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
|
from __future__ import annotations
from collections.abc import Mapping
from typing import Any, Generic
import numpy as np
from xarray.compat.pdcompat import count_not_none
from xarray.computation.apply_ufunc import apply_ufunc
from xarray.core.options import _get_keep_attrs
from xarray.core.types import T_DataWithCoords
from xarray.core.utils import module_available
def _get_alpha(
com: float | None = None,
span: float | None = None,
halflife: float | None = None,
alpha: float | None = None,
) -> float:
"""
Convert com, span, halflife to alpha.
"""
valid_count = count_not_none(com, span, halflife, alpha)
if valid_count > 1:
raise ValueError("com, span, halflife, and alpha are mutually exclusive")
# Convert to alpha
if com is not None:
if com < 0:
raise ValueError("commust satisfy: com>= 0")
return 1 / (com + 1)
elif span is not None:
if span < 1:
raise ValueError("span must satisfy: span >= 1")
return 2 / (span + 1)
elif halflife is not None:
if halflife <= 0:
raise ValueError("halflife must satisfy: halflife > 0")
return 1 - np.exp(np.log(0.5) / halflife)
elif alpha is not None:
if not 0 < alpha <= 1:
raise ValueError("alpha must satisfy: 0 < alpha <= 1")
return alpha
else:
raise ValueError("Must pass one of comass, span, halflife, or alpha")
class RollingExp(Generic[T_DataWithCoords]):
"""
Exponentially-weighted moving window object.
Similar to EWM in pandas
Parameters
----------
obj : Dataset or DataArray
Object to window.
windows : mapping of hashable to int (or float for alpha type)
A mapping from the name of the dimension to create the rolling
exponential window along (e.g. `time`) to the size of the moving window.
window_type : {"span", "com", "halflife", "alpha"}, default: "span"
The format of the previously supplied window. Each is a simple
numerical transformation of the others. Described in detail:
https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.ewm.html
Returns
-------
RollingExp : type of input argument
"""
def __init__(
self,
obj: T_DataWithCoords,
windows: Mapping[Any, int | float],
window_type: str = "span",
min_weight: float = 0.0,
):
if not module_available("numbagg"):
raise ImportError(
"numbagg >= 0.2.1 is required for rolling_exp but currently numbagg is not installed"
)
self.obj: T_DataWithCoords = obj
dim, window = next(iter(windows.items()))
self.dim = dim
self.alpha = _get_alpha(**{window_type: window})
self.min_weight = min_weight
# Don't pass min_weight=0 so we can support older versions of numbagg
kwargs = dict(alpha=self.alpha, axis=-1)
if min_weight > 0:
kwargs["min_weight"] = min_weight
self.kwargs = kwargs
def mean(self, keep_attrs: bool | None = None) -> T_DataWithCoords:
"""
Exponentially weighted moving average.
Parameters
----------
keep_attrs : bool, default: None
If True, the attributes (``attrs``) will be copied from the original
object to the new one. If False, the new object will be returned
without attributes. If None uses the global default.
Examples
--------
>>> da = xr.DataArray([1, 1, 2, 2, 2], dims="x")
>>> da.rolling_exp(x=2, window_type="span").mean()
<xarray.DataArray (x: 5)> Size: 40B
array([1. , 1. , 1.69230769, 1.9 , 1.96694215])
Dimensions without coordinates: x
"""
import numbagg
if keep_attrs is None:
keep_attrs = _get_keep_attrs(default=True)
dim_order = self.obj.dims
return apply_ufunc(
numbagg.move_exp_nanmean,
self.obj,
input_core_dims=[[self.dim]],
kwargs=self.kwargs,
output_core_dims=[[self.dim]],
keep_attrs=keep_attrs,
on_missing_core_dim="copy",
dask="parallelized",
).transpose(*dim_order)
def sum(self, keep_attrs: bool | None = None) -> T_DataWithCoords:
"""
Exponentially weighted moving sum.
Parameters
----------
keep_attrs : bool, default: None
If True, the attributes (``attrs``) will be copied from the original
object to the new one. If False, the new object will be returned
without attributes. If None uses the global default.
Examples
--------
>>> da = xr.DataArray([1, 1, 2, 2, 2], dims="x")
>>> da.rolling_exp(x=2, window_type="span").sum()
<xarray.DataArray (x: 5)> Size: 40B
array([1. , 1.33333333, 2.44444444, 2.81481481, 2.9382716 ])
Dimensions without coordinates: x
"""
import numbagg
if keep_attrs is None:
keep_attrs = _get_keep_attrs(default=True)
dim_order = self.obj.dims
return apply_ufunc(
numbagg.move_exp_nansum,
self.obj,
input_core_dims=[[self.dim]],
kwargs=self.kwargs,
output_core_dims=[[self.dim]],
keep_attrs=keep_attrs,
on_missing_core_dim="copy",
dask="parallelized",
).transpose(*dim_order)
def std(self) -> T_DataWithCoords:
"""
Exponentially weighted moving standard deviation.
`keep_attrs` is always True for this method. Drop attrs separately to remove attrs.
Examples
--------
>>> da = xr.DataArray([1, 1, 2, 2, 2], dims="x")
>>> da.rolling_exp(x=2, window_type="span").std()
<xarray.DataArray (x: 5)> Size: 40B
array([ nan, 0. , 0.67936622, 0.42966892, 0.25389527])
Dimensions without coordinates: x
"""
import numbagg
dim_order = self.obj.dims
return apply_ufunc(
numbagg.move_exp_nanstd,
self.obj,
input_core_dims=[[self.dim]],
kwargs=self.kwargs,
output_core_dims=[[self.dim]],
keep_attrs=True,
on_missing_core_dim="copy",
dask="parallelized",
).transpose(*dim_order)
def var(self) -> T_DataWithCoords:
"""
Exponentially weighted moving variance.
`keep_attrs` is always True for this method. Drop attrs separately to remove attrs.
Examples
--------
>>> da = xr.DataArray([1, 1, 2, 2, 2], dims="x")
>>> da.rolling_exp(x=2, window_type="span").var()
<xarray.DataArray (x: 5)> Size: 40B
array([ nan, 0. , 0.46153846, 0.18461538, 0.06446281])
Dimensions without coordinates: x
"""
dim_order = self.obj.dims
import numbagg
return apply_ufunc(
numbagg.move_exp_nanvar,
self.obj,
input_core_dims=[[self.dim]],
kwargs=self.kwargs,
output_core_dims=[[self.dim]],
keep_attrs=True,
on_missing_core_dim="copy",
dask="parallelized",
).transpose(*dim_order)
def cov(self, other: T_DataWithCoords) -> T_DataWithCoords:
"""
Exponentially weighted moving covariance.
`keep_attrs` is always True for this method. Drop attrs separately to remove attrs.
Examples
--------
>>> da = xr.DataArray([1, 1, 2, 2, 2], dims="x")
>>> da.rolling_exp(x=2, window_type="span").cov(da**2)
<xarray.DataArray (x: 5)> Size: 40B
array([ nan, 0. , 1.38461538, 0.55384615, 0.19338843])
Dimensions without coordinates: x
"""
dim_order = self.obj.dims
import numbagg
return apply_ufunc(
numbagg.move_exp_nancov,
self.obj,
other,
input_core_dims=[[self.dim], [self.dim]],
kwargs=self.kwargs,
output_core_dims=[[self.dim]],
keep_attrs=True,
on_missing_core_dim="copy",
dask="parallelized",
).transpose(*dim_order)
def corr(self, other: T_DataWithCoords) -> T_DataWithCoords:
"""
Exponentially weighted moving correlation.
`keep_attrs` is always True for this method. Drop attrs separately to remove attrs.
Examples
--------
>>> da = xr.DataArray([1, 1, 2, 2, 2], dims="x")
>>> da.rolling_exp(x=2, window_type="span").corr(da.shift(x=1))
<xarray.DataArray (x: 5)> Size: 40B
array([ nan, nan, nan, 0.4330127 , 0.48038446])
Dimensions without coordinates: x
"""
dim_order = self.obj.dims
import numbagg
return apply_ufunc(
numbagg.move_exp_nancorr,
self.obj,
other,
input_core_dims=[[self.dim], [self.dim]],
kwargs=self.kwargs,
output_core_dims=[[self.dim]],
keep_attrs=True,
on_missing_core_dim="copy",
dask="parallelized",
).transpose(*dim_order)
|
