SQL：如何在 jsonb 中找到最常见的键值对？

Question

Given a table where one of the columns has type jsonb, how does one find the most common key-value pairs among all records in that column?

The dict is only one level deep.

Answer 1

If your json has only one level is ease, here is a test code:

-- Test Table:
CREATE TEMP TABLE test (id serial,data jsonb);
-- Test Data
INSERT INTO test(data) VALUES
('{"a":{"a1":{"a2":1}},"b":2,"c":3}'::jsonb),
('{"a":1,"b":2,"c":3}'::jsonb),
('{"a":1}'::jsonb);

To get all keys, there is a function called jsonb_object_keys but only gets the set of keys in the outermost JSON object. So the query to get all the keys is:

SELECT DISTINCT key, COUNT(*) OVER(partition by key)
FROM
(SELECT jsonb_object_keys(data) AS key  FROM test)keys;

The result is:

key | count
----+-------
c   | 2
b   | 2
a   | 3

From that result you can get which key is the most used. f you want to get pairs... you just have to tweak a little this query... It would be better if you give an example data.

If you want to get the keys from a nested json... That's harder. You would have to create a PL code that recursively goes through the JSON to get witch keys are more used.