def test_load_series():
dt = datetime(2019, 1, 29, 15, 15)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, 2, 3],
),
name="mydata",
)
assert (s.load_series(expected_frequency=timedelta(
minutes=15)).loc[dt + timedelta(minutes=30)] == 3)
def test_load_series_with_missing_data():
dt = datetime(2019, 1, 29, 15, 15)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, np.nan, 3],
),
name="mydata",
)
with pytest.raises(NaNData) as e_info:
s.load_series(expected_frequency=timedelta(minutes=15))
assert "Nan values" in str(e_info.value)
def test_load_series_with_frequency_resampling():
dt = datetime(2019, 1, 29, 15, 15)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, 2, 3],
),
name="mydata",
)
series = s.load_series(expected_frequency=timedelta(hours=1))
assert len(series) == 1
assert series[0] == 2 # the mean
def test_load_series_with_expected_time_window():
dt = datetime(2019, 1, 29, 15, 15, tzinfo=pytz.utc)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, 2, 3],
),
name="mydata",
)
assert (s.load_series(
expected_frequency=timedelta(minutes=15),
check_time_window=(dt, dt + timedelta(minutes=30)),
).loc[dt + timedelta(minutes=30)] == 3)
def test_load_series_with_interpolation():
dt = datetime(2019, 1, 29, 15, 15)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, np.nan, 3],
),
name="mydata",
interpolation_config={"method": "time"},
)
series = s.load_series(expected_frequency=timedelta(minutes=15))
assert len(series) == 3
assert series[1] == 2 # the interpolated value
def test_load_series_with_non_existing_custom_frequency_resampling():
dt = datetime(2019, 1, 29, 15, 15)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, 2, 3],
),
name="mydata",
resampling_config={"aggregation": "GGG"},
)
with pytest.raises(IncompatibleModelSpecs) as e_info:
s.load_series(expected_frequency=timedelta(hours=1))
assert "Cannot find resampling aggregation GGG" in str(e_info.value)
def test_load_series_with_non_existing_interpolation():
dt = datetime(2019, 1, 29, 15, 15)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, np.nan, 3],
),
name="mydata",
interpolation_config={"method": "GGG"},
)
with pytest.raises(IncompatibleModelSpecs) as e_info:
s.load_series(expected_frequency=timedelta(minutes=15))
assert "Cannot call interpolate function with arguments {'method': 'GGG'}" in str(
e_info.value)
def test_load_series_with_transformation():
dt = datetime(2019, 1, 29, 15, 15)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, 2, 3],
),
name="mydata",
feature_transformation=MyMultiplicationTransformation(factor=11),
)
assert (s.load_series(expected_frequency=timedelta(
minutes=15)).loc[dt + timedelta(minutes=15)] == 2)
assert (s.load_series(
expected_frequency=timedelta(minutes=15),
transform_features=True).loc[dt + timedelta(minutes=15)] == 2 * 11)
def test_load_series_with_frequency_resampling(down_or_up: str):
dt = datetime(2019, 1, 29, 15, 15)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, 2, 3],
),
name="mydata",
)
series = s.load_series(
expected_frequency=timedelta(hours=1)
if down_or_up == "down"
else timedelta(minutes=5)
)
assert len(series) == 1 if down_or_up == "down" else len(series) == 9
assert series.mean() == 2 # the mean remains the same
def test_load_series_with_larger_expected_time_window():
dt = datetime(2019, 1, 29, 15, 15, tzinfo=pytz.utc)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, 2, 3],
),
name="mydata",
)
with pytest.raises(MissingData) as e_info:
s.load_series(
expected_frequency=timedelta(minutes=15),
check_time_window=(dt - timedelta(minutes=15),
dt + timedelta(minutes=45)),
)
assert "starts too late" in str(e_info.value)
assert "ends too early" in str(e_info.value)
def test_load_series_with_non_existing_custom_frequency_resampling(down_or_up: str):
dt = datetime(2019, 1, 29, 15, 15)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, 2, 3],
),
name="mydata",
resampling_config={f"{down_or_up}sampling_method": "GGG"},
)
with pytest.raises(IncompatibleModelSpecs) as e_info:
s.load_series(
expected_frequency=timedelta(hours=1)
if down_or_up == "down"
else timedelta(minutes=5)
)
assert f"Cannot find {down_or_up}sampling method GGG" in str(e_info.value)
def test_load_series_with_custom_frequency_resampling(down_or_up: str):
dt = datetime(2019, 1, 29, 15, 15)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, 2, 3],
),
name="mydata",
resampling_config={
"downsampling_method": "sum",
"upsampling_method": "reverse_sum",
},
)
series = s.load_series(
expected_frequency=timedelta(hours=1)
if down_or_up == "down"
else timedelta(minutes=5)
)
assert len(series) == 1 if down_or_up == "down" else len(series) == 9
assert sum(series) == 6 # the sum remains the same
def test_load_series_with_instantaneous_measurements(down_or_up: str, exp_series):
""" Test resampling of instantaneous measurements. """
dt = datetime(2019, 1, 29, 15, 30)
s = ObjectSeriesSpecs(
data=pd.Series(
index=pd.date_range(dt, dt + timedelta(minutes=30), freq="15T"),
data=[1, 2, 3],
),
name="mydata",
resampling_config=dict(
downsampling_method="first",
event_resolution=timedelta(hours=0),
),
interpolation_config=dict(method="pad", limit=3),
)
series = s.load_series(
expected_frequency=timedelta(hours=1)
if down_or_up == "down"
else timedelta(minutes=5)
)
assert len(series) == 2 if down_or_up == "down" else len(series) == 7
pd.testing.assert_series_equal(series, exp_series)
def test_load_series_without_datetime_index():
with pytest.raises(Exception) as e_info:
s = ObjectSeriesSpecs(data=pd.Series([1, 2, 3]), name="mydata")
s.load_series(expected_frequency=timedelta(hours=1))
assert "DatetimeIndex" in str(e_info.value)