microimpute 3.1.0__tar.gz → 3.1.2__tar.gz

{microimpute-3.1.0 → microimpute-3.1.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: microimpute
-Version: 3.1.0
+Version: 3.1.2
 Summary: Benchmarking imputation methods for microdata
 Author-email: María Juaristi <juaristi@uni.minerva.edu>, Nikhil Woodruff <nikhil.woodruff@outlook.com>
 Requires-Python: <3.15,>=3.12
@@ -21,7 +21,7 @@ Requires-Dist: psutil
 Provides-Extra: dev
 Requires-Dist: pytest<10.0.0,>=8.0.0; extra == "dev"
 Requires-Dist: pytest-cov<8.0.0,>=6.0.0; extra == "dev"
-Requires-Dist: ruff>=0.9.0; extra == "dev"
+Requires-Dist: ruff<0.17.0,>=0.16.0; extra == "dev"
 Requires-Dist: mypy<2.0.0,>=1.2.3; extra == "dev"
 Requires-Dist: build<2.0.0,>=1.2.0; extra == "dev"
 Requires-Dist: towncrier>=24.8.0; extra == "dev"

{microimpute-3.1.0 → microimpute-3.1.2}/microimpute/models/qrf.py

@@ -50,6 +50,7 @@ class _RandomForestClassifierModel:
         self.var_type = None
         self.categories = None
         self.label_map = None
+        self.feature_columns: List[str] = []
 
     def fit(
         self,
@@ -99,10 +100,24 @@ class _RandomForestClassifierModel:
         fit_kwargs = {}
         if sample_weight is not None:
             fit_kwargs["sample_weight"] = np.asarray(sample_weight, dtype=float)
+        self.feature_columns = list(X.columns)
         self.classifier.fit(X, y_encoded, **fit_kwargs)
 
+    def _align_features(self, X: pd.DataFrame) -> pd.DataFrame:
+        """Reorder prediction features to the fitted sklearn column contract."""
+        if not self.feature_columns:
+            return X
+        missing = [column for column in self.feature_columns if column not in X.columns]
+        if missing:
+            raise ValueError(
+                "Prediction data is missing fitted feature columns for "
+                f"{self.output_column}: {missing}"
+            )
+        return X.loc[:, self.feature_columns]
+
     def predict(self, X: pd.DataFrame, return_probs: bool = False) -> pd.Series:
         """Predict classes or probabilities."""
+        X = self._align_features(X)
         if return_probs:
             probs = self.classifier.predict_proba(X)
             # Return both probabilities and the original category labels
@@ -151,6 +166,7 @@ class _QRFModel:
         self.logger = logger
         self.qrf = None
         self.output_column = None
+        self.feature_columns: List[str] = []
         # Create the RNG once at construction so that repeated predict()
         # calls consume state progressively and return different draws.
         self._rng = np.random.default_rng(self.seed)
@@ -192,8 +208,21 @@ class _QRFModel:
         fit_kwargs = {}
         if sample_weight is not None:
             fit_kwargs["sample_weight"] = np.asarray(sample_weight, dtype=float)
+        self.feature_columns = list(X.columns)
         self.qrf.fit(X, y.values.ravel(), **fit_kwargs)
 
+    def _align_features(self, X: pd.DataFrame) -> pd.DataFrame:
+        """Reorder prediction features to the fitted QRF column contract."""
+        if not self.feature_columns:
+            return X
+        missing = [column for column in self.feature_columns if column not in X.columns]
+        if missing:
+            raise ValueError(
+                "Prediction data is missing fitted feature columns for "
+                f"{self.output_column}: {missing}"
+            )
+        return X.loc[:, self.feature_columns]
+
     def predict(
         self,
         X: pd.DataFrame,
@@ -218,6 +247,8 @@ class _QRFModel:
                 across quantiles for a given row and is used when the caller
                 supplies an explicit ``quantiles`` list.
         """
+        X = self._align_features(X)
+
         # Deterministic path: user asked for a specific quantile — query the
         # QRF directly so that for any row i,
         # prediction(q_low) <= prediction(q_mid) <= prediction(q_high).

{microimpute-3.1.0 → microimpute-3.1.2}/microimpute/models/zero_inflated.py

@@ -232,6 +232,9 @@ class ZeroInflatedImputer(Imputer):
             for v in imputed_variables
             if v in self.numeric_targets or v in constant_numeric_targets
         ]
+        nested_not_numeric_categorical = list(
+            dict.fromkeys([*(not_numeric_categorical or []), *numeric_targets])
+        )
         # Sequential (chained-equations) imputation: condition each numeric
         # target on the original predictors plus the previously-fit numeric
         # targets, so the imputed vector preserves cross-variable joint
@@ -255,6 +258,7 @@ class ZeroInflatedImputer(Imputer):
                 variable=var,
                 regime=regime,
                 y=y,
+                not_numeric_categorical=nested_not_numeric_categorical,
             )
             bundle["predictors"] = list(seq_predictors)
             self._per_variable[var] = bundle
@@ -302,6 +306,7 @@ class ZeroInflatedImputer(Imputer):
         variable: str,
         regime: str,
         y: np.ndarray,
+        not_numeric_categorical: Optional[List[str]] = None,
     ) -> Dict[str, Any]:
         """Fit the gate and base imputer(s) for one numeric target.
 
@@ -317,7 +322,12 @@ class ZeroInflatedImputer(Imputer):
             # No gate; single base imputer on the full training set.
             return {
                 "kind": "single",
-                "base": self._fit_base_single(X_train, predictors, variable),
+                "base": self._fit_base_single(
+                    X_train,
+                    predictors,
+                    variable,
+                    not_numeric_categorical=not_numeric_categorical,
+                ),
             }
 
         if regime == REGIME_ZI_POSITIVE:
@@ -326,7 +336,10 @@ class ZeroInflatedImputer(Imputer):
             clf.fit(X_pred, labels)
             pos_mask = y > self.zero_atol
             pos_base = self._fit_base_single(
-                X_train.loc[pos_mask], predictors, variable
+                X_train.loc[pos_mask],
+                predictors,
+                variable,
+                not_numeric_categorical=not_numeric_categorical,
             )
             return {
                 "kind": "zi_positive",
@@ -340,7 +353,10 @@ class ZeroInflatedImputer(Imputer):
             clf.fit(X_pred, labels)
             neg_mask = y < -self.zero_atol
             neg_base = self._fit_base_single(
-                X_train.loc[neg_mask], predictors, variable
+                X_train.loc[neg_mask],
+                predictors,
+                variable,
+                not_numeric_categorical=not_numeric_categorical,
             )
             return {
                 "kind": "zi_negative",
@@ -360,10 +376,16 @@ class ZeroInflatedImputer(Imputer):
                 "kind": "sign_only",
                 "classifier": clf,
                 "positive_base": self._fit_base_single(
-                    X_train.loc[pos_mask], predictors, variable
+                    X_train.loc[pos_mask],
+                    predictors,
+                    variable,
+                    not_numeric_categorical=not_numeric_categorical,
                 ),
                 "negative_base": self._fit_base_single(
-                    X_train.loc[neg_mask], predictors, variable
+                    X_train.loc[neg_mask],
+                    predictors,
+                    variable,
+                    not_numeric_categorical=not_numeric_categorical,
                 ),
             }
 
@@ -382,10 +404,16 @@ class ZeroInflatedImputer(Imputer):
                 "kind": "three_sign",
                 "classifier": clf,
                 "positive_base": self._fit_base_single(
-                    X_train.loc[pos_mask], predictors, variable
+                    X_train.loc[pos_mask],
+                    predictors,
+                    variable,
+                    not_numeric_categorical=not_numeric_categorical,
                 ),
                 "negative_base": self._fit_base_single(
-                    X_train.loc[neg_mask], predictors, variable
+                    X_train.loc[neg_mask],
+                    predictors,
+                    variable,
+                    not_numeric_categorical=not_numeric_categorical,
                 ),
             }
 
@@ -396,6 +424,7 @@ class ZeroInflatedImputer(Imputer):
         X_train: pd.DataFrame,
         predictors: List[str],
         variable: str,
+        not_numeric_categorical: Optional[List[str]] = None,
     ) -> ImputerResults:
         """Fit a single base Imputer on a (possibly filtered) slice."""
         imputer = self.base_imputer_class(
@@ -406,6 +435,7 @@ class ZeroInflatedImputer(Imputer):
             X_train=X_train,
             predictors=predictors,
             imputed_variables=[variable],
+            not_numeric_categorical=not_numeric_categorical,
         )
 
 

{microimpute-3.1.0 → microimpute-3.1.2}/microimpute.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: microimpute
-Version: 3.1.0
+Version: 3.1.2
 Summary: Benchmarking imputation methods for microdata
 Author-email: María Juaristi <juaristi@uni.minerva.edu>, Nikhil Woodruff <nikhil.woodruff@outlook.com>
 Requires-Python: <3.15,>=3.12
@@ -21,7 +21,7 @@ Requires-Dist: psutil
 Provides-Extra: dev
 Requires-Dist: pytest<10.0.0,>=8.0.0; extra == "dev"
 Requires-Dist: pytest-cov<8.0.0,>=6.0.0; extra == "dev"
-Requires-Dist: ruff>=0.9.0; extra == "dev"
+Requires-Dist: ruff<0.17.0,>=0.16.0; extra == "dev"
 Requires-Dist: mypy<2.0.0,>=1.2.3; extra == "dev"
 Requires-Dist: build<2.0.0,>=1.2.0; extra == "dev"
 Requires-Dist: towncrier>=24.8.0; extra == "dev"

{microimpute-3.1.0 → microimpute-3.1.2}/microimpute.egg-info/requires.txt

@@ -15,7 +15,7 @@ psutil
 [dev]
 pytest<10.0.0,>=8.0.0
 pytest-cov<8.0.0,>=6.0.0
-ruff>=0.9.0
+ruff<0.17.0,>=0.16.0
 mypy<2.0.0,>=1.2.3
 build<2.0.0,>=1.2.0
 towncrier>=24.8.0

{microimpute-3.1.0 → microimpute-3.1.2}/pyproject.toml

@@ -7,7 +7,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "microimpute"
-version = "3.1.0"
+version = "3.1.2"
 description = "Benchmarking imputation methods for microdata"
 readme = "README.md"
 authors = [
@@ -35,7 +35,7 @@ dependencies = [
 dev = [
     "pytest>=8.0.0,<10.0.0",
     "pytest-cov>=6.0.0,<8.0.0",
-    "ruff>=0.9.0",
+    "ruff>=0.16.0,<0.17.0",
     "mypy>=1.2.3,<2.0.0",
     "build>=1.2.0,<2.0.0",
     "towncrier>=24.8.0",
@@ -101,3 +101,9 @@ showcontent = true
 directory = "removed"
 name = "Removed"
 showcontent = true
+
+[tool.ruff]
+# Documentation is prose, not a lint surface. ruff 0.16.0 began formatting
+# Python inside markdown code blocks, which reformatted five docs files that
+# nobody had edited and turned the Lint job red on an untouched checkout.
+extend-exclude = ["docs/**/*.md"]