test_tsio.py 24.6 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))

39
def test_changeset(engine):
40
41
42
43
    # instantiate one time serie handler object
    tso = TimeSerie()

    index = pd.date_range(start=datetime(2017, 1, 1), freq='D', periods=3)
44
    data = [1., 2., 3.]
45

46
47
    with patch('tshistory.tsio.datetime') as mock_date:
        mock_date.now.return_value = datetime(2020, 1, 1)
48
49
50
51
        with engine.connect() as cn:
            with tso.newchangeset(cn, 'babar'):
                tso.insert(cn, pd.Series(data, index=index), 'ts_values')
                tso.insert(cn, pd.Series(['a', 'b', 'c'], index=index), 'ts_othervalues')
52

53
54
55
        g = tso.get_group(engine, 'ts_values')
        g2 = tso.get_group(engine, 'ts_othervalues')
        assert_group_equals(g, g2)
56

57
58
        with pytest.raises(AssertionError):
            tso.insert(engine, pd.Series([2,3,4], index=index), 'ts_values')
59

60
        with engine.connect() as cn:
61
            data.append(data.pop(0))
62
63
            with tso.newchangeset(cn, 'celeste'):
                tso.insert(cn, pd.Series(data, index=index), 'ts_values')
64
                # below should be a noop
65
                tso.insert(cn, pd.Series(['a', 'b', 'c'], index=index), 'ts_othervalues')
66

67
68
69
    g = tso.get_group(engine, 'ts_values')
    assert ['ts_values'] == list(g.keys())

70
    assert_df("""
71
72
73
2017-01-01    2.0
2017-01-02    3.0
2017-01-03    1.0
74
""", tso.get(engine, 'ts_values'))
75

76
    assert_df("""
77
78
79
2017-01-01    a
2017-01-02    b
2017-01-03    c
80
""", tso.get(engine, 'ts_othervalues'))
81

82
83
84
85
86
87
88
89
90
91
92
93
    log = tso.log(engine)
    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

94
95
96
    log = tso.log(engine, names=['ts_othervalues'])
    assert len(log) == 1
    assert log[0]['rev'] == 1
97
    assert log[0]['names'] == ['ts_values', 'ts_othervalues']
98

99
100
101
102
103
104
    log = tso.log(engine, fromrev=2)
    assert len(log) == 1

    log = tso.log(engine, torev=1)
    assert len(log) == 1

105
106
107
108
109
110
111
    info = tso.info(engine)
    assert {
        'changeset count': 2,
        'serie names': ['ts_othervalues', 'ts_values'],
        'series count': 2
    } == info

112

113
114
def test_tstamp_roundtrip(engine):
    tso = TimeSerie()
115
116
    ts = genserie(datetime(2017, 10, 28, 23),
                  'H', 4, tz='UTC')
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
    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)

    tso.insert(engine, ts, 'tztest', 'Babar')
    back = tso.get(engine, 'tztest')

    # 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()


142
def test_differential(engine):
143
144
    # instantiate one time serie handler object
    tso = TimeSerie()
145

146
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 10)
147
    tso.insert(engine, ts_begin, 'ts_test', 'test')
148

149
150
151
    assert tso.exists(engine, 'ts_test')
    assert not tso.exists(engine, 'this_does_not_exist')

152
    assert_df("""
153
154
155
156
157
158
159
160
161
162
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
163
""", tso.get(engine, 'ts_test'))
164
165

    # we should detect the emission of a message
166
    tso.insert(engine, ts_begin, 'ts_test', 'babar')
167

168
    assert_df("""
169
170
171
172
173
174
175
176
177
178
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
179
""", tso.get(engine, 'ts_test'))
180
181
182
183

    ts_slight_variation = ts_begin.copy()
    ts_slight_variation.iloc[3] = 0
    ts_slight_variation.iloc[6] = 0
184
    tso.insert(engine, ts_slight_variation, 'ts_test', 'celeste')
185

186
    assert_df("""
187
188
189
190
191
192
193
194
195
196
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
197
""", tso.get(engine, 'ts_test'))
198

199
    ts_longer = genserie(datetime(2010, 1, 3), 'D', 15)
200
201
202
203
    ts_longer.iloc[1] = 2.48
    ts_longer.iloc[3] = 3.14
    ts_longer.iloc[5] = ts_begin.iloc[7]

204
    tso.insert(engine, ts_longer, 'ts_test', 'test')
205

206
    assert_df("""
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
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
224
""", tso.get(engine, 'ts_test'))
225
226

    # start testing manual overrides
227
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 5, initval=[2])
228
    ts_begin.loc['2010-01-04'] = -1
229
    tso.insert(engine, ts_begin, 'ts_mixte', 'test')
230
231

    # -1 represents bogus upstream data
232
    assert_df("""
233
234
235
236
237
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
238
""", tso.get(engine, 'ts_mixte'))
239
240

    # refresh all the period + 1 extra data point
241
    ts_more = genserie(datetime(2010, 1, 2), 'D', 5, [2])
242
    ts_more.loc['2010-01-04'] = -1
243
    tso.insert(engine, ts_more, 'ts_mixte', 'test')
244

245
    assert_df("""
246
247
248
249
250
251
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
252
""", tso.get(engine, 'ts_mixte'))
253
254

    # just append an extra data point
255
256
    # with no intersection with the previous ts
    ts_one_more = genserie(datetime(2010, 1, 7), 'D', 1, [3])
257
    tso.insert(engine, ts_one_more, 'ts_mixte', 'test')
258

259
    assert_df("""
260
261
262
263
264
265
266
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
267
""", tso.get(engine, 'ts_mixte'))
268

269
    hist = pd.read_sql('select id, parent from timeserie.ts_test order by id',
270
                        engine)
271
    assert_df("""
272
273
274
275
   id  parent
0   1     NaN
1   2     1.0
2   3     2.0
276
""", hist)
277

278
    hist = pd.read_sql('select id, parent from timeserie.ts_mixte order by id',
279
                        engine)
280
    assert_df("""
281
282
283
284
   id  parent
0   1     NaN
1   2     1.0
2   3     2.0
285
""", hist)
286

287
    allts = pd.read_sql("select name, table_name from registry "
288
                        "where name in ('ts_test', 'ts_mixte')",
289
290
                        engine)

291
    assert_df("""
292
293
294
       name          table_name
0   ts_test   timeserie.ts_test
1  ts_mixte  timeserie.ts_mixte
295
""", allts)
296

297
    assert_df("""
298
299
300
301
302
303
304
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
305
306
""", tso.get(engine, 'ts_mixte',
             revision_date=datetime.now()))
307
308
309


def test_bad_import(engine):
310
311
312
    # instantiate one time serie handler object
    tso = TimeSerie()

313
314
315
316
317
    # 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']
318
319
320

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

    # insertion of empty ts
    ts = pd.Series(name='truc', dtype='object')
325
326
    tso.insert(engine, ts, 'empty_ts', 'test')
    assert tso.get(engine, 'empty_ts') is None
327
328
329

    # nan in ts
    # all na
330
    ts = genserie(datetime(2010, 1, 10), 'D', 10, [np.nan], name='truc')
331
332
    tso.insert(engine, ts, 'test_nan', 'test')
    assert tso.get(engine, 'test_nan') is None
333
334
335
336
337

    # 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')
338
339
    tso.insert(engine, ts, 'test_nan', 'test')
    result = tso.get(engine, 'test_nan')
340

341
342
    tso.insert(engine, ts, 'test_nan', 'test')
    result = tso.get(engine, 'test_nan')
343
    assert_df("""
344
345
346
347
348
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
349
""", result)
350
351
352

    # get_ts with name not in database

353
    tso.get(engine, 'inexisting_name', 'test')
354
355
356


def test_revision_date(engine):
357
358
359
    # instantiate one time serie handler object
    tso = TimeSerie()

360
361
    idate1 = datetime(2015, 1, 1, 15, 43, 23)
    with tso.newchangeset(engine, 'test', _insertion_date=idate1):
362

363
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [1], name='truc')
364
365
        tso.insert(engine, ts, 'ts_through_time')
        assert idate1 == tso.latest_insertion_date(engine, 'ts_through_time')
366

367
368
    idate2 = datetime(2015, 1, 2, 15, 43, 23)
    with tso.newchangeset(engine, 'test', _insertion_date=idate2):
369

370
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [2], name='truc')
371
372
        tso.insert(engine, ts, 'ts_through_time')
        assert idate2 == tso.latest_insertion_date(engine, 'ts_through_time')
373

374
375
    idate3 = datetime(2015, 1, 3, 15, 43, 23)
    with tso.newchangeset(engine, 'test', _insertion_date=idate3):
376

377
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [3], name='truc')
378
379
        tso.insert(engine, ts, 'ts_through_time')
        assert idate3 == tso.latest_insertion_date(engine, 'ts_through_time')
380

381
    ts = tso.get(engine, 'ts_through_time')
382

383
    assert_df("""
384
385
386
387
2010-01-04    3.0
2010-01-05    3.0
2010-01-06    3.0
2010-01-07    3.0
388
""", ts)
389

390
391
    ts = tso.get(engine, 'ts_through_time',
                 revision_date=datetime(2015, 1, 2, 18, 43, 23) )
392

393
    assert_df("""
394
395
396
397
2010-01-04    2.0
2010-01-05    2.0
2010-01-06    2.0
2010-01-07    2.0
398
""", ts)
399

400
401
    ts = tso.get(engine, 'ts_through_time',
                 revision_date=datetime(2015, 1, 1, 18, 43, 23))
402

403
    assert_df("""
404
405
406
407
2010-01-04    1.0
2010-01-05    1.0
2010-01-06    1.0
2010-01-07    1.0
408
""", ts)
409

410
411
    ts = tso.get(engine, 'ts_through_time',
                 revision_date=datetime(2014, 1, 1, 18, 43, 23))
412
413
414

    assert ts is None

415
416
417

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

420
    with engine.connect() as cn:
421
        for tscount in range(1, 11):
422
            ts = genserie(datetime(2015, 1, 1), 'D', tscount, [1])
423
            diff = tso.insert(cn, ts, 'growing', 'babar')
424
425
426
427
            assert diff.index[0] == diff.index[-1] == ts.index[-1]

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

429
430
    df = pd.read_sql("select id from timeserie.growing where snapshot is not null",
                     engine)
431
    assert_df("""
432
433
   id
0   1
434
435
436
1   4
2   8
3  10
437
""", df)
438
439

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

453
    df = pd.read_sql("select id, diff, snapshot from timeserie.growing order by id", engine)
454
455
456
    for attr in ('diff', 'snapshot'):
        df[attr] = df[attr].apply(lambda x: 0 if x is None else len(x))

457
    assert_df("""
458
   id  diff  snapshot
Arnaud Campeas's avatar
Arnaud Campeas committed
459
460
461
462
463
464
465
466
467
468
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
469
""", df)
470
471
472
473
474

    table = tso._get_ts_table(engine, 'growing')
    snapid, snap = tso._find_snapshot(engine, table, ())
    assert snapid == 10
    assert (ts == snap).all()
475
476
477
478
479


def test_deletion(engine):
    tso = TimeSerie()

480
481
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 11)
    ts_begin.iloc[-1] = np.nan
482
483
    tso.insert(engine, ts_begin, 'ts_del', 'test')

484
    ts = tso._build_snapshot_upto(engine, tso._get_ts_table(engine, 'ts_del'))
485
    assert ts.iloc[-1] == 9.0
486

487
    ts_begin.iloc[0] = np.nan
488
    ts_begin.iloc[3] = np.nan
489
490
491

    tso.insert(engine, ts_begin, 'ts_del', 'test')

492
    assert_df("""
493
494
495
496
497
498
499
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
500
501
2010-01-10    9.0
""", tso.get(engine, 'ts_del'))
502

503
504
505
506
507
    ts2 = tso.get(engine, 'ts_del',
                 # force snapshot reconstruction feature
                 revision_date=datetime(2038, 1, 1))
    assert (tso.get(engine, 'ts_del') == ts2).all()

508
509
510
511
512
    ts_begin.iloc[0] = 42
    ts_begin.iloc[3] = 23

    tso.insert(engine, ts_begin, 'ts_del', 'test')

513
    assert_df("""
514
515
516
517
518
519
520
521
522
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
523
524
2010-01-10     9.0
""", tso.get(engine, 'ts_del'))
525
526
527

    # now with string!

528
    ts_string = genserie(datetime(2010, 1, 1), 'D', 10, ['machin'])
529
530
531
532
533
534
    tso.insert(engine, ts_string, 'ts_string_del', 'test')

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

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

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

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

562
563
564
565
566
567
    ts_string[ts_string.index] = np.nan
    tso.insert(engine, ts_string, 'ts_string_del', 'test')

    erased = tso.get(engine, 'ts_string_del')
    assert len(erased) == 0

568
569
    # first insertion with only nan

570
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 10, [np.nan])
571
572
573
    tso.insert(engine, ts_begin, 'ts_null', 'test')

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

575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
    # 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)

    tso.insert(engine, ts_repushed, 'ts_repushed', 'test')
    diff = tso.insert(engine, ts_repushed, 'ts_repushed', 'test')
    assert diff is None

598
599
    # there is no difference
    assert 0 == len(tso._compute_diff(ts_repushed, ts_repushed))
600
601
602
603
604
605
606
607
608
609
610
611
612

    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
    diff = tso._compute_diff(ts_repushed, ts_add)

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

618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
    ## full erasing
    # numeric
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 4)
    tso.insert(engine, ts_begin, 'ts_full_del', 'test')

    ts_begin.iloc[:]= np.nan
    tso.insert(engine, ts_begin, 'ts_full_del', 'test')

    ts_end = genserie(datetime(2010, 1, 1), 'D', 4)
    tso.insert(engine, ts_end, 'ts_full_del', 'test')

    # string

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

    ts_begin.iloc[:] = np.nan
    tso.insert(engine, ts_begin, 'ts_full_del_str', 'test')

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

def test_multi_index(engine):
    tso = TimeSerie()

    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)

    tso.insert(engine, ts_multi, 'ts_multi_simple', 'test')

    ts = tso.get(engine, 'ts_multi_simple')
    assert_df("""
                                                    ts_multi_simple
a                   b          c                                   
664
665
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
666
667
668
669
670
671
672
673
674
675
676
677
678
679
""", pd.DataFrame(ts))

    diff = tso.insert(engine, ts_multi, 'ts_multi_simple', 'test')
    assert diff is None

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

    tso.insert(engine, ts_multi_2, 'ts_multi_simple', 'test')
    ts = tso.get(engine, 'ts_multi_simple')

    assert_df("""
                                                    ts_multi_simple
a                   b          c                                   
680
681
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
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
""", 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)

    tso.insert(engine, ts_multi, 'ts_multi', 'test')
    ts = tso.get(engine, 'ts_multi')

    assert_df("""
                                                    ts_multi
a          b                   c                            
713
714
715
716
717
718
719
720
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
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
    """, 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.

    diff = tso.insert(engine, ts_multi_2, 'ts_multi', 'test')
    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

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

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


def test_add_na(engine):
    tso = TimeSerie()

    # 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

    diff = tso.insert(engine, ts_nan, 'ts_add_na', 'test')
    assert diff is None
    result = tso.get(engine, 'ts_add_na')
    assert result is None

    # in case of insertion in existing data
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 5)
    tso.insert(engine, ts_begin, 'ts_add_na', 'test')

    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])

    diff = tso.insert(engine, ts_nan, 'ts_add_na', 'test')
    assert diff is None

    result = tso.get(engine, 'ts_add_na')
    assert len(result) == 5
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


def test_dtype_mismatch(engine):
    tso = TimeSerie()

    tso.insert(engine,
               genserie(datetime(2015, 1, 1), 'D', 11).astype('str'),
               'error1',
               'test')

    with pytest.raises(Exception) as excinfo:
        tso.insert(engine,
                   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)

    tso.insert(engine,
               genserie(datetime(2015, 1, 1), 'D', 11),
               'error2',
               'test')

    with pytest.raises(Exception) as excinfo:
        tso.insert(engine,
                   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)