Unittest continuousDataObject2NpArray fail
photoniker opened this issue · 1 comments
test_continuousDataObject2NpArray of DataObjectNpConversion fails for Python 3.12, Numpy 1.26.
This issue is due to a change in Numpy. If a Numpy array is created from a compatible object of another type (here: a dataObject), Numpy for instance tries to get information about the matrix structure via the optional array_struct member.
If Numpy is able to create its np.ndarray by means of a shallow copy (copy=False), the original source object is stored in the base member of the numpy array. For Numpy < 1.23, the base member was set to a ref-count incremented version of the original dataObject. For Numpy >= 1.23, the base member is now a tuple, that consists of the ref-count incremented dataObject (1st item) and the "PyCapsule" object as 2nd item.
For more information see:
The unittest in itom will be adapted with a version-switch for the numpy version.