A few changes to the tests are necessary for this workflow to pass:
test_index_emptylist is expected to fail because h5pyd returns an empty array with no shape, and numpy returns an empty array with the shape (0, 3). Now, when h5pyd is returning an empty selection, it sets the 'shape' equal to what would result from executing the same (empty) selection on an ndarray.
test_ellipsis and test_tuple under TestScalarArray are expected to raise an error due to H5T_ARRAY only being partially implemented in HSDS. For now, if h5py is being used, these tests manually raise an error to be consistent.
Avoid using is_hdf5 on an ID when using h5py, since the h5py implementation of that function expects a string path, not an integer ID.
A few changes to the tests are necessary for this workflow to pass:
test_index_emptylist
is expected to fail because h5pyd returns an empty array with no shape, and numpy returns an empty array with the shape(0, 3)
. Now, when h5pyd is returning an empty selection, it sets the 'shape' equal to what would result from executing the same (empty) selection on an ndarray.test_ellipsis
andtest_tuple
underTestScalarArray
are expected to raise an error due toH5T_ARRAY
only being partially implemented in HSDS. For now, if h5py is being used, these tests manually raise an error to be consistent.is_hdf5
on an ID when using h5py, since the h5py implementation of that function expects a string path, not an integer ID.