Fix CACG fit array-like sample inputs

src/pyrecest/distributions/hypersphere_subset/complex_angular_central_gaussian_distribution.py

@@ -183,6 +183,14 @@ def estimate_parameter_matrix(Z, n_iterations=100):
         C : array-like of shape (d, d)
             Estimated Hermitian parameter matrix.
         """
+        Z = array(Z)
+        if Z.ndim != 2:
+            raise ValueError("Z must be a two-dimensional array of samples")
+        if Z.shape[0] == 0 or Z.shape[1] == 0:
+            raise ValueError("Z must contain at least one sample and one dimension")
+        if not _to_python_bool(backend_all(isfinite(Z))):
+            raise ValueError("Z must contain only finite values")
+
         N = Z.shape[0]
         D = Z.shape[1]
         C = eye(D, dtype=complex128)

tests/distributions/test_complex_angular_central_gaussian_distribution.py

@@ -227,6 +227,20 @@ def test_fit_returns_distribution(self):
         self.assertIsInstance(dist, ComplexAngularCentralGaussianDistribution)
         self.assertEqual(dist.dim, 2)
 
+    @unittest.skipIf(
+        pyrecest.backend.__backend_name__ == "jax",
+        reason="Not supported on JAX backend",
+    )  # pylint: disable=no-member
+    def test_fit_accepts_array_like_samples(self):
+        """fit() should coerce array-like sample collections before reading shape."""
+        Z = [[1.0 + 0.0j, 0.0 + 0.0j], [0.0 + 0.0j, 1.0 + 0.0j]]
+
+        dist = ComplexAngularCentralGaussianDistribution.fit(Z, n_iterations=1)
+
+        self.assertIsInstance(dist, ComplexAngularCentralGaussianDistribution)
+        self.assertEqual(dist.dim, 2)
+        npt.assert_allclose(real(dist.C), 0.5 * eye(2), atol=1e-12)
+
     @unittest.skipIf(
         pyrecest.backend.__backend_name__ == "jax",
         reason="Not supported on JAX backend",