Release 1.10 plus a version bump

array_api_compat/__init__.py

@@ -17,6 +17,6 @@
 this implementation for the default when working with NumPy arrays.
 
 """
-__version__ = '1.10.0.dev0'
+__version__ = '1.10.1.dev0'
 
 from .common import *  # noqa: F401, F403

cupy-xfails.txt

@@ -173,6 +173,9 @@ array_api_tests/test_special_cases.py::test_unary[trunc(x_i is -0) -> -0]
 array_api_tests/test_fft.py::test_fftn
 array_api_tests/test_fft.py::test_ifftn
 array_api_tests/test_fft.py::test_rfftn
+
+# observed in the 1.10 release process, is likely related to xfails above
+array_api_tests/test_fft.py::test_irfftn
 
 # 2023.12 support
 # cupy.ndaray cannot be specified as `repeats` argument.

docs/changelog.md

@@ -1,5 +1,28 @@
 # Changelog
 
+## 1.10.0 (2024-12-25)
+
+### Major Changes
+
+- New function `is_writeable_array` adds transparent support for readonly
+  arrays, such as JAX arrays or numpy arrays with `.flags.writeable=False`.
+
+- `asarray(..., copy=None)` with `dask` backend always copies, so that
+  `copy=None` and `copy=True` are equivalent for the `dask` backend.
+   This change is made to be forward compatible with the `dask==2024.12`
+   release.
+
+
+### Minor Changes
+
+- `array_namespace` accepts (and ignores) `None` and python scalars (int, float,
+   complex, bool). This change is to simplify downstream adoption, for
+   functions where arguments can be either arrays or scalars.
+
+- `vecdot` conjugates its first argument, as stipulated by the Array API spec.
+  Previously, conjation if the first argument was missing.
+
+
 ## 1.9.1 (2024-10-29)
 
 ### Major Changes