forked from pydata/xarray
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathstrings.py
More file actions
240 lines (181 loc) · 7.92 KB
/
strings.py
File metadata and controls
240 lines (181 loc) · 7.92 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
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
"""Coders for strings."""
from __future__ import annotations
from functools import partial
import numpy as np
from xarray.coding.variables import (
VariableCoder,
lazy_elemwise_func,
pop_to,
safe_setitem,
unpack_for_decoding,
unpack_for_encoding,
)
from xarray.core import indexing
from xarray.core.parallelcompat import get_chunked_array_type, is_chunked_array
from xarray.core.variable import Variable
def create_vlen_dtype(element_type):
if element_type not in (str, bytes):
raise TypeError(f"unsupported type for vlen_dtype: {element_type!r}")
# based on h5py.special_dtype
return np.dtype("O", metadata={"element_type": element_type})
def check_vlen_dtype(dtype):
if dtype.kind != "O" or dtype.metadata is None:
return None
else:
# check xarray (element_type) as well as h5py (vlen)
return dtype.metadata.get("element_type", dtype.metadata.get("vlen"))
def is_unicode_dtype(dtype):
return dtype.kind == "U" or check_vlen_dtype(dtype) == str
def is_bytes_dtype(dtype):
return dtype.kind == "S" or check_vlen_dtype(dtype) == bytes
class EncodedStringCoder(VariableCoder):
"""Transforms between unicode strings and fixed-width UTF-8 bytes."""
def __init__(self, allows_unicode=True):
self.allows_unicode = allows_unicode
def encode(self, variable, name=None):
dims, data, attrs, encoding = unpack_for_encoding(variable)
contains_unicode = is_unicode_dtype(data.dtype)
encode_as_char = encoding.get("dtype") == "S1"
if encode_as_char:
del encoding["dtype"] # no longer relevant
if contains_unicode and (encode_as_char or not self.allows_unicode):
if "_FillValue" in attrs:
raise NotImplementedError(
f"variable {name!r} has a _FillValue specified, but "
"_FillValue is not yet supported on unicode strings: "
"https://github.com/pydata/xarray/issues/1647"
)
string_encoding = encoding.pop("_Encoding", "utf-8")
safe_setitem(attrs, "_Encoding", string_encoding, name=name)
# TODO: figure out how to handle this in a lazy way with dask
data = encode_string_array(data, string_encoding)
return Variable(dims, data, attrs, encoding)
def decode(self, variable, name=None):
dims, data, attrs, encoding = unpack_for_decoding(variable)
if "_Encoding" in attrs:
string_encoding = pop_to(attrs, encoding, "_Encoding")
func = partial(decode_bytes_array, encoding=string_encoding)
data = lazy_elemwise_func(data, func, np.dtype(object))
return Variable(dims, data, attrs, encoding)
def decode_bytes_array(bytes_array, encoding="utf-8"):
# This is faster than using np.char.decode() or np.vectorize()
bytes_array = np.asarray(bytes_array)
decoded = [x.decode(encoding) for x in bytes_array.ravel()]
return np.array(decoded, dtype=object).reshape(bytes_array.shape)
def encode_string_array(string_array, encoding="utf-8"):
string_array = np.asarray(string_array)
encoded = [x.encode(encoding) for x in string_array.ravel()]
return np.array(encoded, dtype=bytes).reshape(string_array.shape)
def ensure_fixed_length_bytes(var):
"""Ensure that a variable with vlen bytes is converted to fixed width."""
dims, data, attrs, encoding = unpack_for_encoding(var)
if check_vlen_dtype(data.dtype) == bytes:
# TODO: figure out how to handle this with dask
data = np.asarray(data, dtype=np.bytes_)
return Variable(dims, data, attrs, encoding)
class CharacterArrayCoder(VariableCoder):
"""Transforms between arrays containing bytes and character arrays."""
def encode(self, variable, name=None):
variable = ensure_fixed_length_bytes(variable)
dims, data, attrs, encoding = unpack_for_encoding(variable)
if data.dtype.kind == "S" and encoding.get("dtype") is not str:
data = bytes_to_char(data)
if "char_dim_name" in encoding.keys():
char_dim_name = encoding.pop("char_dim_name")
else:
char_dim_name = f"string{data.shape[-1]}"
dims = dims + (char_dim_name,)
return Variable(dims, data, attrs, encoding)
def decode(self, variable, name=None):
dims, data, attrs, encoding = unpack_for_decoding(variable)
if data.dtype == "S1" and dims:
encoding["char_dim_name"] = dims[-1]
dims = dims[:-1]
data = char_to_bytes(data)
return Variable(dims, data, attrs, encoding)
def bytes_to_char(arr):
"""Convert numpy/dask arrays from fixed width bytes to characters."""
if arr.dtype.kind != "S":
raise ValueError("argument must have a fixed-width bytes dtype")
if is_chunked_array(arr):
chunkmanager = get_chunked_array_type(arr)
return chunkmanager.map_blocks(
_numpy_bytes_to_char,
arr,
dtype="S1",
chunks=arr.chunks + ((arr.dtype.itemsize,)),
new_axis=[arr.ndim],
)
return _numpy_bytes_to_char(arr)
def _numpy_bytes_to_char(arr):
"""Like netCDF4.stringtochar, but faster and more flexible."""
# ensure the array is contiguous
arr = np.array(arr, copy=False, order="C", dtype=np.bytes_)
return arr.reshape(arr.shape + (1,)).view("S1")
def char_to_bytes(arr):
"""Convert numpy/dask arrays from characters to fixed width bytes."""
if arr.dtype != "S1":
raise ValueError("argument must have dtype='S1'")
if not arr.ndim:
# no dimension to concatenate along
return arr
size = arr.shape[-1]
if not size:
# can't make an S0 dtype
return np.zeros(arr.shape[:-1], dtype=np.bytes_)
if is_chunked_array(arr):
chunkmanager = get_chunked_array_type(arr)
if len(arr.chunks[-1]) > 1:
raise ValueError(
"cannot stacked dask character array with "
f"multiple chunks in the last dimension: {arr}"
)
dtype = np.dtype("S" + str(arr.shape[-1]))
return chunkmanager.map_blocks(
_numpy_char_to_bytes,
arr,
dtype=dtype,
chunks=arr.chunks[:-1],
drop_axis=[arr.ndim - 1],
)
else:
return StackedBytesArray(arr)
def _numpy_char_to_bytes(arr):
"""Like netCDF4.chartostring, but faster and more flexible."""
# based on: http://stackoverflow.com/a/10984878/809705
arr = np.array(arr, copy=False, order="C")
dtype = "S" + str(arr.shape[-1])
return arr.view(dtype).reshape(arr.shape[:-1])
class StackedBytesArray(indexing.ExplicitlyIndexedNDArrayMixin):
"""Wrapper around array-like objects to create a new indexable object where
values, when accessed, are automatically stacked along the last dimension.
>>> indexer = indexing.BasicIndexer((slice(None),))
>>> StackedBytesArray(np.array(["a", "b", "c"], dtype="S1"))[indexer]
array(b'abc', dtype='|S3')
"""
def __init__(self, array):
"""
Parameters
----------
array : array-like
Original array of values to wrap.
"""
if array.dtype != "S1":
raise ValueError(
"can only use StackedBytesArray if argument has dtype='S1'"
)
self.array = indexing.as_indexable(array)
@property
def dtype(self):
return np.dtype("S" + str(self.array.shape[-1]))
@property
def shape(self) -> tuple[int, ...]:
return self.array.shape[:-1]
def __repr__(self):
return f"{type(self).__name__}({self.array!r})"
def __getitem__(self, key):
# require slicing the last dimension completely
key = type(key)(indexing.expanded_indexer(key.tuple, self.array.ndim))
if key.tuple[-1] != slice(None):
raise IndexError("too many indices")
return _numpy_char_to_bytes(self.array[key])