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
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
|
# 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.
"""
Functions to interact with Arrow memory allocated by Arrow Java.
These functions convert the objects holding the metadata, the actual
data is not copied at all.
This will only work with a JVM running in the same process such as provided
through jpype. Modules that talk to a remote JVM like py4j will not work as the
memory addresses reported by them are not reachable in the python process.
"""
import pyarrow as pa
class _JvmBufferNanny:
"""
An object that keeps a org.apache.arrow.memory.ArrowBuf's underlying
memory alive.
"""
ref_manager = None
def __init__(self, jvm_buf):
ref_manager = jvm_buf.getReferenceManager()
# Will raise a java.lang.IllegalArgumentException if the buffer
# is already freed. It seems that exception cannot easily be
# caught...
ref_manager.retain()
self.ref_manager = ref_manager
def __del__(self):
if self.ref_manager is not None:
self.ref_manager.release()
def jvm_buffer(jvm_buf):
"""
Construct an Arrow buffer from org.apache.arrow.memory.ArrowBuf
Parameters
----------
jvm_buf: org.apache.arrow.memory.ArrowBuf
Arrow Buffer representation on the JVM.
Returns
-------
pyarrow.Buffer
Python Buffer that references the JVM memory.
"""
nanny = _JvmBufferNanny(jvm_buf)
address = jvm_buf.memoryAddress()
size = jvm_buf.capacity()
return pa.foreign_buffer(address, size, base=nanny)
def _from_jvm_int_type(jvm_type):
"""
Convert a JVM int type to its Python equivalent.
Parameters
----------
jvm_type : org.apache.arrow.vector.types.pojo.ArrowType$Int
Returns
-------
typ : pyarrow.DataType
"""
bit_width = jvm_type.getBitWidth()
if jvm_type.getIsSigned():
if bit_width == 8:
return pa.int8()
elif bit_width == 16:
return pa.int16()
elif bit_width == 32:
return pa.int32()
elif bit_width == 64:
return pa.int64()
else:
if bit_width == 8:
return pa.uint8()
elif bit_width == 16:
return pa.uint16()
elif bit_width == 32:
return pa.uint32()
elif bit_width == 64:
return pa.uint64()
def _from_jvm_float_type(jvm_type):
"""
Convert a JVM float type to its Python equivalent.
Parameters
----------
jvm_type: org.apache.arrow.vector.types.pojo.ArrowType$FloatingPoint
Returns
-------
typ: pyarrow.DataType
"""
precision = jvm_type.getPrecision().toString()
if precision == 'HALF':
return pa.float16()
elif precision == 'SINGLE':
return pa.float32()
elif precision == 'DOUBLE':
return pa.float64()
def _from_jvm_time_type(jvm_type):
"""
Convert a JVM time type to its Python equivalent.
Parameters
----------
jvm_type: org.apache.arrow.vector.types.pojo.ArrowType$Time
Returns
-------
typ: pyarrow.DataType
"""
time_unit = jvm_type.getUnit().toString()
if time_unit == 'SECOND':
assert jvm_type.getBitWidth() == 32
return pa.time32('s')
elif time_unit == 'MILLISECOND':
assert jvm_type.getBitWidth() == 32
return pa.time32('ms')
elif time_unit == 'MICROSECOND':
assert jvm_type.getBitWidth() == 64
return pa.time64('us')
elif time_unit == 'NANOSECOND':
assert jvm_type.getBitWidth() == 64
return pa.time64('ns')
def _from_jvm_timestamp_type(jvm_type):
"""
Convert a JVM timestamp type to its Python equivalent.
Parameters
----------
jvm_type: org.apache.arrow.vector.types.pojo.ArrowType$Timestamp
Returns
-------
typ: pyarrow.DataType
"""
time_unit = jvm_type.getUnit().toString()
timezone = jvm_type.getTimezone()
if timezone is not None:
timezone = str(timezone)
if time_unit == 'SECOND':
return pa.timestamp('s', tz=timezone)
elif time_unit == 'MILLISECOND':
return pa.timestamp('ms', tz=timezone)
elif time_unit == 'MICROSECOND':
return pa.timestamp('us', tz=timezone)
elif time_unit == 'NANOSECOND':
return pa.timestamp('ns', tz=timezone)
def _from_jvm_date_type(jvm_type):
"""
Convert a JVM date type to its Python equivalent
Parameters
----------
jvm_type: org.apache.arrow.vector.types.pojo.ArrowType$Date
Returns
-------
typ: pyarrow.DataType
"""
day_unit = jvm_type.getUnit().toString()
if day_unit == 'DAY':
return pa.date32()
elif day_unit == 'MILLISECOND':
return pa.date64()
def field(jvm_field):
"""
Construct a Field from a org.apache.arrow.vector.types.pojo.Field
instance.
Parameters
----------
jvm_field: org.apache.arrow.vector.types.pojo.Field
Returns
-------
pyarrow.Field
"""
name = str(jvm_field.getName())
jvm_type = jvm_field.getType()
typ = None
if not jvm_type.isComplex():
type_str = jvm_type.getTypeID().toString()
if type_str == 'Null':
typ = pa.null()
elif type_str == 'Int':
typ = _from_jvm_int_type(jvm_type)
elif type_str == 'FloatingPoint':
typ = _from_jvm_float_type(jvm_type)
elif type_str == 'Utf8':
typ = pa.string()
elif type_str == 'Binary':
typ = pa.binary()
elif type_str == 'FixedSizeBinary':
typ = pa.binary(jvm_type.getByteWidth())
elif type_str == 'Bool':
typ = pa.bool_()
elif type_str == 'Time':
typ = _from_jvm_time_type(jvm_type)
elif type_str == 'Timestamp':
typ = _from_jvm_timestamp_type(jvm_type)
elif type_str == 'Date':
typ = _from_jvm_date_type(jvm_type)
elif type_str == 'Decimal':
typ = pa.decimal128(jvm_type.getPrecision(), jvm_type.getScale())
else:
raise NotImplementedError(
"Unsupported JVM type: {}".format(type_str))
else:
# TODO: The following JVM types are not implemented:
# Struct, List, FixedSizeList, Union, Dictionary
raise NotImplementedError(
"JVM field conversion only implemented for primitive types.")
nullable = jvm_field.isNullable()
jvm_metadata = jvm_field.getMetadata()
if jvm_metadata.isEmpty():
metadata = None
else:
metadata = {str(entry.getKey()): str(entry.getValue())
for entry in jvm_metadata.entrySet()}
return pa.field(name, typ, nullable, metadata)
def schema(jvm_schema):
"""
Construct a Schema from a org.apache.arrow.vector.types.pojo.Schema
instance.
Parameters
----------
jvm_schema: org.apache.arrow.vector.types.pojo.Schema
Returns
-------
pyarrow.Schema
"""
fields = jvm_schema.getFields()
fields = [field(f) for f in fields]
jvm_metadata = jvm_schema.getCustomMetadata()
if jvm_metadata.isEmpty():
metadata = None
else:
metadata = {str(entry.getKey()): str(entry.getValue())
for entry in jvm_metadata.entrySet()}
return pa.schema(fields, metadata)
def array(jvm_array):
"""
Construct an (Python) Array from its JVM equivalent.
Parameters
----------
jvm_array : org.apache.arrow.vector.ValueVector
Returns
-------
array : Array
"""
if jvm_array.getField().getType().isComplex():
minor_type_str = jvm_array.getMinorType().toString()
raise NotImplementedError(
"Cannot convert JVM Arrow array of type {},"
" complex types not yet implemented.".format(minor_type_str))
dtype = field(jvm_array.getField()).type
buffers = [jvm_buffer(buf)
for buf in list(jvm_array.getBuffers(False))]
# If JVM has an empty Vector, buffer list will be empty so create manually
if len(buffers) == 0:
return pa.array([], type=dtype)
length = jvm_array.getValueCount()
null_count = jvm_array.getNullCount()
return pa.Array.from_buffers(dtype, length, buffers, null_count)
def record_batch(jvm_vector_schema_root):
"""
Construct a (Python) RecordBatch from a JVM VectorSchemaRoot
Parameters
----------
jvm_vector_schema_root : org.apache.arrow.vector.VectorSchemaRoot
Returns
-------
record_batch: pyarrow.RecordBatch
"""
pa_schema = schema(jvm_vector_schema_root.getSchema())
arrays = []
for name in pa_schema.names:
arrays.append(array(jvm_vector_schema_root.getVector(name)))
return pa.RecordBatch.from_arrays(
arrays,
pa_schema.names,
metadata=pa_schema.metadata
)
|
