Postgres jsonb null vs empty object query performance

Question

Assuming there's field of key-value data stored in jsonb format, that might not exist for some (few) records, what is preferable, performance wise for that case - null field or empty json object?

Data is to be queried by some key-value criteria not much more difficult than:

SELECT * FROM table WHERE id = :id AND jsondata @> '{"key1":"value1"}

and possibly indexed in future, so the concern is about querying - are null fields about to be skipped faster that ones with empty object?

Answer 1

NULLs are stored as a bitset in the row header, so they can be skipped slightly faster, but if you only have a "few" records with NULL values, the difference is unlikely to be noticeable.

You can also create a partial index to exclude the NULLs.