Fix bug in SplitMatrix matvec

CHANGELOG.rst

@@ -7,6 +7,14 @@
 Changelog
 =========
 
+3.0.6 - 2021-11-11
+------------------
+
+**Bug fix**
+
+- We fixed a bug in :meth:`tabmat.SplitMatrix.matvec`, where incorrect matrix vector products were computed when a ``SplitMatrix`` did not contain any dense components.
+
+
 3.0.5 - 2021-11-05
 ------------------
 
@@ -184,7 +192,7 @@ We are trying to make releases for Windows.
 - Fix a bug in `matvec` for categorical components when the number of categories exceeds the number of rows.
 
 
-0.0.6 - 2020-08-03 
+0.0.6 - 2020-08-03
 ------------------
 
 See git history.

src/tabmat/split_matrix.py

@@ -343,14 +343,15 @@ def matvec(
         # as the target for storing the final output. This reduces the number
         # of output arrays allocated from 2 to 1.
         is_matrix_dense = [isinstance(m, DenseMatrix) for m in self.matrices]
-        dense_matrix_idx = np.argmax(is_matrix_dense)
         if np.any(is_matrix_dense):
+            dense_matrix_idx = np.argmax(is_matrix_dense)
             sub_cols = subset_cols[dense_matrix_idx]
             idx = self.indices[dense_matrix_idx]
             mat = self.matrices[dense_matrix_idx]
             in_vec = v[idx, ...]
             out = np.asarray(mat.matvec(in_vec, sub_cols, out), dtype=out_dtype)
         else:
+            dense_matrix_idx = -1
             out = _prepare_out_array(out, out_shape, out_dtype)
 
         for i, (sub_cols, idx, mat) in enumerate(

tests/test_split_matrix.py

@@ -1,10 +1,12 @@
 from typing import List, Optional, Union
 
 import numpy as np
+import pandas as pd
 import pytest
 import scipy.sparse as sps
 
 import tabmat as tm
+from tabmat import from_pandas
 from tabmat.constructor import _split_sparse_and_dense_parts
 from tabmat.dense_matrix import DenseMatrix
 from tabmat.ext.sparse import csr_dense_sandwich
@@ -237,3 +239,15 @@ def test_init_from_1d():
 
     res = SplitMatrix([m1, m2])
     assert res.shape == (10, 3)
+
+
+@pytest.mark.parametrize("n_rows", [5, 10, 25])
+def test_matvec(n_rows):
+    np.random.seed(1234)
+    n_cols = 2
+    categories = [f"cat_{val}" for val in range(5)]
+    X = pd.DataFrame(np.random.choice(categories, size=(n_rows, n_cols))).astype(
+        "category"
+    )
+    mat = from_pandas(X, cat_threshold=0)
+    np.testing.assert_allclose(mat.matvec(np.array(mat.shape[1] * [1])), n_cols)