def test_datetime(self, instance, regression_beta):
"""
Re-use regression_Beta fixture but cast date as a datetime.
"""
regression_beta['date'] = pandas.to_datetime(regression_beta['date'])
key = Keys.regression_beta(4)
instance.dump(regression_beta, key=key)
loaded = instance.load(key)
pandas.testing.assert_frame_equal(regression_beta, loaded)
def load_transition_date(self, draw_id: int) -> pd.Series:
dates_df = self.regression_marshall.load(key=MKeys.date(draw_id))
dates_df['end_date'] = pd.to_datetime(dates_df['end_date'])
transition_date = dates_df.set_index(
'location_id').sort_index()['end_date'].rename('date')
return transition_date
def test_no_overwriting(self, instance, parameters):
self.assert_no_accidental_overwrites(instance,
parameters,
key=Keys.parameter(4))
def test_beta_scales_marshall(self, instance, beta_scales):
self.assert_load_dump_workflow_correct(
instance,
beta_scales,
key=Keys.beta_scales(scenario='happy', draw_id=4))
def test_date_marshall(self, instance, dates):
self.assert_load_dump_workflow_correct(instance,
dates,
key=Keys.date(4))
def test_CSVMarshall_uses_same_paths_as_paths_module(tmpdir):
"""
Runs the gamut of paths produced by paths.py and compares to CSVMarshall
path computation.
"""
from covid_model_seiir_pipeline.paths import (
RegressionPaths,
ForecastPaths,
)
root = pathlib.Path(tmpdir)
m = CSVMarshall(root)
r = RegressionPaths(root)
f = ForecastPaths(root, scenarios=["happy"])
dargs = {'draw_id': 4}
sdargs = {'scenario': "happy", 'draw_id': 4}
# forecasting/data.py
# load_regression_coefficients
assert r.get_coefficient_file(**dargs) == m.resolve_key(
Keys.coefficient(**dargs))
# load_dates_df
assert r.get_date_file(**dargs) == m.resolve_key(Keys.date(**dargs))
# load_beta_regression
assert r.get_beta_regression_file(**dargs) == m.resolve_key(
Keys.regression_beta(**dargs))
# load_infection_data
assert r.get_data_file(**dargs) == m.resolve_key(
Keys.location_data(**dargs))
# load_beta_params
assert r.get_beta_param_file(**dargs) == m.resolve_key(
Keys.parameter(**dargs))
# save_components
assert f.get_components_path(**sdargs) == m.resolve_key(
Keys.components(**sdargs))
# save_beta_scales
assert f.get_beta_scaling_path(**sdargs) == m.resolve_key(
Keys.beta_scales(**sdargs))
# save_outputs
assert f.get_outputs_path(**sdargs) == m.resolve_key(
Keys.forecast_raw_outputs(**sdargs))
# regression/data.py
# save_beta_param_file
assert r.get_beta_param_file(**dargs) == m.resolve_key(
Keys.parameter(**dargs))
# save_date_file
assert r.get_date_file(**dargs) == m.resolve_key(Keys.date(**dargs))
# save_regression_coefficients
assert r.get_coefficient_file(**dargs) == m.resolve_key(
Keys.coefficient(**dargs))
# save_regression_betas
assert r.get_beta_regression_file(**dargs) == m.resolve_key(
Keys.regression_beta(**dargs))
# save_location_data
assert r.get_data_file(**dargs) == m.resolve_key(
Keys.location_data(**dargs))
def save_regression_betas(self, df: pd.DataFrame, draw_id: int) -> None:
self.regression_marshall.dump(df, key=MKeys.regression_beta(draw_id))
def save_date_file(self, df: pd.DataFrame, draw_id: int) -> None:
self.regression_marshall.dump(df, key=MKeys.date(draw_id))
def save_components(self, forecasts: pd.DataFrame, scenario: str,
draw_id: int, strict: bool):
self.forecast_marshall.dump(forecasts,
key=MKeys.components(scenario=scenario,
draw_id=draw_id),
strict=strict)
def load_beta_params(self, draw_id: int) -> Dict[str, float]:
df = self.regression_marshall.load(key=MKeys.parameter(
draw_id=draw_id))
return df.set_index('params')['values'].to_dict()
def load_raw_covariates(self, scenario: str, draw_id: int):
covariates = self.forecast_marshall.load(
key=MKeys.forecast_raw_covariates(scenario=scenario,
draw_id=draw_id))
covariates['date'] = pd.to_datetime(covariates['date'])
return covariates
def save_raw_covariates(self, covariates: pd.DataFrame, scenario: str,
draw_id: int, strict: bool):
self.forecast_marshall.dump(covariates,
key=MKeys.forecast_raw_covariates(
scenario=scenario, draw_id=draw_id),
strict=strict)
def load_infection_data(self, draw_id: int) -> pd.DataFrame:
infection_data = self.regression_marshall.load(
key=MKeys.location_data(draw_id))
infection_data['date'] = pd.to_datetime(infection_data['date'])
return infection_data
def load_beta_regression(self, draw_id: int) -> pd.DataFrame:
beta_regression = self.regression_marshall.load(
key=MKeys.regression_beta(draw_id))
beta_regression['date'] = pd.to_datetime(beta_regression['date'])
return beta_regression
def load_reimposition_dates(self, scenario: str, reimposition_number: int):
return self.forecast_marshall.load(key=MKeys.reimposition_dates(
scenario=scenario, reimposition_number=reimposition_number))
def save_beta_param_file(self, df: pd.DataFrame, draw_id: int) -> None:
self.regression_marshall.dump(df, key=MKeys.parameter(draw_id))
def load_components(self, scenario: str, draw_id: int) -> pd.DataFrame:
components = self.forecast_marshall.load(
key=MKeys.components(scenario=scenario, draw_id=draw_id))
components['date'] = pd.to_datetime(components['date'])
return components.set_index(['location_id', 'date'])
def save_regression_coefficients(self, coefficients: pd.DataFrame,
draw_id: int) -> None:
self.regression_marshall.dump(coefficients,
key=MKeys.coefficient(draw_id))
def save_beta_scales(self, scales: pd.DataFrame, scenario: str,
draw_id: int):
self.forecast_marshall.dump(scales,
key=MKeys.beta_scales(scenario=scenario,
draw_id=draw_id))
def save_location_data(self, df: pd.DataFrame, draw_id: int) -> None:
# quasi-inverse of load_all_location_data, except types are different
self.regression_marshall.dump(df, key=MKeys.location_data(draw_id))
def load_beta_scales(self, scenario: str, draw_id: int):
return self.forecast_marshall.load(
MKeys.beta_scales(scenario=scenario, draw_id=draw_id))
def test_parameters_marshall(self, instance, parameters):
self.assert_load_dump_workflow_correct(instance,
parameters,
key=Keys.parameter(4))
def save_raw_outputs(self, raw_outputs: pd.DataFrame, scenario: str,
draw_id: int, strict: bool):
self.forecast_marshall.dump(raw_outputs,
key=MKeys.forecast_raw_outputs(
scenario=scenario, draw_id=draw_id),
strict=strict)
def test_coefficients_marshall(self, instance, coefficients):
self.assert_load_dump_workflow_correct(instance,
coefficients,
key=Keys.coefficient(4))
def load_raw_outputs(self, scenario: str, draw_id: int) -> pd.DataFrame:
return self.forecast_marshall.load(
key=MKeys.forecast_raw_outputs(scenario=scenario, draw_id=draw_id))
def test_components_marshall(self, instance, components):
self.assert_load_dump_workflow_correct(
instance,
components,
key=Keys.components(scenario='happy', draw_id=4))
def save_output_miscellaneous(self, output_miscellaneous: pd.DataFrame,
scenario: str, measure: str):
self.forecast_marshall.dump(output_miscellaneous,
key=MKeys.forecast_output_miscellaneous(
scenario=scenario, measure=measure))
def save_reimposition_dates(self, reimposition_dates: pd.DataFrame,
scenario: str, reimposition_number: int):
self.forecast_marshall.dump(
reimposition_dates,
key=MKeys.reimposition_dates(
scenario=scenario, reimposition_number=reimposition_number))
def load_regression_coefficients(self, draw_id: int) -> pd.DataFrame:
return self.regression_marshall.load(MKeys.coefficient(draw_id))