matrix-utils 0.7.0__tar.gz → 0.7.1__tar.gz

{matrix_utils-0.7.0 → matrix_utils-0.7.1}/CHANGELOG.md

@@ -1,5 +1,9 @@
 # Changelog
 
+### [0.7.1] - 2026-03-20
+
+* Fix all elements being masked return sparse matrix instead of array when indexing
+
 ### [0.7.0] - 2026-02-27
 
 * Merge [#33](https://github.com/brightway-lca/matrix_utils/pull/33): Speed up large-index mapping by avoiding repeated unique and using fast integer dedupicator. Thanks @romainsacchi!

{matrix_utils-0.7.0 → matrix_utils-0.7.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: matrix_utils
-Version: 0.7.0
+Version: 0.7.1
 Summary: Tools to create matrices from data packages
 Author-email: Chris Mutel <cmutel@gmail.com>
 Maintainer-email: Chris Mutel <cmutel@gmail.com>

{matrix_utils-0.7.0 → matrix_utils-0.7.1}/matrix_utils/__init__.py

@@ -11,7 +11,7 @@ __all__ = (
     "SparseMatrixDict",
 )
 
-__version__ = "0.7.0"
+__version__ = "0.7.1"
 
 from matrix_utils.array_mapper import ArrayMapper
 from matrix_utils.indexers import CombinatorialIndexer, Proxy, RandomIndexer, SequentialIndexer

{matrix_utils-0.7.0 → matrix_utils-0.7.1}/matrix_utils/array_mapper.py

@@ -26,8 +26,6 @@ class ArrayMapper:
     """  # NOQA: E501
 
     def __init__(self, *, array: np.ndarray, sparse_cutoff: int = 50000, empty_ok: bool = False):
-
-
         self._check_input_array(array)
 
         # `np.unique` can be very slow on large, unsorted integer arrays.
@@ -86,6 +84,8 @@ class ArrayMapper:
 
         result = np.zeros_like(array) - 1
         mask = array <= self.max_value
+        if not mask.any():
+            return result
         # https://numpy.org/doc/stable/reference/generated/numpy.matrix.A1.html
         result[mask] = np.asarray(self.matrix[array[mask], np.zeros_like(array[mask])]).ravel() - 1
         return result

{matrix_utils-0.7.0 → matrix_utils-0.7.1}/matrix_utils.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: matrix_utils
-Version: 0.7.0
+Version: 0.7.1
 Summary: Tools to create matrices from data packages
 Author-email: Chris Mutel <cmutel@gmail.com>
 Maintainer-email: Chris Mutel <cmutel@gmail.com>

{matrix_utils-0.7.0 → matrix_utils-0.7.1}/tests/array_mapper.py

@@ -1,5 +1,6 @@
 import numpy as np
 import pytest
+from scipy import sparse
 
 import matrix_utils.array_mapper as array_mapper_module
 from matrix_utils.array_mapper import ArrayMapper
@@ -92,6 +93,41 @@ def test_reverse_dict():
     assert am.reverse_dict() == {0: 10, 1: 20, 2: 30}
 
 
+def test_csc_empty_fancy_index_returns_sparse():
+    """Document scipy behaviour: indexing a CSC matrix with two empty arrays returns
+    a sparse matrix of shape (1, 0), NOT an empty dense array.
+
+    This is the root cause of the bug fixed in ``map_array``: if this behaviour
+    changes in a future scipy release the guard ``if not mask.any()`` may no
+    longer be necessary, but removing it would still be safe.
+    """
+    m = sparse.coo_matrix(([1], ([2], [0])), shape=(5, 1)).tocsc()
+    empty = np.array([], dtype=np.intp)
+    result = m[empty, empty]
+    assert sparse.issparse(result), (
+        f"Expected a sparse matrix but got {type(result)}. "
+        "scipy may have changed its fancy-indexing behaviour for CSC matrices — "
+        "re-evaluate whether the `if not mask.any()` guard in map_array is still needed."
+    )
+    assert result.shape == (1, 0), f"Expected shape (1, 0) but got {result.shape}."
+
+
+def test_all_values_exceed_max_value():
+    # Regression: when every lookup value exceeds max_value the mask is all-False,
+    # which causes csc_matrix[[], []] to return a (1, 0) sparse matrix whose
+    # np.asarray().ravel() has shape (1,) instead of (0,) — corrupting the result.
+    am = ArrayMapper(array=np.array([1, 2, 3]))
+    result = am.map_array(np.array([100, 200, 300]))
+    assert np.array_equal(result, [-1, -1, -1])
+
+
+def test_mixed_in_range_and_out_of_range():
+    # Some values in range, some beyond max_value — make sure both are handled.
+    am = ArrayMapper(array=np.array([1, 3, 5]))
+    result = am.map_array(np.array([1, 99, 3, 88, 5]))
+    assert np.array_equal(result, [0, -1, 1, -1, 2])
+
+
 def test_large_integer_array_pandas_fallback(monkeypatch):
     # Trigger the large integer optimization branch and force fallback
     inpt = np.tile(np.array([5, 2, 7, 2, 5], dtype=np.int64), 25_000)