Staging
v0.5.1
https://github.com/python/cpython
Raw File
Tip revision: c186b237993ac0c1c6ee1a195754702cc01625d7 authored by Georg Brandl on 02 March 2014, 08:19:03 UTC
Bump to 3.3.5rc2.
Tip revision: c186b23
test_codecencodings_kr.py
#
# test_codecencodings_kr.py
#   Codec encoding tests for ROK encodings.
#

from test import support
from test import multibytecodec_support
import unittest

class Test_CP949(multibytecodec_support.TestBase, unittest.TestCase):
    encoding = 'cp949'
    tstring = multibytecodec_support.load_teststring('cp949')
    codectests = (
        # invalid bytes
        (b"abc\x80\x80\xc1\xc4", "strict",  None),
        (b"abc\xc8", "strict",  None),
        (b"abc\x80\x80\xc1\xc4", "replace", "abc\ufffd\ufffd\uc894"),
        (b"abc\x80\x80\xc1\xc4\xc8", "replace", "abc\ufffd\ufffd\uc894\ufffd"),
        (b"abc\x80\x80\xc1\xc4", "ignore",  "abc\uc894"),
    )

class Test_EUCKR(multibytecodec_support.TestBase, unittest.TestCase):
    encoding = 'euc_kr'
    tstring = multibytecodec_support.load_teststring('euc_kr')
    codectests = (
        # invalid bytes
        (b"abc\x80\x80\xc1\xc4", "strict",  None),
        (b"abc\xc8", "strict",  None),
        (b"abc\x80\x80\xc1\xc4", "replace", 'abc\ufffd\ufffd\uc894'),
        (b"abc\x80\x80\xc1\xc4\xc8", "replace", "abc\ufffd\ufffd\uc894\ufffd"),
        (b"abc\x80\x80\xc1\xc4", "ignore",  "abc\uc894"),

        # composed make-up sequence errors
        (b"\xa4\xd4", "strict", None),
        (b"\xa4\xd4\xa4", "strict", None),
        (b"\xa4\xd4\xa4\xb6", "strict", None),
        (b"\xa4\xd4\xa4\xb6\xa4", "strict", None),
        (b"\xa4\xd4\xa4\xb6\xa4\xd0", "strict", None),
        (b"\xa4\xd4\xa4\xb6\xa4\xd0\xa4", "strict", None),
        (b"\xa4\xd4\xa4\xb6\xa4\xd0\xa4\xd4", "strict", "\uc4d4"),
        (b"\xa4\xd4\xa4\xb6\xa4\xd0\xa4\xd4x", "strict", "\uc4d4x"),
        (b"a\xa4\xd4\xa4\xb6\xa4", "replace", 'a\ufffd'),
        (b"\xa4\xd4\xa3\xb6\xa4\xd0\xa4\xd4", "strict", None),
        (b"\xa4\xd4\xa4\xb6\xa3\xd0\xa4\xd4", "strict", None),
        (b"\xa4\xd4\xa4\xb6\xa4\xd0\xa3\xd4", "strict", None),
        (b"\xa4\xd4\xa4\xff\xa4\xd0\xa4\xd4", "replace", '\ufffd\u6e21\ufffd\u3160\ufffd'),
        (b"\xa4\xd4\xa4\xb6\xa4\xff\xa4\xd4", "replace", '\ufffd\u6e21\ub544\ufffd\ufffd'),
        (b"\xa4\xd4\xa4\xb6\xa4\xd0\xa4\xff", "replace", '\ufffd\u6e21\ub544\u572d\ufffd'),
        (b"\xa4\xd4\xff\xa4\xd4\xa4\xb6\xa4\xd0\xa4\xd4", "replace", '\ufffd\ufffd\ufffd\uc4d4'),
        (b"\xc1\xc4", "strict", "\uc894"),
    )

class Test_JOHAB(multibytecodec_support.TestBase, unittest.TestCase):
    encoding = 'johab'
    tstring = multibytecodec_support.load_teststring('johab')
    codectests = (
        # invalid bytes
        (b"abc\x80\x80\xc1\xc4", "strict",  None),
        (b"abc\xc8", "strict",  None),
        (b"abc\x80\x80\xc1\xc4", "replace", "abc\ufffd\ufffd\ucd27"),
        (b"abc\x80\x80\xc1\xc4\xc8", "replace", "abc\ufffd\ufffd\ucd27\ufffd"),
        (b"abc\x80\x80\xc1\xc4", "ignore",  "abc\ucd27"),
        (b"\xD8abc", "replace",  "\uFFFDabc"),
        (b"\xD8\xFFabc", "replace",  "\uFFFD\uFFFDabc"),
        (b"\x84bxy", "replace",  "\uFFFDbxy"),
        (b"\x8CBxy", "replace",  "\uFFFDBxy"),
    )

def test_main():
    support.run_unittest(__name__)

if __name__ == "__main__":
    test_main()
back to top