You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Following #591, the node supports the new account ID format, which fits into two felts instead of just one. As a consequence, 64-bit long types do not work as a backing type for account IDs anymore, which means we can't use fixed64 on protobuf or u64 for SQL.
Currently, the SQL type for account_id is BLOB which is fine because we can deserialize and serialize as needed, but since the account ID prefix (which fits into 64 bits) is a unique identifier of the account itself, we can separate the ID into two columns (first felt or prefix, and second felt or suffix) in order to speed up lookups and improve indexing. This should also reduce allocations because it means less (de)serialization (would also leverage the fact that AccountId only contains the two felts so instantiation would still be simple). One thing to note is that the prefix being a unique identifier might be relaxed in the future, so this only really applies while this is still the case.
The text was updated successfully, but these errors were encountered:
Following #591, the node supports the new account ID format, which fits into two felts instead of just one. As a consequence, 64-bit long types do not work as a backing type for account IDs anymore, which means we can't use
fixed64
on protobuf oru64
for SQL.Currently, the SQL type for
account_id
isBLOB
which is fine because we can deserialize and serialize as needed, but since the account ID prefix (which fits into 64 bits) is a unique identifier of the account itself, we can separate the ID into two columns (first felt or prefix, and second felt or suffix) in order to speed up lookups and improve indexing. This should also reduce allocations because it means less (de)serialization (would also leverage the fact thatAccountId
only contains the two felts so instantiation would still be simple). One thing to note is that the prefix being a unique identifier might be relaxed in the future, so this only really applies while this is still the case.The text was updated successfully, but these errors were encountered: