Merge pull request #86 from asmeurer/iter_indices-linalg

Use ndindex.iter_indices in _test_stacks in the linalg tests

Author: honno

.github/workflows/numpy.yml

@@ -34,8 +34,9 @@ jobs:
         array_api_tests/test_creation_functions.py::test_linspace
         # https://github.com/numpy/numpy/issues/20870
         array_api_tests/test_data_type_functions.py::test_can_cast
-        # linalg tests generally need more mulling over
-        array_api_tests/test_linalg.py
+        # The return dtype for trace is not consistent in the spec
+        # (https://github.com/data-apis/array-api/issues/202#issuecomment-952529197)
+        array_api_tests/test_linalg.py::test_trace
         # waiting on NumPy to allow/revert distinct NaNs for np.unique
         # https://github.com/numpy/numpy/issues/20326#issuecomment-1012380448
         array_api_tests/test_set_functions.py

array_api_tests/test_linalg.py

@@ -17,8 +17,9 @@
 from hypothesis import assume, given
 from hypothesis.strategies import (booleans, composite, none, tuples, integers,
                                    shared, sampled_from, data, just)
+from ndindex import iter_indices
 
-from .array_helpers import assert_exactly_equal, asarray, equal, zero, infinity
+from .array_helpers import assert_exactly_equal, asarray
 from .hypothesis_helpers import (xps, dtypes, shapes, kwargs, matrix_shapes,
                                  square_matrix_shapes, symmetric_matrices,
                                  positive_definite_matrices, MAX_ARRAY_SIZE,
@@ -43,29 +44,41 @@
 # Standin strategy for not yet implemented tests
 todo = none()
 
-def _test_stacks(f, *args, res=None, dims=2, true_val=None, **kw):
+def _test_stacks(f, *args, res=None, dims=2, true_val=None, matrix_axes=(-2, -1),
+                 assert_equal=assert_exactly_equal, **kw):
     """
     Test that f(*args, **kw) maps across stacks of matrices
 
-    dims is the number of dimensions f should have for a single n x m matrix
-    stack.
+    dims is the number of dimensions f(*args) should have for a single n x m
+    matrix stack.
+
+    matrix_axes are the axes along which matrices (or vectors) are stacked in
+    the input.
+
+    true_val may be a function such that true_val(*x_stacks, **kw) gives the
+    true value for f on a stack.
+
+    res should be the result of f(*args, **kw). It is computed if not passed
+    in.
 
-    true_val may be a function such that true_val(*x_stacks) gives the true
-    value for f on a stack
     """
     if res is None:
         res = f(*args, **kw)
 
-    shape = args[0].shape if len(args) == 1 else broadcast_shapes(*[x.shape
-                                                                    for x in args])
-    for _idx in sh.ndindex(shape[:-2]):
-        idx = _idx + (slice(None),)*dims
-        res_stack = res[idx]
-        x_stacks = [x[_idx + (...,)] for x in args]
+    shapes = [x.shape for x in args]
+
+    for (x_idxes, (res_idx,)) in zip(
+            iter_indices(*shapes, skip_axes=matrix_axes),
+            iter_indices(res.shape, skip_axes=tuple(range(-dims, 0)))):
+        x_idxes = [x_idx.raw for x_idx in x_idxes]
+        res_idx = res_idx.raw
+
+        res_stack = res[res_idx]
+        x_stacks = [x[x_idx] for x, x_idx in zip(args, x_idxes)]
         decomp_res_stack = f(*x_stacks, **kw)
-        assert_exactly_equal(res_stack, decomp_res_stack)
+        assert_equal(res_stack, decomp_res_stack)
         if true_val:
-            assert_exactly_equal(decomp_res_stack, true_val(*x_stacks))
+            assert_equal(decomp_res_stack, true_val(*x_stacks))
 
 def _test_namedtuple(res, fields, func_name):
     """
@@ -452,10 +465,12 @@ def test_slogdet(x):
 
     # Check that when the determinant is 0, the sign and logabsdet are (0,
     # -inf).
-    d = linalg.det(x)
-    zero_det = equal(d, zero(d.shape, d.dtype))
-    assert_exactly_equal(sign[zero_det], zero(sign[zero_det].shape, x.dtype))
-    assert_exactly_equal(logabsdet[zero_det], -infinity(logabsdet[zero_det].shape, x.dtype))
+    # TODO: This test does not necessarily hold exactly. Update it to test it
+    # approximately.
+    # d = linalg.det(x)
+    # zero_det = equal(d, zero(d.shape, d.dtype))
+    # assert_exactly_equal(sign[zero_det], zero(sign[zero_det].shape, x.dtype))
+    # assert_exactly_equal(logabsdet[zero_det], -infinity(logabsdet[zero_det].shape, x.dtype))
 
     # More generally, det(x) should equal sign*exp(logabsdet), but this does
     # not hold exactly due to floating-point loss of precision.
@@ -614,7 +629,7 @@ def true_trace(x_stack):
 
 @given(
     dtypes=mutually_promotable_dtypes(dtypes=dh.numeric_dtypes),
-    shape=shapes(),
+    shape=shapes(min_dims=1),
     data=data(),
 )
 def test_vecdot(dtypes, shape, data):

requirements.txt

@@ -1,4 +1,5 @@
 pytest
 hypothesis>=6.31.1
+ndindex>=1.6
 regex
 removestar