Since PostgreSQL 12 you get the SQL/JSON path language — a jsonpath type plus the @? and @@ operators — for filtering deep inside jsonb with conditions, wildcards, and arithmetic. It's the standardized, expressive alternative to chaining ->/#> operators.
The jsonpath syntax
A path starts at $ (the document root), @ is the current element inside a filter, .* and [*] are wildcards, and ? (...) applies a filter predicate.
'$.address.city' -- a nested field
'$.tags[*]' -- every array element
'$.items[*] ? (@.price > 100)' -- array elements where price > 100
'$.**.id' -- "id" at any depth (recursive wildcard)The two operators: @? and @@
@? asks "does the path match anything?" — it takes a path that selects items and returns true if at least one exists. @@ evaluates a path that is itself a predicate and returns its boolean result.
-- @? : path selects items, true if any match
SELECT * FROM products
WHERE data @? '$.specs.weight ? (@ < 5)';
-- @@ : path is a boolean predicate
SELECT * FROM products
WHERE data @@ '$.specs.weight < 5';Both are GIN-indexable under either operator class — including jsonb_path_ops — so deep filters can use an index just like @>.
Extracting with path functions
The operators answer yes/no; the functions return the matched data:
SELECT jsonb_path_query(data, '$.items[*] ? (@.qty > 1)') FROM orders;
SELECT jsonb_path_exists(data, '$.flags ? (@.beta == true)') FROM accounts;
SELECT jsonb_path_match(data, '$.score > 90') FROM results;jsonb_path_query returns each match as a row; jsonb_path_exists mirrors @?; jsonb_path_match mirrors @@.
Strict vs lax mode
By default jsonpath runs in lax mode, which auto-unwraps arrays and tolerates missing keys (returning no match instead of erroring). Prefix with strict to require the structure to match exactly and surface errors.
SELECT jsonb_path_query('{"a":[1,2,3]}', 'strict $.a[*] ? (@ > 1)'); -- 2, 3