-
-
Notifications
You must be signed in to change notification settings - Fork 1.7k
/
groups.py
427 lines (345 loc) · 13.7 KB
/
groups.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
# Licensed under a 3-clause BSD style license - see LICENSE.rst
import platform
import warnings
import numpy as np
from astropy.utils.exceptions import AstropyUserWarning
from .index import get_index_by_names
__all__ = ["TableGroups", "ColumnGroups"]
def table_group_by(table, keys):
# index copies are unnecessary and slow down _table_group_by
with table.index_mode("discard_on_copy"):
return _table_group_by(table, keys)
def _table_group_by(table, keys):
"""
Get groups for ``table`` on specified ``keys``.
Parameters
----------
table : `Table`
Table to group
keys : str, list of str, `Table`, or Numpy array
Grouping key specifier
Returns
-------
grouped_table : Table object with groups attr set accordingly
"""
from .serialize import represent_mixins_as_columns
from .table import Table
# Pre-convert string to tuple of strings, or Table to the underlying structured array
if isinstance(keys, str):
keys = (keys,)
if isinstance(keys, (list, tuple)):
for name in keys:
if name not in table.colnames:
raise ValueError(f"Table does not have key column {name!r}")
if table.masked and np.any(table[name].mask):
raise ValueError(
f"Missing values in key column {name!r} are not allowed"
)
# Make a column slice of the table without copying
table_keys = table.__class__([table[key] for key in keys], copy=False)
# If available get a pre-existing index for these columns
table_index = get_index_by_names(table, keys)
grouped_by_table_cols = True
elif isinstance(keys, (np.ndarray, Table)):
table_keys = keys
if len(table_keys) != len(table):
raise ValueError(
"Input keys array length {} does not match table length {}".format(
len(table_keys), len(table)
)
)
table_index = None
grouped_by_table_cols = False
else:
raise TypeError(
"Keys input must be string, list, tuple, Table or numpy array, but got {}".format(
type(keys)
)
)
# If there is not already an available index and table_keys is a Table then ensure
# that all cols (including mixins) are in a form that can sorted with the code below.
if not table_index and isinstance(table_keys, Table):
table_keys = represent_mixins_as_columns(table_keys)
# Get the argsort index `idx_sort`, accounting for particulars
try:
# take advantage of index internal sort if possible
if table_index is not None:
idx_sort = table_index.sorted_data()
else:
idx_sort = table_keys.argsort(kind="mergesort")
stable_sort = True
except TypeError:
# Some versions (likely 1.6 and earlier) of numpy don't support
# 'mergesort' for all data types. MacOSX (Darwin) doesn't have a stable
# sort by default, nor does Windows, while Linux does (or appears to).
idx_sort = table_keys.argsort()
stable_sort = platform.system() not in ("Darwin", "Windows")
# Finally do the actual sort of table_keys values
table_keys = table_keys[idx_sort]
# Get all keys
diffs = np.concatenate(([True], table_keys[1:] != table_keys[:-1], [True]))
indices = np.flatnonzero(diffs)
# If the sort is not stable (preserves original table order) then sort idx_sort in
# place within each group.
if not stable_sort:
for i0, i1 in zip(indices[:-1], indices[1:]):
idx_sort[i0:i1].sort()
# Make a new table and set the _groups to the appropriate TableGroups object.
# Take the subset of the original keys at the indices values (group boundaries).
out = table.__class__(table[idx_sort])
out_keys = table_keys[indices[:-1]]
if isinstance(out_keys, Table):
out_keys.meta["grouped_by_table_cols"] = grouped_by_table_cols
out._groups = TableGroups(out, indices=indices, keys=out_keys)
return out
def column_group_by(column, keys):
"""
Get groups for ``column`` on specified ``keys``.
Parameters
----------
column : Column object
Column to group
keys : Table or Numpy array of same length as col
Grouping key specifier
Returns
-------
grouped_column : Column object with groups attr set accordingly
"""
from .serialize import represent_mixins_as_columns
from .table import Table
if isinstance(keys, Table):
keys = represent_mixins_as_columns(keys)
keys = keys.as_array()
if not isinstance(keys, np.ndarray):
raise TypeError(f"Keys input must be numpy array, but got {type(keys)}")
if len(keys) != len(column):
raise ValueError(
"Input keys array length {} does not match column length {}".format(
len(keys), len(column)
)
)
idx_sort = keys.argsort()
keys = keys[idx_sort]
# Get all keys
diffs = np.concatenate(([True], keys[1:] != keys[:-1], [True]))
indices = np.flatnonzero(diffs)
# Make a new column and set the _groups to the appropriate ColumnGroups object.
# Take the subset of the original keys at the indices values (group boundaries).
out = column.__class__(column[idx_sort])
out._groups = ColumnGroups(out, indices=indices, keys=keys[indices[:-1]])
return out
class BaseGroups:
"""
A class to represent groups within a table of heterogeneous data.
- ``keys``: key values corresponding to each group
- ``indices``: index values in parent table or column corresponding to group boundaries
- ``aggregate()``: method to create new table by aggregating within groups
"""
@property
def parent(self):
return (
self.parent_column if isinstance(self, ColumnGroups) else self.parent_table
)
def __iter__(self):
self._iter_index = 0
return self
def next(self):
ii = self._iter_index
if ii < len(self.indices) - 1:
i0, i1 = self.indices[ii], self.indices[ii + 1]
self._iter_index += 1
return self.parent[i0:i1]
else:
raise StopIteration
__next__ = next
def __getitem__(self, item):
parent = self.parent
if isinstance(item, (int, np.integer)):
i0, i1 = self.indices[item], self.indices[item + 1]
out = parent[i0:i1]
out.groups._keys = parent.groups.keys[item]
else:
indices0, indices1 = self.indices[:-1], self.indices[1:]
try:
i0s, i1s = indices0[item], indices1[item]
except Exception as err:
raise TypeError(
"Index item for groups attribute must be a slice, "
"numpy mask or int array"
) from err
mask = np.zeros(len(parent), dtype=bool)
# Is there a way to vectorize this in numpy?
for i0, i1 in zip(i0s, i1s):
mask[i0:i1] = True
out = parent[mask]
out.groups._keys = parent.groups.keys[item]
out.groups._indices = np.concatenate([[0], np.cumsum(i1s - i0s)])
return out
def __repr__(self):
return f"<{self.__class__.__name__} indices={self.indices}>"
def __len__(self):
return len(self.indices) - 1
class ColumnGroups(BaseGroups):
def __init__(self, parent_column, indices=None, keys=None):
self.parent_column = parent_column # parent Column
self.parent_table = parent_column.info.parent_table
self._indices = indices
self._keys = keys
@property
def indices(self):
# If the parent column is in a table then use group indices from table
if self.parent_table:
return self.parent_table.groups.indices
else:
if self._indices is None:
return np.array([0, len(self.parent_column)])
else:
return self._indices
@property
def keys(self):
# If the parent column is in a table then use group indices from table
if self.parent_table:
return self.parent_table.groups.keys
else:
return self._keys
def aggregate(self, func):
from .column import MaskedColumn
i0s, i1s = self.indices[:-1], self.indices[1:]
par_col = self.parent_column
masked = isinstance(par_col, MaskedColumn)
reduceat = hasattr(func, "reduceat")
sum_case = func is np.sum
mean_case = func is np.mean
try:
if not masked and (reduceat or sum_case or mean_case):
if mean_case:
vals = np.add.reduceat(par_col, i0s) / np.diff(self.indices)
else:
if sum_case:
func = np.add
vals = func.reduceat(par_col, i0s)
else:
vals = np.array([func(par_col[i0:i1]) for i0, i1 in zip(i0s, i1s)])
out = par_col.__class__(vals)
except Exception as err:
raise TypeError(
"Cannot aggregate column '{}' with type '{}': {}".format(
par_col.info.name, par_col.info.dtype, err
)
) from err
out_info = out.info
for attr in ("name", "unit", "format", "description", "meta"):
try:
setattr(out_info, attr, getattr(par_col.info, attr))
except AttributeError:
pass
return out
def filter(self, func):
"""
Filter groups in the Column based on evaluating function ``func`` on each
group sub-table.
The function which is passed to this method must accept one argument:
- ``column`` : `Column` object
It must then return either `True` or `False`. As an example, the following
will select all column groups with only positive values::
def all_positive(column):
if np.any(column < 0):
return False
return True
Parameters
----------
func : function
Filter function
Returns
-------
out : Column
New column with the aggregated rows.
"""
mask = np.empty(len(self), dtype=bool)
for i, group_column in enumerate(self):
mask[i] = func(group_column)
return self[mask]
class TableGroups(BaseGroups):
def __init__(self, parent_table, indices=None, keys=None):
self.parent_table = parent_table # parent Table
self._indices = indices
self._keys = keys
@property
def key_colnames(self):
"""
Return the names of columns in the parent table that were used for grouping.
"""
# If the table was grouped by key columns *in* the table then treat those columns
# differently in aggregation. In this case keys will be a Table with
# keys.meta['grouped_by_table_cols'] == True. Keys might not be a Table so we
# need to handle this.
grouped_by_table_cols = getattr(self.keys, "meta", {}).get(
"grouped_by_table_cols", False
)
return self.keys.colnames if grouped_by_table_cols else ()
@property
def indices(self):
if self._indices is None:
return np.array([0, len(self.parent_table)])
else:
return self._indices
def aggregate(self, func):
"""
Aggregate each group in the Table into a single row by applying the reduction
function ``func`` to group values in each column.
Parameters
----------
func : function
Function that reduces an array of values to a single value
Returns
-------
out : Table
New table with the aggregated rows.
"""
i0s = self.indices[:-1]
out_cols = []
parent_table = self.parent_table
for col in parent_table.columns.values():
# For key columns just pick off first in each group since they are identical
if col.info.name in self.key_colnames:
new_col = col.take(i0s)
else:
try:
new_col = col.info.groups.aggregate(func)
except TypeError as err:
warnings.warn(str(err), AstropyUserWarning)
continue
out_cols.append(new_col)
return parent_table.__class__(out_cols, meta=parent_table.meta)
def filter(self, func):
"""
Filter groups in the Table based on evaluating function ``func`` on each
group sub-table.
The function which is passed to this method must accept two arguments:
- ``table`` : `Table` object
- ``key_colnames`` : tuple of column names in ``table`` used as keys for grouping
It must then return either `True` or `False`. As an example, the following
will select all table groups with only positive values in the non-key columns::
def all_positive(table, key_colnames):
colnames = [name for name in table.colnames if name not in key_colnames]
for colname in colnames:
if np.any(table[colname] < 0):
return False
return True
Parameters
----------
func : function
Filter function
Returns
-------
out : Table
New table with the aggregated rows.
"""
mask = np.empty(len(self), dtype=bool)
key_colnames = self.key_colnames
for i, group_table in enumerate(self):
mask[i] = func(group_table, key_colnames)
return self[mask]
@property
def keys(self):
return self._keys