I'm working on a small project which would involve running queries against deeply-nested (up to 5 layers) keys in dictionaries. This would be specifically a read-heavy environment, with only one large set of rites during the app's bootstrapping (nominally once per install). The two options I'm considering for use as database backends right now are PostgreSQL (10.0) and Couchbase (5.0.1).
I'm reasonably familiar with Postgres and its JSONB facilities in terms of performance (although not implementation details), but I'm not familiar enough with Couchbase to draw a concrete comparison, and it's difficult to find this specific comparison in any online benchmarks.
Ignoring issues about the fact that Postgres isn't a document store, could Postgres consistently outperform Couchbase in read queries involving accessing a deeply-nested key/value pair?