How do I generate nested json objects using mysql native json functions?

Question

Using only the native JSON fuctions (no PHP, etc) in MySQL version 5.7.12 (section 13.16 in the manual) I am trying to write a query to generate a JSON document from relational tables that contains a sub object. Given the following example:

CREATE TABLE `parent_table` (
   `id` int(11) NOT NULL,
   `desc` varchar(20) NOT NULL,
   PRIMARY KEY (`id`)
);
CREATE TABLE `child_table` (
   `id` int(11) NOT NULL,
   `parent_id` int(11) NOT NULL,
   `desc` varchar(20) NOT NULL,
   PRIMARY KEY (`id`,`parent_id`)
);
insert `parent_table` values (1,'parent row 1');
insert `child_table` values (1,1,'child row 1');
insert `child_table` values (2,1,'child row 2');

I am trying to generate a JSON document that looks like this:

[{
    "id" : 1,
    "desc" : "parent row 1",
    "child_objects" : [{
            "id" : 1,
            "parent_id" : 1,
            "desc" : "child row 1"
        }, {
            "id" : 2,
            "parent_id" : 1,
            "desc" : "child row 2"
        }
    ]
}]

I am new to MySQL and suspect there is a SQL pattern for generating nested JSON objects from one to many relationships but I'm having trouble finding it.

In Microsoft SQL (which I'm more familiar with) the following works:

select 
 [p].[id]
,[p].[desc]
,(select * from [dbo].[child_table] where [parent_id] = [p].[id] for json auto) AS [child_objects]
from [dbo].[parent_table] [p]
for json path

I attempted to write the equivalent in MySQL as follows:

select json_object(
 'id',p.id 
,'desc',p.`desc`
,'child_objects',(select json_object('id',id,'parent_id',parent_id,'desc',`desc`) 
                  from child_table where parent_id = p.id)
)
from parent_table p;

select json_object(
  'id',p.id 
 ,'desc',p.`desc`
 ,'child_objects',json_array((select json_object('id',id,'parent_id',parent_id,'desc',`desc`) 
                              from child_table where parent_id = p.id))
 )
 from parent_table p

Both attempts fail with the following error:

Error Code: 1242. Subquery returns more than 1 row

Answer 1

The reason you are getting these errors is that the parent json object is not expecting a result set as one of its inputs, you need to have simple object pairs like {name, string} etc bug report - may be available in future functionality... this just means that you need to convert your multi row results into a concatination of results separated by commas and then converted into a json array.

You almost had it with your second example.

You can achieve what you are after with the GROUP_CONCAT function

select json_object(
  'id',p.id 
 ,'desc',p.`desc`
 ,'child_objects',json_array(
                     (select GROUP_CONCAT(
                                 json_object('id',id,'parent_id',parent_id,'desc',`desc`)
                             )   
                      from child_table 
                      where parent_id = p.id))
                   )
 from parent_table p;

This almost works, it ends up treating the subquery as a string which leaves the escape characters in there.

'{"id": 1, 
"desc": "parent row 1", 
"child_objects": 
    ["
    {"id": 1,
     "desc": "child row 1", 
    "parent_id": 1
    },
    {"id": 2, 
    "desc": "child row 2", 
    "parent_id": 1}"
    ]
}'

In order to get this working in an appropriate format, you need to change the way you create the JSON output as follows:

select json_object(
  'id',p.id 
 ,'desc',p.`desc`
 ,'child_objects',(select CAST(CONCAT('[',
                GROUP_CONCAT(
                  JSON_OBJECT(
                    'id',id,'parent_id',parent_id,'desc',`desc`)),
                ']')
         AS JSON) from child_table where parent_id = p.id)

 ) from parent_table p;

This will give you the exact result you require:

'{"id": 1, 
"desc": "parent row 1", 
"child_objects": 
    [{"id": 1, 
    "desc": "child row 1", 
    "parent_id": 1
    }, 
    {"id": 2, 
    "desc": "child row 2", 
    "parent_id": 1
    }]  
}'