Measure RAM usage when doing nothing

Return the parameters for read_many_slices() for all chunks of the selection. Returns ------- Tuple of two 2D arrays, (slices

Resize the dataset, or the specified axis. The rank of the dataset cannot be changed. "size" should be a shape tuple, or if an axis

First index of staged slabs inside self.slabs

Test that an InMemoryArrayDataset built from an array of bare object dtype (stripped of the h5py string metadata) defaults its fillvalue to b"

__array__ changes chunk dtype upon first access.

Test that from_array() accepts anything that implements ArrayProtocol and that it does not coerce it to a numpy array.

Test that the base slabs can be anything that implements ArrayProtocol and that they are not coerced to a numpy array.

Test that h5py AsTypeView is an ArrayProtocol

Test that h5py_astype() returns an ArrayProtocol, also on older h5py versions. TODO delete this test when dropping support for h5py <3.13.

Test that h5py.Dataset is a ArrayProtocol

RawDataView conforms to the writeable MutableArrayProtocol, so that it satisfies the return type of StagedChangesArray.commit()'s ``empty`` callba

Test that the value of __setitem__ can be anything that implements ArrayProtocol

Test workaround to bug when converting from object array of bytes to NpyStrings This issue affected NumPy >=2.0.0,<2.2.3 https://github.com/nu

In object string <-> StringDType conversions the data doesn't change, so the hashes of the swapped base slabs are not invalidated.