diff -r 90d4153728f6 Include/codecs.h --- a/Include/codecs.h Thu Nov 21 23:01:59 2013 -0800 +++ b/Include/codecs.h Fri Nov 22 21:15:49 2013 +1000 @@ -94,6 +94,31 @@ const char *errors ); +#ifndef PY_LIMITED_API +/* Text codec specific encoding and decoding API. + + Checks the encoding against a blacklist of known non-Unicode codecs + before attempting the operation. + + Private for now, we'll likely do something public for 3.5 + + */ + +PyAPI_FUNC(PyObject *) _PyCodec_EncodeText( + PyObject *object, + const char *encoding, + const char *errors + ); + +PyAPI_FUNC(PyObject *) _PyCodec_DecodeText( + PyObject *object, + const char *encoding, + const char *errors + ); +#endif + + + /* --- Codec Lookup APIs -------------------------------------------------- All APIs return a codec object with incremented refcount and are diff -r 90d4153728f6 Lib/codecs.py --- a/Lib/codecs.py Thu Nov 21 23:01:59 2013 -0800 +++ b/Lib/codecs.py Fri Nov 22 21:15:49 2013 +1000 @@ -86,6 +86,16 @@ self.streamreader = streamreader return self + _is_text_encoding = True # Assume codecs are text encodings by default + + @classmethod + def _declare_transform(cls, *args, **kwds): + # Private helper to declare binary and text transforms + # See http://bugs.python.org/issue19619 + self = cls(*args, **kwds) + self._is_text_encoding = False + return self + def __repr__(self): return "<%s.%s object for encoding %s at 0x%x>" % \ (self.__class__.__module__, self.__class__.__name__, diff -r 90d4153728f6 Lib/encodings/base64_codec.py --- a/Lib/encodings/base64_codec.py Thu Nov 21 23:01:59 2013 -0800 +++ b/Lib/encodings/base64_codec.py Fri Nov 22 21:15:49 2013 +1000 @@ -44,7 +44,7 @@ ### encodings module API def getregentry(): - return codecs.CodecInfo( + return codecs.CodecInfo._declare_transform( name='base64', encode=base64_encode, decode=base64_decode, diff -r 90d4153728f6 Lib/encodings/bz2_codec.py --- a/Lib/encodings/bz2_codec.py Thu Nov 21 23:01:59 2013 -0800 +++ b/Lib/encodings/bz2_codec.py Fri Nov 22 21:15:49 2013 +1000 @@ -66,7 +66,7 @@ ### encodings module API def getregentry(): - return codecs.CodecInfo( + return codecs.CodecInfo._declare_transform( name="bz2", encode=bz2_encode, decode=bz2_decode, diff -r 90d4153728f6 Lib/encodings/hex_codec.py --- a/Lib/encodings/hex_codec.py Thu Nov 21 23:01:59 2013 -0800 +++ b/Lib/encodings/hex_codec.py Fri Nov 22 21:15:49 2013 +1000 @@ -44,7 +44,7 @@ ### encodings module API def getregentry(): - return codecs.CodecInfo( + return codecs.CodecInfo._declare_transform( name='hex', encode=hex_encode, decode=hex_decode, diff -r 90d4153728f6 Lib/encodings/quopri_codec.py --- a/Lib/encodings/quopri_codec.py Thu Nov 21 23:01:59 2013 -0800 +++ b/Lib/encodings/quopri_codec.py Fri Nov 22 21:15:49 2013 +1000 @@ -45,7 +45,7 @@ # encodings module API def getregentry(): - return codecs.CodecInfo( + return codecs.CodecInfo._declare_transform( name='quopri', encode=quopri_encode, decode=quopri_decode, diff -r 90d4153728f6 Lib/encodings/rot_13.py --- a/Lib/encodings/rot_13.py Thu Nov 21 23:01:59 2013 -0800 +++ b/Lib/encodings/rot_13.py Fri Nov 22 21:15:49 2013 +1000 @@ -35,7 +35,7 @@ ### encodings module API def getregentry(): - return codecs.CodecInfo( + return codecs.CodecInfo._declare_transform( name='rot-13', encode=Codec().encode, decode=Codec().decode, diff -r 90d4153728f6 Lib/encodings/uu_codec.py --- a/Lib/encodings/uu_codec.py Thu Nov 21 23:01:59 2013 -0800 +++ b/Lib/encodings/uu_codec.py Fri Nov 22 21:15:49 2013 +1000 @@ -88,7 +88,7 @@ ### encodings module API def getregentry(): - return codecs.CodecInfo( + return codecs.CodecInfo._declare_transform( name='uu', encode=uu_encode, decode=uu_decode, diff -r 90d4153728f6 Lib/encodings/zlib_codec.py --- a/Lib/encodings/zlib_codec.py Thu Nov 21 23:01:59 2013 -0800 +++ b/Lib/encodings/zlib_codec.py Fri Nov 22 21:15:49 2013 +1000 @@ -66,7 +66,7 @@ ### encodings module API def getregentry(): - return codecs.CodecInfo( + return codecs.CodecInfo._declare_transform( name='zlib', encode=zlib_encode, decode=zlib_decode, diff -r 90d4153728f6 Lib/test/test_codecs.py --- a/Lib/test/test_codecs.py Thu Nov 21 23:01:59 2013 -0800 +++ b/Lib/test/test_codecs.py Fri Nov 22 21:15:49 2013 +1000 @@ -6,6 +6,7 @@ import sys import unittest import warnings +import encodings from test import support @@ -2381,67 +2382,68 @@ view_decoded = codecs.decode(view, encoding) self.assertEqual(view_decoded, data) - def test_type_error_for_text_input(self): + def test_text_to_binary_blacklists_binary_transforms(self): # Check binary -> binary codecs give a good error for str input bad_input = "bad input type" for encoding in bytes_transform_encodings: with self.subTest(encoding=encoding): - msg = "^encoding with '{}' codec failed".format(encoding) - with self.assertRaisesRegex(TypeError, msg) as failure: + fmt = ( "{!r} is not a text encoding; " + "use codecs.encode\(\) to handle arbitrary codecs") + msg = fmt.format(encoding) + with self.assertRaisesRegex(LookupError, msg) as failure: bad_input.encode(encoding) - self.assertTrue(isinstance(failure.exception.__cause__, - TypeError)) + self.assertIsNone(failure.exception.__cause__) - def test_type_error_for_binary_input(self): - # Check str -> str codec gives a good error for binary input - for bad_input in (b"immutable", bytearray(b"mutable")): - with self.subTest(bad_input=bad_input): - msg = "^decoding with 'rot_13' codec failed" - with self.assertRaisesRegex(AttributeError, msg) as failure: - bad_input.decode("rot_13") - self.assertTrue(isinstance(failure.exception.__cause__, - AttributeError)) + def test_text_to_binary_blacklists_text_transforms(self): + # Check str.encode gives a good error message for str -> str codecs + msg = (r"^'rot_13' is not a text encoding; " + "use codecs.encode\(\) to handle arbitrary codecs") + with self.assertRaisesRegex(LookupError, msg): + "just an example message".encode("rot_13") - def test_custom_zlib_error_is_wrapped(self): - # Check zlib codec gives a good error for malformed input - msg = "^decoding with 'zlib_codec' codec failed" - with self.assertRaisesRegex(Exception, msg) as failure: - b"hello".decode("zlib_codec") - self.assertTrue(isinstance(failure.exception.__cause__, - type(failure.exception))) - - def test_custom_hex_error_is_wrapped(self): - # Check hex codec gives a good error for malformed input - msg = "^decoding with 'hex_codec' codec failed" - with self.assertRaisesRegex(Exception, msg) as failure: - b"hello".decode("hex_codec") - self.assertTrue(isinstance(failure.exception.__cause__, - type(failure.exception))) - - # Unfortunately, the bz2 module throws OSError, which the codec - # machinery currently can't wrap :( - - def test_bad_decoding_output_type(self): + def test_binary_to_text_blacklists_binary_transforms(self): # Check bytes.decode and bytearray.decode give a good error # message for binary -> binary codecs data = b"encode first to ensure we meet any format restrictions" for encoding in bytes_transform_encodings: with self.subTest(encoding=encoding): encoded_data = codecs.encode(data, encoding) - fmt = ("'{}' decoder returned 'bytes' instead of 'str'; " - "use codecs.decode\(\) to decode to arbitrary types") + fmt = (r"{!r} is not a text encoding; " + "use codecs.decode\(\) to handle arbitrary codecs") msg = fmt.format(encoding) - with self.assertRaisesRegex(TypeError, msg): + with self.assertRaisesRegex(LookupError, msg): encoded_data.decode(encoding) - with self.assertRaisesRegex(TypeError, msg): + with self.assertRaisesRegex(LookupError, msg): bytearray(encoded_data).decode(encoding) - def test_bad_encoding_output_type(self): - # Check str.encode gives a good error message for str -> str codecs - msg = ("'rot_13' encoder returned 'str' instead of 'bytes'; " - "use codecs.encode\(\) to encode to arbitrary types") - with self.assertRaisesRegex(TypeError, msg): - "just an example message".encode("rot_13") + def test_binary_to_text_blacklists_text_transforms(self): + # Check str -> str codec gives a good error for binary input + for bad_input in (b"immutable", bytearray(b"mutable")): + with self.subTest(bad_input=bad_input): + msg = (r"^'rot_13' is not a text encoding; " + "use codecs.decode\(\) to handle arbitrary codecs") + with self.assertRaisesRegex(LookupError, msg) as failure: + bad_input.decode("rot_13") + self.assertIsNone(failure.exception.__cause__) + + def test_custom_zlib_error_is_wrapped(self): + # Check zlib codec gives a good error for malformed input + msg = "^decoding with 'zlib_codec' codec failed" + with self.assertRaisesRegex(Exception, msg) as failure: + codecs.decode(b"hello", "zlib_codec") + self.assertIsInstance(failure.exception.__cause__, + type(failure.exception)) + + def test_custom_hex_error_is_wrapped(self): + # Check hex codec gives a good error for malformed input + msg = "^decoding with 'hex_codec' codec failed" + with self.assertRaisesRegex(Exception, msg) as failure: + codecs.decode(b"hello", "hex_codec") + self.assertIsInstance(failure.exception.__cause__, + type(failure.exception)) + + # Unfortunately, the bz2 module throws OSError, which the codec + # machinery currently can't wrap :( # The codec system tries to wrap exceptions in order to ensure the error @@ -2466,27 +2468,27 @@ # case finishes by using the test case repr as the codec name # The codecs module normalizes codec names, although this doesn't # appear to be formally documented... - self.codec_name = repr(self).lower().replace(" ", "-") + self.codec_name = encodings.normalize_encoding(repr(self)).lower() def tearDown(self): _TEST_CODECS.pop(self.codec_name, None) - def set_codec(self, obj_to_raise): - def raise_obj(*args, **kwds): - raise obj_to_raise - codec_info = codecs.CodecInfo(raise_obj, raise_obj, + def set_codec(self, encode, decode): + codec_info = codecs.CodecInfo(encode, decode, name=self.codec_name) _TEST_CODECS[self.codec_name] = codec_info @contextlib.contextmanager def assertWrapped(self, operation, exc_type, msg): - full_msg = "{} with '{}' codec failed \({}: {}\)".format( + full_msg = r"{} with {!r} codec failed \({}: {}\)".format( operation, self.codec_name, exc_type.__name__, msg) with self.assertRaisesRegex(exc_type, full_msg) as caught: yield caught def check_wrapped(self, obj_to_raise, msg, exc_type=RuntimeError): - self.set_codec(obj_to_raise) + def raise_obj(*args, **kwds): + raise obj_to_raise + self.set_codec(raise_obj, raise_obj) with self.assertWrapped("encoding", exc_type, msg): "str_input".encode(self.codec_name) with self.assertWrapped("encoding", exc_type, msg): @@ -2515,23 +2517,17 @@ pass self.check_wrapped(MyRuntimeError(msg), msg, MyRuntimeError) - @contextlib.contextmanager - def assertNotWrapped(self, operation, exc_type, msg_re, msg=None): - if msg is None: - msg = msg_re - with self.assertRaisesRegex(exc_type, msg) as caught: - yield caught - self.assertEqual(str(caught.exception), msg) - - def check_not_wrapped(self, obj_to_raise, msg_re, msg=None): - self.set_codec(obj_to_raise) - with self.assertNotWrapped("encoding", RuntimeError, msg_re, msg): + def check_not_wrapped(self, obj_to_raise, msg): + def raise_obj(*args, **kwds): + raise obj_to_raise + self.set_codec(raise_obj, raise_obj) + with self.assertRaisesRegex(RuntimeError, msg): "str input".encode(self.codec_name) - with self.assertNotWrapped("encoding", RuntimeError, msg_re, msg): + with self.assertRaisesRegex(RuntimeError, msg): codecs.encode("str input", self.codec_name) - with self.assertNotWrapped("decoding", RuntimeError, msg_re, msg): + with self.assertRaisesRegex(RuntimeError, msg): b"bytes input".decode(self.codec_name) - with self.assertNotWrapped("decoding", RuntimeError, msg_re, msg): + with self.assertRaisesRegex(RuntimeError, msg): codecs.decode(b"bytes input", self.codec_name) def test_init_override_is_not_wrapped(self): @@ -2550,29 +2546,56 @@ msg = "This should NOT be wrapped" exc = RuntimeError(msg) exc.attr = 1 - self.check_not_wrapped(exc, msg) + self.check_not_wrapped(exc, "^{}$".format(msg)) def test_non_str_arg_is_not_wrapped(self): self.check_not_wrapped(RuntimeError(1), "1") def test_multiple_args_is_not_wrapped(self): - msg_re = "\('a', 'b', 'c'\)" - msg = "('a', 'b', 'c')" - self.check_not_wrapped(RuntimeError('a', 'b', 'c'), msg_re, msg) + msg_re = r"^\('a', 'b', 'c'\)$" + self.check_not_wrapped(RuntimeError('a', 'b', 'c'), msg_re) # http://bugs.python.org/issue19609 def test_codec_lookup_failure_not_wrapped(self): - msg = "unknown encoding: %s" % self.codec_name + msg = "^unknown encoding: {}$".format(self.codec_name) # The initial codec lookup should not be wrapped - with self.assertNotWrapped("encoding", LookupError, msg): + with self.assertRaisesRegex(LookupError, msg): "str input".encode(self.codec_name) - with self.assertNotWrapped("encoding", LookupError, msg): + with self.assertRaisesRegex(LookupError, msg): codecs.encode("str input", self.codec_name) - with self.assertNotWrapped("decoding", LookupError, msg): + with self.assertRaisesRegex(LookupError, msg): b"bytes input".decode(self.codec_name) - with self.assertNotWrapped("decoding", LookupError, msg): + with self.assertRaisesRegex(LookupError, msg): codecs.decode(b"bytes input", self.codec_name) + def test_unflagged_non_text_codec_handling(self): + # The stdlib non-text codecs are now marked so they're + # pre-emptively skipped by the text model related methods + # However, third party codecs won't be flagged, so we still make + # sure the case where an inappropriate output type is produced is + # handled appropriately + def encode_to_str(*args, **kwds): + return "not bytes!", 0 + def decode_to_bytes(*args, **kwds): + return b"not str!", 0 + self.set_codec(encode_to_str, decode_to_bytes) + # No input or output type checks on the codecs module functions + encoded = codecs.encode(None, self.codec_name) + self.assertEqual(encoded, "not bytes!") + decoded = codecs.decode(None, self.codec_name) + self.assertEqual(decoded, b"not str!") + # Text model methods should complain + fmt = (r"^{!r} encoder returned 'str' instead of 'bytes'; " + "use codecs.encode\(\) to encode to arbitrary types$") + msg = fmt.format(self.codec_name) + with self.assertRaisesRegex(TypeError, msg): + "str_input".encode(self.codec_name) + fmt = (r"^{!r} decoder returned 'bytes' instead of 'str'; " + "use codecs.decode\(\) to decode to arbitrary types$") + msg = fmt.format(self.codec_name) + with self.assertRaisesRegex(TypeError, msg): + b"bytes input".decode(self.codec_name) + @unittest.skipUnless(sys.platform == 'win32', diff -r 90d4153728f6 Objects/unicodeobject.c --- a/Objects/unicodeobject.c Thu Nov 21 23:01:59 2013 -0800 +++ b/Objects/unicodeobject.c Fri Nov 22 21:15:49 2013 +1000 @@ -3044,7 +3044,7 @@ buffer = PyMemoryView_FromBuffer(&info); if (buffer == NULL) goto onError; - unicode = PyCodec_Decode(buffer, encoding, errors); + unicode = _PyCodec_DecodeText(buffer, encoding, errors); if (unicode == NULL) goto onError; if (!PyUnicode_Check(unicode)) { @@ -3410,7 +3410,7 @@ } /* Encode via the codec registry */ - v = PyCodec_Encode(unicode, encoding, errors); + v = _PyCodec_EncodeText(unicode, encoding, errors); if (v == NULL) return NULL; diff -r 90d4153728f6 Python/codecs.c --- a/Python/codecs.c Thu Nov 21 23:01:59 2013 -0800 +++ b/Python/codecs.c Fri Nov 22 21:15:49 2013 +1000 @@ -353,18 +353,15 @@ errors is passed to the encoder factory as argument if non-NULL. */ -PyObject *PyCodec_Encode(PyObject *object, - const char *encoding, - const char *errors) +static PyObject * +_PyCodec_EncodeInternal(PyObject *object, + PyObject *encoder, + const char *encoding, + const char *errors) { - PyObject *encoder = NULL; PyObject *args = NULL, *result = NULL; PyObject *v = NULL; - encoder = PyCodec_Encoder(encoding); - if (encoder == NULL) - goto onError; - args = args_tuple(object, errors); if (args == NULL) goto onError; @@ -402,18 +399,15 @@ errors is passed to the decoder factory as argument if non-NULL. */ -PyObject *PyCodec_Decode(PyObject *object, - const char *encoding, - const char *errors) +static PyObject * +_PyCodec_DecodeInternal(PyObject *object, + PyObject *decoder, + const char *encoding, + const char *errors) { - PyObject *decoder = NULL; PyObject *args = NULL, *result = NULL; PyObject *v; - decoder = PyCodec_Decoder(encoding); - if (decoder == NULL) - goto onError; - args = args_tuple(object, errors); if (args == NULL) goto onError; @@ -445,6 +439,118 @@ return NULL; } +/* Generic encoding/decoding API */ +PyObject *PyCodec_Encode(PyObject *object, + const char *encoding, + const char *errors) +{ + PyObject *encoder; + + encoder = PyCodec_Encoder(encoding); + if (encoder == NULL) + return NULL; + + return _PyCodec_EncodeInternal(object, encoder, encoding, errors); +} + +PyObject *PyCodec_Decode(PyObject *object, + const char *encoding, + const char *errors) +{ + PyObject *decoder; + + decoder = PyCodec_Decoder(encoding); + if (decoder == NULL) + return NULL; + + return _PyCodec_DecodeInternal(object, decoder, encoding, errors); +} + +/* Text encoding/decoding API */ +static +PyObject *codec_getitem_checked(const char *encoding, + const char *operation_name, + int index) +{ + _Py_IDENTIFIER(_is_text_encoding); + PyObject *codec; + PyObject *attr; + PyObject *v; + int is_text_codec; + + codec = _PyCodec_Lookup(encoding); + if (codec == NULL) + return NULL; + + /* Backwards compatibility: assume any raw tuple describes a text + * encoding, and the same for anything lacking the private + * attribute. + */ + if (!PyTuple_CheckExact(codec)) { + attr = _PyObject_GetAttrId(codec, &PyId__is_text_encoding); + if (attr == NULL) { + if (PyErr_ExceptionMatches(PyExc_AttributeError)) { + PyErr_Clear(); + } else { + Py_DECREF(codec); + return NULL; + } + } else { + is_text_codec = PyObject_IsTrue(attr); + Py_DECREF(attr); + if (!is_text_codec) { + Py_DECREF(codec); + PyErr_Format(PyExc_LookupError, + "'%.400s' is not a text encoding; " + "use codecs.%s() to handle arbitrary codecs", + encoding, operation_name); + return NULL; + } + } + } + + v = PyTuple_GET_ITEM(codec, index); + Py_DECREF(codec); + Py_INCREF(v); + return v; +} + +static PyObject * _PyCodec_TextEncoder(const char *encoding) +{ + return codec_getitem_checked(encoding, "encode", 0); +} + +static PyObject * _PyCodec_TextDecoder(const char *encoding) +{ + return codec_getitem_checked(encoding, "decode", 1); +} + +PyObject *_PyCodec_EncodeText(PyObject *object, + const char *encoding, + const char *errors) +{ + PyObject *encoder; + + encoder = _PyCodec_TextEncoder(encoding); + if (encoder == NULL) + return NULL; + + return _PyCodec_EncodeInternal(object, encoder, encoding, errors); +} + +PyObject *_PyCodec_DecodeText(PyObject *object, + const char *encoding, + const char *errors) +{ + PyObject *decoder; + + decoder = _PyCodec_TextDecoder(encoding); + if (decoder == NULL) + return NULL; + + return _PyCodec_DecodeInternal(object, decoder, encoding, errors); +} + /* Register the error handling callback function error under the name name. This function will be called by the codec when it encounters an unencodable characters/undecodable bytes and doesn't know the