Continue work related to storing binary serialized operations in the HAF database
There is done a prototype: https://gitlab.syncad.com/hive/haf/-/tree/tm-ops-as-hive-operation where HAF operations are stored in the binary serialized form instead of JSON. The benefit of it is smaller database storage size like also faster data dump during replay.
We need to confirm:
-
all HAF APIs (sql functions) are taking SQL type pointing the operation. Also improve their test coverage to claim they work correctly -
once again measure performance during write (replay up to head block) -
verify read performance where JSON form of operations must be rebuilt and returned to API call.
Maybe we should consider creation of official SQL operation type (exposed to APPs) to avoid dummy conversions to json.
@mtyszczak will consult details related to above implementation