test_tsio.py 38.2 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.testutil import assert_group_equals, genserie, assert_df
14

15
DATADIR = Path(__file__).parent / 'data'
16

Aurélien Campéas's avatar
Aurélien Campéas committed
17

18
def test_changeset(engine, tsh):
19
    index = pd.date_range(start=datetime(2017, 1, 1), freq='D', periods=3)
20
    data = [1., 2., 3.]
21

22
23
    with patch('tshistory.tsio.datetime') as mock_date:
        mock_date.now.return_value = datetime(2020, 1, 1)
24
        with engine.connect() as cn:
25
            with tsh.newchangeset(cn, 'babar'):
26
                tsh.insert(cn, pd.Series(data, index=index), 'ts_values', author='WONTBEUSED')
27
                tsh.insert(cn, pd.Series(['a', 'b', 'c'], index=index), 'ts_othervalues')
28

29
30
31
        # bogus author won't show up
        assert tsh.log(engine)[0]['author'] == 'babar'

32
33
        g = tsh.get_group(engine, 'ts_values')
        g2 = tsh.get_group(engine, 'ts_othervalues')
34
        assert_group_equals(g, g2)
35

36
        with pytest.raises(AssertionError):
37
            tsh.insert(engine, pd.Series([2, 3, 4], index=index), 'ts_values')
38

39
        with engine.connect() as cn:
40
            data.append(data.pop(0))
41
42
            with tsh.newchangeset(cn, 'celeste'):
                tsh.insert(cn, pd.Series(data, index=index), 'ts_values')
43
                # below should be a noop
44
                tsh.insert(cn, pd.Series(['a', 'b', 'c'], index=index), 'ts_othervalues')
45

46
    g = tsh.get_group(engine, 'ts_values')
47
48
    assert ['ts_values'] == list(g.keys())

49
    assert_df("""
50
51
52
2017-01-01    2.0
2017-01-02    3.0
2017-01-03    1.0
53
""", tsh.get(engine, 'ts_values'))
54

55
    assert_df("""
56
57
58
2017-01-01    a
2017-01-02    b
2017-01-03    c
59
""", tsh.get(engine, 'ts_othervalues'))
60

61
    log = tsh.log(engine, names=['ts_values', 'ts_othervalues'])
62
63
64
65
    assert [
        {'author': 'babar',
         'rev': 1,
         'date': datetime(2020, 1, 1, 0, 0),
66
         'meta': {},
67
68
69
         'names': ['ts_values', 'ts_othervalues']},
        {'author': 'celeste',
         'rev': 2,
70
         'meta': {},
71
72
73
74
         'date': datetime(2020, 1, 1, 0, 0),
         'names': ['ts_values']}
    ] == log

75
    log = tsh.log(engine, names=['ts_othervalues'])
76
77
    assert len(log) == 1
    assert log[0]['rev'] == 1
78
    assert log[0]['names'] == ['ts_values', 'ts_othervalues']
79

80
    log = tsh.log(engine, fromrev=2)
81
82
    assert len(log) == 1

83
    log = tsh.log(engine, torev=1)
84
85
    assert len(log) == 1

86
    info = tsh.info(engine)
87
88
89
90
91
92
    assert {
        'changeset count': 2,
        'serie names': ['ts_othervalues', 'ts_values'],
        'series count': 2
    } == info

93

94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
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))
        df = pd.read_sql("select id, diff, snapshot from xserie order by id", cn)
        for attr in ('diff', 'snapshot'):
            df[attr] = df[attr].apply(lambda x: False if x is None else True)

    assert_df("""
id   diff  snapshot
0   1  False      True
1   2   True      True
""", 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
204
205
206
207
208
209
210
211
212
213

    # 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()


214
def test_differential(engine, tsh):
215
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 10)
216
    tsh.insert(engine, ts_begin, 'ts_test', 'test')
217

218
219
    assert tsh.exists(engine, 'ts_test')
    assert not tsh.exists(engine, 'this_does_not_exist')
220

221
    assert_df("""
222
223
224
225
226
227
228
229
230
231
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
232
""", tsh.get(engine, 'ts_test'))
233
234

    # we should detect the emission of a message
235
    tsh.insert(engine, ts_begin, 'ts_test', 'babar')
236

237
    assert_df("""
238
239
240
241
242
243
244
245
246
247
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
248
""", tsh.get(engine, 'ts_test'))
249
250
251
252

    ts_slight_variation = ts_begin.copy()
    ts_slight_variation.iloc[3] = 0
    ts_slight_variation.iloc[6] = 0
253
    tsh.insert(engine, ts_slight_variation, 'ts_test', 'celeste')
254

255
    assert_df("""
256
257
258
259
260
261
262
263
264
265
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
266
""", tsh.get(engine, 'ts_test'))
267

268
    ts_longer = genserie(datetime(2010, 1, 3), 'D', 15)
269
270
271
272
    ts_longer.iloc[1] = 2.48
    ts_longer.iloc[3] = 3.14
    ts_longer.iloc[5] = ts_begin.iloc[7]

273
    tsh.insert(engine, ts_longer, 'ts_test', 'test')
274

275
    assert_df("""
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
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
293
""", tsh.get(engine, 'ts_test'))
294
295

    # start testing manual overrides
296
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 5, initval=[2])
297
    ts_begin.loc['2010-01-04'] = -1
298
    tsh.insert(engine, ts_begin, 'ts_mixte', 'test')
299
300

    # -1 represents bogus upstream data
301
    assert_df("""
302
303
304
305
306
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
307
""", tsh.get(engine, 'ts_mixte'))
308
309

    # refresh all the period + 1 extra data point
310
    ts_more = genserie(datetime(2010, 1, 2), 'D', 5, [2])
311
    ts_more.loc['2010-01-04'] = -1
312
    tsh.insert(engine, ts_more, 'ts_mixte', 'test')
313

314
    assert_df("""
315
316
317
318
319
320
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
321
""", tsh.get(engine, 'ts_mixte'))
322
323

    # just append an extra data point
324
325
    # with no intersection with the previous ts
    ts_one_more = genserie(datetime(2010, 1, 7), 'D', 1, [3])
326
    tsh.insert(engine, ts_one_more, 'ts_mixte', 'test')
327

328
    assert_df("""
329
330
331
332
333
334
335
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
336
""", tsh.get(engine, 'ts_mixte'))
337

338
    with engine.connect() as cn:
339
        cn.execute('set search_path to "{0}.timeserie", {0}, public'.format(tsh.namespace))
340
341
342
        hist = pd.read_sql('select id, parent from ts_test order by id',
                           cn)
        assert_df("""
343
344
345
346
   id  parent
0   1     NaN
1   2     1.0
2   3     2.0
347
""", hist)
348

349
350
351
        hist = pd.read_sql('select id, parent from ts_mixte order by id',
                           cn)
        assert_df("""
352
353
354
355
   id  parent
0   1     NaN
1   2     1.0
2   3     2.0
356
""", hist)
357

358
359
360
        allts = pd.read_sql("select name, table_name from registry "
                            "where name in ('ts_test', 'ts_mixte')",
                            cn)
361

362
363
        assert_df("""
name              table_name
364
365
366
0   ts_test   {0}.timeserie.ts_test
1  ts_mixte  {0}.timeserie.ts_mixte
""".format(tsh.namespace), allts)
367

368
        assert_df("""
369
370
371
372
373
374
375
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
376
""", tsh.get(cn, 'ts_mixte',
377
             revision_date=datetime.now()))
378
379


380
def test_bad_import(engine, tsh):
381
    # the data were parsed as date by pd.read_json()
Aurélien Campéas's avatar
Aurélien Campéas committed
382
    df_result = pd.read_csv(str(DATADIR / 'test_data.csv'))
383
384
385
    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']
386

387
388
    tsh.insert(engine, ts, 'SND_SC', 'test')
    result = tsh.get(engine, 'SND_SC')
389
    assert result.dtype == 'float64'
390
391
392

    # insertion of empty ts
    ts = pd.Series(name='truc', dtype='object')
393
394
    tsh.insert(engine, ts, 'empty_ts', 'test')
    assert tsh.get(engine, 'empty_ts') is None
395
396
397

    # nan in ts
    # all na
398
    ts = genserie(datetime(2010, 1, 10), 'D', 10, [np.nan], name='truc')
399
400
    tsh.insert(engine, ts, 'test_nan', 'test')
    assert tsh.get(engine, 'test_nan') is None
401
402
403
404
405

    # 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')
406
407
    tsh.insert(engine, ts, 'test_nan', 'test')
    result = tsh.get(engine, 'test_nan')
408

409
410
    tsh.insert(engine, ts, 'test_nan', 'test')
    result = tsh.get(engine, 'test_nan')
411
    assert_df("""
412
413
414
415
416
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
417
""", result)
418
419
420

    # get_ts with name not in database

421
    tsh.get(engine, 'inexisting_name', 'test')
422
423


424
def test_revision_date(engine, tsh):
425
426
427
428
429
430
431
    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')

432
    idate1 = datetime(2015, 1, 1, 15, 43, 23)
433
    with tsh.newchangeset(engine, 'test', _insertion_date=idate1):
434

435
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [1], name='truc')
436
437
        tsh.insert(engine, ts, 'ts_through_time')
        assert idate1 == tsh.latest_insertion_date(engine, 'ts_through_time')
438

439
    idate2 = datetime(2015, 1, 2, 15, 43, 23)
440
    with tsh.newchangeset(engine, 'test', _insertion_date=idate2):
441

442
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [2], name='truc')
443
444
        tsh.insert(engine, ts, 'ts_through_time')
        assert idate2 == tsh.latest_insertion_date(engine, 'ts_through_time')
445

446
    idate3 = datetime(2015, 1, 3, 15, 43, 23)
447
    with tsh.newchangeset(engine, 'test', _insertion_date=idate3):
448

449
        ts = genserie(datetime(2010, 1, 4), 'D', 4, [3], name='truc')
450
451
        tsh.insert(engine, ts, 'ts_through_time')
        assert idate3 == tsh.latest_insertion_date(engine, 'ts_through_time')
452

453
    ts = tsh.get(engine, 'ts_through_time')
454

455
    assert_df("""
456
457
458
459
2010-01-04    3.0
2010-01-05    3.0
2010-01-06    3.0
2010-01-07    3.0
460
""", ts)
461

462
    ts = tsh.get(engine, 'ts_through_time',
Aurélien Campéas's avatar
Aurélien Campéas committed
463
                 revision_date=datetime(2015, 1, 2, 18, 43, 23))
464

465
    assert_df("""
466
467
468
469
2010-01-04    2.0
2010-01-05    2.0
2010-01-06    2.0
2010-01-07    2.0
470
""", ts)
471

472
    ts = tsh.get(engine, 'ts_through_time',
473
                 revision_date=datetime(2015, 1, 1, 18, 43, 23))
474

475
    assert_df("""
476
477
478
479
2010-01-04    1.0
2010-01-05    1.0
2010-01-06    1.0
2010-01-07    1.0
480
""", ts)
481

482
    ts = tsh.get(engine, 'ts_through_time',
483
                 revision_date=datetime(2014, 1, 1, 18, 43, 23))
484
485
486

    assert ts is None

487

488
def test_snapshots(engine, tsh):
489
    baseinterval = tsh._snapshot_interval
490
    tsh._snapshot_interval = 4
491

492
    with engine.connect() as cn:
493
        for tscount in range(1, 11):
494
            ts = genserie(datetime(2015, 1, 1), 'D', tscount, [1])
495
            diff = tsh.insert(cn, ts, 'growing', 'babar')
496
497
            assert diff.index[0] == diff.index[-1] == ts.index[-1]

498
    diff = tsh.insert(engine, ts, 'growing', 'babar')
499
    assert diff is None
500

501
    with engine.connect() as cn:
502
        cn.execute('set search_path to "{}.timeserie"'.format(tsh.namespace))
503
504
505
        df = pd.read_sql("select id from growing where snapshot is not null",
                         cn)
        assert_df("""
506
507
   id
0   1
508
509
510
1   4
2   8
3  10
511
""", df)
512

513
514
        ts = tsh.get(cn, 'growing')
        assert_df("""
515
516
517
518
519
520
521
522
523
524
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
525
""", ts)
526

527
528
529
        df = pd.read_sql("select id, diff, snapshot from growing order by id", cn)
        for attr in ('diff', 'snapshot'):
            df[attr] = df[attr].apply(lambda x: 0 if x is None else len(x))
530

531
        assert_df("""
532
533
534
535
536
537
538
539
540
541
542
id  diff  snapshot
0   1     0        35
1   2    36         0
2   3    36         0
3   4    36        47
4   5    36         0
5   6    36         0
6   7    36         0
7   8    36        59
8   9    36         0
9  10    36        67
543
""", df)
544

545
546
    table = tsh._get_ts_table(engine, 'growing')
    snapid, snap = tsh._find_snapshot(engine, table, ())
547
548
    assert snapid == 10
    assert (ts == snap).all()
549
    tsh._snapshot_interval = baseinterval
550
551


552
def test_deletion(engine, tsh):
553
554
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 11)
    ts_begin.iloc[-1] = np.nan
555
    tsh.insert(engine, ts_begin, 'ts_del', 'test')
556

557
    ts = tsh._build_snapshot_upto(engine, tsh._get_ts_table(engine, 'ts_del'))
558
    assert ts.iloc[-1] == 9.0
559

560
    ts_begin.iloc[0] = np.nan
561
    ts_begin.iloc[3] = np.nan
562

563
    tsh.insert(engine, ts_begin, 'ts_del', 'test')
564

565
    assert_df("""
566
567
568
569
570
571
572
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
573
2010-01-10    9.0
574
""", tsh.get(engine, 'ts_del'))
575

576
    ts2 = tsh.get(engine, 'ts_del',
577
578
                 # force snapshot reconstruction feature
                 revision_date=datetime(2038, 1, 1))
579
    assert (tsh.get(engine, 'ts_del') == ts2).all()
580

581
582
583
    ts_begin.iloc[0] = 42
    ts_begin.iloc[3] = 23

584
    tsh.insert(engine, ts_begin, 'ts_del', 'test')
585

586
    assert_df("""
587
588
589
590
591
592
593
594
595
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
596
2010-01-10     9.0
597
""", tsh.get(engine, 'ts_del'))
598
599
600

    # now with string!

601
    ts_string = genserie(datetime(2010, 1, 1), 'D', 10, ['machin'])
602
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
603
604
605
606

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

607
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
608
    assert_df("""
609
610
611
612
613
614
615
616
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
617
""", tsh.get(engine, 'ts_string_del'))
618
619
620
621

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

622
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
623
    assert_df("""
624
625
626
627
628
629
630
631
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
632
2010-01-10    machin
633
""", tsh.get(engine, 'ts_string_del'))
634

635
    ts_string[ts_string.index] = np.nan
636
    tsh.insert(engine, ts_string, 'ts_string_del', 'test')
637

638
    erased = tsh.get(engine, 'ts_string_del')
639
640
    assert len(erased) == 0

641
642
    # first insertion with only nan

643
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 10, [np.nan])
644
    tsh.insert(engine, ts_begin, 'ts_null', 'test')
645

646
    assert tsh.get(engine, 'ts_null') is None
647

648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
    # 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)

667
668
    tsh.insert(engine, ts_repushed, 'ts_repushed', 'test')
    diff = tsh.insert(engine, ts_repushed, 'ts_repushed', 'test')
669
670
    assert diff is None

671
    # there is no difference
672
    assert 0 == len(tsh._compute_diff(ts_repushed, ts_repushed))
673
674
675
676
677

    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
678
    diff = tsh._compute_diff(ts_repushed, ts_add)
679
680
681
682
683
684

    assert_df("""
2010-01-02     1.0
2010-01-03     2.0
2010-01-09     NaN
2010-01-12    11.0
685
2010-01-13    12.0""", diff.sort_index())
686
687
688
689
    # value on nan => value
    # nan on value => nan
    # nan on nan => Nothing
    # nan on nothing=> Nothing
690

Aurélien Campéas's avatar
Aurélien Campéas committed
691
    # full erasing
692
693
    # numeric
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 4)
694
    tsh.insert(engine, ts_begin, 'ts_full_del', 'test')
695

Aurélien Campéas's avatar
Aurélien Campéas committed
696
    ts_begin.iloc[:] = np.nan
697
    tsh.insert(engine, ts_begin, 'ts_full_del', 'test')
698
699

    ts_end = genserie(datetime(2010, 1, 1), 'D', 4)
700
    tsh.insert(engine, ts_end, 'ts_full_del', 'test')
701
702
703
704

    # string

    ts_begin = genserie(datetime(2010, 1, 1), 'D', 4, ['text'])
705
    tsh.insert(engine, ts_begin, 'ts_full_del_str', 'test')
706
707

    ts_begin.iloc[:] = np.nan
708
    tsh.insert(engine, ts_begin, 'ts_full_del_str', 'test')
709
710

    ts_end = genserie(datetime(2010, 1, 1), 'D', 4, ['text'])
711
    tsh.insert(engine, ts_end, 'ts_full_del_str', 'test')
712

Aurélien Campéas's avatar
Aurélien Campéas committed
713

714
def test_multi_index(engine, tsh):
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
    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)

730
    tsh.insert(engine, ts_multi, 'ts_multi_simple', 'test')
731

732
    ts = tsh.get(engine, 'ts_multi_simple')
733
734
735
    assert_df("""
                                                    ts_multi_simple
a                   b          c                                   
736
737
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
738
739
""", pd.DataFrame(ts))

740
    diff = tsh.insert(engine, ts_multi, 'ts_multi_simple', 'test')
741
742
743
744
745
    assert diff is None

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

746
747
    tsh.insert(engine, ts_multi_2, 'ts_multi_simple', 'test')
    ts = tsh.get(engine, 'ts_multi_simple')
748
749
750
751

    assert_df("""
                                                    ts_multi_simple
a                   b          c                                   
752
753
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
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
""", 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)

779
780
    tsh.insert(engine, ts_multi, 'ts_multi', 'test')
    ts = tsh.get(engine, 'ts_multi')
781
782
783
784

    assert_df("""
                                                    ts_multi
a          b                   c                            
785
786
787
788
789
790
791
792
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
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
    """, 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.

815
    diff = tsh.insert(engine, ts_multi_2, 'ts_multi', 'test')
816
817
818
819
820
821
822
823
824
825
826
827
828
829
    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

830
    ts = tsh.get(engine, 'ts_multi')
831
832
833
    assert_df("""
                                                    ts_multi
a          b                   c                            
834
835
836
837
838
839
840
841
842
843
844
845
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
846
847
848
        """, pd.DataFrame(ts.sort_index()))

    # the result ts have now 3 values for each point in 'a'
849
850


851
def test_get_history(engine, tsh):
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
    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',
868
         'meta': {},
869
870
871
872
         'date': datetime(2017, 2, 1, 0, 0),
         'names': ['smallserie']
        },
        {'author': 'aurelien.campeas@pythonian.fr',
873
         'meta': {},
874
875
876
877
         'date': datetime(2017, 2, 2, 0, 0),
         'names': ['smallserie']
        },
        {'author': 'aurelien.campeas@pythonian.fr',
878
         'meta': {},
879
880
881
882
883
884
         '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')
885
    assert histts.name == 'smallserie'
886
887
888
889
890
891
892
893
894
895
896

    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)

897
898
899
900
901
902
903
904
    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)

905
    for idate in histts.index.get_level_values('insertion_date').unique():
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
        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)

945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
    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, 31))
    assert tsc is None

965
966
967
968
969
970
971
972
973
974
975
976
977
    # restrictions on value dates
    tsc = tsh.get_history(engine, 'smallserie',
                          from_value_date=datetime(2017, 1, 1),
                          to_value_date=datetime(2017, 1, 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
2017-02-03      2017-01-01    0.0
                2017-01-02    1.0
""", tsc)

978
979
980
981
982
983
984
985
986
987
    diffs = tsh.get_history(engine, 'smallserie',
                            diffmode=True,
                            from_value_date=datetime(2017, 1, 1),
                            to_value_date=datetime(2017, 1, 2))
    assert_df("""
insertion_date  value_date
2017-02-01      2017-01-01    0.0
2017-02-02      2017-01-02    1.0
""", diffs)

988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
    tsc = tsh.get_history(engine, 'smallserie',
                          from_value_date=datetime(2017, 1, 2))
    assert_df("""
insertion_date  value_date
2017-02-02      2017-01-02    1.0
2017-02-03      2017-01-02    1.0
                2017-01-03    2.0
""", tsc)

    tsc = tsh.get_history(engine, 'smallserie',
                          to_value_date=datetime(2017, 1, 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
2017-02-03      2017-01-01    0.0
                2017-01-02    1.0
""", tsc)

1008

1009
def test_add_na(engine, tsh):
1010
1011
1012
1013
1014
    # 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

1015
    diff = tsh.insert(engine, ts_nan, 'ts_add_na', 'test')
1016
    assert diff is None
1017
    result = tsh.get(engine, 'ts_add_na')
1018
1019
1020
1021
    assert result is None

    # in case of insertion in existing data
    ts_begin = genserie(datetime(2010, 1, 1), 'D', 5)
1022
    tsh.insert(engine, ts_begin, 'ts_add_na', 'test')
1023
1024
1025
1026
1027

    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])

1028
    diff = tsh.insert(engine, ts_nan, 'ts_add_na', 'test')
1029
1030
    assert diff is None

1031
    result = tsh.get(engine, 'ts_add_na')
1032
    assert len(result) == 5
1033
1034


1035
def test_dtype_mismatch(engine, tsh):
1036
    tsh.insert(engine,
1037
1038
1039
1040
1041
               genserie(datetime(2015, 1, 1), 'D', 11).astype('str'),
               'error1',
               'test')

    with pytest.raises(Exception) as excinfo:
1042
        tsh.insert(engine,
1043
1044
1045
1046
1047
1048
                   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)

1049
    tsh.insert(engine,
1050
1051
1052
1053
1054
               genserie(datetime(2015, 1, 1), 'D', 11),
               'error2',
               'test')

    with pytest.raises(Exception) as excinfo:
1055
        tsh.insert(engine,
1056
1057
1058
1059
1060
                   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)
1061
1062


1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
def test_precision(engine, tsh):

    floaty = 0.123456789123456789
    ts = genserie(datetime(2015, 1, 1), 'D', 5, initval=[floaty])

    tsh.insert(engine, ts, 'precision', 'test')
    ts_round = tsh.get(engine, 'precision')
    assert 0.12345678912346 == ts_round.iloc[0]

    diff = tsh.insert(engine, ts_round, 'precision', 'test')
    assert diff is None # the roundtriped series does not produce a diff when reinserted

    diff = tsh.insert(engine, ts, 'precision', 'test') # neither does the original series
    assert diff is None


1079
@pytest.mark.perf
1080
1081
def test_bigdata(engine, tracker, ptsh):
    tsh = ptsh
1082
1083
    def create_data():
        for year in range(2015, 2020):