test_tsio.py 42.4 KB
Newer Older
1
# coding: utf-8
2
from datetime import datetime, timedelta
3
from time import time
4
from dateutil import parser
5
import calendar
6

Aurélien Campéas's avatar
Aurélien Campéas committed
7
from pathlib2 import Path
8
9
import pandas as pd
import numpy as np
10
import pytest
11
from mock import patch
12

13
from tshistory.snapshot import Snapshot
14
from tshistory.testutil import assert_group_equals, genserie, assert_df
15

16
DATADIR = Path(__file__).parent / 'data'
17

Aurélien Campéas's avatar
Aurélien Campéas committed
18

19
def test_changeset(engine, tsh):
20
    index = pd.date_range(start=datetime(2017, 1, 1), freq='D', periods=3)
21
    data = [1., 2., 3.]
22

23
24
    with patch('tshistory.tsio.datetime') as mock_date:
        mock_date.now.return_value = datetime(2020, 1, 1)
25
        with engine.connect() as cn:
26
            with tsh.newchangeset(cn, 'babar'):
27
                tsh.insert(cn, pd.Series(data, index=index), 'ts_values', author='WONTBEUSED')
28
                tsh.insert(cn, pd.Series(['a', 'b', 'c'], index=index), 'ts_othervalues')
29

30
31
32
        # bogus author won't show up
        assert tsh.log(engine)[0]['author'] == 'babar'

33
34
        g = tsh.get_group(engine, 'ts_values')
        g2 = tsh.get_group(engine, 'ts_othervalues')
35
        assert_group_equals(g, g2)
36

37
        with pytest.raises(AssertionError):
38
            tsh.insert(engine, pd.Series([2, 3, 4], index=index), 'ts_values')
39

40
        with engine.connect() as cn:
41
            data.append(data.pop(0))
42
43
            with tsh.newchangeset(cn, 'celeste'):
                tsh.insert(cn, pd.Series(data, index=index), 'ts_values')
44
                # below should be a noop
45
                tsh.insert(cn, pd.Series(['a', 'b', 'c'], index=index), 'ts_othervalues')
46

47
    g = tsh.get_group(engine, 'ts_values')
48
49
    assert ['ts_values'] == list(g.keys())

50
    assert_df("""
51
52
53
2017-01-01    2.0
2017-01-02    3.0
2017-01-03    1.0
54
""", tsh.get(engine, 'ts_values'))
55

56
    assert_df("""
57
58
59
2017-01-01    a
2017-01-02    b
2017-01-03    c
60
""", tsh.get(engine, 'ts_othervalues'))
61

62
    log = tsh.log(engine, names=['ts_values', 'ts_othervalues'])
63
64
65
66
    assert [
        {'author': 'babar',
         'rev': 1,
         'date': datetime(2020, 1, 1, 0, 0),
67
         'meta': {},
68
69
70
         'names': ['ts_values', 'ts_othervalues']},
        {'author': 'celeste',
         'rev': 2,
71
         'meta': {},
72
73
74
75
         'date': datetime(2020, 1, 1, 0, 0),
         'names': ['ts_values']}
    ] == log

76
    log = tsh.log(engine, names=['ts_othervalues'])
77
78
    assert len(log) == 1
    assert log[0]['rev'] == 1
79
    assert log[0]['names'] == ['ts_values', 'ts_othervalues']
80

81
    log = tsh.log(engine, fromrev=2)
82
83
    assert len(log) == 1

84
    log = tsh.log(engine, torev=1)
85
86
    assert len(log) == 1

87
    info = tsh.info(engine)
88
89
90
91
92
93
    assert {
        'changeset count': 2,
        'serie names': ['ts_othervalues', 'ts_values'],
        'series count': 2
    } == info

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
def test_strip(engine, tsh):
    for i in range(1, 5):
        pubdate = datetime(2017, 1, i)
        ts = genserie(datetime(2017, 1, 10), 'H', 1 + i)
        with tsh.newchangeset(engine, 'babar', _insertion_date=pubdate):
            tsh.insert(engine, ts, 'xserie')
        # also insert something completely unrelated
        tsh.insert(engine, genserie(datetime(2018, 1, 1), 'D', 1 + i), 'yserie', 'celeste')

    csida = tsh.changeset_at(engine, 'xserie', datetime(2017, 1, 3))
    assert csida is not None
    csidb = tsh.changeset_at(engine, 'xserie', datetime(2017, 1, 3, 1), mode='before')
    csidc = tsh.changeset_at(engine, 'xserie', datetime(2017, 1, 3, 1), mode='after')
    assert csidb < csida < csidc

    log = tsh.log(engine, names=['xserie', 'yserie'])
    assert [(idx, l['author']) for idx, l in enumerate(log, start=1)
    ] == [
        (1, 'babar'),
        (2, 'celeste'),
        (3, 'babar'),
        (4, 'celeste'),
        (5, 'babar'),
        (6, 'celeste'),
        (7, 'babar'),
        (8, 'celeste')
    ]

    h = tsh.get_history(engine, 'xserie')
    assert_df("""
insertion_date  value_date         
2017-01-01      2017-01-10 00:00:00    0.0
                2017-01-10 01:00:00    1.0
2017-01-02      2017-01-10 00:00:00    0.0
                2017-01-10 01:00:00    1.0
                2017-01-10 02:00:00    2.0
2017-01-03      2017-01-10 00:00:00    0.0
                2017-01-10 01:00:00    1.0
                2017-01-10 02:00:00    2.0
                2017-01-10 03:00:00    3.0
2017-01-04      2017-01-10 00:00:00    0.0
                2017-01-10 01:00:00    1.0
                2017-01-10 02:00:00    2.0
                2017-01-10 03:00:00    3.0
                2017-01-10 04:00:00    4.0
""", h)

    csid = tsh.changeset_at(engine, 'xserie', datetime(2017, 1, 3))
    with engine.connect() as cn:
        tsh.strip(cn, 'xserie', csid)

    assert_df("""
insertion_date  value_date         
2017-01-01      2017-01-10 00:00:00    0.0
                2017-01-10 01:00:00    1.0
2017-01-02      2017-01-10 00:00:00    0.0
                2017-01-10 01:00:00    1.0
                2017-01-10 02:00:00    2.0
""", tsh.get_history(engine, 'xserie'))

    assert_df("""
2017-01-10 00:00:00    0.0
2017-01-10 01:00:00    1.0
2017-01-10 02:00:00    2.0
""", tsh.get(engine, 'xserie'))

    # internal structure is ok
    with engine.connect() as cn:
        cn.execute('set search_path to "{}.timeserie"'.format(tsh.namespace))
164
165
        df = pd.read_sql("select id, diff from xserie order by id", cn)
        df['diff'] = df['diff'].apply(lambda x: False if x is None else True)
166
167

    assert_df("""
168
169
170
id   diff
0   1  False
1   2   True
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
""", df)

    log = tsh.log(engine, names=['xserie', 'yserie'])
    # 5 and 7 have disappeared
    assert [l['author'] for l in log
    ] == ['babar', 'celeste', 'babar', 'celeste', 'celeste', 'celeste']

    log = tsh.log(engine, stripped=True, names=['xserie', 'yserie'])
    assert [list(l['meta'].values())[0][:-1] + 'X' for l in log if l['meta']
    ] == [
        'got stripped from X',
        'got stripped from X'
    ]


186
def test_tstamp_roundtrip(engine, tsh):
187
188
    ts = genserie(datetime(2017, 10, 28, 23),
                  'H', 4, tz='UTC')
189
190
191
192
193
194
195
196
197
198
    ts.index = ts.index.tz_convert('Europe/Paris')

    assert_df("""
2017-10-29 01:00:00+02:00    0
2017-10-29 02:00:00+02:00    1
2017-10-29 02:00:00+01:00    2
2017-10-29 03:00:00+01:00    3
Freq: H
    """, ts)

199
200
    tsh.insert(engine, ts, 'tztest', 'Babar')
    back = tsh.get(engine, 'tztest')
201
202
203

    # though un localized we understand it's been normalized to utc
    assert_df("""
204
205
206
207
2017-10-28 23:00:00+00:00    0.0
2017-10-29 00:00:00+00:00    1.0
2017-10-29 01:00:00+00:00    2.0
2017-10-29 02:00:00+00:00    3.0
208
209
210
""", back)

    assert (ts.index == back.index).all()
211
    assert str(back.index.dtype) == 'datetime64[ns, UTC]'
212
213


214
215
def test_multi_index_aware(engine, tsh):
    ts_multi_aware = genserie(
216
217
218
        start=pd.Timestamp(
            2017, 10, 28, 23
        ).tz_localize('UTC').tz_convert('Europe/Paris'),
219
220
        freq=['15T', '30T', '60T'],
        repeat=10,
221
        tz='Europe/Paris',
222
223
224
225
        name='ts_multi_aware',
    )
    ts_multi_aware.index.rename(['a', 'b', 'c'], inplace=True)

226
227
228
229
230
231
232
233
234
235
236
237
238
239
    assert_df("""
a                          b                          c                        
2017-10-29 01:00:00+02:00  2017-10-29 01:00:00+02:00  2017-10-29 01:00:00+02:00    0
2017-10-29 01:15:00+02:00  2017-10-29 01:30:00+02:00  2017-10-29 02:00:00+02:00    1
2017-10-29 01:30:00+02:00  2017-10-29 02:00:00+02:00  2017-10-29 02:00:00+01:00    2
2017-10-29 01:45:00+02:00  2017-10-29 02:30:00+02:00  2017-10-29 03:00:00+01:00    3
2017-10-29 02:00:00+02:00  2017-10-29 02:00:00+01:00  2017-10-29 04:00:00+01:00    4
2017-10-29 02:15:00+02:00  2017-10-29 02:30:00+01:00  2017-10-29 05:00:00+01:00    5
2017-10-29 02:30:00+02:00  2017-10-29 03:00:00+01:00  2017-10-29 06:00:00+01:00    6
2017-10-29 02:45:00+02:00  2017-10-29 03:30:00+01:00  2017-10-29 07:00:00+01:00    7
2017-10-29 02:00:00+01:00  2017-10-29 04:00:00+01:00  2017-10-29 08:00:00+01:00    8
2017-10-29 02:15:00+01:00  2017-10-29 04:30:00+01:00  2017-10-29 09:00:00+01:00    9
""", ts_multi_aware)

240
241
242
243
    tsh.insert(engine, ts_multi_aware, 'ts_multi_aware', 'test')
    ts_aware = tsh.get(engine, 'ts_multi_aware')

    assert_df("""
244
ts_multi_aware
245
a                         b                         c                                        
246
247
248
249
250
251
252
253
254
255
2017-10-28 23:00:00+00:00 2017-10-28 23:00:00+00:00 2017-10-28 23:00:00+00:00             0.0
2017-10-28 23:15:00+00:00 2017-10-28 23:30:00+00:00 2017-10-29 00:00:00+00:00             1.0
2017-10-28 23:30:00+00:00 2017-10-29 00:00:00+00:00 2017-10-29 01:00:00+00:00             2.0
2017-10-28 23:45:00+00:00 2017-10-29 00:30:00+00:00 2017-10-29 02:00:00+00:00             3.0
2017-10-29 00:00:00+00:00 2017-10-29 01:00:00+00:00 2017-10-29 03:00:00+00:00             4.0
2017-10-29 00:15:00+00:00 2017-10-29 01:30:00+00:00 2017-10-29 04:00:00+00:00             5.0
2017-10-29 00:30:00+00:00 2017-10-29 02:00:00+00:00 2017-10-29 05:00:00+00:00             6.0
2017-10-29 00:45:00+00:00 2017-10-29 02:30:00+00:00 2017-10-29 06:00:00+00:00             7.0
2017-10-29 01:00:00+00:00 2017-10-29 03:00:00+00:00 2017-10-29 07:00:00+00:00             8.0
2017-10-29 01:15:00+00:00 2017-10-29 03:30:00+00:00 2017-10-29 08:00:00+00:00             9.0
256
257
258
    """, pd.DataFrame(ts_aware.sort_index()))
    # Note: the columnns are returned according to the alphabetic order

259
260
261
262
263
264
265
266
267
268
269
270
    ts = tsh.get(engine, 'ts_multi_aware',
                 from_value_date=pd.Timestamp(2017, 10, 29, 0).tz_localize('UTC'),
                 to_value_date=pd.Timestamp(2017, 10, 29, 1).tz_localize('UTC'))
    assert_df("""
a                          b                          c                        
2017-10-29 00:00:00+00:00  2017-10-29 01:00:00+00:00  2017-10-29 03:00:00+00:00    4.0
2017-10-29 00:15:00+00:00  2017-10-29 01:30:00+00:00  2017-10-29 04:00:00+00:00    5.0
2017-10-29 00:30:00+00:00  2017-10-29 02:00:00+00:00  2017-10-29 05:00:00+00:00    6.0
2017-10-29 00:45:00+00:00  2017-10-29 02:30:00+00:00  2017-10-29 06:00:00+00:00    7.0
2017-10-29 01:00:00+00:00  2017-10-29 03:00:00+00:00  2017-10-29 07:00:00+00:00    8.0
    """, ts)

271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
    ts = genserie(datetime(2010, 1, 1), 'D', 10)
    with pytest.raises(Exception) as err:
        tsh.insert(engine, ts, 'ts_multi_aware', 'test')
    assert err.value.args[0] == 'Incompatible index types'

    ts = genserie(
        start=pd.Timestamp(
            2017, 10, 28, 23
        ).tz_localize('UTC').tz_convert('Europe/Paris'),
        freq=['15T', '30T'],
        repeat=10,
        tz='Europe/Paris',
        name='ts_multi_aware',
    )
    ts.index.rename(['a', 'b'], inplace=True)

    with pytest.raises(Exception) as err:
        tsh.insert(engine, ts, 'ts_multi_aware', 'test')
    assert err.value.args[0] == "Incompatible multi indexes: ['a', 'b', 'c'] vs ['a', 'b']"

291

292
def test_differential(engine, tsh):
293
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 10)
294
    tsh.insert(engine, ts_begin, 'ts_test', 'test')
295

296
297
    assert tsh.exists(engine, 'ts_test')
    assert not tsh.exists(engine, 'this_does_not_exist')
298

299
    assert_df("""
300
301
302
303
304
305
306
307
308
309
2010-01-01    0.0
2010-01-02    1.0
2010-01-03    2.0
2010-01-04    3.0
2010-01-05    4.0
2010-01-06    5.0
2010-01-07    6.0
2010-01-08    7.0
2010-01-09    8.0
2010-01-10    9.0
310
""", tsh.get(engine, 'ts_test'))
311
312

    # we should detect the emission of a message
313
    tsh.insert(engine, ts_begin, 'ts_test', 'babar')
314

315
    assert_df("""
316
317
318
319
320
321
322
323
324
325
2010-01-01    0.0
2010-01-02    1.0
2010-01-03    2.0
2010-01-04    3.0
2010-01-05    4.0
2010-01-06    5.0
2010-01-07    6.0
2010-01-08    7.0
2010-01-09    8.0
2010-01-10    9.0
326
""", tsh.get(engine, 'ts_test'))
327
328
329
330

    ts_slight_variation = ts_begin.copy()
    ts_slight_variation.iloc[3] = 0
    ts_slight_variation.iloc[6] = 0
331
    tsh.insert(engine, ts_slight_variation, 'ts_test', 'celeste')
332

333
    assert_df("""
334
335
336
337
338
339
340
341
342
343
2010-01-01    0.0
2010-01-02    1.0
2010-01-03    2.0
2010-01-04    0.0
2010-01-05    4.0
2010-01-06    5.0
2010-01-07    0.0
2010-01-08    7.0
2010-01-09    8.0
2010-01-10    9.0
344
""", tsh.get(engine, 'ts_test'))
345

346
    ts_longer = genserie(datetime(2010, 1, 3), 'D', 15)
347
348
349
350
    ts_longer.iloc[1] = 2.48
    ts_longer.iloc[3] = 3.14
    ts_longer.iloc[5] = ts_begin.iloc[7]

351
    tsh.insert(engine, ts_longer, 'ts_test', 'test')
352

353
    assert_df("""
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
2010-01-01     0.00
2010-01-02     1.00
2010-01-03     0.00
2010-01-04     2.48
2010-01-05     2.00
2010-01-06     3.14
2010-01-07     4.00
2010-01-08     7.00
2010-01-09     6.00
2010-01-10     7.00
2010-01-11     8.00
2010-01-12     9.00
2010-01-13    10.00
2010-01-14    11.00
2010-01-15    12.00
2010-01-16    13.00
2010-01-17    14.00
371
""", tsh.get(engine, 'ts_test'))
372
373

    # start testing manual overrides
374
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 5, initval=[2])
375
    ts_begin.loc['2010-01-04'] = -1
376
    tsh.insert(engine, ts_begin, 'ts_mixte', 'test')
377
378

    # -1 represents bogus upstream data
379
    assert_df("""
380
381
382
383
384
2010-01-01    2.0
2010-01-02    2.0
2010-01-03    2.0
2010-01-04   -1.0
2010-01-05    2.0
385
""", tsh.get(engine, 'ts_mixte'))
386
387

    # refresh all the period + 1 extra data point
388
    ts_more = genserie(datetime(2010, 1, 2), 'D', 5, [2])
389
    ts_more.loc['2010-01-04'] = -1
390
    tsh.insert(engine, ts_more, 'ts_mixte', 'test')
391

392
    assert_df("""
393
394
395
396
397
398
2010-01-01    2.0
2010-01-02    2.0
2010-01-03    2.0
2010-01-04   -1.0
2010-01-05    2.0
2010-01-06    2.0
399
""", tsh.get(engine, 'ts_mixte'))
400
401

    # just append an extra data point
402
403
    # with no intersection with the previous ts
    ts_one_more = genserie(datetime(2010, 1, 7), 'D', 1, [3])
404
    tsh.insert(engine, ts_one_more, 'ts_mixte', 'test')
405

406
    assert_df("""
407
408
409
410
411
412
413
2010-01-01    2.0
2010-01-02    2.0
2010-01-03    2.0
2010-01-04   -1.0
2010-01-05    2.0
2010-01-06    2.0
2010-01-07    3.0
414
""", tsh.get(engine, 'ts_mixte'))
415

416
    with engine.connect() as cn:
417
        cn.execute('set search_path to "{0}.timeserie", {0}, public'.format(tsh.namespace))
418
419
420
        allts = pd.read_sql("select name, table_name from registry "
                            "where name in ('ts_test', 'ts_mixte')",
                            cn)
421

422
423
        assert_df("""
name              table_name
424
425
426
0   ts_test   {0}.timeserie.ts_test
1  ts_mixte  {0}.timeserie.ts_mixte
""".format(tsh.namespace), allts)
427

428
        assert_df("""
429
430
431
432
433
434
435
2010-01-01    2.0
2010-01-02    2.0
2010-01-03    2.0
2010-01-04   -1.0
2010-01-05    2.0
2010-01-06    2.0
2010-01-07    3.0
436
""", tsh.get(cn, 'ts_mixte',
437
             revision_date=datetime.now()))
438
439


440
def test_bad_import(engine, tsh):
441
    # the data were parsed as date by pd.read_json()
Aurélien Campéas's avatar
Aurélien Campéas committed
442
    df_result = pd.read_csv(str(DATADIR / 'test_data.csv'))
443
444
445
    df_result['Gas Day'] = df_result['Gas Day'].apply(parser.parse, dayfirst=True, yearfirst=False)
    df_result.set_index('Gas Day', inplace=True)
    ts = df_result['SC']
446

447
448
    tsh.insert(engine, ts, 'SND_SC', 'test')
    result = tsh.get(engine, 'SND_SC')
449
    assert result.dtype == 'float64'
450
451
452

    # insertion of empty ts
    ts = pd.Series(name='truc', dtype='object')
453
454
    tsh.insert(engine, ts, 'empty_ts', 'test')
    assert tsh.get(engine, 'empty_ts') is None
455
456
457

    # nan in ts
    # all na
458
    ts = genserie(datetime(2010, 1, 10), 'D', 10, [np.nan], name='truc')
459
460
    tsh.insert(engine, ts, 'test_nan', 'test')
    assert tsh.get(engine, 'test_nan') is None
461
462
463
464
465

    # mixe na
    ts = pd.Series([np.nan] * 5 + [3] * 5,
                   index=pd.date_range(start=datetime(2010, 1, 10),
                                       freq='D', periods=10), name='truc')
466
467
    tsh.insert(engine, ts, 'test_nan', 'test')
    result = tsh.get(engine, 'test_nan')
468

469
470
    tsh.insert(engine, ts, 'test_nan', 'test')
    result = tsh.get(engine, 'test_nan')
471
    assert_df("""
472
473
474
475
476
2010-01-15    3.0
2010-01-16    3.0
2010-01-17    3.0
2010-01-18    3.0
2010-01-19    3.0
477
""", result)
478
479
480

    # get_ts with name not in database

481
    tsh.get(engine, 'inexisting_name', 'test')
482
483


484
def test_revision_date(engine, tsh):
485
    # we prepare a good joke for the end of the test
486
487
    ival = Snapshot._interval
    Snapshot._interval = 3
488
489
490
491
492
493
494
495

    for i in range(1, 5):
        with engine.connect() as cn:
            with tsh.newchangeset(cn, 'test',
                                  _insertion_date=datetime(2016, 1, i)):
                tsh.insert(cn, genserie(datetime(2017, 1, i), 'D', 3, [i]), 'revdate')

    # end of prologue, now some real meat
496
497
498
499
500
501
502
    idate0 = datetime(2015, 1, 1, 0, 0, 0)
    with tsh.newchangeset(engine, 'test', _insertion_date=idate0):

        ts = genserie(datetime(2010, 1, 4), 'D', 4, [0], name='truc')
        tsh.insert(engine, ts, 'ts_through_time')
        assert idate0 == tsh.latest_insertion_date(engine, 'ts_through_time')

503
    idate1 = datetime(2015, 1, 1, 15, 43, 23)
504
    with tsh.newchangeset(engine, 'test', _insertion_date=idate1):
505

506
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [1], name='truc')
507
508
        tsh.insert(engine, ts, 'ts_through_time')
        assert idate1 == tsh.latest_insertion_date(engine, 'ts_through_time')
509

510
    idate2 = datetime(2015, 1, 2, 15, 43, 23)
511
    with tsh.newchangeset(engine, 'test', _insertion_date=idate2):
512

513
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [2], name='truc')
514
515
        tsh.insert(engine, ts, 'ts_through_time')
        assert idate2 == tsh.latest_insertion_date(engine, 'ts_through_time')
516

517
    idate3 = datetime(2015, 1, 3, 15, 43, 23)
518
    with tsh.newchangeset(engine, 'test', _insertion_date=idate3):
519

520
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [3], name='truc')
521
522
        tsh.insert(engine, ts, 'ts_through_time')
        assert idate3 == tsh.latest_insertion_date(engine, 'ts_through_time')
523

524
    ts = tsh.get(engine, 'ts_through_time')
525

526
    assert_df("""
527
528
529
530
2010-01-04    3.0
2010-01-05    3.0
2010-01-06    3.0
2010-01-07    3.0
531
""", ts)
532

533
    ts = tsh.get(engine, 'ts_through_time',
Aurélien Campéas's avatar
Aurélien Campéas committed
534
                 revision_date=datetime(2015, 1, 2, 18, 43, 23))
535

536
    assert_df("""
537
538
539
540
2010-01-04    2.0
2010-01-05    2.0
2010-01-06    2.0
2010-01-07    2.0
541
""", ts)
542

543
    ts = tsh.get(engine, 'ts_through_time',
544
                 revision_date=datetime(2015, 1, 1, 18, 43, 23))
545

546
    assert_df("""
547
548
549
550
2010-01-04    1.0
2010-01-05    1.0
2010-01-06    1.0
2010-01-07    1.0
551
""", ts)
552

553
    ts = tsh.get(engine, 'ts_through_time',
554
                 revision_date=datetime(2014, 1, 1, 18, 43, 23))
555
556
557

    assert ts is None

558
559
560
561
562
563
564
565
566
567
    # epilogue: back to the revdate issue
    assert_df("""
2017-01-01    1.0
2017-01-02    2.0
2017-01-03    3.0
2017-01-04    4.0
2017-01-05    4.0
2017-01-06    4.0
""", tsh.get(engine, 'revdate'))

568
    oldstate = tsh.get(engine, 'revdate', revision_date=datetime(2016, 1, 2))
569
570
571
    assert_df("""
2017-01-01    1.0
2017-01-02    2.0
572
573
574
2017-01-03    2.0
2017-01-04    2.0
""", oldstate)
575

576
    Snapshot._interval = ival
577

578

579
def test_snapshots(engine, tsh):
580
581
    baseinterval = Snapshot._interval
    Snapshot._interval = 4
582

583
    with engine.connect() as cn:
584
        for tscount in range(1, 11):
585
            ts = genserie(datetime(2015, 1, 1), 'D', tscount, [1])
586
            diff = tsh.insert(cn, ts, 'growing', 'babar')
587
588
            assert diff.index[0] == diff.index[-1] == ts.index[-1]

589
    diff = tsh.insert(engine, ts, 'growing', 'babar')
590
    assert diff is None
591

592
    with engine.connect() as cn:
593
594
595
596
597
598
599
600
601
602
        cn.execute('set search_path to "{}.snapshot"'.format(tsh.namespace))
#         df = pd.read_sql("select cset from growing",
#                          cn)
#         assert_df("""
# cset
# 0     1
# 1     4
# 2     8
# 3    10
# """, df)
603

604
605
        ts = tsh.get(cn, 'growing')
        assert_df("""
606
607
608
609
610
611
612
613
614
615
2015-01-01    1.0
2015-01-02    1.0
2015-01-03    1.0
2015-01-04    1.0
2015-01-05    1.0
2015-01-06    1.0
2015-01-07    1.0
2015-01-08    1.0
2015-01-09    1.0
2015-01-10    1.0
616
""", ts)
617

618
619
        df = pd.read_sql("select id, chunk from growing order by id", cn)
        df['chunk'] = df['chunk'].apply(lambda x: 0 if x is None else len(x))
620

621
        assert_df("""
622
623
624
625
626
id  chunk
0   1     35
1   4     47
2   8     59
3  10     67
627
""", df)
628

629
630
    # table = tsh._get_ts_table(engine, 'growing')
    _, snap = Snapshot(engine, tsh, 'growing').find()
631
    assert (ts == snap).all()
632
    Snapshot._interval = baseinterval
633
634


635
def test_deletion(engine, tsh):
636
637
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 11)
    ts_begin.iloc[-1] = np.nan
638
    tsh.insert(engine, ts_begin, 'ts_del', 'test')
639

640
    ts = Snapshot(engine, tsh, 'ts_del').build_upto()
641
    assert ts.iloc[-1] == 9.0
642

643
    ts_begin.iloc[0] = np.nan
644
    ts_begin.iloc[3] = np.nan
645

646
    tsh.insert(engine, ts_begin, 'ts_del', 'test')
647

648
    assert_df("""
649
650
651
652
653
654
655
2010-01-02    1.0
2010-01-03    2.0
2010-01-05    4.0
2010-01-06    5.0
2010-01-07    6.0
2010-01-08    7.0
2010-01-09    8.0
656
2010-01-10    9.0
657
""", tsh.get(engine, 'ts_del'))
658

659
    ts2 = tsh.get(engine, 'ts_del',
660
661
                 # force snapshot reconstruction feature
                 revision_date=datetime(2038, 1, 1))
662
    assert (tsh.get(engine, 'ts_del') == ts2).all()
663

664
665
666
    ts_begin.iloc[0] = 42
    ts_begin.iloc[3] = 23

667
    tsh.insert(engine, ts_begin, 'ts_del', 'test')
668

669
    assert_df("""
670
671
672
673
674
675
676
677
678
2010-01-01    42.0
2010-01-02     1.0
2010-01-03     2.0
2010-01-04    23.0
2010-01-05     4.0
2010-01-06     5.0
2010-01-07     6.0
2010-01-08     7.0
2010-01-09     8.0
679
2010-01-10     9.0
680
""", tsh.get(engine, 'ts_del'))
681
682
683

    # now with string!

684
    ts_string = genserie(datetime(2010, 1, 1), 'D', 10, ['machin'])
685
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
686
687
688
689

    ts_string[4] = None
    ts_string[5] = None

690
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
691
    assert_df("""
692
693
694
695
696
697
698
699
2010-01-01    machin
2010-01-02    machin
2010-01-03    machin
2010-01-04    machin
2010-01-07    machin
2010-01-08    machin
2010-01-09    machin
2010-01-10    machin
700
""", tsh.get(engine, 'ts_string_del'))
701
702
703
704

    ts_string[4] = 'truc'
    ts_string[6] = 'truc'

705
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
706
    assert_df("""
707
708
709
710
711
712
713
714
2010-01-01    machin
2010-01-02    machin
2010-01-03    machin
2010-01-04    machin
2010-01-05      truc
2010-01-07      truc
2010-01-08    machin
2010-01-09    machin
715
2010-01-10    machin
716
""", tsh.get(engine, 'ts_string_del'))
717

718
    ts_string[ts_string.index] = np.nan
719
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
720

721
    erased = tsh.get(engine, 'ts_string_del')
722
723
    assert len(erased) == 0

724
725
    # first insertion with only nan

726
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 10, [np.nan])
727
    tsh.insert(engine, ts_begin, 'ts_null', 'test')
728

729
    assert tsh.get(engine, 'ts_null') is None
730

731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
    # exhibit issue with nans handling
    ts_repushed = genserie(datetime(2010, 1, 1), 'D', 11)
    ts_repushed[0:3] = np.nan

    assert_df("""
2010-01-01     NaN
2010-01-02     NaN
2010-01-03     NaN
2010-01-04     3.0
2010-01-05     4.0
2010-01-06     5.0
2010-01-07     6.0
2010-01-08     7.0
2010-01-09     8.0
2010-01-10     9.0
2010-01-11    10.0
Freq: D
""", ts_repushed)

750
751
    tsh.insert(engine, ts_repushed, 'ts_repushed', 'test')
    diff = tsh.insert(engine, ts_repushed, 'ts_repushed', 'test')
752
753
    assert diff is None

754
    # there is no difference
755
    assert 0 == len(tsh.diff(ts_repushed, ts_repushed))
756
757
758
759
760

    ts_add = genserie(datetime(2010, 1, 1), 'D', 15)
    ts_add.iloc[0] = np.nan
    ts_add.iloc[13:] = np.nan
    ts_add.iloc[8] = np.nan
761
    diff = tsh.diff(ts_repushed, ts_add)
762
763
764
765
766
767

    assert_df("""
2010-01-02     1.0
2010-01-03     2.0
2010-01-09     NaN
2010-01-12    11.0
768
2010-01-13    12.0""", diff.sort_index())
769
770
771
772
    # value on nan => value
    # nan on value => nan
    # nan on nan => Nothing
    # nan on nothing=> Nothing
773

Aurélien Campéas's avatar
Aurélien Campéas committed
774
    # full erasing
775
776
    # numeric
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 4)
777
    tsh.insert(engine, ts_begin, 'ts_full_del', 'test')
778

Aurélien Campéas's avatar
Aurélien Campéas committed
779
    ts_begin.iloc[:] = np.nan
780
    tsh.insert(engine, ts_begin, 'ts_full_del', 'test')
781
782

    ts_end = genserie(datetime(2010, 1, 1), 'D', 4)
783
    tsh.insert(engine, ts_end, 'ts_full_del', 'test')
784
785
786
787

    # string

    ts_begin = genserie(datetime(2010, 1, 1), 'D', 4, ['text'])
788
    tsh.insert(engine, ts_begin, 'ts_full_del_str', 'test')
789
790

    ts_begin.iloc[:] = np.nan
791
    tsh.insert(engine, ts_begin, 'ts_full_del_str', 'test')
792
793

    ts_end = genserie(datetime(2010, 1, 1), 'D', 4, ['text'])
794
    tsh.insert(engine, ts_end, 'ts_full_del_str', 'test')
795

Aurélien Campéas's avatar
Aurélien Campéas committed
796

797
def test_multi_index(engine, tsh):
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
    appdate_0 = pd.DatetimeIndex(start=datetime(2015, 1, 1),
                                 end=datetime(2015, 1, 2),
                                 freq='D').values
    pubdate_0 = [pd.datetime(2015, 1, 11, 12, 0, 0)] * 2
    insertion_date_0 = [pd.datetime(2015, 1, 11, 12, 30, 0)] * 2

    multi = [
        appdate_0,
        np.array(pubdate_0),
        np.array(insertion_date_0)
    ]

    ts_multi = pd.Series(range(2), index=multi)
    ts_multi.index.rename(['b', 'c', 'a'], inplace=True)

813
    tsh.insert(engine, ts_multi, 'ts_multi_simple', 'test')
814

815
    ts = tsh.get(engine, 'ts_multi_simple')
816
817
818
    assert_df("""
                                                    ts_multi_simple
a                   b          c                                   
819
820
2015-01-11 12:30:00 2015-01-01 2015-01-11 12:00:00              0.0
                    2015-01-02 2015-01-11 12:00:00              1.0
821
822
""", pd.DataFrame(ts))

823
    diff = tsh.insert(engine, ts_multi, 'ts_multi_simple', 'test')
824
825
826
827
828
    assert diff is None

    ts_multi_2 = pd.Series([0, 2], index=multi)
    ts_multi_2.index.rename(['b', 'c', 'a'], inplace=True)

829
830
    tsh.insert(engine, ts_multi_2, 'ts_multi_simple', 'test')
    ts = tsh.get(engine, 'ts_multi_simple')
831
832
833
834

    assert_df("""
                                                    ts_multi_simple
a                   b          c                                   
835
836
2015-01-11 12:30:00 2015-01-01 2015-01-11 12:00:00              0.0
                    2015-01-02 2015-01-11 12:00:00              2.0
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
""", pd.DataFrame(ts))

    # bigger ts
    appdate_0 = pd.DatetimeIndex(start=datetime(2015, 1, 1),
                                 end=datetime(2015, 1, 4),
                                 freq='D').values
    pubdate_0 = [pd.datetime(2015, 1, 11, 12, 0, 0)] * 4
    insertion_date_0 = [pd.datetime(2015, 1, 11, 12, 30, 0)] * 4

    appdate_1 = pd.DatetimeIndex(start=datetime(2015, 1, 1),
                                 end=datetime(2015, 1, 4),
                                 freq='D').values

    pubdate_1 = [pd.datetime(2015, 1, 21, 12, 0, 0)] * 4
    insertion_date_1 = [pd.datetime(2015, 1, 21, 12, 30, 0)] * 4

    multi = [
        np.concatenate([appdate_0, appdate_1]),
        np.array(pubdate_0 + pubdate_1),
        np.array(insertion_date_0 + insertion_date_1)
    ]

    ts_multi = pd.Series(range(8), index=multi)
    ts_multi.index.rename(['a', 'c', 'b'], inplace=True)

862
863
    tsh.insert(engine, ts_multi, 'ts_multi', 'test')
    ts = tsh.get(engine, 'ts_multi')
864
865
866
867

    assert_df("""
                                                    ts_multi
a          b                   c                            
868
869
870
871
872
873
874
875
2015-01-01 2015-01-11 12:30:00 2015-01-11 12:00:00       0.0
           2015-01-21 12:30:00 2015-01-21 12:00:00       4.0
2015-01-02 2015-01-11 12:30:00 2015-01-11 12:00:00       1.0
           2015-01-21 12:30:00 2015-01-21 12:00:00       5.0
2015-01-03 2015-01-11 12:30:00 2015-01-11 12:00:00       2.0
           2015-01-21 12:30:00 2015-01-21 12:00:00       6.0
2015-01-04 2015-01-11 12:30:00 2015-01-11 12:00:00       3.0
           2015-01-21 12:30:00 2015-01-21 12:00:00       7.0
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
    """, pd.DataFrame(ts.sort_index()))
    # Note: the columnns are returned according to the alphabetic order

    appdate_2 = pd.DatetimeIndex(start=datetime(2015, 1, 1),
                                 end=datetime(2015, 1, 4),
                                 freq='D').values
    pubdate_2 = [pd.datetime(2015, 1, 31, 12, 0, 0)] * 4
    insertion_date_2 = [pd.datetime(2015, 1, 31, 12, 30, 0)] * 4

    multi_2 = [
        np.concatenate([appdate_1, appdate_2]),
        np.array(pubdate_1 + pubdate_2),
        np.array(insertion_date_1 + insertion_date_2)
    ]

    ts_multi_2 = pd.Series([4] * 8, index=multi_2)
    ts_multi_2.index.rename(['a', 'c', 'b'], inplace=True)

    # A second ts is inserted with some index in common with the first
    # one: appdate_1, pubdate_1,and insertion_date_1. The value is set
    # at 4, which matches the previous value of the "2015-01-01" point.

898
    diff = tsh.insert(engine, ts_multi_2, 'ts_multi', 'test')
899
900
901
902
903
904
905
906
907
908
909
910
911
912
    assert_df("""
                                                    ts_multi
a          b                   c                            
2015-01-01 2015-01-31 12:30:00 2015-01-31 12:00:00       4.0
2015-01-02 2015-01-21 12:30:00 2015-01-21 12:00:00       4.0
           2015-01-31 12:30:00 2015-01-31 12:00:00       4.0
2015-01-03 2015-01-21 12:30:00 2015-01-21 12:00:00       4.0
           2015-01-31 12:30:00 2015-01-31 12:00:00       4.0
2015-01-04 2015-01-21 12:30:00 2015-01-21 12:00:00       4.0
           2015-01-31 12:30:00 2015-01-31 12:00:00       4.0
        """, pd.DataFrame(diff.sort_index()))
    # the differential skips a value for "2015-01-01"
    # which does not change from the previous ts

913
    ts = tsh.get(engine, 'ts_multi')
914
915
916
    assert_df("""
                                                    ts_multi
a          b                   c                            
917
918
919
920
921
922
923
924
925
926
927
928
2015-01-01 2015-01-11 12:30:00 2015-01-11 12:00:00       0.0
           2015-01-21 12:30:00 2015-01-21 12:00:00       4.0
           2015-01-31 12:30:00 2015-01-31 12:00:00       4.0
2015-01-02 2015-01-11 12:30:00 2015-01-11 12:00:00       1.0
           2015-01-21 12:30:00 2015-01-21 12:00:00       4.0
           2015-01-31 12:30:00 2015-01-31 12:00:00       4.0
2015-01-03 2015-01-11 12:30:00 2015-01-11 12:00:00       2.0
           2015-01-21 12:30:00 2015-01-21 12:00:00       4.0
           2015-01-31 12:30:00 2015-01-31 12:00:00       4.0
2015-01-04 2015-01-11 12:30:00 2015-01-11 12:00:00       3.0
           2015-01-21 12:30:00 2015-01-21 12:00:00       4.0
           2015-01-31 12:30:00 2015-01-31 12:00:00       4.0
929
930
931
        """, pd.DataFrame(ts.sort_index()))

    # the result ts have now 3 values for each point in 'a'
932
933


934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
def test_multi_index_get_history(engine, tsh):
    appdate = pd.DatetimeIndex(
        start=datetime(2015, 1, 1),
        end=datetime(2015, 1, 2),
        freq='D'
    ).values
    forecast_date = [pd.Timestamp(2015, 1, 11, 12, 0, 0)] * 2
    multi = [
        appdate,
        np.array(forecast_date),
    ]

    ts_multi = pd.Series(range(2), index=multi)
    ts_multi.index.rename(['app_date', 'fc_date'], inplace=True)

    tsh.insert(engine, ts_multi, 'ts_mi', 'Babar',
               _insertion_date=pd.datetime(2015, 1, 11, 12, 30, 0))


    ts = tsh.get_history(engine, 'ts_mi')
    assert_df("""
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
1001
1002
1003
1004
insertion_date       app_date    fc_date            
2015-01-11 12:30:00  2015-01-01  2015-01-11 12:00:00    0.0
                     2015-01-02  2015-01-11 12:00:00    1.0
""", ts)

    ts = tsh.get_history(engine, 'ts_mi', diffmode=True)

    assert_df("""
insertion_date       app_date    fc_date            
2015-01-11 12:30:00  2015-01-01  2015-01-11 12:00:00    0.0
                     2015-01-02  2015-01-11 12:00:00    1.0
""", ts)

    # new forecast
    appdate = pd.DatetimeIndex(
        start=datetime(2015, 1, 1),
        end=datetime(2015, 1, 2),
        freq='D'
    ).values
    forecast_date = [pd.Timestamp(2015, 1, 11, 13, 0, 0)] * 2
    multi = [
        appdate,
        np.array(forecast_date),
    ]

    ts_multi = pd.Series((x+.1 for x in range(2)), index=multi)
    ts_multi.index.rename(['app_date', 'fc_date'], inplace=True)

    tsh.insert(engine, ts_multi, 'ts_mi', 'Babar',
               _insertion_date=pd.datetime(2015, 1, 11, 13, 30, 0))

    ts = tsh.get_history(engine, 'ts_mi')
    assert_df("""
insertion_date       app_date    fc_date            
2015-01-11 12:30:00  2015-01-01  2015-01-11 12:00:00    0.0
                     2015-01-02  2015-01-11 12:00:00    1.0
2015-01-11 13:30:00  2015-01-01  2015-01-11 12:00:00    0.0
                                 2015-01-11 13:00:00    0.1
                     2015-01-02  2015-01-11 12:00:00    1.0
                                 2015-01-11 13:00:00    1.1
""", ts)

    ts = tsh.get_history(engine, 'ts_mi', diffmode=True)
    assert_df("""
insertion_date       app_date    fc_date            
2015-01-11 12:30:00  2015-01-01  2015-01-11 12:00:00    0.0
                     2015-01-02  2015-01-11 12:00:00    1.0
2015-01-11 13:30:00  2015-01-01  2015-01-11 13:00:00    0.1
                     2015-01-02  2015-01-11 13:00:00    1.1
""", ts)
1005
1006


1007
def test_get_history(engine, tsh):
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
    for numserie in (1, 2, 3):
        with engine.connect() as cn:
            with tsh.newchangeset(cn, 'aurelien.campeas@pythonian.fr',
                                  _insertion_date=datetime(2017, 2, numserie)):
                tsh.insert(cn, genserie(datetime(2017, 1, 1), 'D', numserie), 'smallserie')

    ts = tsh.get(engine, 'smallserie')
    assert_df("""
2017-01-01    0.0
2017-01-02    1.0
2017-01-03    2.0
""", ts)

    logs = tsh.log(engine, names=['smallserie'])
    assert [
        {'author': 'aurelien.campeas@pythonian.fr',
1024
         'meta': {},
1025
1026
1027
1028
         'date': datetime(2017, 2, 1, 0, 0),
         'names': ['smallserie']
        },
        {'author': 'aurelien.campeas@pythonian.fr',
1029
         'meta': {},
1030
1031
1032
1033
         'date': datetime(2017, 2, 2, 0, 0),
         'names': ['smallserie']
        },
        {'author': 'aurelien.campeas@pythonian.fr',
1034
         'meta': {},
1035
1036
1037
1038
1039
1040
         'date': datetime(2017, 2, 3, 0, 0),
         'names': ['smallserie']
        }
    ] == [{k: v for k, v in log.items() if k != 'rev'}
          for log in logs]
    histts = tsh.get_history(engine, 'smallserie')
1041
    assert histts.name == 'smallserie'
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052

    assert_df("""
insertion_date  value_date
2017-02-01      2017-01-01    0.0
2017-02-02      2017-01-01    0.0
                2017-01-02    1.0
2017-02-03      2017-01-01    0.0
                2017-01-02    1.0
                2017-01-03    2.0
""", histts)

1053
1054
1055
1056
1057
1058
1059
1060
    diffs = tsh.get_history(engine, 'smallserie', diffmode=True)
    assert_df("""
insertion_date  value_date
2017-02-01      2017-01-01    0.0
2017-02-02      2017-01-02    1.0
2017-02-03      2017-01-03    2.0
""", diffs)

1061
    for idate in histts.index.get_level_values('insertion_date').unique():
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
        with engine.connect() as cn:
            with tsh.newchangeset(cn, 'aurelien.campeas@pythonian.f',
                                  _insertion_date=idate):
                tsh.insert(cn, histts[idate], 'smallserie2')

    # this is perfectly round-tripable
    assert (tsh.get(engine, 'smallserie2') == ts).all()
    assert (tsh.get_history(engine, 'smallserie2') == histts).all()

    # get history ranges
    tsa = tsh.get_history(engine, 'smallserie',
                          from_insertion_date=datetime(2017, 2, 2))
    assert_df("""
insertion_date  value_date
2017-02-02      2017-01-01    0.0
                2017-01-02    1.0
2017-02-03      2017-01-01    0.0
                2017-01-02    1.0
                2017-01-03    2.0
""", tsa)

    tsb = tsh.get_history(engine, 'smallserie',
                          to_insertion_date=datetime(2017, 2, 2))
    assert_df("""
insertion_date  value_date
2017-02-01      2017-01-01    0.0
2017-02-02      2017-01-01    0.0
                2017-01-02    1.0
""", tsb)

    tsc = tsh.get_history(engine, 'smallserie',
                          from_insertion_date=datetime(2017, 2, 2),
                          to_insertion_date=datetime(2017, 2, 2))
    assert_df("""
insertion_date  value_date
2017-02-02      2017-01-01    0.0
                2017-01-02    1.0
""", tsc)

1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
    tsc = tsh.get_history(engine, 'smallserie',
                          from_insertion_date=datetime(2017, 2, 4),
                          to_insertion_date=datetime(2017, 2, 4))
    assert tsc is None

    tsc = tsh.get_history(engine, 'smallserie',
                          from_insertion_date=datetime(2016, 2, 1),
                          to_insertion_date=datetime(2017, 2, 2))
    assert_df("""
insertion_date  value_date
2017-02-01      2017-01-01    0.0
2017-02-02      2017-01-01    0.0
                2017-01-02    1.0
""", tsc)

    tsc = tsh.get_history(engine, 'smallserie',
                          from_insertion_date=datetime(2016, 2, 1),
                          to_insertion_date=datetime(2016, 12