diff options
Diffstat (limited to 'python/google/protobuf/internal/encoder.py')
| -rwxr-xr-x | python/google/protobuf/internal/encoder.py | 120 |
1 files changed, 111 insertions, 9 deletions
diff --git a/python/google/protobuf/internal/encoder.py b/python/google/protobuf/internal/encoder.py index aa05d5b..38a5138 100755 --- a/python/google/protobuf/internal/encoder.py +++ b/python/google/protobuf/internal/encoder.py @@ -1,6 +1,6 @@ # Protocol Buffers - Google's data interchange format # Copyright 2008 Google Inc. All rights reserved. -# http://code.google.com/p/protobuf/ +# https://developers.google.com/protocol-buffers/ # # Redistribution and use in source and binary forms, with or without # modification, are permitted provided that the following conditions are @@ -28,6 +28,10 @@ # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +#PY25 compatible for GAE. +# +# Copyright 2009 Google Inc. All Rights Reserved. + """Code for encoding protocol message primitives. Contains the logic for encoding every logical protocol field type @@ -67,9 +71,17 @@ sizer rather than when calling them. In particular: __author__ = 'kenton@google.com (Kenton Varda)' import struct +import sys ##PY25 +_PY2 = sys.version_info[0] < 3 ##PY25 from google.protobuf.internal import wire_format +# This will overflow and thus become IEEE-754 "infinity". We would use +# "float('inf')" but it doesn't work on Windows pre-Python-2.6. +_POS_INF = 1e10000 +_NEG_INF = -_POS_INF + + def _VarintSize(value): """Compute the size of a varint value.""" if value <= 0x7f: return 1 @@ -334,7 +346,8 @@ def MessageSetItemSizer(field_number): def _VarintEncoder(): """Return an encoder for a basic varint value (does not include tag).""" - local_chr = chr + local_chr = _PY2 and chr or (lambda x: bytes((x,))) ##PY25 +##!PY25 local_chr = chr if bytes is str else lambda x: bytes((x,)) def EncodeVarint(write, value): bits = value & 0x7f value >>= 7 @@ -351,7 +364,8 @@ def _SignedVarintEncoder(): """Return an encoder for a basic signed varint value (does not include tag).""" - local_chr = chr + local_chr = _PY2 and chr or (lambda x: bytes((x,))) ##PY25 +##!PY25 local_chr = chr if bytes is str else lambda x: bytes((x,)) def EncodeSignedVarint(write, value): if value < 0: value += (1 << 64) @@ -376,7 +390,8 @@ def _VarintBytes(value): pieces = [] _EncodeVarint(pieces.append, value) - return "".join(pieces) + return "".encode("latin1").join(pieces) ##PY25 +##!PY25 return b"".join(pieces) def TagBytes(field_number, wire_type): @@ -502,6 +517,90 @@ def _StructPackEncoder(wire_type, format): return SpecificEncoder +def _FloatingPointEncoder(wire_type, format): + """Return a constructor for an encoder for float fields. + + This is like StructPackEncoder, but catches errors that may be due to + passing non-finite floating-point values to struct.pack, and makes a + second attempt to encode those values. + + Args: + wire_type: The field's wire type, for encoding tags. + format: The format string to pass to struct.pack(). + """ + + b = _PY2 and (lambda x:x) or (lambda x:x.encode('latin1')) ##PY25 + value_size = struct.calcsize(format) + if value_size == 4: + def EncodeNonFiniteOrRaise(write, value): + # Remember that the serialized form uses little-endian byte order. + if value == _POS_INF: + write(b('\x00\x00\x80\x7F')) ##PY25 +##!PY25 write(b'\x00\x00\x80\x7F') + elif value == _NEG_INF: + write(b('\x00\x00\x80\xFF')) ##PY25 +##!PY25 write(b'\x00\x00\x80\xFF') + elif value != value: # NaN + write(b('\x00\x00\xC0\x7F')) ##PY25 +##!PY25 write(b'\x00\x00\xC0\x7F') + else: + raise + elif value_size == 8: + def EncodeNonFiniteOrRaise(write, value): + if value == _POS_INF: + write(b('\x00\x00\x00\x00\x00\x00\xF0\x7F')) ##PY25 +##!PY25 write(b'\x00\x00\x00\x00\x00\x00\xF0\x7F') + elif value == _NEG_INF: + write(b('\x00\x00\x00\x00\x00\x00\xF0\xFF')) ##PY25 +##!PY25 write(b'\x00\x00\x00\x00\x00\x00\xF0\xFF') + elif value != value: # NaN + write(b('\x00\x00\x00\x00\x00\x00\xF8\x7F')) ##PY25 +##!PY25 write(b'\x00\x00\x00\x00\x00\x00\xF8\x7F') + else: + raise + else: + raise ValueError('Can\'t encode floating-point values that are ' + '%d bytes long (only 4 or 8)' % value_size) + + def SpecificEncoder(field_number, is_repeated, is_packed): + local_struct_pack = struct.pack + if is_packed: + tag_bytes = TagBytes(field_number, wire_format.WIRETYPE_LENGTH_DELIMITED) + local_EncodeVarint = _EncodeVarint + def EncodePackedField(write, value): + write(tag_bytes) + local_EncodeVarint(write, len(value) * value_size) + for element in value: + # This try/except block is going to be faster than any code that + # we could write to check whether element is finite. + try: + write(local_struct_pack(format, element)) + except SystemError: + EncodeNonFiniteOrRaise(write, element) + return EncodePackedField + elif is_repeated: + tag_bytes = TagBytes(field_number, wire_type) + def EncodeRepeatedField(write, value): + for element in value: + write(tag_bytes) + try: + write(local_struct_pack(format, element)) + except SystemError: + EncodeNonFiniteOrRaise(write, element) + return EncodeRepeatedField + else: + tag_bytes = TagBytes(field_number, wire_type) + def EncodeField(write, value): + write(tag_bytes) + try: + write(local_struct_pack(format, value)) + except SystemError: + EncodeNonFiniteOrRaise(write, value) + return EncodeField + + return SpecificEncoder + + # ==================================================================== # Here we declare an encoder constructor for each field type. These work # very similarly to sizer constructors, described earlier. @@ -525,15 +624,17 @@ Fixed32Encoder = _StructPackEncoder(wire_format.WIRETYPE_FIXED32, '<I') Fixed64Encoder = _StructPackEncoder(wire_format.WIRETYPE_FIXED64, '<Q') SFixed32Encoder = _StructPackEncoder(wire_format.WIRETYPE_FIXED32, '<i') SFixed64Encoder = _StructPackEncoder(wire_format.WIRETYPE_FIXED64, '<q') -FloatEncoder = _StructPackEncoder(wire_format.WIRETYPE_FIXED32, '<f') -DoubleEncoder = _StructPackEncoder(wire_format.WIRETYPE_FIXED64, '<d') +FloatEncoder = _FloatingPointEncoder(wire_format.WIRETYPE_FIXED32, '<f') +DoubleEncoder = _FloatingPointEncoder(wire_format.WIRETYPE_FIXED64, '<d') def BoolEncoder(field_number, is_repeated, is_packed): """Returns an encoder for a boolean field.""" - false_byte = chr(0) - true_byte = chr(1) +##!PY25 false_byte = b'\x00' +##!PY25 true_byte = b'\x01' + false_byte = '\x00'.encode('latin1') ##PY25 + true_byte = '\x01'.encode('latin1') ##PY25 if is_packed: tag_bytes = TagBytes(field_number, wire_format.WIRETYPE_LENGTH_DELIMITED) local_EncodeVarint = _EncodeVarint @@ -669,7 +770,8 @@ def MessageSetItemEncoder(field_number): } } """ - start_bytes = "".join([ + start_bytes = "".encode("latin1").join([ ##PY25 +##!PY25 start_bytes = b"".join([ TagBytes(1, wire_format.WIRETYPE_START_GROUP), TagBytes(2, wire_format.WIRETYPE_VARINT), _VarintBytes(field_number), |