| #!/usr/bin/env python |
| # |
| # test_multibytecodec_support.py |
| # Common Unittest Routines for CJK codecs |
| # |
| # $CJKCodecs: test_multibytecodec_support.py,v 1.6 2004/06/19 06:09:55 perky Exp $ |
| |
| import sys, codecs, os.path |
| import unittest |
| from test import test_support |
| from StringIO import StringIO |
| |
| __cjkcodecs__ = 0 # define this as 0 for python |
| |
| class TestBase: |
| encoding = '' # codec name |
| codec = None # codec tuple (with 4 elements) |
| tstring = '' # string to test StreamReader |
| |
| codectests = None # must set. codec test tuple |
| 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 |
| |
| def setUp(self): |
| if self.codec is None: |
| self.codec = codecs.lookup(self.encoding) |
| self.encode, self.decode, self.reader, self.writer = self.codec |
| |
| def test_chunkcoding(self): |
| for native, utf8 in zip(*[StringIO(f).readlines() |
| for f in self.tstring]): |
| u = self.decode(native)[0] |
| self.assertEqual(u, utf8.decode('utf-8')) |
| if self.roundtriptest: |
| self.assertEqual(native, self.encode(u)[0]) |
| |
| def test_errorhandle(self): |
| for source, scheme, expected in self.codectests: |
| if type(source) == type(''): |
| func = self.decode |
| else: |
| func = self.encode |
| if expected: |
| result = func(source, scheme)[0] |
| self.assertEqual(result, expected) |
| else: |
| self.assertRaises(UnicodeError, func, source, scheme) |
| |
| if sys.hexversion >= 0x02030000: |
| def test_xmlcharrefreplace(self): |
| if self.has_iso10646: |
| return |
| |
| s = u"\u0b13\u0b23\u0b60 nd eggs" |
| self.assertEqual( |
| self.encode(s, "xmlcharrefreplace")[0], |
| "ଓଣୠ nd eggs" |
| ) |
| |
| def test_customreplace(self): |
| if self.has_iso10646: |
| return |
| |
| import htmlentitydefs |
| |
| names = {} |
| for (key, value) in htmlentitydefs.entitydefs.items(): |
| if len(value)==1: |
| names[value.decode('latin-1')] = self.decode(key)[0] |
| else: |
| names[unichr(int(value[2:-1]))] = self.decode(key)[0] |
| |
| 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(u"&%s;" % names[c]) |
| except KeyError: |
| l.append(u"&#%d;" % ord(c)) |
| return (u"".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" |
| sout = "«ℜ» = ⟨ሴ⟩" |
| self.assertEqual(self.encode(sin, |
| "test.xmlcharnamereplace")[0], sout) |
| |
| def test_streamreader(self): |
| UTF8Writer = codecs.getwriter('utf-8') |
| for name in ["read", "readline", "readlines"]: |
| for sizehint in [None, -1] + range(1, 33) + \ |
| [64, 128, 256, 512, 1024]: |
| istream = self.reader(StringIO(self.tstring[0])) |
| ostream = UTF8Writer(StringIO()) |
| func = getattr(istream, name) |
| while 1: |
| data = func(sizehint) |
| if not data: |
| break |
| if name == "readlines": |
| ostream.writelines(data) |
| else: |
| ostream.write(data) |
| |
| self.assertEqual(ostream.getvalue(), self.tstring[1]) |
| |
| def test_streamwriter(self): |
| if __cjkcodecs__: |
| readfuncs = ('read', 'readline', 'readlines') |
| else: |
| # standard utf8 codec has broken readline and readlines. |
| readfuncs = ('read',) |
| UTF8Reader = codecs.getreader('utf-8') |
| for name in readfuncs: |
| for sizehint in [None] + range(1, 33) + \ |
| [64, 128, 256, 512, 1024]: |
| istream = UTF8Reader(StringIO(self.tstring[1])) |
| ostream = self.writer(StringIO()) |
| func = getattr(istream, name) |
| while 1: |
| if sizehint is not None: |
| data = func(sizehint) |
| else: |
| data = func() |
| |
| if not data: |
| break |
| if name == "readlines": |
| ostream.writelines(data) |
| else: |
| ostream.write(data) |
| |
| self.assertEqual(ostream.getvalue(), self.tstring[0]) |
| |
| if len(u'\U00012345') == 2: # ucs2 build |
| _unichr = unichr |
| def unichr(v): |
| if v >= 0x10000: |
| return _unichr(0xd800 + ((v - 0x10000) >> 10)) + \ |
| _unichr(0xdc00 + ((v - 0x10000) & 0x3ff)) |
| else: |
| return _unichr(v) |
| _ord = ord |
| def ord(c): |
| if len(c) == 2: |
| return 0x10000 + ((_ord(c[0]) - 0xd800) << 10) + \ |
| (ord(c[1]) - 0xdc00) |
| else: |
| return _ord(c) |
| |
| class TestBase_Mapping(unittest.TestCase): |
| pass_enctest = [] |
| pass_dectest = [] |
| supmaps = [] |
| |
| def __init__(self, *args, **kw): |
| unittest.TestCase.__init__(self, *args, **kw) |
| if not os.path.exists(self.mapfilename): |
| raise test_support.TestSkipped('%s not found, download from %s' % |
| (self.mapfilename, self.mapfileurl)) |
| |
| def test_mapping_file(self): |
| unichrs = lambda s: u''.join(map(unichr, map(eval, s.split('+')))) |
| urt_wa = {} |
| |
| for line in open(self.mapfilename): |
| if not line: |
| break |
| data = line.split('#')[0].strip().split() |
| if len(data) != 2: |
| continue |
| |
| csetval = eval(data[0]) |
| if csetval <= 0x7F: |
| csetch = chr(csetval & 0xff) |
| elif csetval >= 0x1000000: |
| csetch = chr(csetval >> 24) + chr((csetval >> 16) & 0xff) + \ |
| chr((csetval >> 8) & 0xff) + chr(csetval & 0xff) |
| elif csetval >= 0x10000: |
| csetch = chr(csetval >> 16) + \ |
| chr((csetval >> 8) & 0xff) + chr(csetval & 0xff) |
| elif csetval >= 0x100: |
| csetch = chr(csetval >> 8) + chr(csetval & 0xff) |
| else: |
| continue |
| |
| unich = unichrs(data[1]) |
| if ord(unich) == 0xfffd or urt_wa.has_key(unich): |
| continue |
| urt_wa[unich] = csetch |
| |
| self._testpoint(csetch, unich) |
| |
| def test_mapping_supplemental(self): |
| for mapping in self.supmaps: |
| self._testpoint(*mapping) |
| |
| def _testpoint(self, csetch, unich): |
| 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) |
| |
| def load_teststring(encoding): |
| if __cjkcodecs__: |
| etxt = open(os.path.join('sampletexts', encoding) + '.txt').read() |
| utxt = open(os.path.join('sampletexts', encoding) + '.utf8').read() |
| return (etxt, utxt) |
| else: |
| from test import cjkencodings_test |
| return cjkencodings_test.teststring[encoding] |
| |
| def register_skip_expected(*cases): |
| for case in cases: # len(cases) must be 1 at least. |
| for path in [os.path.curdir, os.path.pardir]: |
| fn = os.path.join(path, case.mapfilename) |
| if os.path.exists(fn): |
| case.mapfilename = fn |
| break |
| else: |
| sys.modules[case.__module__].skip_expected = True |
| break |
| else: |
| sys.modules[case.__module__].skip_expected = False |