-
-
Notifications
You must be signed in to change notification settings - Fork 1k
/
rolling.py
1258 lines (1073 loc) · 43.4 KB
/
rolling.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
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
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
from __future__ import annotations
import functools
import itertools
import math
import warnings
from collections.abc import Hashable, Iterator, Mapping
from typing import TYPE_CHECKING, Any, Callable, Generic, TypeVar
import numpy as np
from packaging.version import Version
from xarray.core import dtypes, duck_array_ops, pycompat, utils
from xarray.core.arithmetic import CoarsenArithmetic
from xarray.core.options import OPTIONS, _get_keep_attrs
from xarray.core.pycompat import is_duck_dask_array
from xarray.core.types import CoarsenBoundaryOptions, SideOptions, T_Xarray
from xarray.core.utils import either_dict_or_kwargs, module_available
try:
import bottleneck
except ImportError:
# use numpy methods instead
bottleneck = None
if TYPE_CHECKING:
from xarray.core.dataarray import DataArray
from xarray.core.dataset import Dataset
RollingKey = Any
_T = TypeVar("_T")
_ROLLING_REDUCE_DOCSTRING_TEMPLATE = """\
Reduce this object's data windows by applying `{name}` along its dimension.
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.
**kwargs : dict
Additional keyword arguments passed on to `{name}`.
Returns
-------
reduced : same type as caller
New object with `{name}` applied along its rolling dimension.
"""
class Rolling(Generic[T_Xarray]):
"""A object that implements the moving window pattern.
See Also
--------
xarray.Dataset.groupby
xarray.DataArray.groupby
xarray.Dataset.rolling
xarray.DataArray.rolling
"""
__slots__ = ("obj", "window", "min_periods", "center", "dim")
_attributes = ("window", "min_periods", "center", "dim")
dim: list[Hashable]
window: list[int]
center: list[bool]
obj: T_Xarray
min_periods: int
def __init__(
self,
obj: T_Xarray,
windows: Mapping[Any, int],
min_periods: int | None = None,
center: bool | Mapping[Any, bool] = False,
) -> None:
"""
Moving window object.
Parameters
----------
obj : Dataset or DataArray
Object to window.
windows : mapping of hashable to int
A mapping from the name of the dimension to create the rolling
window along (e.g. `time`) to the size of the moving window.
min_periods : int or None, default: None
Minimum number of observations in window required to have a value
(otherwise result is NA). The default, None, is equivalent to
setting min_periods equal to the size of the window.
center : bool or dict-like Hashable to bool, default: False
Set the labels at the center of the window. If dict-like, set this
property per rolling dimension.
Returns
-------
rolling : type of input argument
"""
self.dim = []
self.window = []
for d, w in windows.items():
self.dim.append(d)
if w <= 0:
raise ValueError("window must be > 0")
self.window.append(w)
self.center = self._mapping_to_list(center, default=False)
self.obj = obj
missing_dims = tuple(dim for dim in self.dim if dim not in self.obj.dims)
if missing_dims:
# NOTE: we raise KeyError here but ValueError in Coarsen.
raise KeyError(
f"Window dimensions {missing_dims} not found in {self.obj.__class__.__name__} "
f"dimensions {tuple(self.obj.dims)}"
)
# attributes
if min_periods is not None and min_periods <= 0:
raise ValueError("min_periods must be greater than zero or None")
self.min_periods = (
math.prod(self.window) if min_periods is None else min_periods
)
def __repr__(self) -> str:
"""provide a nice str repr of our rolling object"""
attrs = [
"{k}->{v}{c}".format(k=k, v=w, c="(center)" if c else "")
for k, w, c in zip(self.dim, self.window, self.center)
]
return "{klass} [{attrs}]".format(
klass=self.__class__.__name__, attrs=",".join(attrs)
)
def __len__(self) -> int:
return math.prod(self.obj.sizes[d] for d in self.dim)
@property
def ndim(self) -> int:
return len(self.dim)
def _reduce_method( # type: ignore[misc]
name: str, fillna: Any, rolling_agg_func: Callable | None = None
) -> Callable[..., T_Xarray]:
"""Constructs reduction methods built on a numpy reduction function (e.g. sum),
a numbagg reduction function (e.g. move_sum), a bottleneck reduction function
(e.g. move_sum), or a Rolling reduction (_mean).
The logic here for which function to run is quite diffuse, across this method &
_array_reduce. Arguably we could refactor this. But one constraint is that we
need context of xarray options, of the functions each library offers, of
the array (e.g. dtype).
"""
if rolling_agg_func:
array_agg_func = None
else:
array_agg_func = getattr(duck_array_ops, name)
bottleneck_move_func = getattr(bottleneck, "move_" + name, None)
if module_available("numbagg"):
import numbagg
numbagg_move_func = getattr(numbagg, "move_" + name, None)
else:
numbagg_move_func = None
def method(self, keep_attrs=None, **kwargs):
keep_attrs = self._get_keep_attrs(keep_attrs)
return self._array_reduce(
array_agg_func=array_agg_func,
bottleneck_move_func=bottleneck_move_func,
numbagg_move_func=numbagg_move_func,
rolling_agg_func=rolling_agg_func,
keep_attrs=keep_attrs,
fillna=fillna,
**kwargs,
)
method.__name__ = name
method.__doc__ = _ROLLING_REDUCE_DOCSTRING_TEMPLATE.format(name=name)
return method
def _mean(self, keep_attrs, **kwargs):
result = self.sum(keep_attrs=False, **kwargs) / duck_array_ops.astype(
self.count(keep_attrs=False), dtype=self.obj.dtype, copy=False
)
if keep_attrs:
result.attrs = self.obj.attrs
return result
_mean.__doc__ = _ROLLING_REDUCE_DOCSTRING_TEMPLATE.format(name="mean")
argmax = _reduce_method("argmax", dtypes.NINF)
argmin = _reduce_method("argmin", dtypes.INF)
max = _reduce_method("max", dtypes.NINF)
min = _reduce_method("min", dtypes.INF)
prod = _reduce_method("prod", 1)
sum = _reduce_method("sum", 0)
mean = _reduce_method("mean", None, _mean)
std = _reduce_method("std", None)
var = _reduce_method("var", None)
median = _reduce_method("median", None)
def _counts(self, keep_attrs: bool | None) -> T_Xarray:
raise NotImplementedError()
def count(self, keep_attrs: bool | None = None) -> T_Xarray:
keep_attrs = self._get_keep_attrs(keep_attrs)
rolling_count = self._counts(keep_attrs=keep_attrs)
enough_periods = rolling_count >= self.min_periods
return rolling_count.where(enough_periods)
count.__doc__ = _ROLLING_REDUCE_DOCSTRING_TEMPLATE.format(name="count")
def _mapping_to_list(
self,
arg: _T | Mapping[Any, _T],
default: _T | None = None,
allow_default: bool = True,
allow_allsame: bool = True,
) -> list[_T]:
if utils.is_dict_like(arg):
if allow_default:
return [arg.get(d, default) for d in self.dim]
for d in self.dim:
if d not in arg:
raise KeyError(f"Argument has no dimension key {d}.")
return [arg[d] for d in self.dim]
if allow_allsame: # for single argument
return [arg] * self.ndim # type: ignore[list-item] # no check for negatives
if self.ndim == 1:
return [arg] # type: ignore[list-item] # no check for negatives
raise ValueError(f"Mapping argument is necessary for {self.ndim}d-rolling.")
def _get_keep_attrs(self, keep_attrs):
if keep_attrs is None:
keep_attrs = _get_keep_attrs(default=True)
return keep_attrs
class DataArrayRolling(Rolling["DataArray"]):
__slots__ = ("window_labels",)
def __init__(
self,
obj: DataArray,
windows: Mapping[Any, int],
min_periods: int | None = None,
center: bool | Mapping[Any, bool] = False,
) -> None:
"""
Moving window object for DataArray.
You should use DataArray.rolling() method to construct this object
instead of the class constructor.
Parameters
----------
obj : DataArray
Object to window.
windows : mapping of hashable to int
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.
min_periods : int, default: None
Minimum number of observations in window required to have a value
(otherwise result is NA). The default, None, is equivalent to
setting min_periods equal to the size of the window.
center : bool, default: False
Set the labels at the center of the window.
Returns
-------
rolling : type of input argument
See Also
--------
xarray.DataArray.rolling
xarray.DataArray.groupby
xarray.Dataset.rolling
xarray.Dataset.groupby
"""
super().__init__(obj, windows, min_periods=min_periods, center=center)
# TODO legacy attribute
self.window_labels = self.obj[self.dim[0]]
def __iter__(self) -> Iterator[tuple[DataArray, DataArray]]:
if self.ndim > 1:
raise ValueError("__iter__ is only supported for 1d-rolling")
dim0 = self.dim[0]
window0 = int(self.window[0])
offset = (window0 + 1) // 2 if self.center[0] else 1
stops = np.arange(offset, self.obj.sizes[dim0] + offset)
starts = stops - window0
starts[: window0 - offset] = 0
for label, start, stop in zip(self.window_labels, starts, stops):
window = self.obj.isel({dim0: slice(start, stop)})
counts = window.count(dim=[dim0])
window = window.where(counts >= self.min_periods)
yield (label, window)
def construct(
self,
window_dim: Hashable | Mapping[Any, Hashable] | None = None,
stride: int | Mapping[Any, int] = 1,
fill_value: Any = dtypes.NA,
keep_attrs: bool | None = None,
**window_dim_kwargs: Hashable,
) -> DataArray:
"""
Convert this rolling object to xr.DataArray,
where the window dimension is stacked as a new dimension
Parameters
----------
window_dim : Hashable or dict-like to Hashable, optional
A mapping from dimension name to the new window dimension names.
stride : int or mapping of int, default: 1
Size of stride for the rolling window.
fill_value : default: dtypes.NA
Filling value to match the dimension size.
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.
**window_dim_kwargs : Hashable, optional
The keyword arguments form of ``window_dim`` {dim: new_name, ...}.
Returns
-------
DataArray that is a view of the original array. The returned array is
not writeable.
Examples
--------
>>> da = xr.DataArray(np.arange(8).reshape(2, 4), dims=("a", "b"))
>>> rolling = da.rolling(b=3)
>>> rolling.construct("window_dim")
<xarray.DataArray (a: 2, b: 4, window_dim: 3)>
array([[[nan, nan, 0.],
[nan, 0., 1.],
[ 0., 1., 2.],
[ 1., 2., 3.]],
<BLANKLINE>
[[nan, nan, 4.],
[nan, 4., 5.],
[ 4., 5., 6.],
[ 5., 6., 7.]]])
Dimensions without coordinates: a, b, window_dim
>>> rolling = da.rolling(b=3, center=True)
>>> rolling.construct("window_dim")
<xarray.DataArray (a: 2, b: 4, window_dim: 3)>
array([[[nan, 0., 1.],
[ 0., 1., 2.],
[ 1., 2., 3.],
[ 2., 3., nan]],
<BLANKLINE>
[[nan, 4., 5.],
[ 4., 5., 6.],
[ 5., 6., 7.],
[ 6., 7., nan]]])
Dimensions without coordinates: a, b, window_dim
"""
return self._construct(
self.obj,
window_dim=window_dim,
stride=stride,
fill_value=fill_value,
keep_attrs=keep_attrs,
**window_dim_kwargs,
)
def _construct(
self,
obj: DataArray,
window_dim: Hashable | Mapping[Any, Hashable] | None = None,
stride: int | Mapping[Any, int] = 1,
fill_value: Any = dtypes.NA,
keep_attrs: bool | None = None,
**window_dim_kwargs: Hashable,
) -> DataArray:
from xarray.core.dataarray import DataArray
keep_attrs = self._get_keep_attrs(keep_attrs)
if window_dim is None:
if len(window_dim_kwargs) == 0:
raise ValueError(
"Either window_dim or window_dim_kwargs need to be specified."
)
window_dim = {d: window_dim_kwargs[str(d)] for d in self.dim}
window_dims = self._mapping_to_list(
window_dim, allow_default=False, allow_allsame=False
)
strides = self._mapping_to_list(stride, default=1)
window = obj.variable.rolling_window(
self.dim, self.window, window_dims, self.center, fill_value=fill_value
)
attrs = obj.attrs if keep_attrs else {}
result = DataArray(
window,
dims=obj.dims + tuple(window_dims),
coords=obj.coords,
attrs=attrs,
name=obj.name,
)
return result.isel({d: slice(None, None, s) for d, s in zip(self.dim, strides)})
def reduce(
self, func: Callable, keep_attrs: bool | None = None, **kwargs: Any
) -> DataArray:
"""Reduce the items in this group by applying `func` along some
dimension(s).
Parameters
----------
func : callable
Function which can be called in the form
`func(x, **kwargs)` to return the result of collapsing an
np.ndarray over an the rolling dimension.
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.
**kwargs : dict
Additional keyword arguments passed on to `func`.
Returns
-------
reduced : DataArray
Array with summarized data.
Examples
--------
>>> da = xr.DataArray(np.arange(8).reshape(2, 4), dims=("a", "b"))
>>> rolling = da.rolling(b=3)
>>> rolling.construct("window_dim")
<xarray.DataArray (a: 2, b: 4, window_dim: 3)>
array([[[nan, nan, 0.],
[nan, 0., 1.],
[ 0., 1., 2.],
[ 1., 2., 3.]],
<BLANKLINE>
[[nan, nan, 4.],
[nan, 4., 5.],
[ 4., 5., 6.],
[ 5., 6., 7.]]])
Dimensions without coordinates: a, b, window_dim
>>> rolling.reduce(np.sum)
<xarray.DataArray (a: 2, b: 4)>
array([[nan, nan, 3., 6.],
[nan, nan, 15., 18.]])
Dimensions without coordinates: a, b
>>> rolling = da.rolling(b=3, min_periods=1)
>>> rolling.reduce(np.nansum)
<xarray.DataArray (a: 2, b: 4)>
array([[ 0., 1., 3., 6.],
[ 4., 9., 15., 18.]])
Dimensions without coordinates: a, b
"""
keep_attrs = self._get_keep_attrs(keep_attrs)
rolling_dim = {
d: utils.get_temp_dimname(self.obj.dims, f"_rolling_dim_{d}")
for d in self.dim
}
# save memory with reductions GH4325
fillna = kwargs.pop("fillna", dtypes.NA)
if fillna is not dtypes.NA:
obj = self.obj.fillna(fillna)
else:
obj = self.obj
windows = self._construct(
obj, rolling_dim, keep_attrs=keep_attrs, fill_value=fillna
)
dim = list(rolling_dim.values())
result = windows.reduce(func, dim=dim, keep_attrs=keep_attrs, **kwargs)
# Find valid windows based on count.
counts = self._counts(keep_attrs=False)
return result.where(counts >= self.min_periods)
def _counts(self, keep_attrs: bool | None) -> DataArray:
"""Number of non-nan entries in each rolling window."""
rolling_dim = {
d: utils.get_temp_dimname(self.obj.dims, f"_rolling_dim_{d}")
for d in self.dim
}
# We use False as the fill_value instead of np.nan, since boolean
# array is faster to be reduced than object array.
# The use of skipna==False is also faster since it does not need to
# copy the strided array.
dim = list(rolling_dim.values())
counts = (
self.obj.notnull(keep_attrs=keep_attrs)
.rolling(
{d: w for d, w in zip(self.dim, self.window)},
center={d: self.center[i] for i, d in enumerate(self.dim)},
)
.construct(rolling_dim, fill_value=False, keep_attrs=keep_attrs)
.sum(dim=dim, skipna=False, keep_attrs=keep_attrs)
)
return counts
def _numbagg_reduce(self, func, keep_attrs, **kwargs):
# Some of this is copied from `_bottleneck_reduce`, we could reduce this as part
# of a wider refactor.
axis = self.obj.get_axis_num(self.dim[0])
padded = self.obj.variable
if self.center[0]:
if is_duck_dask_array(padded.data):
# workaround to make the padded chunk size larger than
# self.window - 1
shift = -(self.window[0] + 1) // 2
offset = (self.window[0] - 1) // 2
valid = (slice(None),) * axis + (
slice(offset, offset + self.obj.shape[axis]),
)
else:
shift = (-self.window[0] // 2) + 1
valid = (slice(None),) * axis + (slice(-shift, None),)
padded = padded.pad({self.dim[0]: (0, -shift)}, mode="constant")
if is_duck_dask_array(padded.data) and False:
raise AssertionError("should not be reachable")
else:
values = func(
padded.data,
window=self.window[0],
min_count=self.min_periods,
axis=axis,
)
if self.center[0]:
values = values[valid]
attrs = self.obj.attrs if keep_attrs else {}
return self.obj.__class__(
values, self.obj.coords, attrs=attrs, name=self.obj.name
)
def _bottleneck_reduce(self, func, keep_attrs, **kwargs):
# bottleneck doesn't allow min_count to be 0, although it should
# work the same as if min_count = 1
# Note bottleneck only works with 1d-rolling.
if self.min_periods is not None and self.min_periods == 0:
min_count = 1
else:
min_count = self.min_periods
axis = self.obj.get_axis_num(self.dim[0])
padded = self.obj.variable
if self.center[0]:
if is_duck_dask_array(padded.data):
# workaround to make the padded chunk size larger than
# self.window - 1
shift = -(self.window[0] + 1) // 2
offset = (self.window[0] - 1) // 2
valid = (slice(None),) * axis + (
slice(offset, offset + self.obj.shape[axis]),
)
else:
shift = (-self.window[0] // 2) + 1
valid = (slice(None),) * axis + (slice(-shift, None),)
padded = padded.pad({self.dim[0]: (0, -shift)}, mode="constant")
if is_duck_dask_array(padded.data):
raise AssertionError("should not be reachable")
else:
values = func(
padded.data, window=self.window[0], min_count=min_count, axis=axis
)
if self.center[0]:
values = values[valid]
attrs = self.obj.attrs if keep_attrs else {}
return self.obj.__class__(
values, self.obj.coords, attrs=attrs, name=self.obj.name
)
def _array_reduce(
self,
array_agg_func,
bottleneck_move_func,
numbagg_move_func,
rolling_agg_func,
keep_attrs,
fillna,
**kwargs,
):
if "dim" in kwargs:
warnings.warn(
f"Reductions are applied along the rolling dimension(s) "
f"'{self.dim}'. Passing the 'dim' kwarg to reduction "
f"operations has no effect.",
DeprecationWarning,
stacklevel=3,
)
del kwargs["dim"]
if (
OPTIONS["use_numbagg"]
and module_available("numbagg")
and pycompat.mod_version("numbagg") >= Version("0.6.3")
and numbagg_move_func is not None
# TODO: we could at least allow this for the equivalent of `apply_ufunc`'s
# "parallelized". `rolling_exp` does this, as an example (but rolling_exp is
# much simpler)
and not is_duck_dask_array(self.obj.data)
# Numbagg doesn't handle object arrays and generally has dtype consistency,
# so doesn't deal well with bool arrays which are expected to change type.
and self.obj.data.dtype.kind not in "ObMm"
# TODO: we could also allow this, probably as part of a refactoring of this
# module, so we can use the machinery in `self.reduce`.
and self.ndim == 1
):
import numbagg
# Numbagg has a default ddof of 1. I (@max-sixty) think we should make
# this the default in xarray too, but until we do, don't use numbagg for
# std and var unless ddof is set to 1.
if (
numbagg_move_func not in [numbagg.move_std, numbagg.move_var]
or kwargs.get("ddof") == 1
):
return self._numbagg_reduce(
numbagg_move_func, keep_attrs=keep_attrs, **kwargs
)
if (
OPTIONS["use_bottleneck"]
and bottleneck_move_func is not None
and not is_duck_dask_array(self.obj.data)
and self.ndim == 1
):
# TODO: re-enable bottleneck with dask after the issues
# underlying https://github.com/pydata/xarray/issues/2940 are
# fixed.
return self._bottleneck_reduce(
bottleneck_move_func, keep_attrs=keep_attrs, **kwargs
)
if rolling_agg_func:
return rolling_agg_func(self, keep_attrs=self._get_keep_attrs(keep_attrs))
if fillna is not None:
if fillna is dtypes.INF:
fillna = dtypes.get_pos_infinity(self.obj.dtype, max_for_int=True)
elif fillna is dtypes.NINF:
fillna = dtypes.get_neg_infinity(self.obj.dtype, min_for_int=True)
kwargs.setdefault("skipna", False)
kwargs.setdefault("fillna", fillna)
return self.reduce(array_agg_func, keep_attrs=keep_attrs, **kwargs)
class DatasetRolling(Rolling["Dataset"]):
__slots__ = ("rollings",)
def __init__(
self,
obj: Dataset,
windows: Mapping[Any, int],
min_periods: int | None = None,
center: bool | Mapping[Any, bool] = False,
) -> None:
"""
Moving window object for Dataset.
You should use Dataset.rolling() method to construct this object
instead of the class constructor.
Parameters
----------
obj : Dataset
Object to window.
windows : mapping of hashable to int
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.
min_periods : int, default: None
Minimum number of observations in window required to have a value
(otherwise result is NA). The default, None, is equivalent to
setting min_periods equal to the size of the window.
center : bool or mapping of hashable to bool, default: False
Set the labels at the center of the window.
Returns
-------
rolling : type of input argument
See Also
--------
xarray.Dataset.rolling
xarray.DataArray.rolling
xarray.Dataset.groupby
xarray.DataArray.groupby
"""
super().__init__(obj, windows, min_periods, center)
# Keep each Rolling object as a dictionary
self.rollings = {}
for key, da in self.obj.data_vars.items():
# keeps rollings only for the dataset depending on self.dim
dims, center = [], {}
for i, d in enumerate(self.dim):
if d in da.dims:
dims.append(d)
center[d] = self.center[i]
if dims:
w = {d: windows[d] for d in dims}
self.rollings[key] = DataArrayRolling(da, w, min_periods, center)
def _dataset_implementation(self, func, keep_attrs, **kwargs):
from xarray.core.dataset import Dataset
keep_attrs = self._get_keep_attrs(keep_attrs)
reduced = {}
for key, da in self.obj.data_vars.items():
if any(d in da.dims for d in self.dim):
reduced[key] = func(self.rollings[key], keep_attrs=keep_attrs, **kwargs)
else:
reduced[key] = self.obj[key].copy()
# we need to delete the attrs of the copied DataArray
if not keep_attrs:
reduced[key].attrs = {}
attrs = self.obj.attrs if keep_attrs else {}
return Dataset(reduced, coords=self.obj.coords, attrs=attrs)
def reduce(
self, func: Callable, keep_attrs: bool | None = None, **kwargs: Any
) -> DataArray:
"""Reduce the items in this group by applying `func` along some
dimension(s).
Parameters
----------
func : callable
Function which can be called in the form
`func(x, **kwargs)` to return the result of collapsing an
np.ndarray over an the rolling dimension.
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.
**kwargs : dict
Additional keyword arguments passed on to `func`.
Returns
-------
reduced : DataArray
Array with summarized data.
"""
return self._dataset_implementation(
functools.partial(DataArrayRolling.reduce, func=func),
keep_attrs=keep_attrs,
**kwargs,
)
def _counts(self, keep_attrs: bool | None) -> Dataset:
return self._dataset_implementation(
DataArrayRolling._counts, keep_attrs=keep_attrs
)
def _array_reduce(
self,
array_agg_func,
bottleneck_move_func,
rolling_agg_func,
keep_attrs,
**kwargs,
):
return self._dataset_implementation(
functools.partial(
DataArrayRolling._array_reduce,
array_agg_func=array_agg_func,
bottleneck_move_func=bottleneck_move_func,
rolling_agg_func=rolling_agg_func,
),
keep_attrs=keep_attrs,
**kwargs,
)
def construct(
self,
window_dim: Hashable | Mapping[Any, Hashable] | None = None,
stride: int | Mapping[Any, int] = 1,
fill_value: Any = dtypes.NA,
keep_attrs: bool | None = None,
**window_dim_kwargs: Hashable,
) -> Dataset:
"""
Convert this rolling object to xr.Dataset,
where the window dimension is stacked as a new dimension
Parameters
----------
window_dim : str or mapping, optional
A mapping from dimension name to the new window dimension names.
Just a string can be used for 1d-rolling.
stride : int, optional
size of stride for the rolling window.
fill_value : Any, default: dtypes.NA
Filling value to match the dimension size.
**window_dim_kwargs : {dim: new_name, ...}, optional
The keyword arguments form of ``window_dim``.
Returns
-------
Dataset with variables converted from rolling object.
"""
from xarray.core.dataset import Dataset
keep_attrs = self._get_keep_attrs(keep_attrs)
if window_dim is None:
if len(window_dim_kwargs) == 0:
raise ValueError(
"Either window_dim or window_dim_kwargs need to be specified."
)
window_dim = {d: window_dim_kwargs[str(d)] for d in self.dim}
window_dims = self._mapping_to_list(
window_dim, allow_default=False, allow_allsame=False
)
strides = self._mapping_to_list(stride, default=1)
dataset = {}
for key, da in self.obj.data_vars.items():
# keeps rollings only for the dataset depending on self.dim
dims = [d for d in self.dim if d in da.dims]
if dims:
wi = {d: window_dims[i] for i, d in enumerate(self.dim) if d in da.dims}
st = {d: strides[i] for i, d in enumerate(self.dim) if d in da.dims}
dataset[key] = self.rollings[key].construct(
window_dim=wi,
fill_value=fill_value,
stride=st,
keep_attrs=keep_attrs,
)
else:
dataset[key] = da.copy()
# as the DataArrays can be copied we need to delete the attrs
if not keep_attrs:
dataset[key].attrs = {}
# Need to stride coords as well. TODO: is there a better way?
coords = self.obj.isel(
{d: slice(None, None, s) for d, s in zip(self.dim, strides)}
).coords
attrs = self.obj.attrs if keep_attrs else {}
return Dataset(dataset, coords=coords, attrs=attrs)
class Coarsen(CoarsenArithmetic, Generic[T_Xarray]):
"""A object that implements the coarsen.
See Also
--------
Dataset.coarsen
DataArray.coarsen
"""
__slots__ = (
"obj",
"boundary",
"coord_func",
"windows",
"side",
"trim_excess",
)
_attributes = ("windows", "side", "trim_excess")
obj: T_Xarray
windows: Mapping[Hashable, int]
side: SideOptions | Mapping[Hashable, SideOptions]
boundary: CoarsenBoundaryOptions
coord_func: Mapping[Hashable, str | Callable]
def __init__(
self,
obj: T_Xarray,
windows: Mapping[Any, int],
boundary: CoarsenBoundaryOptions,
side: SideOptions | Mapping[Any, SideOptions],
coord_func: str | Callable | Mapping[Any, str | Callable],
) -> None:
"""
Moving window object.
Parameters
----------
obj : Dataset or DataArray
Object to window.
windows : mapping of hashable to int
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.
boundary : {"exact", "trim", "pad"}
If 'exact', a ValueError will be raised if dimension size is not a
multiple of window size. If 'trim', the excess indexes are trimmed.
If 'pad', NA will be padded.
side : 'left' or 'right' or mapping from dimension to 'left' or 'right'
coord_func : function (name) or mapping from coordinate name to function (name).
Returns
-------
coarsen
"""
self.obj = obj
self.windows = windows
self.side = side
self.boundary = boundary
missing_dims = tuple(dim for dim in windows.keys() if dim not in self.obj.dims)
if missing_dims:
raise ValueError(
f"Window dimensions {missing_dims} not found in {self.obj.__class__.__name__} "
f"dimensions {tuple(self.obj.dims)}"
)
if utils.is_dict_like(coord_func):
coord_func_map = coord_func
else:
coord_func_map = {d: coord_func for d in self.obj.dims}
for c in self.obj.coords:
if c not in coord_func_map:
coord_func_map[c] = duck_array_ops.mean # type: ignore[index]
self.coord_func = coord_func_map
def _get_keep_attrs(self, keep_attrs):
if keep_attrs is None:
keep_attrs = _get_keep_attrs(default=True)
return keep_attrs
def __repr__(self) -> str:
"""provide a nice str repr of our coarsen object"""
attrs = [
f"{k}->{getattr(self, k)}"
for k in self._attributes
if getattr(self, k, None) is not None
]
return "{klass} [{attrs}]".format(
klass=self.__class__.__name__, attrs=",".join(attrs)
)
def construct(
self,
window_dim=None,
keep_attrs=None,
**window_dim_kwargs,
) -> T_Xarray:
"""
Convert this Coarsen object to a DataArray or Dataset,
where the coarsening dimension is split or reshaped to two
new dimensions.
Parameters
----------
window_dim: mapping
A mapping from existing dimension name to new dimension names.
The size of the second dimension will be the length of the
coarsening window.
keep_attrs: bool, optional
Preserve attributes if True