Fix NumPy 2.0 Warning

`np.array(..., copy=False)` to `np.asarray(..., copy=False)`.

Bump requirements to NumPy 2.0+: we so far supported NumPy 1 & 2
in parallel, but NumPy 2 is widely enough deployed to require it now.

Author: ax3l

README.md

@@ -54,7 +54,7 @@ pyAMReX depends on the following popular third party software.
 - [AMReX *development*](https://amrex-codes.github.io): we automatically download and compile a copy of AMReX
 - [pybind11](https://github.com/pybind/pybind11/) 2.13.0+: we automatically download and compile a copy of pybind11 ([new BSD](https://github.com/pybind/pybind11/blob/master/LICENSE))
   - [Python](https://python.org) 3.9+
-  - [Numpy](https://numpy.org) 1.15+
+  - [NumPy](https://numpy.org) 2.0+
 
 Optional dependencies include:
 - [mpi4py](https://mpi4py.readthedocs.io) 2.1+: for multi-node and/or multi-GPU execution

docs/source/install/dependencies.rst

@@ -13,7 +13,7 @@ Please see installation instructions below.
 - `pybind11 2.13.0+ <https://github.com/pybind/pybind11/>`__: we automatically download and compile a copy
 - `Python 3.9+ <https://www.python.org>`__
 
-  - `numpy 1.15+ <https://numpy.org>`__
+  - `NumPy 2.0+ <https://numpy.org>`__
 
 Optional dependencies include:
 

requirements.txt

@@ -1 +1 @@
-numpy>=1.15
+numpy>=2.0

src/amrex/extensions/Array4.py

@@ -40,7 +40,7 @@ def array4_to_numpy(self, copy=False, order="F"):
         # This supports a device-to-host copy.
         data = self.to_host()
     else:
-        data = np.array(self, copy=False)
+        data = np.asarray(self, copy=False)
 
     if order == "F":
         return data.T

src/amrex/extensions/ArrayOfStructs.py

@@ -35,9 +35,9 @@ def aos_to_numpy(self, copy=False):
         # todo: validate of the to_host() returned object
         #       lifetime is always managed correctly by
         #       Python's GC - otherwise copy twice via copy=True
-        return np.array(self.to_host(), copy=False)
+        return np.asarray(self.to_host(), copy=False)
     else:
-        return np.array(self, copy=False)
+        return np.asarray(self, copy=False)
 
 
 def aos_to_cupy(self, copy=False):

src/amrex/extensions/MultiFab.py

@@ -610,7 +610,7 @@ def __setitem__(self, index, value):
         # (it needs to be 4-D).
         # This converts value to an array if needed, and the [...] grabs a view so
         # that the shape change below doesn't affect value.
-        value3d = np.array(value)[...]
+        value3d = np.asarray(value)[...]
         global_shape = list(value3d.shape)
         # The shape of 1 is added for the extra dimensions and when index is an integer
         # (in which case the dimension was not in the input array).

src/amrex/extensions/PODVector.py

@@ -32,9 +32,9 @@ def podvector_to_numpy(self, copy=False):
             # todo: validate of the to_host() returned object
             #       lifetime is always managed correctly by
             #       Python's GC - otherwise copy twice via copy=True
-            return np.array(self.to_host(), copy=False)
+            return np.asarray(self.to_host(), copy=False)
         else:
-            return np.array(self, copy=False)
+            return np.asarray(self, copy=False)
     else:
         raise ValueError("Vector is empty.")
 

src/amrex/extensions/SmallMatrix.py

@@ -35,9 +35,9 @@ def smallmatrix_to_numpy(self, copy=False, order="F"):
     import numpy as np
 
     if copy:
-        data = np.array(self, copy=True)
+        data = np.asarray(self, copy=True)
     else:
-        data = np.array(self, copy=False)
+        data = np.asarray(self, copy=False)
 
     # TODO: Check self.order == "F" ?
     if order == "F":