diff --git a/swh/core/api/serializers.py b/swh/core/api/serializers.py index 58dbc4f..9d7c9ec 100644 --- a/swh/core/api/serializers.py +++ b/swh/core/api/serializers.py @@ -1,245 +1,250 @@ # Copyright (C) 2015-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import base64 import datetime import json import traceback import types from uuid import UUID import arrow import iso8601 import msgpack from typing import Any, Dict, Union, Tuple from requests import Response ENCODERS = [ (arrow.Arrow, 'arrow', arrow.Arrow.isoformat), (datetime.datetime, 'datetime', datetime.datetime.isoformat), (datetime.timedelta, 'timedelta', lambda o: { 'days': o.days, 'seconds': o.seconds, 'microseconds': o.microseconds, }), (UUID, 'uuid', str), # Only for JSON: (bytes, 'bytes', lambda o: base64.b85encode(o).decode('ascii')), ] DECODERS = { 'arrow': arrow.get, 'datetime': lambda d: iso8601.parse_date(d, default_timezone=None), 'timedelta': lambda d: datetime.timedelta(**d), 'uuid': UUID, # Only for JSON: 'bytes': base64.b85decode, } def encode_data_client(data: Any, extra_encoders=None) -> bytes: try: return msgpack_dumps(data, extra_encoders=extra_encoders) except OverflowError as e: raise ValueError('Limits were reached. Please, check your input.\n' + str(e)) def decode_response(response: Response, extra_decoders=None) -> Any: content_type = response.headers['content-type'] if content_type.startswith('application/x-msgpack'): r = msgpack_loads(response.content, extra_decoders=extra_decoders) elif content_type.startswith('application/json'): r = json_loads(response.text, extra_decoders=extra_decoders) elif content_type.startswith('text/'): r = response.text else: raise ValueError('Wrong content type `%s` for API response' % content_type) return r class SWHJSONEncoder(json.JSONEncoder): """JSON encoder for data structures generated by Software Heritage. This JSON encoder extends the default Python JSON encoder and adds awareness for the following specific types: - bytes (get encoded as a Base85 string); - datetime.datetime (get encoded as an ISO8601 string). Non-standard types get encoded as a a dictionary with two keys: - swhtype with value 'bytes' or 'datetime'; - d containing the encoded value. SWHJSONEncoder also encodes arbitrary iterables as a list (allowing serialization of generators). Caveats: Limitations in the JSONEncoder extension mechanism prevent us from "escaping" dictionaries that only contain the swhtype and d keys, and therefore arbitrary data structures can't be round-tripped through SWHJSONEncoder and SWHJSONDecoder. """ def __init__(self, extra_encoders=None, **kwargs): super().__init__(**kwargs) self.encoders = ENCODERS if extra_encoders: self.encoders += extra_encoders def default(self, o: Any ) -> Union[Dict[str, Union[Dict[str, int], str]], list]: for (type_, type_name, encoder) in self.encoders: if isinstance(o, type_): return { 'swhtype': type_name, 'd': encoder(o), } try: return super().default(o) except TypeError as e: try: iterable = iter(o) except TypeError: raise e from None else: return list(iterable) class SWHJSONDecoder(json.JSONDecoder): """JSON decoder for data structures encoded with SWHJSONEncoder. This JSON decoder extends the default Python JSON decoder, allowing the decoding of: - bytes (encoded as a Base85 string); - datetime.datetime (encoded as an ISO8601 string). Non-standard types must be encoded as a a dictionary with exactly two keys: - swhtype with value 'bytes' or 'datetime'; - d containing the encoded value. To limit the impact our encoding, if the swhtype key doesn't contain a known value, the dictionary is decoded as-is. """ def __init__(self, extra_decoders=None, **kwargs): super().__init__(**kwargs) self.decoders = DECODERS if extra_decoders: self.decoders = {**self.decoders, **extra_decoders} def decode_data(self, o: Any) -> Any: if isinstance(o, dict): if set(o.keys()) == {'d', 'swhtype'}: if o['swhtype'] == 'bytes': return base64.b85decode(o['d']) decoder = self.decoders.get(o['swhtype']) if decoder: return decoder(self.decode_data(o['d'])) return {key: self.decode_data(value) for key, value in o.items()} if isinstance(o, list): return [self.decode_data(value) for value in o] else: return o def raw_decode(self, s: str, idx: int = 0) -> Tuple[Any, int]: data, index = super().raw_decode(s, idx) return self.decode_data(data), index def json_dumps(data: Any, extra_encoders=None) -> str: return json.dumps(data, cls=SWHJSONEncoder, extra_encoders=extra_encoders) def json_loads(data: str, extra_decoders=None) -> Any: return json.loads(data, cls=SWHJSONDecoder, extra_decoders=extra_decoders) def msgpack_dumps(data: Any, extra_encoders=None) -> bytes: """Write data as a msgpack stream""" encoders = ENCODERS if extra_encoders: encoders += extra_encoders def encode_types(obj): if isinstance(obj, types.GeneratorType): return list(obj) for (type_, type_name, encoder) in encoders: if isinstance(obj, type_): return { b'swhtype': type_name, b'd': encoder(obj), } return obj return msgpack.packb(data, use_bin_type=True, default=encode_types) def msgpack_loads(data: bytes, extra_decoders=None) -> Any: - """Read data as a msgpack stream""" + """Read data as a msgpack stream. + + .. Caution:: + This function is used by swh.journal to decode the contents of the + journal. This function **must** be kept backwards-compatible. + """ decoders = DECODERS if extra_decoders: decoders = {**decoders, **extra_decoders} def decode_types(obj): # Support for current encodings if set(obj.keys()) == {b'd', b'swhtype'}: decoder = decoders.get(obj[b'swhtype']) if decoder: return decoder(obj[b'd']) # Support for legacy encodings if b'__datetime__' in obj and obj[b'__datetime__']: return iso8601.parse_date(obj[b's'], default_timezone=None) if b'__uuid__' in obj and obj[b'__uuid__']: return UUID(obj[b's']) if b'__timedelta__' in obj and obj[b'__timedelta__']: return datetime.timedelta(**obj[b's']) if b'__arrow__' in obj and obj[b'__arrow__']: return arrow.get(obj[b's']) # Fallthrough return obj try: try: return msgpack.unpackb(data, raw=False, object_hook=decode_types, strict_map_key=False) except TypeError: # msgpack < 0.6.0 return msgpack.unpackb(data, raw=False, object_hook=decode_types) except TypeError: # msgpack < 0.5.2 return msgpack.unpackb(data, encoding='utf-8', object_hook=decode_types) def exception_to_dict(exception): tb = traceback.format_exception(None, exception, exception.__traceback__) return { 'exception': { 'type': type(exception).__name__, 'args': exception.args, 'message': str(exception), 'traceback': tb, } }