Add freeze test.

environment-win.yml

@@ -16,6 +16,9 @@ dependencies:
   - ruff
   - setuptools_scm
 
+  # for test data
+  - scikit-learn
+
   # build tools
   - c-compiler
   - cxx-compiler

environment.yml

@@ -15,6 +15,9 @@ dependencies:
   - ruff
   - setuptools_scm
 
+  # for test data
+  - scikit-learn
+
   # build tools
   - c-compiler
   - cxx-compiler
@@ -31,4 +34,4 @@ dependencies:
   - seaborn-base
   - sphinx
   - sphinx_rtd_theme
-  - sphinxcontrib-apidoc
+  - sphinxcontrib-apidoc

tests/test_categorical_matrix.py

@@ -2,6 +2,7 @@
 import pandas as pd
 import pytest
 
+import tabmat as tm
 from tabmat.categorical_matrix import CategoricalMatrix
 
 
@@ -71,3 +72,23 @@ def test_categorical_indexing(drop_first):
     mat = CategoricalMatrix(catvec, drop_first=drop_first)
     expected = pd.get_dummies(catvec, drop_first=drop_first).to_numpy()[:, [0, 1]]
     np.testing.assert_allclose(mat[:, [0, 1]].A, expected)
+
+
+# The only categorical variable (zipcode) has 70 levels in the data; the test
+# fails when we set the cat treshold to 70.
+@pytest.mark.parametrize("cat_threshold", [70, 71])
+def test_freeze(cat_threshold):
+    pytest.importorskip("sklearn")
+
+    from sklearn.datasets import fetch_openml
+
+    house_data = fetch_openml(name="house_sales", version=3, as_frame=True)
+    X = house_data.data
+
+    weights = np.ones(len(X)) / len(X)
+
+    mat = tm.from_pandas(X, cat_threshold=cat_threshold)
+
+    np.testing.assert_array_equal(
+        mat.transpose_matvec(weights), mat.transpose_matvec(weights)
+    )