forked from apache/arrow
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathserialization.py
More file actions
217 lines (164 loc) · 7.11 KB
/
Copy pathserialization.py
File metadata and controls
217 lines (164 loc) · 7.11 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied. See the License for the
# specific language governing permissions and limitations
# under the License.
from collections import OrderedDict, defaultdict
import six
import sys
import numpy as np
from pyarrow.compat import builtin_pickle
from pyarrow.lib import (SerializationContext, _default_serialization_context,
py_buffer)
try:
import cloudpickle
except ImportError:
cloudpickle = builtin_pickle
# ----------------------------------------------------------------------
# Set up serialization for numpy with dtype object (primitive types are
# handled efficiently with Arrow's Tensor facilities, see
# python_to_arrow.cc)
def _serialize_numpy_array_list(obj):
if obj.dtype.str != '|O':
# Make the array c_contiguous if necessary so that we can call change
# the view.
if not obj.flags.c_contiguous:
obj = np.ascontiguousarray(obj)
return obj.view('uint8'), obj.dtype.str
else:
return obj.tolist(), obj.dtype.str
def _deserialize_numpy_array_list(data):
if data[1] != '|O':
assert data[0].dtype == np.uint8
return data[0].view(data[1])
else:
return np.array(data[0], dtype=np.dtype(data[1]))
def _pickle_to_buffer(x):
pickled = builtin_pickle.dumps(x, protocol=builtin_pickle.HIGHEST_PROTOCOL)
return py_buffer(pickled)
def _load_pickle_from_buffer(data):
as_memoryview = memoryview(data)
if six.PY2:
return builtin_pickle.loads(as_memoryview.tobytes())
else:
return builtin_pickle.loads(as_memoryview)
# ----------------------------------------------------------------------
# pandas-specific serialization matters
def _register_custom_pandas_handlers(context):
# ARROW-1784, faster path for pandas-only visibility
try:
import pandas as pd
except ImportError:
return
import pyarrow.pandas_compat as pdcompat
sparse_type_error_msg = (
'{0} serialization is not supported.\n'
'Note that {0} is planned to be deprecated '
'in pandas future releases.\n'
'See https://github.com/pandas-dev/pandas/issues/19239 '
'for more information.'
)
def _serialize_pandas_dataframe(obj):
if isinstance(obj, pd.SparseDataFrame):
raise NotImplementedError(
sparse_type_error_msg.format('SparseDataFrame')
)
return pdcompat.dataframe_to_serialized_dict(obj)
def _deserialize_pandas_dataframe(data):
return pdcompat.serialized_dict_to_dataframe(data)
def _serialize_pandas_series(obj):
if isinstance(obj, pd.SparseSeries):
raise NotImplementedError(
sparse_type_error_msg.format('SparseSeries')
)
return _serialize_pandas_dataframe(pd.DataFrame({obj.name: obj}))
def _deserialize_pandas_series(data):
deserialized = _deserialize_pandas_dataframe(data)
return deserialized[deserialized.columns[0]]
context.register_type(
pd.Series, 'pd.Series',
custom_serializer=_serialize_pandas_series,
custom_deserializer=_deserialize_pandas_series)
context.register_type(
pd.Index, 'pd.Index',
custom_serializer=_pickle_to_buffer,
custom_deserializer=_load_pickle_from_buffer)
context.register_type(
pd.DataFrame, 'pd.DataFrame',
custom_serializer=_serialize_pandas_dataframe,
custom_deserializer=_deserialize_pandas_dataframe)
def register_torch_serialization_handlers(serialization_context):
# ----------------------------------------------------------------------
# Set up serialization for pytorch tensors
try:
import torch
def _serialize_torch_tensor(obj):
return obj.numpy()
def _deserialize_torch_tensor(data):
return torch.from_numpy(data)
for t in [torch.FloatTensor, torch.DoubleTensor, torch.HalfTensor,
torch.ByteTensor, torch.CharTensor, torch.ShortTensor,
torch.IntTensor, torch.LongTensor, torch.Tensor]:
serialization_context.register_type(
t, "torch." + t.__name__,
custom_serializer=_serialize_torch_tensor,
custom_deserializer=_deserialize_torch_tensor)
except ImportError:
# no torch
pass
def register_default_serialization_handlers(serialization_context):
# ----------------------------------------------------------------------
# Set up serialization for primitive datatypes
# TODO(pcm): This is currently a workaround until arrow supports
# arbitrary precision integers. This is only called on long integers,
# see the associated case in the append method in python_to_arrow.cc
serialization_context.register_type(
int, "int",
custom_serializer=lambda obj: str(obj),
custom_deserializer=lambda data: int(data))
if (sys.version_info < (3, 0)):
serialization_context.register_type(
long, "long", # noqa: F821
custom_serializer=lambda obj: str(obj),
custom_deserializer=lambda data: long(data)) # noqa: F821
def _serialize_ordered_dict(obj):
return list(obj.keys()), list(obj.values())
def _deserialize_ordered_dict(data):
return OrderedDict(zip(data[0], data[1]))
serialization_context.register_type(
OrderedDict, "OrderedDict",
custom_serializer=_serialize_ordered_dict,
custom_deserializer=_deserialize_ordered_dict)
def _serialize_default_dict(obj):
return list(obj.keys()), list(obj.values()), obj.default_factory
def _deserialize_default_dict(data):
return defaultdict(data[2], zip(data[0], data[1]))
serialization_context.register_type(
defaultdict, "defaultdict",
custom_serializer=_serialize_default_dict,
custom_deserializer=_deserialize_default_dict)
serialization_context.register_type(
type(lambda: 0), "function",
pickle=True)
serialization_context.register_type(type, "type", pickle=True)
serialization_context.register_type(
np.ndarray, 'np.array',
custom_serializer=_serialize_numpy_array_list,
custom_deserializer=_deserialize_numpy_array_list)
_register_custom_pandas_handlers(serialization_context)
def default_serialization_context():
context = SerializationContext()
register_default_serialization_handlers(context)
return context
register_default_serialization_handlers(_default_serialization_context)