def test_incremental_df_right_closed(self):
length = 100
data = [x for x in range(length)]
df = _make_df_with_one_col(data)
print(df.shape)
params = dku_timeseries.WindowAggregatorParams(window_width=3,
closed_option='right',
window_type='gaussian')
window_aggregator = dku_timeseries.WindowAggregator(params)
output_df = window_aggregator.compute(df, TIME_COL)
ground_truth = [0, 0, 0, 1, 2, 3, 4, 5, 6, 7, 8]
for x, y in zip(output_df[DATA_COL + '_min'][1:], ground_truth[1:]):
assert output_df[DATA_COL][x] == y
def test_extrema_without_neighbors(self):
length = 100
data = [x for x in range(length)]
df = _make_df_with_one_col(data)
window_aggregator = dku_timeseries.WindowAggregator(
dku_timeseries.WindowAggregatorParams(window_unit='milliseconds'))
params = dku_timeseries.ExtremaExtractorParams(
window_aggregator=window_aggregator)
extrema_extractor = dku_timeseries.ExtremaExtractor(params)
output_df = extrema_extractor.compute(df, TIME_COL, DATA_COL)
# only have DATE_TIME col and DATA_COL of the extrema, no stats because no neighbors
assert output_df.shape == (1, 2)
assert output_df[DATA_COL][0] == 99
def test_group_window_time_unit(self):
start_time_1 = pd.Timestamp('20190131 01:59:00').tz_localize('CET')
start_time_2 = pd.Timestamp('20190131 02:00:00').tz_localize('CET')
start_time_list = [start_time_1, start_time_2]
len1 = 100
len2 = 10
data1 = range(len1)
data2 = range(len2)
data_list = [data1, data2]
period1 = pd.DateOffset(seconds=1)
period2 = pd.DateOffset(seconds=1)
period_list = [period1, period2]
df_list = []
for group_id, data, period, start_time in zip(range(len(data_list)),
data_list, period_list,
start_time_list):
group_name = 'group_{}'.format(group_id)
temp_df = _make_df_with_one_col(data,
period=period,
start_time=start_time)
temp_df.loc[:, GROUP_COL] = group_name
df_list.append(temp_df)
df = pd.concat(df_list, axis=0)
params = dku_timeseries.WindowAggregatorParams(window_width=3,
closed_option='left',
window_type=None)
window_aggregator = dku_timeseries.WindowAggregator(params)
output_df = window_aggregator.compute(df,
datetime_column=TIME_COL,
groupby_columns=[GROUP_COL])
ground_truth = [np.NaN, 0, 0, 0, 1, 2, 3, 4, 5, 6]
output_0 = output_df.groupby(GROUP_COL).get_group(
'group_0').data_col_min.values[:10]
assert math.isnan(output_0[0])
assert np.array_equal(output_0[1:], ground_truth[1:])
output_1 = output_df.groupby(GROUP_COL).get_group(
'group_1').data_col_min.values[:10]
assert math.isnan(output_1[0])
assert np.array_equal(output_1[1:], ground_truth[1:])
def test_group_extrema_without_neighbors(self):
start_time_1 = pd.Timestamp('20190131 01:59:00').tz_localize('CET')
start_time_2 = pd.Timestamp('20190131 02:00:00').tz_localize('CET')
start_time_list = [start_time_1, start_time_2]
len1 = 100
len2 = 10
data1 = range(len1)
data2 = range(len2)
data_list = [data1, data2]
period1 = pd.DateOffset(seconds=1)
period2 = pd.DateOffset(seconds=1)
period_list = [period1, period2]
df_list = []
for group_id, data, period, start_time in zip(range(len(data_list)),
data_list, period_list,
start_time_list):
group_name = 'group_{}'.format(group_id)
temp_df = _make_df_with_one_col(data,
period=period,
start_time=start_time)
temp_df[GROUP_COL] = group_name
df_list.append(temp_df)
df = pd.concat(df_list, axis=0)
window_aggregator = dku_timeseries.WindowAggregator(
dku_timeseries.WindowAggregatorParams(window_unit='milliseconds'))
params = dku_timeseries.ExtremaExtractorParams(
window_aggregator=window_aggregator)
extrema_extractor = dku_timeseries.ExtremaExtractor(params)
output_df = extrema_extractor.compute(df,
TIME_COL,
DATA_COL,
groupby_columns=[GROUP_COL])
assert output_df.shape == (2, 3)
assert np.array_equal(output_df[DATA_COL], [99, 9])
def _make_window_aggregator():
params = _make_window_aggregator_params()
return dku_timeseries.WindowAggregator(params)
def _make_window_aggregator(window_width=1):
params = _make_window_aggregator_params(window_width)
return dku_timeseries.WindowAggregator(params)