[BUG] Investigate another way to map functions and objects from the pipeline.

[jcfaracco]

Describe the bug
Apparently, there is a high probability of collision by using a hash function from python.

To Reproduce
This is hard to reproduce, but it happens.

Expected behavior
If you are passing to the pipeline two objects that came from the same class but they are different instances, they should be threatened as separated stanzas.

Additional context
One possible solution is to use uuid module and functions instead.

[otavioon]

Could you please add the file location where hash is calculated? It would be here:

dasf-core/dasf/pipeline/pipeline.py

Line 57 in 57113f8

key = hash(obj)

?

[otavioon]

uuid is a good alternative, but we will lose the determinism. The same object instance will have two different ids. Maybe we can look at dask's tokenize function, the function used for hashing and creating tasks names, and implement something similar...?

[jcfaracco]

I merged the last changes and it generated a new task for each hierarchy. This needs to be rebased if we are planning to support operators with uuids.