There is no easy way. Per documentation:
GIN indexes can be used to efficiently search for keys or key/value
pairs occurring within a large number of jsonb documents (datums)
Bold emphasis mine. There is no index over all values. (Those can have non-compatible data types!) If you do not know the name(s) of all key(s) you have to inspect all JSON values in every row.
If there are just two keys like you demonstrate (or just a few well-kown keys), it's still easy enough:
SELECT *
FROM users
WHERE data->>'property_a' = 'b2' OR
data->>'property_b' = 'b2';
Can be supported with a simple expression index:
CREATE INDEX foo_idx ON users ((data->>'property_a'), (data->>'property_b'))
Or with a GIN index:
SELECT *
FROM users
WHERE data @> '{"property_a": "b2"}' OR
data @> '{"property_b": "b2"}'
CREATE INDEX bar_idx ON users USING gin (data jsonb_path_ops);
If you don't know all key names, things get more complicated ...
You could use jsonb_each()
or jsonb_each_text()
to unnest all values into a set and then check with an ANY
construct:
SELECT *
FROM users
WHERE jsonb '"b2"' = ANY (SELECT (jsonb_each(data)).value);
Or
...
WHERE 'b2' = ANY (SELECT (jsonb_each_text(data)).value);
db<>fiddle here
But there is no index support for the last one. You could instead extract all values into and array and create an expression index on that, and match that expression in queries with array operators ...
Related: