Type erasure in GenericStorage

[XzzX]

Depending on the storage backend type information and more might get lost. How to deal with this when loading?

@pyiron/storage

[XzzX]

We can extend every type to be a group/dict. This way we can store additional attributes like type, strides, ... However, this will have some overhead for plain types that might have been correct. What do you think?

[ligerzero-ai]

every basic type you mean? or do you mean really every single possible conceivable datatype then undergoes a transformation that we do to it?

[XzzX]

Everything that might get lost when stored in for example json format.

[XzzX]

Another way would be to provide an additional parameter cls: type to the load API providing the type we want to get back.

[XzzX]

@pyiron/storage Please think about it.

Currently, I am in favour of cls: type.