MySQL 如何在 JSON 字段中搜索与 JSON 值相交的所有行

Question

Assume that I have a MySQL table with a JSON field, with the structure & values shown below:

CREATE TABLE `projects` (
  `project_ids` json DEFAULT NULL
) ENGINE=InnoDB;

INSERT INTO `projects` (`project_ids`)
VALUES
    ('[1, 2, 3]'),
    ('[1, 2]'),
    ('[2]'),
    ('[1]'),
    ('[2, 3]');

I would like to get all rows that have ANY of the JSON values that match in a SELECT query.

For example:

SELECT * FROM projects WHERE JSON_CONTAINS(project_ids, '[1,2]');

This yields 2 rows:

[1, 2, 3]
[1, 2]

However, I would like it to yield all of the rows, because every row has at least a "1" or a "2" in the JSON. In other words, I'm looking for a query that will return rows where the JSON intersection is non-empty.

Thus, I WANT it to return:

[1, 2, 3]
[1, 2]
[2]
[1]
[2, 3]

Please ignore the larger question of whether it's even a good idea to use JSON fields over foreign keys, etc. Assume that I have to use a JSON field.

Answer 1

OR is the simplest method:

SELECT *
FROM projects
WHERE JSON_CONTAINS(project_ids, '[1]') OR
      JSON_CONTAINS(project_ids, '[2]') ;

But if you want to pass in the JSON array, use JSON_OVERLAPS() :

SELECT *
FROM projects
WHERE JSON_OVERLAPS(project_ids, '[1,2]') ;

Here is a db<>fiddle.

Answer 2

The answer from @GordonLinoff gives the correct result, but it's worth mentioning that most solutions that use JSON functions cannot be optimized with indexes. You should consider what happens as the table gets larger. Every query will do a table-scan, and that becomes worse and worse for performance as the table grows.

EXPLAIN shows this:

+----+-------------+----------+------------+------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table    | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra       |
+----+-------------+----------+------------+------+---------------+------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | projects | NULL       | ALL  | NULL          | NULL | NULL    | NULL |    1 |   100.00 | Using where |
+----+-------------+----------+------------+------+---------------+------+---------+------+------+----------+-------------+

(the rows column will be as large as the number of rows in the table)

Instead, an easier way to implement this kind of query is to store the data in normal rows and columns, not JSON. Store one project_id per row.

Then the query becomes much more clear, and it can be optimized with an index:

SELECT * FROM projects WHERE project_id IN (1, 2);

I understand you mentioned in your question that we should take as a given that using JSON is a requirement. You have a solution to get the result you described. I just want to make clear to other readers:

If your query references a JSON column in any clause other than the select-list, it's a sign that you should store the data in a normalized form, not in JSON.