know how to handle %r" % exc)
codecs.register_error("test.replacing", replacing)
with test.support.check_warnings():
# unicode-internal has been deprecated
for (encoding, data) in baddata:
with self.assertRaises(TypeError):
data.decode(encoding, "test.replacing")
def mutating(exc):
if isinstance(exc, UnicodeDecodeError):
exc.object[:] = b""
return ("\u4242", 0)
else:
raise TypeError("don't know how to handle %r" % exc)
codecs.register_error("test.mutating", mutating)
# If the decoder doesn't pick up the modified input the following
# will lead to an endless loop
with test.support.check_warnings():
# unicode-internal has been deprecated
for (encoding, data) in baddata:
with self.assertRaises(TypeError):
data.decode(encoding, "test.replacing")
def test_fake_error_class(self):
handlers = [
codecs.strict_errors,
codecs.ignore_errors,
codecs.replace_errors,
codecs.backslashreplace_errors,
codecs.namereplace_errors,
codecs.xmlcharrefreplace_errors,
codecs.lookup_error('surrogateescape'),
codecs.lookup_error('surrogatepass'),
]
for cls in UnicodeEncodeError, UnicodeDecodeError, UnicodeTranslateError:
class FakeUnicodeError(str):
__class__ = cls
for handler in handlers:
with self.subTest(handler=handler, error_class=cls):
self.assertRaises(TypeError, handler, FakeUnicodeError())
class FakeUnicodeError(Exception):
__class__ = cls
for handler in handlers:
with self.subTest(handler=handler, error_class=cls):
with self.assertRaises((TypeError, FakeUnicodeError)):
handler(FakeUnicodeError())
if __name__ == "__main__":
unittest.main()