unitx 0.16.1__tar.gz → 0.16.2__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (177) hide show
  1. {unitx-0.16.1 → unitx-0.16.2}/.github/consumers/arrays.py +17 -1
  2. {unitx-0.16.1 → unitx-0.16.2}/CHANGELOG.md +12 -0
  3. {unitx-0.16.1 → unitx-0.16.2}/PKG-INFO +1 -1
  4. {unitx-0.16.1 → unitx-0.16.2}/docs/numpy/selection.md +18 -0
  5. {unitx-0.16.1 → unitx-0.16.2}/docs/numpy/statistics.md +30 -0
  6. {unitx-0.16.1 → unitx-0.16.2}/docs/numpy/writes.md +1 -1
  7. {unitx-0.16.1 → unitx-0.16.2}/docs/numpy.md +8 -7
  8. {unitx-0.16.1 → unitx-0.16.2}/pyproject.toml +1 -1
  9. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_numerics/numpy.py +32 -18
  10. {unitx-0.16.1 → unitx-0.16.2}/tests/numpy_typing/check_numpy.py +10 -1
  11. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_affine.py +2 -0
  12. unitx-0.16.2/tests/test_numpy_nan_reductions.py +260 -0
  13. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_outputs.py +1 -1
  14. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_products.py +23 -23
  15. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_statistics.py +13 -3
  16. {unitx-0.16.1 → unitx-0.16.2}/.github/consumers/pyrightconfig.json +0 -0
  17. {unitx-0.16.1 → unitx-0.16.2}/.github/consumers/scalars.py +0 -0
  18. {unitx-0.16.1 → unitx-0.16.2}/.github/workflows/ci.yml +0 -0
  19. {unitx-0.16.1 → unitx-0.16.2}/.github/workflows/pages.yml +0 -0
  20. {unitx-0.16.1 → unitx-0.16.2}/.github/workflows/release.yml +0 -0
  21. {unitx-0.16.1 → unitx-0.16.2}/.gitignore +0 -0
  22. {unitx-0.16.1 → unitx-0.16.2}/.readthedocs.yaml +0 -0
  23. {unitx-0.16.1 → unitx-0.16.2}/CONTRIBUTING.md +0 -0
  24. {unitx-0.16.1 → unitx-0.16.2}/LICENSE +0 -0
  25. {unitx-0.16.1 → unitx-0.16.2}/README.md +0 -0
  26. {unitx-0.16.1 → unitx-0.16.2}/benchmarks/README.md +0 -0
  27. {unitx-0.16.1 → unitx-0.16.2}/benchmarks/numpy_writes.py +0 -0
  28. {unitx-0.16.1 → unitx-0.16.2}/benchmarks/scalars.py +0 -0
  29. {unitx-0.16.1 → unitx-0.16.2}/conftest.py +0 -0
  30. {unitx-0.16.1 → unitx-0.16.2}/docs/README.md +0 -0
  31. {unitx-0.16.1 → unitx-0.16.2}/docs/affine.md +0 -0
  32. {unitx-0.16.1 → unitx-0.16.2}/docs/architecture.md +0 -0
  33. {unitx-0.16.1 → unitx-0.16.2}/docs/catalogues/README.md +0 -0
  34. {unitx-0.16.1 → unitx-0.16.2}/docs/catalogues/cgs.md +0 -0
  35. {unitx-0.16.1 → unitx-0.16.2}/docs/catalogues/imperial.md +0 -0
  36. {unitx-0.16.1 → unitx-0.16.2}/docs/catalogues/si.md +0 -0
  37. {unitx-0.16.1 → unitx-0.16.2}/docs/custom-units.md +0 -0
  38. {unitx-0.16.1 → unitx-0.16.2}/docs/display.md +0 -0
  39. {unitx-0.16.1 → unitx-0.16.2}/docs/images/blackbody.svg +0 -0
  40. {unitx-0.16.1 → unitx-0.16.2}/docs/javascripts/mermaid.mjs +0 -0
  41. {unitx-0.16.1 → unitx-0.16.2}/docs/logarithmic.md +0 -0
  42. {unitx-0.16.1 → unitx-0.16.2}/docs/migration.md +0 -0
  43. {unitx-0.16.1 → unitx-0.16.2}/docs/numpy/calculus.md +0 -0
  44. {unitx-0.16.1 → unitx-0.16.2}/docs/requirements.txt +0 -0
  45. {unitx-0.16.1 → unitx-0.16.2}/docs/scalars.md +0 -0
  46. {unitx-0.16.1 → unitx-0.16.2}/docs/serialization.md +0 -0
  47. {unitx-0.16.1 → unitx-0.16.2}/docs/stdlib.md +0 -0
  48. {unitx-0.16.1 → unitx-0.16.2}/docs/typing.md +0 -0
  49. {unitx-0.16.1 → unitx-0.16.2}/docs/units.md +0 -0
  50. {unitx-0.16.1 → unitx-0.16.2}/examples/README.md +0 -0
  51. {unitx-0.16.1 → unitx-0.16.2}/examples/arrhenius_fit.ipynb +0 -0
  52. {unitx-0.16.1 → unitx-0.16.2}/examples/blackbody.ipynb +0 -0
  53. {unitx-0.16.1 → unitx-0.16.2}/examples/damped_oscillator.ipynb +0 -0
  54. {unitx-0.16.1 → unitx-0.16.2}/examples/gaussian_wavepacket.ipynb +0 -0
  55. {unitx-0.16.1 → unitx-0.16.2}/examples/impedance.md +0 -0
  56. {unitx-0.16.1 → unitx-0.16.2}/examples/impedance.py +0 -0
  57. {unitx-0.16.1 → unitx-0.16.2}/examples/optical_attenuation.ipynb +0 -0
  58. {unitx-0.16.1 → unitx-0.16.2}/examples/pendulum_uncertainty.ipynb +0 -0
  59. {unitx-0.16.1 → unitx-0.16.2}/examples/smith_chart.ipynb +0 -0
  60. {unitx-0.16.1 → unitx-0.16.2}/mkdocs.yml +0 -0
  61. {unitx-0.16.1 → unitx-0.16.2}/pyrightconfig.json +0 -0
  62. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/__init__.py +0 -0
  63. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_constnamespace.py +0 -0
  64. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_coordinates/__init__.py +0 -0
  65. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_coordinates/affine.py +0 -0
  66. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_coordinates/definition.py +0 -0
  67. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_coordinates/logarithmic.py +0 -0
  68. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_coordinates/logarithms.py +0 -0
  69. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_coordinates/points.py +0 -0
  70. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_coordinates/resolve.py +0 -0
  71. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_electromagnetism/__init__.py +0 -0
  72. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_electromagnetism/coefficients.py +0 -0
  73. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_electromagnetism/conversion.py +0 -0
  74. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_formatting/__init__.py +0 -0
  75. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_formatting/render.py +0 -0
  76. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_formatting/unicode.py +0 -0
  77. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_immutable.py +0 -0
  78. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_numerics/__init__.py +0 -0
  79. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_numerics/conversion.py +0 -0
  80. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_numerics/numpy_typing.py +0 -0
  81. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_numerics/scaling.py +0 -0
  82. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_operands.py +0 -0
  83. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_registry/__init__.py +0 -0
  84. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_registry/builtin_units.py +0 -0
  85. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_registry/prefixes.py +0 -0
  86. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/_registry/reconstruction.py +0 -0
  87. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/affinequantity.py +0 -0
  88. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/affineunit.py +0 -0
  89. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/cgs/__init__.py +0 -0
  90. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/cgs/units.py +0 -0
  91. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/dimension.py +0 -0
  92. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/exceptions.py +0 -0
  93. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/factor.py +0 -0
  94. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/imperial/__init__.py +0 -0
  95. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/imperial/units.py +0 -0
  96. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/logquantity.py +0 -0
  97. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/logunit.py +0 -0
  98. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/math.py +0 -0
  99. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/monomial.py +0 -0
  100. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/py.typed +0 -0
  101. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/quantity.py +0 -0
  102. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/serialization.py +0 -0
  103. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/si/__init__.py +0 -0
  104. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/si/bases.py +0 -0
  105. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/si/constants.py +0 -0
  106. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/si/units.py +0 -0
  107. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/singleunit.py +0 -0
  108. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/time.py +0 -0
  109. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/unit.py +0 -0
  110. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/unitcatalog.py +0 -0
  111. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/unitdefinition.py +0 -0
  112. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/unitparser.py +0 -0
  113. {unitx-0.16.1 → unitx-0.16.2}/src/unitx/unitregistry.py +0 -0
  114. {unitx-0.16.1 → unitx-0.16.2}/tests/__init__.py +0 -0
  115. {unitx-0.16.1 → unitx-0.16.2}/tests/_notebooks.py +0 -0
  116. {unitx-0.16.1 → unitx-0.16.2}/tests/numpy_typing/check_asarray.py +0 -0
  117. {unitx-0.16.1 → unitx-0.16.2}/tests/numpy_typing/check_cgs.py +0 -0
  118. {unitx-0.16.1 → unitx-0.16.2}/tests/numpy_typing/check_logarithmic.py +0 -0
  119. {unitx-0.16.1 → unitx-0.16.2}/tests/numpy_typing/check_math.py +0 -0
  120. {unitx-0.16.1 → unitx-0.16.2}/tests/test_affine.py +0 -0
  121. {unitx-0.16.1 → unitx-0.16.2}/tests/test_affinequantity.py +0 -0
  122. {unitx-0.16.1 → unitx-0.16.2}/tests/test_builtin_definitions.py +0 -0
  123. {unitx-0.16.1 → unitx-0.16.2}/tests/test_catalog_collection.py +0 -0
  124. {unitx-0.16.1 → unitx-0.16.2}/tests/test_cgs.py +0 -0
  125. {unitx-0.16.1 → unitx-0.16.2}/tests/test_cgs_numpy.py +0 -0
  126. {unitx-0.16.1 → unitx-0.16.2}/tests/test_cgs_systems.py +0 -0
  127. {unitx-0.16.1 → unitx-0.16.2}/tests/test_decibel_call.py +0 -0
  128. {unitx-0.16.1 → unitx-0.16.2}/tests/test_definition_scope.py +0 -0
  129. {unitx-0.16.1 → unitx-0.16.2}/tests/test_dimension.py +0 -0
  130. {unitx-0.16.1 → unitx-0.16.2}/tests/test_display.py +0 -0
  131. {unitx-0.16.1 → unitx-0.16.2}/tests/test_examples.py +0 -0
  132. {unitx-0.16.1 → unitx-0.16.2}/tests/test_factor.py +0 -0
  133. {unitx-0.16.1 → unitx-0.16.2}/tests/test_immutable.py +0 -0
  134. {unitx-0.16.1 → unitx-0.16.2}/tests/test_impedance_example.py +0 -0
  135. {unitx-0.16.1 → unitx-0.16.2}/tests/test_imperial.py +0 -0
  136. {unitx-0.16.1 → unitx-0.16.2}/tests/test_import.py +0 -0
  137. {unitx-0.16.1 → unitx-0.16.2}/tests/test_linear_boundaries.py +0 -0
  138. {unitx-0.16.1 → unitx-0.16.2}/tests/test_logarithmic_model.py +0 -0
  139. {unitx-0.16.1 → unitx-0.16.2}/tests/test_math.py +0 -0
  140. {unitx-0.16.1 → unitx-0.16.2}/tests/test_math_numpy.py +0 -0
  141. {unitx-0.16.1 → unitx-0.16.2}/tests/test_monomial.py +0 -0
  142. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numeric.py +0 -0
  143. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy.py +0 -0
  144. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_differentiation.py +0 -0
  145. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_integration.py +0 -0
  146. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_interpolation.py +0 -0
  147. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_multi_outputs.py +0 -0
  148. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_mutation.py +0 -0
  149. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_nan_statistics.py +0 -0
  150. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_selection.py +0 -0
  151. {unitx-0.16.1 → unitx-0.16.2}/tests/test_numpy_weighted_statistics.py +0 -0
  152. {unitx-0.16.1 → unitx-0.16.2}/tests/test_quantity.py +0 -0
  153. {unitx-0.16.1 → unitx-0.16.2}/tests/test_registry.py +0 -0
  154. {unitx-0.16.1 → unitx-0.16.2}/tests/test_registry_scope.py +0 -0
  155. {unitx-0.16.1 → unitx-0.16.2}/tests/test_serialization.py +0 -0
  156. {unitx-0.16.1 → unitx-0.16.2}/tests/test_si.py +0 -0
  157. {unitx-0.16.1 → unitx-0.16.2}/tests/test_singleunit.py +0 -0
  158. {unitx-0.16.1 → unitx-0.16.2}/tests/test_siunitx_display.py +0 -0
  159. {unitx-0.16.1 → unitx-0.16.2}/tests/test_time.py +0 -0
  160. {unitx-0.16.1 → unitx-0.16.2}/tests/test_typing.py +0 -0
  161. {unitx-0.16.1 → unitx-0.16.2}/tests/test_unicode.py +0 -0
  162. {unitx-0.16.1 → unitx-0.16.2}/tests/test_unicode_units.py +0 -0
  163. {unitx-0.16.1 → unitx-0.16.2}/tests/test_unit.py +0 -0
  164. {unitx-0.16.1 → unitx-0.16.2}/tests/test_unit_definition.py +0 -0
  165. {unitx-0.16.1 → unitx-0.16.2}/tests/test_unitparser.py +0 -0
  166. {unitx-0.16.1 → unitx-0.16.2}/tests/test_unitregistry.py +0 -0
  167. {unitx-0.16.1 → unitx-0.16.2}/tests/test_warnings.py +0 -0
  168. {unitx-0.16.1 → unitx-0.16.2}/tests/typing/check_affine.py +0 -0
  169. {unitx-0.16.1 → unitx-0.16.2}/tests/typing/check_cgs.py +0 -0
  170. {unitx-0.16.1 → unitx-0.16.2}/tests/typing/check_impedance_example.py +0 -0
  171. {unitx-0.16.1 → unitx-0.16.2}/tests/typing/check_imperial.py +0 -0
  172. {unitx-0.16.1 → unitx-0.16.2}/tests/typing/check_logarithmic.py +0 -0
  173. {unitx-0.16.1 → unitx-0.16.2}/tests/typing/check_math_time.py +0 -0
  174. {unitx-0.16.1 → unitx-0.16.2}/tests/typing/check_quantity.py +0 -0
  175. {unitx-0.16.1 → unitx-0.16.2}/tests/typing/check_registry.py +0 -0
  176. {unitx-0.16.1 → unitx-0.16.2}/tests/typing/check_serialization.py +0 -0
  177. {unitx-0.16.1 → unitx-0.16.2}/tests/typing/check_si.py +0 -0
@@ -94,11 +94,27 @@ def check_arrays(values: NDArray[np.float64]) -> None:
94
94
  np.testing.assert_allclose(median_target.value, 2000)
95
95
 
96
96
  gapped = Quantity(np.array([1.0, np.nan, 3]), si.m)
97
- for operation in (np.nanmean, np.nanmedian, np.nanstd, np.nanvar):
97
+ for operation in (
98
+ np.nanmean,
99
+ np.nanmedian,
100
+ np.nanstd,
101
+ np.nanvar,
102
+ np.nanmin,
103
+ np.nanmax,
104
+ np.nansum,
105
+ ):
98
106
  result = cast(Quantity[Array], operation(gapped))
99
107
  assert isinstance(result, Quantity)
100
108
  assert result.unit == (si.m**2 if operation is np.nanvar else si.m)
101
109
  np.testing.assert_allclose(result.value, operation(gapped.value))
110
+ product = cast(Quantity[Array], np.nanprod(gapped))
111
+ assert isinstance(product, Quantity) and product.unit == si.m**2
112
+ np.testing.assert_allclose(product.value, 3)
113
+ for operation in (np.nanargmin, np.nanargmax):
114
+ indices = np.empty((), dtype=np.intp)
115
+ assert operation(gapped, out=indices) is indices
116
+ assert_type(indices.dtype, np.dtype[np.intp])
117
+ np.testing.assert_array_equal(indices, operation(gapped.value))
102
118
  for operation, probability in ((np.nanquantile, 0.5), (np.nanpercentile, 50)):
103
119
  result = cast(Quantity[Array], operation(gapped, probability))
104
120
  assert isinstance(result, Quantity) and result.unit == si.m
@@ -2,6 +2,18 @@
2
2
 
3
3
  ## [Unreleased]
4
4
 
5
+ ## [0.16.2] - 2026-10-25
6
+
7
+ ### Added
8
+
9
+ - Support NumPy `nanmin`, `nanmax`, `nanargmin` and `nanargmax`, preserving
10
+ units or point coordinates for extrema and returning bare indices. Retain
11
+ NumPy's all-NaN warnings/errors and stage output writes before mutation.
12
+ - Support `nansum` with the existing sum unit, initial and mask rules, and
13
+ `nanprod` with unit powers determined by selected non-NaN element counts.
14
+ Dimensional product slices require equal counts; dimensionless products
15
+ normalize to coherent values and may have different counts.
16
+
5
17
  ## [0.16.1] - 2026-10-25
6
18
 
7
19
  ### Added
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.5
2
2
  Name: unitx
3
- Version: 0.16.1
3
+ Version: 0.16.2
4
4
  Summary: Physical units, quantities, and unit conversion with exact unit algebra, type hints, and NumPy integration.
5
5
  Project-URL: Homepage, https://github.com/DaitTHU/unitx
6
6
  Project-URL: Documentation, https://daitthu.github.io/unitx/
@@ -50,9 +50,27 @@ Empty reductions, axes, NaNs and `keepdims` follow NumPy. These reductions
50
50
  write the exact result shape to `out`; elementwise selection and clipping
51
51
  allow the ufunc output broadcasting rules.
52
52
 
53
+ `np.nanmin/max` share the same unit, point, mask, initial and output rules,
54
+ while omitting NaNs. An all-NaN slice returns NaN and emits RuntimeWarning;
55
+ an empty reduction still needs an initial value. A warning promoted to an
56
+ exception leaves `out` unchanged.
57
+
58
+ `np.argmin/argmax` and `np.nanargmin/nanargmax` return bare indices and accept
59
+ one integer axis or `axis=None`, `keepdims`, and a bare integer ndarray `out`
60
+ of exactly the result shape. They also accept affine and logarithmic points
61
+ and operate on stored readings. The NaN variants omit NaNs and raise
62
+ ValueError for all-NaN slices before writing `out`. NumPy's caveat remains:
63
+ `nanargmin` results cannot be trusted for slices containing only NaNs and
64
+ positive infinity; the corresponding `nanargmax` case is NaNs and negative
65
+ infinity. unitx delegates these numerical choices to NumPy.
66
+
53
67
  See NumPy's [where](https://numpy.org/doc/stable/reference/generated/numpy.where.html),
54
68
  [clip](https://numpy.org/doc/stable/reference/generated/numpy.clip.html) and
55
69
  [minimum](https://numpy.org/doc/stable/reference/generated/numpy.minimum.html)
56
70
  references for the underlying numerical behavior.
71
+ See also [nanmin](https://numpy.org/doc/stable/reference/generated/numpy.nanmin.html),
72
+ [nanargmin](https://numpy.org/doc/stable/reference/generated/numpy.nanargmin.html)
73
+ and [nanargmax](https://numpy.org/doc/stable/reference/generated/numpy.nanargmax.html)
74
+ for NaN and infinity behavior.
57
75
 
58
76
  [NumPy overview and support matrix](../numpy.md) · [Documentation index](../README.md)
@@ -191,6 +191,22 @@ described here and uses the installed NumPy kernel.
191
191
 
192
192
  ## NaN statistics
193
193
 
194
+ `nansum` omits NaN samples and otherwise shares `sum`'s unit, `axis`, `dtype`,
195
+ `keepdims`, `where`, scalar `initial` and staged `out` rules. All-NaN, empty
196
+ and fully masked slices return zero in the input unit, or the converted
197
+ initial when supplied. Affine and logarithmic point sums remain undefined.
198
+
199
+ `nanprod` shares `prod`'s options and initial-unit rules. For dimensional
200
+ input U, its unit is `U**N * initial.unit`, where N counts elements selected
201
+ by `where` that are **not NaN**. Zero and infinity still count. All result
202
+ slices must have equal N, even when a zero initial makes their magnitudes
203
+ equal; otherwise DimensionError is raised before writing `out`. All-NaN,
204
+ empty and fully masked slices have N = 0, with value 1 and unit `Unit()`
205
+ before applying initial. When the output has no elements, use `prod`'s
206
+ nominal-extent convention. Dimensionless inputs normalize to coherent values
207
+ first, so unequal non-NaN counts are allowed. Point products remain undefined.
208
+ NaN initial values are passed through to NumPy; they are not omitted.
209
+
194
210
  `nanmean`, `nanmedian`, `nanstd`, `nanvar`, `nanquantile`, and `nanpercentile`
195
211
  omit NaN samples. Their units, supported options, parameter normalization and
196
212
  staged `out` writes match the corresponding functions above. Sample units,
@@ -235,11 +251,25 @@ assert variances.unit == Unit('m2')
235
251
  np.testing.assert_allclose(variances.value, [2, 2])
236
252
  medians = np.nanpercentile(gapped, 50, axis=1)
237
253
  np.testing.assert_allclose(medians.value, [2, 6])
254
+ totals = np.nansum(gapped, axis=1)
255
+ assert totals.unit == Unit('m')
256
+ np.testing.assert_array_equal(totals.value, [4, 12])
257
+ areas = np.nanprod(gapped, axis=1)
258
+ assert areas.unit == Unit('m2')
259
+ np.testing.assert_array_equal(areas.value, [3, 35])
260
+ minima = np.nanmin(gapped, axis=1)
261
+ assert minima.unit == Unit('m')
262
+ np.testing.assert_array_equal(minima.value, [1, 5])
263
+ np.testing.assert_array_equal(np.nanargmax(gapped, axis=1), [2, 2])
238
264
  ```
239
265
 
240
266
  See [NumPy nanmean](https://numpy.org/doc/stable/reference/generated/numpy.nanmean.html),
241
267
  [nanstd](https://numpy.org/doc/stable/reference/generated/numpy.nanstd.html) and
242
268
  [nanquantile](https://numpy.org/doc/stable/reference/generated/numpy.nanquantile.html)
243
269
  for the numerical definitions and warnings.
270
+ The [nansum](https://numpy.org/doc/stable/reference/generated/numpy.nansum.html)
271
+ and [nanprod](https://numpy.org/doc/stable/reference/generated/numpy.nanprod.html)
272
+ kernels determine NaN identities and accumulator behavior, with unitx's
273
+ additional single-unit rules above.
244
274
 
245
275
  [NumPy overview and support matrix](../numpy.md) · [Documentation index](../README.md)
@@ -125,7 +125,7 @@ length target. `Unit` factors participate in the conversion. Every numeric
125
125
  result, including a coherent or scaled dimensionless result, requires a Quantity
126
126
  target; writing it to a bare ndarray would discard its unit. Boolean predicates
127
127
  and comparisons instead accept only a bare boolean ndarray output.
128
- `argmin` and `argmax` accept only a bare integer ndarray output, with the
128
+ `argmin`, `argmax`, `nanargmin` and `nanargmax` accept only a bare integer ndarray output, with the
129
129
  same exact shape requirement as other function outputs.
130
130
 
131
131
  Final casting uses `casting='same_kind'` by default, or the explicitly requested
@@ -116,18 +116,18 @@ functions, including exp/log. Python scalar numerical semantics remain native.
116
116
  | --- | --- | --- |
117
117
  | reshape, transpose | Preserve unit | shape/order/copy; axes |
118
118
  | concatenate, stack | Convert to first Quantity's unit; bare arrays remain dimensionless | axis, dtype, casting, out; dtype and out are mutually exclusive |
119
- | sum | Preserve unit | axis, dtype, keepdims, initial, where, out |
119
+ | sum, nansum | Preserve unit | axis, dtype, keepdims, initial, where, out |
120
120
  | mean, nanmean | Preserve unit or point coordinate | axis, dtype, keepdims, where, out |
121
- | min/amin, max/amax | Preserve unit or point coordinate | axis, keepdims, initial, where, out |
121
+ | min/amin/nanmin, max/amax/nanmax | Preserve unit or point coordinate | axis, keepdims, initial, where, out |
122
122
  | where | Select between compatible quantities or points | condition, x, y, all positional; no out |
123
123
  | clip | Preserve input unit or point coordinate | a_min/a_max or min/max, out; ufunc call options |
124
- | prod | Product unit rules below | axis, dtype, keepdims, initial, where, out |
124
+ | prod, nanprod | Product unit rules below; nanprod counts selected non-NaN elements | axis, dtype, keepdims, initial, where, out |
125
125
  | median, nanmedian | Preserve unit or point coordinate | axis, keepdims, out; overwrite_input=False |
126
126
  | std/nanstd, var/nanvar | Difference unit / square difference unit | axis, dtype, ddof, keepdims, where, out |
127
127
  | average | Preserve sample unit; returned weight sum is a dimensionless Quantity | axis, weights, returned, keepdims; linear Quantity only |
128
128
  | quantile, percentile, nanquantile, nanpercentile | Preserve sample unit; dimensionless q and weights | axis, out, method, keepdims, weights; overwrite_input=False; linear Quantity only |
129
129
  | cumsum | Preserve unit | axis, dtype, out |
130
- | argmin, argmax | Bare indices | axis, keepdims, out |
130
+ | argmin/nanargmin, argmax/nanargmax | Bare indices | axis, keepdims, out |
131
131
  | interp | Align x to xp's unit; retain fp's unit | finite strictly increasing 1-D xp; scalar left/right, period=None; no out |
132
132
  | trapezoid | Multiply ordinate and coordinate units | x, scalar dx, axis; no out |
133
133
  | diff | Preserve input unit | n, axis, prepend, append; no out |
@@ -138,9 +138,9 @@ Supply the first magnitude argument positionally; concatenate/stack take a
138
138
  sequence. Options follow each supported signature, including positional out
139
139
  checks. Function and reduction outputs follow the rules below;
140
140
  overwrite_input=True is rejected. Reduction/statistical `where` is supported
141
- for sum, mean/nanmean, std/nanstd, var/nanvar, add.reduce, prod, multiply.reduce,
142
- and min/max; `initial` is supported for sum/add.reduce, prod/multiply.reduce,
143
- and min/max. Precomputed `mean`,
141
+ for sum/nansum, mean/nanmean, std/nanstd, var/nanvar, add.reduce, prod/nanprod,
142
+ multiply.reduce, and min/max/nanmin/nanmax; `initial` is supported for
143
+ sum/nansum/add.reduce, prod/nanprod/multiply.reduce, and min/max/nanmin/nanmax. Precomputed `mean`,
144
144
  `correction`, and other unsupported keywords raise TypeError.
145
145
  Indexed assignment and augmented arithmetic on ndarrays follow the fixed-target
146
146
  rules below. Python and NumPy scalar magnitudes retain ordinary result/rebinding
@@ -150,6 +150,7 @@ behavior for augmented arithmetic.
150
150
 
151
151
  See [selection and clipping](numpy/selection.md) for `where`, `clip`,
152
152
  `minimum/maximum`, and `min/max`, including coordinate and bound rules.
153
+ The NaN extrema and index functions share those coordinate rules and omit NaNs.
153
154
 
154
155
  ## Indexed assignment and augmented arithmetic
155
156
 
@@ -4,7 +4,7 @@ build-backend = 'hatchling.build'
4
4
 
5
5
  [project]
6
6
  name = 'unitx'
7
- version = '0.16.1'
7
+ version = '0.16.2'
8
8
  description = 'Physical units, quantities, and unit conversion with exact unit algebra, type hints, and NumPy integration.'
9
9
  readme = 'README.md'
10
10
  requires-python = '>=3.12'
@@ -403,7 +403,7 @@ def _dispatch_ufunc(
403
403
  }:
404
404
  return NotImplemented
405
405
  handler = _sum if ufunc is np.add else _prod
406
- return handler(inputs[0], **{'axis': 0, **kwargs})
406
+ return handler(ufunc.reduce, inputs[0], **{'axis': 0, **kwargs})
407
407
  if method == 'accumulate' and ufunc is np.add:
408
408
  if kwargs.keys() - {'axis', 'dtype'}:
409
409
  return NotImplemented
@@ -606,6 +606,8 @@ def dispatch_function(
606
606
  np.transpose,
607
607
  np.argmin,
608
608
  np.argmax,
609
+ np.nanargmin,
610
+ np.nanargmax,
609
611
  np.median,
610
612
  np.mean,
611
613
  np.std,
@@ -618,6 +620,8 @@ def dispatch_function(
618
620
  np.max,
619
621
  np.amin,
620
622
  np.amax,
623
+ np.nanmin,
624
+ np.nanmax,
621
625
  ):
622
626
  raise TypeError(
623
627
  'This NumPy function is undefined for affine points; convert explicitly.'
@@ -961,6 +965,7 @@ def _quantile(
961
965
 
962
966
 
963
967
  def _sum(
968
+ func: Any,
964
969
  q: Any,
965
970
  axis: Any = None,
966
971
  dtype: Any = None,
@@ -975,33 +980,34 @@ def _sum(
975
980
  initial_value = None
976
981
  if initial is not None:
977
982
  if not _is_supported_operand(initial):
978
- raise TypeError('initial must be a numeric scalar or scalar Quantity.')
983
+ error_message = 'initial must be a numeric scalar or scalar Quantity.'
984
+ raise TypeError(error_message)
979
985
  initial_value, initial_unit = _linear_parts(initial)
980
986
  if np.ndim(initial_value) != 0:
981
- raise TypeError('initial must be a numeric scalar or scalar Quantity.')
987
+ error_message = 'initial must be a numeric scalar or scalar Quantity.'
988
+ raise TypeError(error_message)
982
989
  if not compatible(initial_unit, unit):
983
990
  # Bare zero is the additive identity in any dimension. A Quantity
984
991
  # zero still has its declared dimension and must be compatible.
985
992
  if isinstance(initial, (Quantity, _Point)) or initial != 0:
986
- raise DimensionError('initial has an incompatible dimension.')
993
+ error_message = 'initial has an incompatible dimension.'
994
+ raise DimensionError(error_message)
987
995
  else:
988
996
  initial_value = scale_magnitude(initial_value, initial_unit.factor / unit.factor)
989
997
  result = _wrap(
990
- np.add.reduce(
991
- value, axis=axis, dtype=dtype, keepdims=keepdims, initial=initial_value, where=where
992
- ),
998
+ func(value, axis=axis, dtype=dtype, keepdims=keepdims, initial=initial_value, where=where),
993
999
  unit,
994
1000
  )
995
1001
  return _write_output(result, out)
996
1002
 
997
1003
 
998
1004
  def _prod(
1005
+ func: Any,
999
1006
  q: Any,
1000
1007
  axis: Any = None,
1001
1008
  dtype: Any = None,
1002
1009
  out: Any = None,
1003
1010
  keepdims: bool = False,
1004
- *,
1005
1011
  initial: Any = 1,
1006
1012
  where: Any = True,
1007
1013
  ) -> Any:
@@ -1010,10 +1016,12 @@ def _prod(
1010
1016
  initial_value, initial_unit = None, Unit()
1011
1017
  if initial is not None:
1012
1018
  if not _is_supported_operand(initial):
1013
- raise TypeError('initial must be a numeric scalar or scalar Quantity.')
1019
+ error_message = 'initial must be a numeric scalar or scalar Quantity.'
1020
+ raise TypeError(error_message)
1014
1021
  initial_value, initial_unit = _linear_parts(initial)
1015
1022
  if np.ndim(initial_value) != 0:
1016
- raise TypeError('initial must be a numeric scalar or scalar Quantity.')
1023
+ error_message = 'initial must be a numeric scalar or scalar Quantity.'
1024
+ raise TypeError(error_message)
1017
1025
 
1018
1026
  value, unit = _linear_parts(q)
1019
1027
  mask = np.broadcast_to(where, np.shape(value))
@@ -1021,7 +1029,7 @@ def _prod(
1021
1029
  if unit.dimension == DIMENSIONLESS:
1022
1030
  value = scale_magnitude(value, unit.factor, where=mask)
1023
1031
  unit = Unit()
1024
- result = np.multiply.reduce(
1032
+ result = func(
1025
1033
  value, axis=axis, dtype=dtype, keepdims=keepdims, initial=initial_value, where=where
1026
1034
  )
1027
1035
  if unit.dimension != DIMENSIONLESS:
@@ -1029,16 +1037,17 @@ def _prod(
1029
1037
  axes = range(len(shape)) if axis is None else axis if isinstance(axis, tuple) else (axis,)
1030
1038
  # NumPy has already validated axes, including its scalar 0/-1 special case.
1031
1039
  count = math.prod(shape[i] for i in axes) if shape else 1
1040
+ if func is np.nanprod:
1041
+ mask = mask & ~np.isnan(value)
1032
1042
  if np.ndim(where) == 0:
1033
1043
  count = count if bool(where) else 0
1034
- else:
1044
+ if np.ndim(where) != 0 or func is np.nanprod:
1035
1045
  counts = np.sum(mask, axis=axis)
1036
1046
  if np.size(counts):
1037
1047
  count = int(np.asarray(counts).flat[0])
1038
1048
  if np.any(counts != count):
1039
- raise DimensionError(
1040
- 'Product reductions require equal element counts per result.'
1041
- )
1049
+ error_message = 'Product reductions require equal element counts per result.'
1050
+ raise DimensionError(error_message)
1042
1051
  # With no output elements, retain the nominal reduction extent.
1043
1052
  unit = unit**count
1044
1053
  assert isinstance(unit, Unit) and isinstance(initial_unit, Unit)
@@ -1191,7 +1200,6 @@ def _extremum(
1191
1200
  axis: Any = None,
1192
1201
  out: Any = None,
1193
1202
  keepdims: bool = False,
1194
- *,
1195
1203
  initial: Any = _UNSET,
1196
1204
  where: Any = True,
1197
1205
  ) -> Any:
@@ -1210,15 +1218,19 @@ _QUANTITY_FUNCTIONS = {
1210
1218
  np.transpose: _transpose,
1211
1219
  np.argmin: partial(_argmin_max, np.argmin),
1212
1220
  np.argmax: partial(_argmin_max, np.argmax),
1221
+ np.nanargmin: partial(_argmin_max, np.nanargmin),
1222
+ np.nanargmax: partial(_argmin_max, np.nanargmax),
1213
1223
  np.median: partial(_median, np.median),
1214
1224
  np.nanmedian: partial(_median, np.nanmedian),
1215
1225
  np.cumsum: _cumsum,
1216
- np.prod: _prod,
1226
+ np.prod: partial(_prod, np.multiply.reduce),
1227
+ np.nanprod: partial(_prod, np.nanprod),
1217
1228
  np.std: partial(_std_var, np.std),
1218
1229
  np.var: partial(_std_var, np.var),
1219
1230
  np.nanstd: partial(_std_var, np.nanstd),
1220
1231
  np.nanvar: partial(_std_var, np.nanvar),
1221
- np.sum: _sum,
1232
+ np.sum: partial(_sum, np.add.reduce),
1233
+ np.nansum: partial(_sum, np.nansum),
1222
1234
  np.mean: partial(_mean, np.mean),
1223
1235
  np.nanmean: partial(_mean, np.nanmean),
1224
1236
  np.average: _average,
@@ -1230,4 +1242,6 @@ _QUANTITY_FUNCTIONS = {
1230
1242
  np.max: partial(_extremum, np.max),
1231
1243
  np.amin: partial(_extremum, np.amin),
1232
1244
  np.amax: partial(_extremum, np.amax),
1245
+ np.nanmin: partial(_extremum, np.nanmin),
1246
+ np.nanmax: partial(_extremum, np.nanmax),
1233
1247
  }
@@ -125,8 +125,16 @@ def output_contracts(a: npt.NDArray[np.float64]) -> None:
125
125
  q.__array_ufunc__(
126
126
  np.add, 'reduce', q, out=(scalar_target,), where=mask, initial=Quantity(2, 'cm')
127
127
  )
128
- for operation in (np.sum, np.mean, np.std, np.nanmean, np.nanstd):
128
+ for operation in (np.sum, np.nansum, np.mean, np.std, np.nanmean, np.nanstd):
129
129
  q.__array_function__(operation, (Quantity,), (q,), {'where': mask, 'out': scalar_target})
130
+ for operation in (np.nanmin, np.nanmax):
131
+ q.__array_function__(
132
+ operation, (Quantity,), (q,), {'initial': Quantity(0, 'cm'), 'out': scalar_target}
133
+ )
134
+ for operation in (np.nanargmin, np.nanargmax):
135
+ indices = np.empty((), dtype=np.intp)
136
+ q.__array_function__(operation, (Quantity,), (q,), {'out': indices})
137
+ assert_type(indices.dtype, np.dtype[np.intp])
130
138
  q.__array_function__(np.nanmedian, (Quantity,), (q,), {'out': scalar_target})
131
139
  for operation in (np.var, np.nanvar):
132
140
  q.__array_function__(
@@ -144,6 +152,7 @@ def output_contracts(a: npt.NDArray[np.float64]) -> None:
144
152
  (q,),
145
153
  {'initial': Quantity(2, 's'), 'out': Quantity(np.empty(()), 'm2 s')},
146
154
  )
155
+ q.__array_function__(np.nanprod, (Quantity,), (q,), {'where': mask, 'out': product_target})
147
156
  assert_type(product_target.value.dtype, np.dtype[np.float64])
148
157
  assert_type(scalar_target.value.dtype, np.dtype[np.float64])
149
158
  assert_type(widened.value, Array)
@@ -71,8 +71,10 @@ def test_numpy_add_sub_and_comparison() -> None:
71
71
  'func',
72
72
  [
73
73
  np.sum,
74
+ np.nansum,
74
75
  np.cumsum,
75
76
  np.prod,
77
+ np.nanprod,
76
78
  np.exp,
77
79
  np.log,
78
80
  np.sin,
@@ -0,0 +1,260 @@
1
+ """NaN extrema, indices, sums and products retain their unit contracts."""
2
+
3
+ from __future__ import annotations
4
+
5
+ import warnings
6
+ from collections.abc import Callable
7
+ from typing import TYPE_CHECKING, Any
8
+
9
+ import pytest
10
+
11
+ np = pytest.importorskip('numpy')
12
+
13
+ from unitx import AffineQuantity, LogQuantity, Quantity, Unit # noqa: E402
14
+ from unitx.exceptions import DimensionError # noqa: E402
15
+ from unitx.factor import Factor # noqa: E402
16
+ from unitx.imperial import imperial # noqa: E402
17
+ from unitx.si import si # noqa: E402
18
+
19
+ if TYPE_CHECKING:
20
+ from numpy.typing import DTypeLike
21
+
22
+
23
+ EXTREMA = [np.nanmin, np.nanmax]
24
+ INDICES = [np.nanargmin, np.nanargmax]
25
+
26
+
27
+ @pytest.mark.parametrize('operation', [*EXTREMA, np.nansum])
28
+ @pytest.mark.parametrize('axis', [None, 0, -1, (), (0, -1)])
29
+ @pytest.mark.parametrize('dtype', [np.int32, np.float32, np.longdouble, np.complex128])
30
+ @pytest.mark.parametrize('unit', [Unit('cm', scale=Factor(2)), Unit('km/m')])
31
+ def test_numpy_nan_reduction_axes(
32
+ operation: Callable[..., Any],
33
+ axis: int | tuple[int, ...] | None,
34
+ dtype: DTypeLike,
35
+ unit: Unit,
36
+ ) -> None:
37
+ values = np.arange(48, dtype=dtype).reshape(2, 3, 8)[..., ::2]
38
+ if values.dtype.kind in 'fc':
39
+ values[0, :, 0] = np.nan
40
+ saved = values.copy()
41
+ values.flags.writeable = False
42
+ samples = Quantity(values, unit)
43
+ for keepdims in (False, True):
44
+ with warnings.catch_warnings():
45
+ warnings.simplefilter('ignore', RuntimeWarning) # axis=() retains all-NaN slices.
46
+ result = operation(samples, axis=axis, keepdims=keepdims)
47
+ expected = operation(values, axis=axis, keepdims=keepdims)
48
+ assert result.unit == unit
49
+ np.testing.assert_array_equal(result.value, expected)
50
+ assert np.asarray(result.value).dtype == np.asarray(expected).dtype
51
+ np.testing.assert_array_equal(values, saved)
52
+
53
+
54
+ @pytest.mark.parametrize('operation', EXTREMA)
55
+ def test_numpy_nan_extremum_initial(operation: Callable[..., Any]) -> None:
56
+ samples = Quantity(np.array([[np.nan, 2, 3], [4, np.nan, 6]]), 'm')
57
+ mask = np.array([True, True, False])
58
+ target = Quantity(np.empty((2, 1)), 'cm')
59
+ expected = operation(samples.value, 1, None, True, 2.5, mask)
60
+ assert operation(samples, 1, target, True, Quantity(250, 'cm'), mask) is target
61
+ np.testing.assert_allclose(target.value, expected * 100)
62
+ empty = Quantity(np.empty((2, 0)), 'm')
63
+ np.testing.assert_array_equal(operation(empty, axis=1, initial=Quantity(2, 'm')).value, 2)
64
+
65
+
66
+ @pytest.mark.parametrize('operation', EXTREMA)
67
+ def test_numpy_nan_extremum_warning(operation: Callable[..., Any]) -> None:
68
+ samples = Quantity(np.full((2, 3), np.nan), 'm')
69
+ with pytest.warns(RuntimeWarning, match='All-NaN'):
70
+ result = operation(samples, axis=1)
71
+ assert result.unit == si.m
72
+ np.testing.assert_array_equal(result.value, np.nan)
73
+ target = Quantity(np.full(2, -123.0), 'cm')
74
+ with warnings.catch_warnings():
75
+ warnings.simplefilter('error', RuntimeWarning)
76
+ with pytest.raises(RuntimeWarning, match='All-NaN'):
77
+ operation(samples, axis=1, out=target)
78
+ np.testing.assert_array_equal(target.value, -123)
79
+ with pytest.raises(ValueError, match='zero-size'):
80
+ operation(Quantity(np.empty((2, 0)), 'm'), axis=1, out=target)
81
+ np.testing.assert_array_equal(target.value, -123)
82
+
83
+
84
+ @pytest.mark.parametrize('operation', [*EXTREMA, *INDICES])
85
+ @pytest.mark.parametrize(
86
+ 'point_type,coordinate,target_coordinate',
87
+ [
88
+ (AffineQuantity, si.degC, imperial.degF),
89
+ (LogQuantity, si.dBm, si.dBW),
90
+ ],
91
+ )
92
+ def test_numpy_nan_selection_points(
93
+ operation: Callable[..., Any], point_type: type, coordinate: Any, target_coordinate: Any
94
+ ) -> None:
95
+ samples = point_type(np.array([[np.nan, 0, 20], [30, 50, np.nan]]), coordinate)
96
+ expected = operation(samples.value, axis=1)
97
+ if operation in INDICES:
98
+ target = np.full(2, -123, dtype=np.intp)
99
+ result = operation(samples, axis=1)
100
+ np.testing.assert_array_equal(result, expected)
101
+ assert operation(samples, 1, target) is target
102
+ np.testing.assert_array_equal(target, expected)
103
+ else:
104
+ result = operation(samples, axis=1)
105
+ assert type(result) is point_type and result.unit is coordinate
106
+ np.testing.assert_array_equal(result.value, expected)
107
+ target = point_type(np.empty(2), target_coordinate)
108
+ assert operation(samples, axis=1, out=target) is target
109
+ np.testing.assert_allclose(target.value, result.value_in(target_coordinate))
110
+ initial = point_type(10, coordinate).to(target_coordinate)
111
+ bounded = operation(samples, axis=1, initial=initial, where=True)
112
+ np.testing.assert_allclose(bounded.value, operation(samples.value, axis=1, initial=10))
113
+
114
+
115
+ @pytest.mark.parametrize('operation', INDICES)
116
+ @pytest.mark.parametrize('axis', [None, 0, -1])
117
+ def test_numpy_nan_index_axes(operation: Callable[..., Any], axis: int | None) -> None:
118
+ values = np.array([[np.nan, 2, np.inf, 2], [-np.inf, 5, 5, np.nan]])[:, ::-1]
119
+ samples = Quantity(values, 'km/m')
120
+ for keepdims in (False, True):
121
+ expected = operation(values, axis=axis, keepdims=keepdims)
122
+ target = np.full(np.shape(expected), -123, dtype=np.intp)
123
+ assert operation(samples, axis, target, keepdims=keepdims) is target
124
+ np.testing.assert_array_equal(target, expected)
125
+
126
+
127
+ @pytest.mark.parametrize('operation', INDICES)
128
+ @pytest.mark.parametrize('values', [np.full((2, 3), np.nan), np.empty((2, 0))])
129
+ def test_numpy_nan_index_invalid_slice(operation: Callable[..., Any], values: Any) -> None:
130
+ target = np.full(2, -123, dtype=np.intp)
131
+ with pytest.raises(ValueError):
132
+ operation(Quantity(values, 'm'), axis=1, out=target)
133
+ np.testing.assert_array_equal(target, -123)
134
+
135
+
136
+ @pytest.mark.parametrize('operation', EXTREMA)
137
+ @pytest.mark.parametrize('initial', [0, Quantity(0, 's'), 0 @ si.degC])
138
+ def test_numpy_nan_extremum_invalid_initial(operation: Callable[..., Any], initial: Any) -> None:
139
+ target = Quantity(np.array(-123.0), 'm')
140
+ with pytest.raises((TypeError, DimensionError)):
141
+ operation(Quantity(np.array([1.0, np.nan]), 'm'), initial=initial, out=target)
142
+ assert target.value == -123
143
+
144
+
145
+ @pytest.mark.parametrize('operation', EXTREMA)
146
+ def test_numpy_nan_extremum_invalid_convention(operation: Callable[..., Any]) -> None:
147
+ samples = LogQuantity(np.array([np.nan, 10.0]), si.dBW)
148
+ target = LogQuantity(np.array(-123.0), si.dBm)
149
+ with pytest.raises(DimensionError, match='same convention'):
150
+ operation(samples, initial=0 @ si.dB(si.W, kind='amplitude'), out=target)
151
+ assert target.value == -123
152
+
153
+
154
+ @pytest.mark.parametrize('operation', [*EXTREMA, *INDICES, np.nansum, np.nanprod])
155
+ @pytest.mark.parametrize('invalid', ['shape', 'dtype', 'readonly', 'wrapper'])
156
+ def test_numpy_nan_reduction_invalid_out(operation: Callable[..., Any], invalid: str) -> None:
157
+ samples = Quantity(np.array([[np.nan, 1, 3], [4, np.nan, 6]]), 'm')
158
+ indices = operation in INDICES
159
+ dtype = np.intp if indices else np.float64
160
+ unit = si.m**2 if operation is np.nanprod else si.m
161
+ storage = np.full(3 if invalid == 'shape' else 2, -123, dtype=dtype)
162
+ if invalid == 'dtype':
163
+ storage = storage.astype(float if indices else int)
164
+ if invalid == 'readonly':
165
+ storage.flags.writeable = False
166
+ if (invalid == 'wrapper') == indices:
167
+ target = Quantity(storage, unit)
168
+ else:
169
+ target = storage
170
+ with pytest.raises((TypeError, ValueError)):
171
+ operation(samples, axis=1, out=target)
172
+ np.testing.assert_array_equal(storage, -123)
173
+
174
+
175
+ @pytest.mark.parametrize('operation', [*EXTREMA, *INDICES, np.nansum, np.nanprod])
176
+ def test_numpy_nan_reduction_out_alias(operation: Callable[..., Any]) -> None:
177
+ dtype = np.intp if operation in INDICES else np.float64
178
+ storage = np.array([[1, 3, 2], [4, 6, 5]], dtype=dtype)
179
+ if operation not in INDICES:
180
+ storage[:, -1] = np.nan
181
+ saved = storage.copy()
182
+ samples = Quantity(storage, Unit('m', scale=Factor(2)))
183
+ unit = samples.unit**2 if operation is np.nanprod else samples.unit
184
+ target = storage[:, 0] if operation in INDICES else Quantity(storage[:, 0], unit)
185
+ expected = operation(saved, axis=1)
186
+ assert operation(samples, axis=1, out=target) is target
187
+ saved[:, 0] = expected
188
+ np.testing.assert_array_equal(storage, saved)
189
+
190
+
191
+ def test_numpy_nansum_mask_and_empty() -> None:
192
+ samples = Quantity(np.array([[1.0, np.nan, 3], [np.nan, 5, 7]]), 'm')
193
+ mask = np.array([[True, True, False], [False, True, True]])
194
+ target = Quantity(np.empty(2), 'cm')
195
+ assert np.nansum(samples, 1, np.float32, target, False, Quantity(50, 'cm'), mask) is target
196
+ np.testing.assert_array_equal(target.value, [150, 1250])
197
+ for values in (np.full((2, 3), np.nan), np.empty((2, 0))):
198
+ result = np.nansum(Quantity(values, 'm'), axis=1)
199
+ assert result.unit == si.m
200
+ np.testing.assert_array_equal(result.value, 0)
201
+
202
+
203
+ @pytest.mark.parametrize('axis,count', [(None, 4), (0, 1), (-1, 2), ((0, 1), 4)])
204
+ def test_numpy_nanprod_counts(axis: int | tuple[int, ...] | None, count: int) -> None:
205
+ values = np.array([[2.0, np.nan, 3, np.nan], [np.nan, 4, np.nan, 5]])
206
+ samples = Quantity(values, Unit('cm', scale=Factor(2)))
207
+ for keepdims in (False, True):
208
+ result = np.nanprod(samples, axis=axis, keepdims=keepdims, initial=Quantity(2, 's'))
209
+ assert result.unit == samples.unit**count * si.s
210
+ np.testing.assert_array_equal(
211
+ result.value, np.nanprod(values, axis=axis, keepdims=keepdims, initial=2)
212
+ )
213
+ target = Quantity(np.empty(2), 'm2 s')
214
+ assert np.nanprod(samples, 1, None, target, False, Quantity(2, 's'), True) is target
215
+ np.testing.assert_allclose(target.value, np.nanprod(values, axis=1) * 0.02**2 * 2)
216
+
217
+
218
+ def test_numpy_nanprod_mask() -> None:
219
+ samples = Quantity(np.array([[2.0, np.nan, 3], [np.nan, 4, 5]]), 'm')
220
+ mask = np.array([True, True, False])
221
+ result = np.nanprod(samples, axis=1, where=mask)
222
+ assert result.unit == si.m
223
+ np.testing.assert_array_equal(result.value, [2, 4])
224
+ samples = Quantity(np.array([[0.0, np.nan, 3], [np.inf, 4, np.nan]]), 'm')
225
+ assert np.nanprod(samples, axis=1).unit == si.m**2 # Zero and infinity still count.
226
+
227
+
228
+ @pytest.mark.parametrize(
229
+ 'values,axis',
230
+ [
231
+ (np.full((2, 3), np.nan), 1),
232
+ (np.array(np.nan), None),
233
+ (np.array(np.nan), 0),
234
+ (np.array(np.nan), -1),
235
+ (np.array(np.nan), ()),
236
+ (np.empty((2, 0)), 1),
237
+ ],
238
+ )
239
+ def test_numpy_nanprod_identity(values: Any, axis: int | tuple[int, ...] | None) -> None:
240
+ result = np.nanprod(Quantity(values, 'm'), axis=axis, initial=Quantity(2, 's'))
241
+ assert result.unit == si.s
242
+ np.testing.assert_array_equal(result.value, 2)
243
+
244
+
245
+ @pytest.mark.parametrize('unit', ['m', 'dB'])
246
+ @pytest.mark.parametrize('axis', [1, ()])
247
+ def test_numpy_nanprod_invalid_counts(unit: str, axis: int | tuple[int, ...]) -> None:
248
+ samples = Quantity(np.array([[2.0, np.nan, 3], [4, 5, 6]]), unit)
249
+ target = Quantity(np.full((2,) if axis == 1 else (2, 3), -123.0), unit)
250
+ with pytest.raises(DimensionError, match='equal element counts'):
251
+ np.nanprod(samples, axis=axis, initial=0, out=target)
252
+ np.testing.assert_array_equal(target.value, -123)
253
+
254
+
255
+ @pytest.mark.parametrize('unit', ['km/m', 'deg', ''])
256
+ def test_numpy_nanprod_dimensionless(unit: str) -> None:
257
+ samples = Quantity(np.array([[2.0, np.nan, 3], [4, 5, 6]]), unit)
258
+ result = np.nanprod(samples, axis=1, initial=Quantity(2, 's'))
259
+ assert result.unit == si.s
260
+ np.testing.assert_allclose(result.value, np.nanprod(samples.value_in(''), axis=1, initial=2))