-
-
Notifications
You must be signed in to change notification settings - Fork 1k
/
dtypes.py
195 lines (147 loc) · 5.13 KB
/
dtypes.py
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
from __future__ import annotations
import functools
from typing import Any
import numpy as np
from xarray.core import utils
# Use as a sentinel value to indicate a dtype appropriate NA value.
NA = utils.ReprObject("<NA>")
@functools.total_ordering
class AlwaysGreaterThan:
def __gt__(self, other):
return True
def __eq__(self, other):
return isinstance(other, type(self))
@functools.total_ordering
class AlwaysLessThan:
def __lt__(self, other):
return True
def __eq__(self, other):
return isinstance(other, type(self))
# Equivalence to np.inf (-np.inf) for object-type
INF = AlwaysGreaterThan()
NINF = AlwaysLessThan()
# Pairs of types that, if both found, should be promoted to object dtype
# instead of following NumPy's own type-promotion rules. These type promotion
# rules match pandas instead. For reference, see the NumPy type hierarchy:
# https://numpy.org/doc/stable/reference/arrays.scalars.html
PROMOTE_TO_OBJECT: tuple[tuple[type[np.generic], type[np.generic]], ...] = (
(np.number, np.character), # numpy promotes to character
(np.bool_, np.character), # numpy promotes to character
(np.bytes_, np.str_), # numpy promotes to unicode
)
def maybe_promote(dtype: np.dtype) -> tuple[np.dtype, Any]:
"""Simpler equivalent of pandas.core.common._maybe_promote
Parameters
----------
dtype : np.dtype
Returns
-------
dtype : Promoted dtype that can hold missing values.
fill_value : Valid missing value for the promoted dtype.
"""
# N.B. these casting rules should match pandas
dtype_: np.typing.DTypeLike
fill_value: Any
if np.issubdtype(dtype, np.floating):
dtype_ = dtype
fill_value = np.nan
elif np.issubdtype(dtype, np.timedelta64):
# See https://github.com/numpy/numpy/issues/10685
# np.timedelta64 is a subclass of np.integer
# Check np.timedelta64 before np.integer
fill_value = np.timedelta64("NaT")
dtype_ = dtype
elif np.issubdtype(dtype, np.integer):
dtype_ = np.float32 if dtype.itemsize <= 2 else np.float64
fill_value = np.nan
elif np.issubdtype(dtype, np.complexfloating):
dtype_ = dtype
fill_value = np.nan + np.nan * 1j
elif np.issubdtype(dtype, np.datetime64):
dtype_ = dtype
fill_value = np.datetime64("NaT")
else:
dtype_ = object
fill_value = np.nan
dtype_out = np.dtype(dtype_)
fill_value = dtype_out.type(fill_value)
return dtype_out, fill_value
NAT_TYPES = {np.datetime64("NaT").dtype, np.timedelta64("NaT").dtype}
def get_fill_value(dtype):
"""Return an appropriate fill value for this dtype.
Parameters
----------
dtype : np.dtype
Returns
-------
fill_value : Missing value corresponding to this dtype.
"""
_, fill_value = maybe_promote(dtype)
return fill_value
def get_pos_infinity(dtype, max_for_int=False):
"""Return an appropriate positive infinity for this dtype.
Parameters
----------
dtype : np.dtype
max_for_int : bool
Return np.iinfo(dtype).max instead of np.inf
Returns
-------
fill_value : positive infinity value corresponding to this dtype.
"""
if issubclass(dtype.type, np.floating):
return np.inf
if issubclass(dtype.type, np.integer):
if max_for_int:
return np.iinfo(dtype).max
else:
return np.inf
if issubclass(dtype.type, np.complexfloating):
return np.inf + 1j * np.inf
return INF
def get_neg_infinity(dtype, min_for_int=False):
"""Return an appropriate positive infinity for this dtype.
Parameters
----------
dtype : np.dtype
min_for_int : bool
Return np.iinfo(dtype).min instead of -np.inf
Returns
-------
fill_value : positive infinity value corresponding to this dtype.
"""
if issubclass(dtype.type, np.floating):
return -np.inf
if issubclass(dtype.type, np.integer):
if min_for_int:
return np.iinfo(dtype).min
else:
return -np.inf
if issubclass(dtype.type, np.complexfloating):
return -np.inf - 1j * np.inf
return NINF
def is_datetime_like(dtype):
"""Check if a dtype is a subclass of the numpy datetime types"""
return np.issubdtype(dtype, np.datetime64) or np.issubdtype(dtype, np.timedelta64)
def result_type(
*arrays_and_dtypes: np.typing.ArrayLike | np.typing.DTypeLike,
) -> np.dtype:
"""Like np.result_type, but with type promotion rules matching pandas.
Examples of changed behavior:
number + string -> object (not string)
bytes + unicode -> object (not unicode)
Parameters
----------
*arrays_and_dtypes : list of arrays and dtypes
The dtype is extracted from both numpy and dask arrays.
Returns
-------
numpy.dtype for the result.
"""
types = {np.result_type(t).type for t in arrays_and_dtypes}
for left, right in PROMOTE_TO_OBJECT:
if any(issubclass(t, left) for t in types) and any(
issubclass(t, right) for t in types
):
return np.dtype(object)
return np.result_type(*arrays_and_dtypes)