Details
-
Enabler
-
Not Assigned
-
None
-
Data Processing
-
-
-
3
-
3
-
13.333
-
Team_YANDA
-
Sprint 5
-
-
-
-
9.1
-
Stories Completed, Integrated, Solution Intent Updated, BDD Testing Passes (no errors), Outcomes Reviewed, NFRS met, Demonstrated, Satisfies Acceptance Criteria, Accepted by FO
-
-
SPO-878
Description
So far data is exchanged only via tensors and (pickled) metadata. The more appropriate way to approach this is to use Arrow tables for representing such data. This should both give us options for rich access interfaces (such as Awkward array) as well as enable more portability, especially to non-Python environments.
Note that https://gitlab.com/ska-telescope/sdp-dal-prototype at this point supports tables, but not table slices. Might mean that for visibility streaming a (time) table needs to get created for every chunk.