def test__all_SetupSteps_for_EvaluationModel_population_slices(
fixture__population_slice_setup_steps,
fixture__population_slice__expected_columns,
tmpdir,
):
population_slice_generator = PopulationSliceGenerator(
setup_steps_by_date={
pd.Timestamp("2016-01-01"): fixture__population_slice_setup_steps
},
start=pd.Timestamp("2016-07-01"),
end=pd.Timestamp("2016-09-30"),
)
data_handler = ModelDataHandler(
database_type="sqlite", location=tmpdir, name="jobpath_evaluation",
)
evaluation_model = EvaluationModel(
data_handler=data_handler,
population_slice_generator=population_slice_generator,
)
evaluation_model.add_population_slices()
results = evaluation_model.population_slices[
PopulationSliceID(date=pd.Timestamp("2016-07-01"))
]
assert set(results.data.columns) == set(fixture__population_slice__expected_columns)
# Manually check how many people are on LR and eligible
assert len(results.data) == 315654
assert len(results.data[results.data["eligible_population"]]) == 86240
def test__ModelDataHandler__init(tmpdir):
"""Simple test to make sure everything gets initiated correctly
"""
results = ModelDataHandler(database_type="sqlite",
location=tmpdir,
name="test")
assert results.engine.name == sa.create_engine(
f"sqlite:///{tmpdir}/test.db").name
def test__ModelDataHandler__run__new(fixture__setup_steps_by_date,
fixture__population_slice_generator,
tmpdir):
data_path = f"sqlite:///{tmpdir}/test.db"
data_handler = ModelDataHandler(data_path)
population_slice_generator = fixture__population_slice_generator
results = {
population_slice.id: population_slice
for population_slice in population_slice_generator.run(data_handler)
}
key = PopulationSliceID(date=pd.Timestamp("2016-07-01", freq="QS-JAN"))
assert results[key].data.shape == (
90,
5,
)
def test__ModelDataHandler__read(fixture__population_slice, tmpdir):
data_path = f"sqlite:///{tmpdir}/test.db"
population_slice = fixture__population_slice
data_handler = ModelDataHandler(data_path)
data_handler.write(
data_type=population_slice.class_name,
data_id=population_slice.id,
data=population_slice.data,
index=False,
)
display(population_slice.data)
results = data_handler.read(
data_type=population_slice.class_name,
data_id=population_slice.id,
)
display(results)
assert results.shape == population_slice.data.shape
def test__ModelDataHandler__run__existing(fixture__setup_steps_by_date,
fixture__population_slice_generator,
tmpdir):
data_path = f"sqlite:///{tmpdir}/test.db"
data_handler = ModelDataHandler(data_path)
population_slice_generator = fixture__population_slice_generator
# First iteration should run setup_steps then write to storage
first_population_slices = {
population_slice.id: population_slice
for population_slice in population_slice_generator.run(data_handler)
}
# Second iteration should just read from storage
second_population_slices = {
population_slice.id: population_slice
for population_slice in population_slice_generator.run(data_handler)
}
key = PopulationSliceID(date=pd.Timestamp("2016-07-01", freq="QS-JAN"))
assert (len(first_population_slices[key].data) == len(
second_population_slices[key].data))
def test__all_SetupSteps_for_EvaluationModel_treatment_periods(
fixture__population_slice_setup_steps,
fixture__treatment_period_setup_steps,
fixture__treatment_period_expected_columns,
tmpdir,
):
data_handler = ModelDataHandler(
database_type="sqlite", location=tmpdir, name="jobpath_evaluation",
)
population_slice_generator = PopulationSliceGenerator(
setup_steps_by_date={
pd.Timestamp("2016-01-01"): fixture__population_slice_setup_steps
},
start=pd.Timestamp("2016-07-01"),
end=pd.Timestamp("2016-09-30"),
)
treatment_period_generator = TreatmentPeriodGenerator(
setup_steps_by_date={
pd.Timestamp("2016-07-01"): fixture__treatment_period_setup_steps
},
end=pd.Period("2016-09"),
)
evaluation_model = EvaluationModel(
data_handler=data_handler,
population_slice_generator=population_slice_generator,
treatment_period_generator=treatment_period_generator,
)
evaluation_model.add_population_slices()
evaluation_model.add_treatment_periods()
results_id = TreatmentPeriodID(
population_slice_id=PopulationSliceID(date=pd.Timestamp("2016-07-01")),
time_period=pd.Period("2016-09"),
)
results = evaluation_model.treatment_periods[results_id]
assert set(results.data.columns) == set(fixture__treatment_period_expected_columns)
# Manually check how many people are on LR and eligible
assert len(results.data[results.data["eligible_population"]]) < len(results.data)
assert len(results.data[results.data["eligible_population"]]) == 64333
assert len(results.data[results.data["jobpath_starts"]]) == 4273
def test__ModelDataHandler__write__new(fixture__population_slice, tmpdir):
"""Given a population_slice instance that's not saved, save it correctly.
-- fixture__population_slice returns a population_slice with data.shape (10, 5).
-- data_handler.write() saves this to table PopulationSlice, adding date column.
-- Reading the PopulationSlice table back directly from database, dropping date...
-- ...should return a df with the same shape as population_slice.data
"""
data_path = f"sqlite:///{tmpdir}/test.db"
population_slice = fixture__population_slice
data_handler = ModelDataHandler(data_path)
data_handler.write(
data_type=population_slice.class_name,
data_id=population_slice.id,
data=population_slice.data,
)
engine = sa.create_engine(data_path)
df = pd.read_sql("PopulationSlice", con=engine)
results = df.loc[df["data_id_date"] == str(
population_slice.id.date.date())].drop(["data_id_date", "index"],
axis="columns")
display(results)
display(population_slice.data)
assert results.shape == population_slice.data.shape
def test__ModelDataHandler__write__overwrite(fixture__population_slice,
tmpdir):
"""Given a population_slice that overwrites an old one, save it correctly.
"""
data_path = f"sqlite:///{tmpdir}/test.db"
population_slice = fixture__population_slice
data_handler = ModelDataHandler(data_path)
# Write first version of data
data_handler.write(
data_type=population_slice.class_name,
data_id=population_slice.id,
data=population_slice.data,
)
# Change the data - it's ok for now to assume same number of columns!
population_slice.data = pd.DataFrame(np.random.randint(100,
200,
size=(20, 4)),
columns=list("ABCD"))
population_slice.data["date"] = pd.date_range("2016-01-01",
periods=8,
freq="QS")[0]
# Now write the changed data to database
data_handler.write(
data_type=population_slice.class_name,
data_id=population_slice.id,
data=population_slice.data,
)
engine = sa.create_engine(data_path)
df = pd.read_sql("PopulationSlice", con=engine)
display(df)
display(df.info())
results = df.loc[df["data_id_date"] == population_slice.id.date].drop(
["data_id_date", "index"], axis="columns")
assert results.shape == population_slice.data.shape
AgeEligible,
ClaimCodeEligible,
ClaimDurationEligible,
OnLES,
OnJobPath,
JobPathStartedEndedSamePeriod,
EligiblePopulation,
JobPathStarts,
EvaluationGroup,
)
from evaluation_jp.data import ModelDataHandler
evaluation_model = EvaluationModel(
data_handler=ModelDataHandler(
database_type="sqlite",
location="//cskma0294/f/Evaluations/JobPath",
name="jobpath_evaluation",
),
population_slice_generator=PopulationSliceGenerator(
start=pd.Timestamp("2016-01-01"),
end=pd.Timestamp("2017-12-31"),
freq="QS",
setup_steps_by_date={
pd.Timestamp("2016-01-01"):
SetupSteps(steps=[
LiveRegisterPopulation(
columns_by_type={
"lr_code": "category",
"clm_comm_date": "datetime64",
"JobPath_Flag": "boolean",
"date_of_birth": "datetime64",