Skip to content

Commit

Permalink
Parametrize exception test fixtures
Browse files Browse the repository at this point in the history 
Parametrized `test_nonunique_parameter_names_value_error`,
`test_timeseries_parameter_has_insufficient_columns_value_error`, and
`test_timeseries_parameter_has_insufficient_dates_value_error` test
fixtures. And moved long comment about dead code in the insufficient
dates ValueError to the main issue.
  • Loading branch information
@TimothyWillard
TimothyWillard committed Aug 6, 2024
1 parent 9a381d5 commit 33dbf17
Showing 1 changed file with 144 additions and 86 deletions.
230 changes: 144 additions & 86 deletions flepimop/gempyor_pkg/tests/parameters/test_parameters_class.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -107,10 +107,12 @@ def has_timeseries_parameter(self) -> bool:
return True
return False

def get_timeseries_df(self, param_name: str) -> pd.DataFrame:
def get_timeseries_df(
self, param_name: str, subset_by_subpops: bool = True
) -> pd.DataFrame:
df = self._timeseries_dfs.get(param_name)
if df is not None:
return df.copy()
return df[self.subpop_names].copy() if subset_by_subpops else df.copy()
conf = self.config.get(param_name, {})
df_file = conf.get("timeseries")
if df_file is None:
Expand All @@ -120,9 +122,8 @@ def get_timeseries_df(self, param_name: str) -> pd.DataFrame:
df = pd.read_csv(df_file, index_col=None)
df["date"] = pd.to_datetime(df["date"])
df = df.set_index("date")
df = df[self.subpop_names]
self._timeseries_dfs[param_name] = df
return df.copy()
return df[self.subpop_names].copy() if subset_by_subpops else df.copy()


def fixed_three_valid_parameter_factory(tmp_path: pathlib.Path) -> MockParametersInput:
Expand Down Expand Up @@ -165,103 +166,160 @@ def valid_parameters_factory(tmp_path: pathlib.Path) -> MockParametersInput:
)


def nonunique_invalid_parameter_factory(tmp_path: pathlib.Path) -> MockParametersInput:
return MockParametersInput(
config={
"sigma": {"value": 0.1},
"eta": {"value": 0.2},
"SIGMA": {"value": 0.3},
},
ti=date(2024, 1, 1),
tf=date(2024, 1, 3),
subpop_names=["1", "2", "3"],
)


def insufficient_columns_parameter_factory(
tmp_path: pathlib.Path,
) -> MockParametersInput:
df = pd.DataFrame(
data={
"date": pd.date_range(date(2024, 1, 1), date(2024, 1, 5)),
"1": [1.2, 2.3, 3.4, 4.5, 5.6],
"2": [2.3, 3.4, 4.5, 5.6, 6.7],
}
)
tmp_file = tmp_path / f"{uuid4().hex}.csv"
df.to_csv(tmp_file, index=False)
return MockParametersInput(
config={"sigma": {"timeseries": str(tmp_file.absolute())}},
ti=date(2024, 1, 1),
tf=date(2024, 1, 5),
subpop_names=["1", "2", "3"],
)


def insufficient_dates_parameter_factory(tmp_path: pathlib.Path) -> MockParametersInput:
df = pd.DataFrame(
data={
"date": pd.date_range(date(2024, 1, 1), date(2024, 1, 5)),
"1": [1.2, 2.3, 3.4, 4.5, 5.6],
"2": [2.3, 3.4, 4.5, 5.6, 6.7],
}
)
tmp_file = tmp_path / f"{uuid4().hex}.csv"
df.to_csv(tmp_file, index=False)
return MockParametersInput(
config={"sigma": {"timeseries": str(tmp_file.absolute())}},
ti=date(2024, 1, 1),
tf=date(2024, 1, 6),
subpop_names=["1", "2"],
)


class TestParameters:
def test_nonunique_parameter_names_value_error(self) -> None:
duplicated_parameters = create_confuse_subview_from_dict(
"parameters",
{"sigma": {"value": 0.1}, "gamma": {"value": 0.2}, "GAMMA": {"value": 0.3}},
)
@pytest.mark.parametrize("factory", [(nonunique_invalid_parameter_factory)])
def test_nonunique_parameter_names_value_error(
self,
tmp_path: pathlib.Path,
factory: Callable[[pathlib.Path], MockParametersInput],
) -> None:
mock_inputs = factory(tmp_path)
with pytest.raises(
ValueError,
match=(
r"Parameters of the SEIR model have the same name "
r"\(remember that case is not sufficient\!\)"
),
):
Parameters(
duplicated_parameters,
ti=date(2024, 1, 1),
tf=date(2024, 12, 31),
subpop_names=["1", "2"],
)
mock_inputs.create_parameters_instance()

def test_timeseries_parameter_has_insufficient_columns_value_error(self) -> None:
param_df = pd.DataFrame(
data={
"date": pd.date_range(date(2024, 1, 1), date(2024, 1, 5)),
"1": [1.2, 2.3, 3.4, 4.5, 5.6],
"2": [2.3, 3.4, 4.5, 5.6, 6.7],
}
)
with NamedTemporaryFile(suffix=".csv") as temp_file:
param_df.to_csv(temp_file.name, index=False)
invalid_timeseries_parameters = create_confuse_subview_from_dict(
"parameters", {"sigma": {"timeseries": temp_file.name}}
)
with pytest.raises(
ValueError,
match=(
rf"ERROR loading file {temp_file.name} for parameter sigma\: "
rf"the number of non 'date'\s+columns are 2, expected 3 "
rf"\(the number of subpops\) or one\."
),
):
Parameters(
invalid_timeseries_parameters,
ti=date(2024, 1, 1),
tf=date(2024, 1, 5),
subpop_names=["1", "2", "3"],
@pytest.mark.parametrize("factory", [(insufficient_columns_parameter_factory)])
def test_timeseries_parameter_has_insufficient_columns_value_error(
self,
tmp_path: pathlib.Path,
factory: Callable[[pathlib.Path], MockParametersInput],
) -> None:
mock_inputs = factory(tmp_path)
tmp_file = None
for param_name, conf in mock_inputs.config.items():
if "timeseries" in conf:
df = mock_inputs.get_timeseries_df(param_name, subset_by_subpops=False)
actual_columns = len(df.columns)
if (
actual_columns != mock_inputs.number_of_subpops()
and actual_columns != 1
):
tmp_file = conf.get("timeseries")
break
if tmp_file is None:
raise RuntimeError(
(
"The given factory does not produce a timeseries "
"with an insufficient number of columns."
)
)
with pytest.raises(
ValueError,
match=(
rf"^ERROR loading file {tmp_file} for parameter sigma\: the number of "
rf"non 'date'\s+columns are {actual_columns}, expected "
rf"{mock_inputs.number_of_subpops()} \(the number of subpops\) or "
rf"one\.$"
),
):
mock_inputs.create_parameters_instance()

@pytest.mark.parametrize(
"start_date,end_date,timeseries_df",
[(date(2024, 1, 1), date(2024, 1, 6), MockData.simple_timeseries_param_df)],
)
@pytest.mark.parametrize("factory", [(insufficient_dates_parameter_factory)])
def test_timeseries_parameter_has_insufficient_dates_value_error(
self, start_date: date, end_date: date, timeseries_df: pd.DataFrame
self,
tmp_path: pathlib.Path,
factory: Callable[[pathlib.Path], MockParametersInput],
) -> None:
# First way to get at this error, purely a length difference
with NamedTemporaryFile(suffix=".csv") as temp_file:
timeseries_df.to_csv(temp_file.name, index=False)
invalid_timeseries_parameters = create_confuse_subview_from_dict(
"parameters", {"sigma": {"timeseries": temp_file.name}}
)
timeseries_start_date = timeseries_df["date"].dt.date.min()
timeseries_end_date = timeseries_df["date"].dt.date.max()
subpop_names = [c for c in timeseries_df.columns.to_list() if c != "date"]
with pytest.raises(
ValueError,
match=(
rf"ERROR loading file {temp_file.name} for parameter sigma\:\s+"
rf"the \'date\' entries of the provided file do not include all the"
rf" days specified to be modeled by\s+the config\. the provided "
rf"file includes 5 days between {timeseries_start_date}"
rf"( 00\:00\:00)? to {timeseries_end_date}( 00\:00\:00)?,\s+while "
rf"there are 6 days in the config time span of {start_date}->"
rf"{end_date}\. The file must contain entries for the\s+the exact "
rf"start and end dates from the config\. "
),
):
Parameters(
invalid_timeseries_parameters,
ti=start_date,
tf=end_date,
subpop_names=subpop_names,
mock_inputs = factory(tmp_path)

tmp_file = None
for param_name, conf in mock_inputs.config.items():
if "timeseries" in conf:
df = mock_inputs.get_timeseries_df(param_name)
timeseries_start_date = df.index.to_series().dt.date.min()
timeseries_end_date = df.index.to_series().dt.date.max()
if (
(timeseries_start_date > mock_inputs.ti)
or (timeseries_end_date < mock_inputs.tf)
or (
not pd.date_range(mock_inputs.ti, mock_inputs.tf)
.isin(df.index)
.all()
)
):
tmp_file = conf.get("timeseries")
break

if tmp_file is None:
raise RuntimeError(
(
"The given factory does not produce a timeseries with an "
"insufficient date range."
)
)

# TODO: I'm not sure how to get to the second pathway to this error message.
# 1) We subset the read in dataframe to `ti` to `tf` so if the dataframe goes
# from 2024-01-01 through 2024-01-05 and the given date range is 2024-01-02
# through 2024-01-06 the dataframe's date range will be subsetted to 2024-01-02
# through 2024-01-05 which is a repeat of the above.
# 2) Because of the subsetting you can't provide anything except a monotonic
# increasing sequence of dates, pandas only allows subsetting on ordered date
# indexes so you'll get a different error.
# 3) If you provide a monotonic increasing sequence of dates but 'reverse' `ti`
# and `tf` you get no errors (which I think is also bad) because the slice
# operation returns an empty dataframe with the right columns & index and the
# `pd.date_range` function only creates monotonic increasing sequences and
# 0 == 0.
file_days = (timeseries_end_date - timeseries_start_date).days + 1
with pytest.raises(
ValueError,
match=(
rf"^ERROR loading file {tmp_file} for parameter sigma\:\s+the \'date\' "
rf"entries of the provided file do not include all the days specified "
rf"to be modeled by\s+the config\. the provided file includes "
rf"{(timeseries_end_date - timeseries_start_date).days + 1} days "
rf"between {timeseries_start_date}( 00\:00\:00)? to "
rf"{timeseries_end_date}( 00\:00\:00)?,\s+while there are "
rf"{mock_inputs.number_of_days()} days in the config time span of "
rf"{mock_inputs.ti}->{mock_inputs.tf}\. The file must contain entries "
rf"for the\s+the exact start and end dates from the config\. $"
),
):
mock_inputs.create_parameters_instance()

@pytest.mark.parametrize(
"factory",
Expand Down

0 comments on commit 33dbf17

Please sign in to comment.