I have simplified the case and tried to look from a list of json items (please see test table content). I think I have managed to unpack values in such a way that also partial matching is possible. However, the end result has two "value" named columns even if I have tried to rename them.
How can I rename the columns coming from jsonb_ functions? Or is there a better way of unpacking the list and items?
select * from test; id | main ----+-------------------------- 1 | [{"A": "b"}, {"B": "b"}] (1 row)
select * from test, jsonb_array_elements(test.main) as m, jsonb_each(value); id | main | value | key | value ----+--------------------------+------------+-----+------- 1 | [{"A": "b"}, {"B": "b"}] | {"A": "b"} | A | "b" 1 | [{"A": "b"}, {"B": "b"}] | {"B": "b"} | B | "b" (2 rows)
Thanks
Performance of this will be awful. Unfortunately, FWICT there is no
way to do partial string matches against json in a indexed way without some serious elbow grease. For full key-value matching though you're good to go.
Do index this operation, the best option today will revolve around the pg_trgm module. It can optimize like expressions against text. Performance of this is a mixed bag although upcoming pg_trgm enhancements should make it a lot faster. merlin