Linting

nannyml/drift/univariate/calculator.py

@@ -264,8 +264,9 @@ def _fit(self, reference_data: pd.DataFrame, *args, **kwargs) -> UnivariateDrift
 
         _list_missing(self.column_names, reference_data)
 
-
-        self.continuous_column_names, self.categorical_column_names = self._split_continuous_and_categorical(reference_data)
+        self.continuous_column_names, self.categorical_column_names = self._split_continuous_and_categorical(
+            reference_data
+        )
 
         timestamps = reference_data[self.timestamp_column_name] if self.timestamp_column_name else None
         for column_name in self.continuous_column_names:
@@ -394,7 +395,7 @@ def _calculate(self, data: pd.DataFrame, *args, **kwargs) -> Result:
             self.result.analysis_data = data.copy()
 
         return self.result
-    
+
     def _split_continuous_and_categorical(self, data: pd.DataFrame) -> Tuple[List[str], List[str]]:
         """Splits the features in the data set into continuous and categorical features."""
         treat_as_numerical_set, treat_as_categorical_set = set(self.treat_as_numerical), set(self.treat_as_categorical)
@@ -404,22 +405,20 @@ def _split_continuous_and_categorical(self, data: pd.DataFrame) -> Tuple[List[st
         treat_as_numerical_set = treat_as_numerical_set - invalid_continuous_column_names
         if invalid_continuous_column_names:
             self._logger.info(
-                f"ignoring 'treat_as_numerical' values {list(invalid_continuous_column_names)} because they were not in "
-                f"listed column names"
+                f"ignoring 'treat_as_numerical' values {list(invalid_continuous_column_names)} because "
+                f"they were not in listed column names"
             )
 
         invalid_categorical_column_names = treat_as_categorical_set - column_names_set
         treat_as_categorical_set = treat_as_categorical_set - invalid_categorical_column_names
         if invalid_categorical_column_names:
             self._logger.info(
-                f"ignoring 'treat_as_categorical' values {list(invalid_categorical_column_names)} because they were not in "
-                f"listed column names"
+                f"ignoring 'treat_as_categorical' values {list(invalid_categorical_column_names)} because "
+                f"they were not in listed column names"
             )
-        
+
         unspecified_columns = column_names_set - treat_as_numerical_set - treat_as_categorical_set
-        continuous_column_names, categorical_column_names = _split_features_by_type(
-            data, unspecified_columns
-        )
+        continuous_column_names, categorical_column_names = _split_features_by_type(data, unspecified_columns)
 
         continuous_column_names = continuous_column_names + list(treat_as_numerical_set)
         categorical_column_names = categorical_column_names + list(treat_as_categorical_set)

nannyml/drift/univariate/methods.py

@@ -29,7 +29,7 @@
 from scipy.stats import chi2_contingency, ks_2samp, wasserstein_distance
 
 from nannyml._typing import Self
-from nannyml.base import _column_is_categorical, _remove_nans
+from nannyml.base import _remove_nans
 from nannyml.chunk import Chunker
 from nannyml.exceptions import InvalidArgumentsException, NotFittedException
 from nannyml.thresholds import Threshold, calculate_threshold_values
@@ -290,7 +290,7 @@ def _calculate(self, data: pd.Series):
         data = _remove_nans(data)
         if data.empty:
             return np.nan
-        
+
         len_data = len(data)
         data_proba_in_bins = np.histogram(data, bins=self._bins)[0] / len_data
 
@@ -346,7 +346,7 @@ def _calculate(self, data: pd.Series):
         data = _remove_nans(data)
         if data.empty:
             return np.nan
-        
+
         data_unique, data_counts = np.unique(data, return_counts=True)
         data_counts_dic = dict(zip(data_unique, data_counts))
         data_count_on_ref_bins = [data_counts_dic[key] if key in data_counts_dic else 0 for key in self._bins]
@@ -361,6 +361,7 @@ def _calculate(self, data: pd.Series):
 
         return distance
 
+
 @MethodFactory.register(key='kolmogorov_smirnov', feature_type=FeatureType.CONTINUOUS)
 class KolmogorovSmirnovStatistic(Method):
     """Calculates the Kolmogorov-Smirnov d-stat.
@@ -734,7 +735,7 @@ def _fit(self, reference_data: pd.Series, timestamps: Optional[pd.Series] = None
         reference_proba_in_bins = np.histogram(reference_data, bins=bins)[0] / len_reference
         self._bins = bins
         self._reference_proba_in_bins = reference_proba_in_bins
-        
+
         return self
 
     def _calculate(self, data: pd.Series):
@@ -753,6 +754,7 @@ def _calculate(self, data: pd.Series):
 
         return distance
 
+
 @MethodFactory.register(key='hellinger', feature_type=FeatureType.CATEGORICAL)
 class CategoricalHellingerDistance(Method):
     """Calculates the Hellinger Distance between two distributions."""
@@ -794,7 +796,7 @@ def _calculate(self, data: pd.Series):
         if data.empty:
             return np.nan
         reference_proba_in_bins = copy(self._reference_proba_in_bins)
-        
+
         data_unique, data_counts = np.unique(data, return_counts=True)
         data_counts_dic = dict(zip(data_unique, data_counts))
         data_count_on_ref_bins = [data_counts_dic[key] if key in data_counts_dic else 0 for key in self._bins]
@@ -808,4 +810,3 @@ def _calculate(self, data: pd.Series):
         distance = np.sqrt(np.sum((np.sqrt(reference_proba_in_bins) - np.sqrt(data_proba_in_bins)) ** 2)) / np.sqrt(2)
 
         return distance
-

tests/drift/test_drift.py

@@ -14,13 +14,13 @@
 from nannyml._typing import Key, Result, Self
 from nannyml.base import Abstract1DResult, AbstractCalculator
 from nannyml.chunk import CountBasedChunker, DefaultChunker, PeriodBasedChunker, SizeBasedChunker
+from nannyml.datasets import load_synthetic_car_loan_dataset
 from nannyml.drift.multivariate.data_reconstruction import DataReconstructionDriftCalculator
 from nannyml.drift.multivariate.domain_classifier import DomainClassifierCalculator
 from nannyml.drift.univariate.calculator import DEFAULT_THRESHOLDS, UnivariateDriftCalculator
 from nannyml.exceptions import InvalidArgumentsException
 from nannyml.performance_estimation.confidence_based import CBPE
 from nannyml.thresholds import ConstantThreshold, StandardDeviationThreshold
-from nannyml.datasets import load_synthetic_car_loan_dataset
 
 
 @pytest.fixture(scope="module")
@@ -183,9 +183,7 @@ def test_base_drift_calculator_uses_default_chunker_when_no_chunker_specified(sa
 
 
 @pytest.mark.parametrize('column_names, expected', [('f1', ['f1']), (['f1', 'f2'], ['f1', 'f2'])])
-def test_univariate_drift_calculator_create_with_single_or_list_of_column_names(  # noqa: D103
-    column_names, expected
-):
+def test_univariate_drift_calculator_create_with_single_or_list_of_column_names(column_names, expected):  # noqa: D103
     calc = UnivariateDriftCalculator(
         column_names=column_names,
         timestamp_column_name='timestamp',
@@ -264,9 +262,7 @@ def test_univariate_drift_calculator_treat_as_categorical_for_continuous_column(
     assert sorted(calc.categorical_column_names) == expected_categorical
 
 
-def test_univariate_drift_calculator_treat_as_categorical_for_categorical_column(  # noqa: D103
-    sample_drift_data
-):
+def test_univariate_drift_calculator_treat_as_categorical_for_categorical_column(sample_drift_data):  # noqa: D103
     calc = UnivariateDriftCalculator(
         column_names=['f1', 'f2', 'f3', 'f4'],
         treat_as_categorical='f3',
@@ -281,9 +277,7 @@ def test_univariate_drift_calculator_treat_as_categorical_for_categorical_column
     assert sorted(calc.categorical_column_names) == expected_categorical
 
 
-def test_univariate_drift_calculator_treat_as_for_non_existing_column(  # noqa: D103
-    sample_drift_data, caplog
-):
+def test_univariate_drift_calculator_treat_as_for_non_existing_column(sample_drift_data, caplog):  # noqa: D103
     caplog.set_level(logging.INFO)
 
     calc = UnivariateDriftCalculator(
@@ -300,8 +294,13 @@ def test_univariate_drift_calculator_treat_as_for_non_existing_column(  # noqa:
     assert sorted(calc.continuous_column_names) == expected_continuous
     assert sorted(calc.categorical_column_names) == expected_categorical
 
-    assert "ignoring 'treat_as_categorical' values ['foo'] because they were not in listed column names" in caplog.messages
-    assert "ignoring 'treat_as_numerical' values ['bar'] because they were not in listed column names" in caplog.messages
+    assert (
+        "ignoring 'treat_as_categorical' values ['foo'] because they were not in listed column names" in caplog.messages
+    )
+    assert (
+        "ignoring 'treat_as_numerical' values ['bar'] because they were not in listed column names" in caplog.messages
+    )
+
 
 def test_univariate_drift_calculator_without_custom_thresholds():  # noqa: D103
     sut = UnivariateDriftCalculator(
@@ -602,7 +601,7 @@ def test_base_drift_calculator_given_non_empty_features_list_should_only_calcula
 
 # See https://github.com/NannyML/nannyml/issues/192
 def test_univariate_drift_calculator_returns_distinct_but_consistent_results_when_reused(  # noqa: D103
-    sample_drift_data
+    sample_drift_data,
 ):
     ref_data = sample_drift_data.loc[sample_drift_data['period'] == 'reference']
     sut = UnivariateDriftCalculator(
@@ -871,7 +870,7 @@ def test_input_dataframes_are_not_altered_by_univ_calculator():  # noqa: D103
         'loan_length',
         'repaid_loan_on_prev_car',
         'size_of_downpayment',
-        'driver_tenure'
+        'driver_tenure',
     ]
     calc = UnivariateDriftCalculator(
         column_names=feature_column_names,
@@ -897,7 +896,7 @@ def test_input_dataframes_are_not_altered_by_dre_calculator():  # noqa: D103
         'loan_length',
         'repaid_loan_on_prev_car',
         'size_of_downpayment',
-        'driver_tenure'
+        'driver_tenure',
     ]
     calc = DataReconstructionDriftCalculator(
         column_names=feature_column_names,
@@ -920,12 +919,10 @@ def test_input_dataframes_are_not_altered_by_dc_calculator():  # noqa: D103
         'loan_length',
         'repaid_loan_on_prev_car',
         'size_of_downpayment',
-        'driver_tenure'
+        'driver_tenure',
     ]
     calc = DomainClassifierCalculator(
-        feature_column_names=feature_column_names,
-        timestamp_column_name='timestamp',
-        chunk_number=1
+        feature_column_names=feature_column_names, timestamp_column_name='timestamp', chunk_number=1
     )
     calc.fit(reference2)
     results = calc.calculate(monitored2)  # noqa: F841