Merge pull request #100 from asmeurer/release

1.5 release

Author: asmeurer

CHANGELOG.md

@@ -1,6 +1,52 @@
-# 1.4.1 (2024-01-18)
+# Array API Compat Changelog
 
-## Minor Changes
+## 1.5 (2024-03-07)
+
+### Major Changes
+
+- Add support for Dask ([@lithomas1](https://github.com/lithomas1)).
+
+- Add support for JAX. Note that unlike other array libraries,
+  array-api-compat does not contain any wrappers for JAX functions. All JAX
+  array API support is in JAX itself. Thus, there is no `array_api_compat.jax`
+  submodule, and `array_namespace(<JAX array>)` returns the
+  `jax.experimental.array_api` module.
+
+- The functions `is_numpy_array(x)`, `is_cupy_array(x)`, `is_torch_array(x)`,
+  `is_dask_array(x)`, `is_jax_array(x)` are now part of the public
+  `array_api_compat` API.
+
+- Add wrappers for the `fft` extension module for NumPy, CuPy, and PyTorch.
+
+### Minor Changes
+
+- Allow `'2022.12'` as the `api_version` in `array_namespace()`. `'2021.12'`
+  is also supported but will issue a warning since the returned namespace will
+  still be a 2022.12 compliant one.
+
+- Add wrapper for numpy.linalg.solve, which broadcasts the inputs according to
+  the standard.
+
+- Add wrappers for various PyTorch linalg functions.
+
+- Fix a bug with `numpy.linalg.vector_norm(keepdims=True)`.
+
+- BREAKING: Update `vecdot` wrappers to apply `axes` before broadcasting, not
+  after. This matches the updated 2023.12 standard wording, and also the
+  behavior of the new `numpy.vecdot` gufunc in NumPy 2.0.
+
+- Fix some linalg functions which were supposed to be in both the main
+  namespace and the linalg extension namespace.
+
+- Add Ruff to CI. ([@adonath](https://github.com/adonath))
+
+- Test that internal definitions of `__all__` are self-consistent, which
+  should help to avoid issues where wrappers are accidentally not exported to
+  the compat namespaces properly.
+
+## 1.4.1 (2024-01-18)
+
+### Minor Changes
 
 - Add support for the upcoming NumPy 2.0 release.
 
@@ -17,14 +63,14 @@
 - Fix linalg.cholesky returning the conjugate of the expected upper
   decomposition for numpy and cupy.
 
-# 1.4 (2023-09-13)
+## 1.4 (2023-09-13)
 
-## Major Changes
+### Major Changes
 
 - Releases are now made with GitHub Actions (thanks
   [@matthewfeickert](https://github.com/matthewfeickert)).
 
-## Minor Changes
+### Minor Changes
 
 - Fix `torch.result_type()` cross-kind promotion
   ([@lucascolley](https://github.com/lucascolley)).
@@ -34,42 +80,42 @@
 - Add requires-python metadata to the package
   ([@matthewfeickert](https://github.com/matthewfeickert)).
 
-# 1.3 (2023-06-20)
+## 1.3 (2023-06-20)
 
-## Major Changes
+### Major Changes
 
 - Add [2022.12](https://data-apis.org/array-api/2022.12/) standard support.
   This includes things like adding complex dtype support, adding the new
   `take` function, and various minor changes in the specification.
 
-## Minor Changes
+### Minor Changes
 
 - Support `"cpu"` in CuPy `to_device()`.
 
 - Return a new array in NumPy/CuPy `reshape(copy=False)`.
 
 - Fix signatures for PyTorch `broadcast_to` and `permute_dims`.
 
-# 1.2 (2023-04-03)
+## 1.2 (2023-04-03)
 
-## Major Changes
+### Major Changes
 
 - Support the linalg extension in the `array_api_compat.torch` namespace.
 
 - Add `isdtype()`.
 
-## Minor Changes
+### Minor Changes
 
 - Fix the `k` keyword argument to `tril` and `triu` in `torch`.
 
-# 1.1.1 (2023-03-10)
+## 1.1.1 (2023-03-10)
 
-## Major Changes
+### Major Changes
 
 - Rename `get_namespace()` to `array_namespace()` (`get_namespace()` is
   maintained as a backwards compatible alias).
 
-## Minor Changes
+### Minor Changes
 
 - The minimum supported NumPy version is now 1.21. Fixed a few issues with
   NumPy 1.21 (with `unique_*` and `asarray`), although there are also a few
@@ -91,9 +137,9 @@
 - Add `torch` wrappers for `ones`, `empty`, and `zeros` so that `shape` can be
   passed as a keyword argument.
 
-# 1.1 (2023-02-24)
+## 1.1 (2023-02-24)
 
-## Major Changes
+### Major Changes
 
 - Added support for PyTorch.
 
@@ -104,14 +150,14 @@
 - All wrapper functions that wrap existing library functions now pass through
   arbitrary `**kwargs`.
 
-## Minor Changes
+### Minor Changes
 
 - Added CI to run against the [array API testsuite](https://github.com/data-apis/array-api-tests).
 
 - Fix `sort(stable=False)` and `argsort(stable=False)` with CuPy.
 
-# 1.0 (2022-12-05)
+## 1.0 (2022-12-05)
 
-## Major Changes
+### Major Changes
 
 - Initial release. Includes support for NumPy and CuPy.

array_api_compat/__init__.py

@@ -17,6 +17,6 @@
 this implementation for the default when working with NumPy arrays.
 
 """
-__version__ = '1.4.1'
+__version__ = '1.5'
 
 from .common import *  # noqa: F401, F403

array_api_compat/common/_helpers.py

@@ -325,4 +325,4 @@ def size(x):
     "to_device",
 ]
 
-_all_ignore = ['sys', 'math', 'inspect']
+_all_ignore = ['sys', 'math', 'inspect', 'warnings']

cupy-xfails.txt

@@ -44,6 +44,9 @@ array_api_tests/test_searching_functions.py::test_argmin
 array_api_tests/test_statistical_functions.py::test_min
 array_api_tests/test_statistical_functions.py::test_max
 
+# prod() sometimes doesn't give nan for 0*overflow
+array_api_tests/test_statistical_functions.py::test_prod
+
 # testsuite incorrectly thinks meshgrid doesn't have indexing argument
 # (https://github.com/data-apis/array-api-tests/issues/171)
 array_api_tests/test_signatures.py::test_func_signature[meshgrid]

numpy-1-21-xfails.txt

@@ -122,6 +122,7 @@ array_api_tests/test_operators_and_elementwise_functions.py::test_divide[__trued
 array_api_tests/test_operators_and_elementwise_functions.py::test_divide[__truediv__(x1, x2)]
 array_api_tests/test_operators_and_elementwise_functions.py::test_divide[divide(x1, x2)]
 array_api_tests/test_operators_and_elementwise_functions.py::test_equal[equal(x1, x2)]
+array_api_tests/test_operators_and_elementwise_functions.py::test_equal[__eq__(x1, x2)]
 array_api_tests/test_operators_and_elementwise_functions.py::test_floor_divide[__floordiv__(x, s)]
 array_api_tests/test_operators_and_elementwise_functions.py::test_floor_divide[__floordiv__(x1, x2)]
 array_api_tests/test_operators_and_elementwise_functions.py::test_floor_divide[__ifloordiv__(x, s)]

tests/test_array_namespace.py

@@ -1,5 +1,6 @@
 import subprocess
 import sys
+import warnings
 
 import numpy as np
 import pytest
@@ -57,13 +58,24 @@ def test_array_namespace_errors():
     pytest.raises(TypeError, lambda: array_namespace((x, x)))
     pytest.raises(TypeError, lambda: array_namespace(x, (x, x)))
 
-
 def test_array_namespace_errors_torch():
     y = torch.asarray([1, 2])
     x = np.asarray([1, 2])
     pytest.raises(TypeError, lambda: array_namespace(x, y))
-    pytest.raises(ValueError, lambda: array_namespace(x, api_version="2022.12"))
 
+def test_api_version():
+    x = np.asarray([1, 2])
+    np_ = import_("numpy", wrapper=True)
+    assert array_namespace(x, api_version="2022.12") == np_
+    assert array_namespace(x, api_version=None) == np_
+    assert array_namespace(x) == np_
+    # Should issue a warning
+    with warnings.catch_warnings(record=True) as w:
+        assert array_namespace(x, api_version="2021.12") == np_
+        assert len(w) == 1
+        assert "2021.12" in str(w[0].message)
+
+    pytest.raises(ValueError, lambda: array_namespace(x, api_version="2020.12"))
 
 def test_get_namespace():
     # Backwards compatible wrapper