test_tsio.py 27.4 KB
Newer Older
1
2
3
4
5
6
7
# coding: utf-8
from pathlib import Path
from datetime import datetime
from dateutil import parser

import pandas as pd
import numpy as np
8
import pytest
9
from mock import patch
10

11
from tshistory.tsio import TimeSerie
12
13
14
15

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


16
17
18
19
20
21
22
def assert_group_equals(g1, g2):
    for (n1, s1), (n2, s2) in zip(sorted(g1.items()),
                                  sorted(g2.items())):
        assert n1 == n2
        assert s1.equals(s2)


23
24
25
26
def assert_df(expected, df):
    assert expected.strip() == df.to_string().strip()


27
28
29
30
31
32
33
34
35
36
37
38
def genserie(start, freq, repeat, initval=None, tz=None, name=None):
    if initval is None:
        values = range(repeat)
    else:
        values = initval * repeat
    return pd.Series(values,
                     name=name,
                     index=pd.date_range(start=start,
                                         freq=freq,
                                         periods=repeat,
                                         tz=tz))

Aurélien Campéas's avatar
Aurélien Campéas committed
39

40
def test_changeset(engine):
41
    # instantiate one time serie handler object
42
    tsh = TimeSerie()
43
44

    index = pd.date_range(start=datetime(2017, 1, 1), freq='D', periods=3)
45
    data = [1., 2., 3.]
46

47
48
    with patch('tshistory.tsio.datetime') as mock_date:
        mock_date.now.return_value = datetime(2020, 1, 1)
49
        with engine.connect() as cn:
50
51
52
            with tsh.newchangeset(cn, 'babar'):
                tsh.insert(cn, pd.Series(data, index=index), 'ts_values')
                tsh.insert(cn, pd.Series(['a', 'b', 'c'], index=index), 'ts_othervalues')
53

54
55
        g = tsh.get_group(engine, 'ts_values')
        g2 = tsh.get_group(engine, 'ts_othervalues')
56
        assert_group_equals(g, g2)
57

58
        with pytest.raises(AssertionError):
59
            tsh.insert(engine, pd.Series([2, 3, 4], index=index), 'ts_values')
60

61
        with engine.connect() as cn:
62
            data.append(data.pop(0))
63
64
            with tsh.newchangeset(cn, 'celeste'):
                tsh.insert(cn, pd.Series(data, index=index), 'ts_values')
65
                # below should be a noop
66
                tsh.insert(cn, pd.Series(['a', 'b', 'c'], index=index), 'ts_othervalues')
67

68
    g = tsh.get_group(engine, 'ts_values')
69
70
    assert ['ts_values'] == list(g.keys())

71
    assert_df("""
72
73
74
2017-01-01    2.0
2017-01-02    3.0
2017-01-03    1.0
75
""", tsh.get(engine, 'ts_values'))
76

77
    assert_df("""
78
79
80
2017-01-01    a
2017-01-02    b
2017-01-03    c
81
""", tsh.get(engine, 'ts_othervalues'))
82

83
    log = tsh.log(engine)
84
85
86
87
88
89
90
91
92
93
94
    assert [
        {'author': 'babar',
         'rev': 1,
         'date': datetime(2020, 1, 1, 0, 0),
         'names': ['ts_values', 'ts_othervalues']},
        {'author': 'celeste',
         'rev': 2,
         'date': datetime(2020, 1, 1, 0, 0),
         'names': ['ts_values']}
    ] == log

95
    log = tsh.log(engine, names=['ts_othervalues'])
96
97
    assert len(log) == 1
    assert log[0]['rev'] == 1
98
    assert log[0]['names'] == ['ts_values', 'ts_othervalues']
99

100
    log = tsh.log(engine, fromrev=2)
101
102
    assert len(log) == 1

103
    log = tsh.log(engine, torev=1)
104
105
    assert len(log) == 1

106
    info = tsh.info(engine)
107
108
109
110
111
112
    assert {
        'changeset count': 2,
        'serie names': ['ts_othervalues', 'ts_values'],
        'series count': 2
    } == info

113

114
def test_tstamp_roundtrip(engine):
115
    tsh = TimeSerie()
116
117
    ts = genserie(datetime(2017, 10, 28, 23),
                  'H', 4, tz='UTC')
118
119
120
121
122
123
124
125
126
127
    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)

128
129
    tsh.insert(engine, ts, 'tztest', 'Babar')
    back = tsh.get(engine, 'tztest')
130
131
132
133
134
135
136
137
138
139
140
141
142

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

    back.index = back.index.tz_localize('UTC')
    assert (ts.index == back.index).all()


143
def test_differential(engine):
144
    # instantiate one time serie handler object
145
    tsh = TimeSerie()
146

147
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 10)
148
    tsh.insert(engine, ts_begin, 'ts_test', 'test')
149

150
151
    assert tsh.exists(engine, 'ts_test')
    assert not tsh.exists(engine, 'this_does_not_exist')
152

153
    assert_df("""
154
155
156
157
158
159
160
161
162
163
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
164
""", tsh.get(engine, 'ts_test'))
165
166

    # we should detect the emission of a message
167
    tsh.insert(engine, ts_begin, 'ts_test', 'babar')
168

169
    assert_df("""
170
171
172
173
174
175
176
177
178
179
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
180
""", tsh.get(engine, 'ts_test'))
181
182
183
184

    ts_slight_variation = ts_begin.copy()
    ts_slight_variation.iloc[3] = 0
    ts_slight_variation.iloc[6] = 0
185
    tsh.insert(engine, ts_slight_variation, 'ts_test', 'celeste')
186

187
    assert_df("""
188
189
190
191
192
193
194
195
196
197
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
198
""", tsh.get(engine, 'ts_test'))
199

200
    ts_longer = genserie(datetime(2010, 1, 3), 'D', 15)
201
202
203
204
    ts_longer.iloc[1] = 2.48
    ts_longer.iloc[3] = 3.14
    ts_longer.iloc[5] = ts_begin.iloc[7]

205
    tsh.insert(engine, ts_longer, 'ts_test', 'test')
206

207
    assert_df("""
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
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
225
""", tsh.get(engine, 'ts_test'))
226
227

    # start testing manual overrides
228
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 5, initval=[2])
229
    ts_begin.loc['2010-01-04'] = -1
230
    tsh.insert(engine, ts_begin, 'ts_mixte', 'test')
231
232

    # -1 represents bogus upstream data
233
    assert_df("""
234
235
236
237
238
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
239
""", tsh.get(engine, 'ts_mixte'))
240
241

    # refresh all the period + 1 extra data point
242
    ts_more = genserie(datetime(2010, 1, 2), 'D', 5, [2])
243
    ts_more.loc['2010-01-04'] = -1
244
    tsh.insert(engine, ts_more, 'ts_mixte', 'test')
245

246
    assert_df("""
247
248
249
250
251
252
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
253
""", tsh.get(engine, 'ts_mixte'))
254
255

    # just append an extra data point
256
257
    # with no intersection with the previous ts
    ts_one_more = genserie(datetime(2010, 1, 7), 'D', 1, [3])
258
    tsh.insert(engine, ts_one_more, 'ts_mixte', 'test')
259

260
    assert_df("""
261
262
263
264
265
266
267
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
268
""", tsh.get(engine, 'ts_mixte'))
269

270
    hist = pd.read_sql('select id, parent from timeserie.ts_test order by id',
Aurélien Campéas's avatar
Aurélien Campéas committed
271
                       engine)
272
    assert_df("""
273
274
275
276
   id  parent
0   1     NaN
1   2     1.0
2   3     2.0
277
""", hist)
278

279
    hist = pd.read_sql('select id, parent from timeserie.ts_mixte order by id',
Aurélien Campéas's avatar
Aurélien Campéas committed
280
                       engine)
281
    assert_df("""
282
283
284
285
   id  parent
0   1     NaN
1   2     1.0
2   3     2.0
286
""", hist)
287

288
    allts = pd.read_sql("select name, table_name from registry "
289
                        "where name in ('ts_test', 'ts_mixte')",
290
291
                        engine)

292
    assert_df("""
293
294
295
       name          table_name
0   ts_test   timeserie.ts_test
1  ts_mixte  timeserie.ts_mixte
296
""", allts)
297

298
    assert_df("""
299
300
301
302
303
304
305
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
306
""", tsh.get(engine, 'ts_mixte',
307
             revision_date=datetime.now()))
308
309
310


def test_bad_import(engine):
311
    # instantiate one time serie handler object
312
    tsh = TimeSerie()
313

314
315
316
317
318
    # the data were parsed as date by pd.read_json()
    df_result = pd.read_csv(DATADIR / 'test_data.csv')
    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']
319

320
321
    tsh.insert(engine, ts, 'SND_SC', 'test')
    result = tsh.get(engine, 'SND_SC')
322
    assert result.dtype == 'float64'
323
324
325

    # insertion of empty ts
    ts = pd.Series(name='truc', dtype='object')
326
327
    tsh.insert(engine, ts, 'empty_ts', 'test')
    assert tsh.get(engine, 'empty_ts') is None
328
329
330

    # nan in ts
    # all na
331
    ts = genserie(datetime(2010, 1, 10), 'D', 10, [np.nan], name='truc')
332
333
    tsh.insert(engine, ts, 'test_nan', 'test')
    assert tsh.get(engine, 'test_nan') is None
334
335
336
337
338

    # 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')
339
340
    tsh.insert(engine, ts, 'test_nan', 'test')
    result = tsh.get(engine, 'test_nan')
341

342
343
    tsh.insert(engine, ts, 'test_nan', 'test')
    result = tsh.get(engine, 'test_nan')
344
    assert_df("""
345
346
347
348
349
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
350
""", result)
351
352
353

    # get_ts with name not in database

354
    tsh.get(engine, 'inexisting_name', 'test')
355
356
357


def test_revision_date(engine):
358
    # instantiate one time serie handler object
359
    tsh = TimeSerie()
360

361
    idate1 = datetime(2015, 1, 1, 15, 43, 23)
362
    with tsh.newchangeset(engine, 'test', _insertion_date=idate1):
363

364
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [1], name='truc')
365
366
        tsh.insert(engine, ts, 'ts_through_time')
        assert idate1 == tsh.latest_insertion_date(engine, 'ts_through_time')
367

368
    idate2 = datetime(2015, 1, 2, 15, 43, 23)
369
    with tsh.newchangeset(engine, 'test', _insertion_date=idate2):
370

371
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [2], name='truc')
372
373
        tsh.insert(engine, ts, 'ts_through_time')
        assert idate2 == tsh.latest_insertion_date(engine, 'ts_through_time')
374

375
    idate3 = datetime(2015, 1, 3, 15, 43, 23)
376
    with tsh.newchangeset(engine, 'test', _insertion_date=idate3):
377

378
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [3], name='truc')
379
380
        tsh.insert(engine, ts, 'ts_through_time')
        assert idate3 == tsh.latest_insertion_date(engine, 'ts_through_time')
381

382
    ts = tsh.get(engine, 'ts_through_time')
383

384
    assert_df("""
385
386
387
388
2010-01-04    3.0
2010-01-05    3.0
2010-01-06    3.0
2010-01-07    3.0
389
""", ts)
390

391
    ts = tsh.get(engine, 'ts_through_time',
Aurélien Campéas's avatar
Aurélien Campéas committed
392
                 revision_date=datetime(2015, 1, 2, 18, 43, 23))
393

394
    assert_df("""
395
396
397
398
2010-01-04    2.0
2010-01-05    2.0
2010-01-06    2.0
2010-01-07    2.0
399
""", ts)
400

401
    ts = tsh.get(engine, 'ts_through_time',
402
                 revision_date=datetime(2015, 1, 1, 18, 43, 23))
403

404
    assert_df("""
405
406
407
408
2010-01-04    1.0
2010-01-05    1.0
2010-01-06    1.0
2010-01-07    1.0
409
""", ts)
410

411
    ts = tsh.get(engine, 'ts_through_time',
412
                 revision_date=datetime(2014, 1, 1, 18, 43, 23))
413
414
415

    assert ts is None

416
417

def test_snapshots(engine):
418
419
    tsh = TimeSerie()
    tsh._snapshot_interval = 4
420

421
    with engine.connect() as cn:
422
        for tscount in range(1, 11):
423
            ts = genserie(datetime(2015, 1, 1), 'D', tscount, [1])
424
            diff = tsh.insert(cn, ts, 'growing', 'babar')
425
426
            assert diff.index[0] == diff.index[-1] == ts.index[-1]

427
    diff = tsh.insert(engine, ts, 'growing', 'babar')
428
    assert diff is None
429

430
431
    df = pd.read_sql("select id from timeserie.growing where snapshot is not null",
                     engine)
432
    assert_df("""
433
434
   id
0   1
435
436
437
1   4
2   8
3  10
438
""", df)
439

440
    ts = tsh.get(engine, 'growing')
441
    assert_df("""
442
443
444
445
446
447
448
449
450
451
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
452
""", ts)
453

454
    df = pd.read_sql("select id, diff, snapshot from timeserie.growing order by id", engine)
455
456
457
    for attr in ('diff', 'snapshot'):
        df[attr] = df[attr].apply(lambda x: 0 if x is None else len(x))

458
    assert_df("""
459
   id  diff  snapshot
Arnaud Campeas's avatar
Arnaud Campeas committed
460
461
462
463
464
465
466
467
468
469
0   1     0        32
1   2    32         0
2   3    32         0
3   4    32       125
4   5    32         0
5   6    32         0
6   7    32         0
7   8    32       249
8   9    32         0
9  10    32       311
470
""", df)
471

472
473
    table = tsh._get_ts_table(engine, 'growing')
    snapid, snap = tsh._find_snapshot(engine, table, ())
474
475
    assert snapid == 10
    assert (ts == snap).all()
476
477
478


def test_deletion(engine):
479
    tsh = TimeSerie()
480

481
482
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 11)
    ts_begin.iloc[-1] = np.nan
483
    tsh.insert(engine, ts_begin, 'ts_del', 'test')
484

485
    ts = tsh._build_snapshot_upto(engine, tsh._get_ts_table(engine, 'ts_del'))
486
    assert ts.iloc[-1] == 9.0
487

488
    ts_begin.iloc[0] = np.nan
489
    ts_begin.iloc[3] = np.nan
490

491
    tsh.insert(engine, ts_begin, 'ts_del', 'test')
492

493
    assert_df("""
494
495
496
497
498
499
500
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
501
2010-01-10    9.0
502
""", tsh.get(engine, 'ts_del'))
503

504
    ts2 = tsh.get(engine, 'ts_del',
505
506
                 # force snapshot reconstruction feature
                 revision_date=datetime(2038, 1, 1))
507
    assert (tsh.get(engine, 'ts_del') == ts2).all()
508

509
510
511
    ts_begin.iloc[0] = 42
    ts_begin.iloc[3] = 23

512
    tsh.insert(engine, ts_begin, 'ts_del', 'test')
513

514
    assert_df("""
515
516
517
518
519
520
521
522
523
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
524
2010-01-10     9.0
525
""", tsh.get(engine, 'ts_del'))
526
527
528

    # now with string!

529
    ts_string = genserie(datetime(2010, 1, 1), 'D', 10, ['machin'])
530
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
531
532
533
534

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

535
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
536
    assert_df("""
537
538
539
540
541
542
543
544
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
545
""", tsh.get(engine, 'ts_string_del'))
546
547
548
549

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

550
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
551
    assert_df("""
552
553
554
555
556
557
558
559
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
560
2010-01-10    machin
561
""", tsh.get(engine, 'ts_string_del'))
562

563
    ts_string[ts_string.index] = np.nan
564
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
565

566
    erased = tsh.get(engine, 'ts_string_del')
567
568
    assert len(erased) == 0

569
570
    # first insertion with only nan

571
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 10, [np.nan])
572
    tsh.insert(engine, ts_begin, 'ts_null', 'test')
573

574
    assert tsh.get(engine, 'ts_null') is None
575

576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
    # 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)

595
596
    tsh.insert(engine, ts_repushed, 'ts_repushed', 'test')
    diff = tsh.insert(engine, ts_repushed, 'ts_repushed', 'test')
597
598
    assert diff is None

599
    # there is no difference
600
    assert 0 == len(tsh._compute_diff(ts_repushed, ts_repushed))
601
602
603
604
605

    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
606
    diff = tsh._compute_diff(ts_repushed, ts_add)
607
608
609
610
611
612

    assert_df("""
2010-01-02     1.0
2010-01-03     2.0
2010-01-09     NaN
2010-01-12    11.0
613
2010-01-13    12.0""", diff.sort_index())
614
615
616
617
    # value on nan => value
    # nan on value => nan
    # nan on nan => Nothing
    # nan on nothing=> Nothing
618

Aurélien Campéas's avatar
Aurélien Campéas committed
619
    # full erasing
620
621
    # numeric
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 4)
622
    tsh.insert(engine, ts_begin, 'ts_full_del', 'test')
623

Aurélien Campéas's avatar
Aurélien Campéas committed
624
    ts_begin.iloc[:] = np.nan
625
    tsh.insert(engine, ts_begin, 'ts_full_del', 'test')
626
627

    ts_end = genserie(datetime(2010, 1, 1), 'D', 4)
628
    tsh.insert(engine, ts_end, 'ts_full_del', 'test')
629
630
631
632

    # string

    ts_begin = genserie(datetime(2010, 1, 1), 'D', 4, ['text'])
633
    tsh.insert(engine, ts_begin, 'ts_full_del_str', 'test')
634
635

    ts_begin.iloc[:] = np.nan
636
    tsh.insert(engine, ts_begin, 'ts_full_del_str', 'test')
637
638

    ts_end = genserie(datetime(2010, 1, 1), 'D', 4, ['text'])
639
    tsh.insert(engine, ts_end, 'ts_full_del_str', 'test')
640

Aurélien Campéas's avatar
Aurélien Campéas committed
641

642
def test_multi_index(engine):
643
    tsh = TimeSerie()
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659

    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)

660
    tsh.insert(engine, ts_multi, 'ts_multi_simple', 'test')
661

662
    ts = tsh.get(engine, 'ts_multi_simple')
663
664
665
    assert_df("""
                                                    ts_multi_simple
a                   b          c                                   
666
667
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
668
669
""", pd.DataFrame(ts))

670
    diff = tsh.insert(engine, ts_multi, 'ts_multi_simple', 'test')
671
672
673
674
675
    assert diff is None

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

676
677
    tsh.insert(engine, ts_multi_2, 'ts_multi_simple', 'test')
    ts = tsh.get(engine, 'ts_multi_simple')
678
679
680
681

    assert_df("""
                                                    ts_multi_simple
a                   b          c                                   
682
683
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
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
""", 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)

709
710
    tsh.insert(engine, ts_multi, 'ts_multi', 'test')
    ts = tsh.get(engine, 'ts_multi')
711
712
713
714

    assert_df("""
                                                    ts_multi
a          b                   c                            
715
716
717
718
719
720
721
722
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
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
    """, 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.

745
    diff = tsh.insert(engine, ts_multi_2, 'ts_multi', 'test')
746
747
748
749
750
751
752
753
754
755
756
757
758
759
    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

760
    ts = tsh.get(engine, 'ts_multi')
761
762
763
    assert_df("""
                                                    ts_multi
a          b                   c                            
764
765
766
767
768
769
770
771
772
773
774
775
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
776
777
778
        """, pd.DataFrame(ts.sort_index()))

    # the result ts have now 3 values for each point in 'a'
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
def test_get_history(engine):
    tsh = TimeSerie()

    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',
         'date': datetime(2017, 2, 1, 0, 0),
         'names': ['smallserie']
        },
        {'author': 'aurelien.campeas@pythonian.fr',
         'date': datetime(2017, 2, 2, 0, 0),
         'names': ['smallserie']
        },
        {'author': 'aurelien.campeas@pythonian.fr',
         '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')

    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)

    for idx, idate in enumerate(histts.groupby('insertion_date').groups):
        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)


866
def test_add_na(engine):
867
    tsh = TimeSerie()
868
869
870
871
872
873

    # a serie of NaNs won't be insert in base
    # in case of first insertion
    ts_nan = genserie(datetime(2010, 1, 1), 'D', 5)
    ts_nan[[True] * len(ts_nan)] = np.nan

874
    diff = tsh.insert(engine, ts_nan, 'ts_add_na', 'test')
875
    assert diff is None
876
    result = tsh.get(engine, 'ts_add_na')
877
878
879
880
    assert result is None

    # in case of insertion in existing data
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 5)
881
    tsh.insert(engine, ts_begin, 'ts_add_na', 'test')
882
883
884
885
886

    ts_nan = genserie(datetime(2010, 1, 6), 'D', 5)
    ts_nan[[True] * len(ts_nan)] = np.nan
    ts_nan = pd.concat([ts_begin, ts_nan])

887
    diff = tsh.insert(engine, ts_nan, 'ts_add_na', 'test')
888
889
    assert diff is None

890
    result = tsh.get(engine, 'ts_add_na')
891
    assert len(result) == 5
892
893
894


def test_dtype_mismatch(engine):
895
    tsh = TimeSerie()
896

897
    tsh.insert(engine,
898
899
900
901
902
               genserie(datetime(2015, 1, 1), 'D', 11).astype('str'),
               'error1',
               'test')

    with pytest.raises(Exception) as excinfo:
903
        tsh.insert(engine,
904
905
906
907
908
909
                   genserie(datetime(2015, 1, 1), 'D', 11),
                   'error1',
                   'test')

    assert 'Type error when inserting error1, new type is float64, type in base is object' == str(excinfo.value)

910
    tsh.insert(engine,
911
912
913
914
915
               genserie(datetime(2015, 1, 1), 'D', 11),
               'error2',
               'test')

    with pytest.raises(Exception) as excinfo:
916
        tsh.insert(engine,
917
918
919
920
921
                   genserie(datetime(2015, 1, 1), 'D', 11).astype('str'),
                   'error2',
                   'test')

    assert 'Type error when inserting error2, new type is object, type in base is float64' == str(excinfo.value)