From bd9add4557ec0213adf1bd691f5e8dca9970b99e Mon Sep 17 00:00:00 2001
From: Martin Stancsics <martin.stancsics@quantco.com>
Date: Thu, 3 Aug 2023 15:30:16 +0200
Subject: [PATCH 1/2] Do not use _set_up_... in covariance_matrix

---
 src/glum/_glm.py | 37 +++++++++++++++++++++----------------
 1 file changed, 21 insertions(+), 16 deletions(-)

diff --git a/src/glum/_glm.py b/src/glum/_glm.py
index db8a2086..55171258 100644
--- a/src/glum/_glm.py
+++ b/src/glum/_glm.py
@@ -1456,26 +1456,31 @@ def covariance_matrix(
            Cambridge university press
 
         """
-        (
-            X,
-            y,
-            sample_weight,
-            offset,
-            sum_weights,
-            P1,
-            P2,
-        ) = self._set_up_and_check_fit_args(
+
+        if isinstance(X, pd.DataFrame) and hasattr(self, "feature_dtypes_"):
+            X = _align_df_categories(X, self.feature_dtypes_)
+
+        X, y = check_X_y_tabmat_compliant(
             X,
             y,
-            sample_weight,
-            offset,
-            solver=self.solver,
-            force_all_finite=self.force_all_finite,
+            accept_sparse=["csr", "csc", "coo"],
+            dtype="numeric",
+            copy=self._should_copy_X(),
+            ensure_2d=True,
+            allow_nd=False,
+            drop_first=self.drop_first,
         )
 
-        # Here we don't want sample_weight to be normalized to sum up to 1
-        # We want sample_weight to sum up to the number of samples
-        sample_weight = sample_weight * sum_weights
+        if isinstance(X, np.ndarray):
+            X = tm.DenseMatrix(X)
+        if sparse.issparse(X) and not isinstance(X, tm.SparseMatrix):
+            X = tm.SparseMatrix(X)
+
+        sample_weight = _check_weights(
+            sample_weight, y.shape[0], X.dtype, force_all_finite=self.force_all_finite
+        )
+        sum_weights = np.sum(sample_weight)
+        offset = _check_offset(offset, y.shape[0], X.dtype)
 
         mu = self.predict(X, offset=offset) if mu is None else np.asanyarray(mu)
 

From 93c0b25cb9b5b4b051e51df9a953ae369c52a06f Mon Sep 17 00:00:00 2001
From: Martin Stancsics <martin.stancsics@quantco.com>
Date: Thu, 3 Aug 2023 16:15:03 +0200
Subject: [PATCH 2/2] Add changelog entry

---
 CHANGELOG.rst | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/CHANGELOG.rst b/CHANGELOG.rst
index b63f3376..508b78e9 100644
--- a/CHANGELOG.rst
+++ b/CHANGELOG.rst
@@ -14,6 +14,10 @@ Changelog
 
 - Added the complementary log-log (`cloglog`) link function.
 
+**Bug fix**
+
+- Fixed :meth:`~glum.GeneralizedLinearRegressorBase.covariance_matrix` mutating feature names when called with a data frame. See `here <https://github.com/Quantco/glum/issues/669>`_.
+
 **Other changes:**
 
 - When computing the covariance matrix, check for ill-conditionedness for all types of input. Furthermore, do it in a more efficient way.