新特性解读 | MySQL 8.0 json到表的转换

原创： 杨涛涛

---

我们知道，JSON是一种轻量级的数据交互的格式，大部分NO SQL数据库的存储都用JSON。MySQL从5.7开始支持JSON格式的数据存储，并且新增了很多JSON相关函数。MySQL 8.0 又带来了一个新的把JSON转换为TABLE的函数JSON_TABLE，实现了JSON到表的转换。

举例一

我们看下简单的例子：

简单定义一个两级JSON 对象

mysql> set @ytt='{"name":[{"a":"ytt","b":"action"}, {"a":"dble","b":"shard"},{"a":"mysql","b":"oracle"}]}';
Query OK, 0 rows affected (0.00 sec)

第一级：

mysql> select json_keys(@ytt);
+-----------------+
| json_keys(@ytt) |
+-----------------+
| ["name"] |
+-----------------+
1 row in set (0.00 sec)

第二级：

mysql> select json_keys(@ytt,'$.name[0]');
+-----------------------------+
| json_keys(@ytt,'$.name[0]') |
+-----------------------------+
| ["a", "b"] |
+-----------------------------+
1 row in set (0.00 sec)

我们使用MySQL 8.0 的JSON_TABLE 来转换 @ytt。

mysql> select * from json_table(@ytt,'$.name[*]' columns (f1 varchar(10) path '$.a', f2 varchar(10) path '$.b')) as tt;

+-------+--------+
| f1 | f2 |
+-------+--------+
| ytt | action |
| dble | shard |
| mysql | oracle |
+-------+--------+
3 rows in set (0.00 sec)

举例二

再来一个复杂点的例子，用的是EXPLAIN 的JSON结果集。

JSON 串 @json_str1。

set @json_str1 = ' {
    "query_block": {
      "select_id": 1,
      "cost_info": {
        "query_cost": "1.00"
    },
    "table": {
      "table_name": "bigtable",
      "access_type": "const",
      "possible_keys": [
        "id"
    ],
     "key": "id",
     "used_key_parts": [
      "id"
    ],
     "key_length": "8",
     "ref": [
      "const"
    ],
     "rows_examined_per_scan": 1,
     "rows_produced_per_join": 1,
     "filtered": "100.00",
     "cost_info": {
       "read_cost": "0.00",
       "eval_cost": "0.20",
       "prefix_cost": "0.00",
       "data_read_per_join": "176"
   },
     "used_columns": [
       "id",
       "log_time",
       "str1",
       "str2"
     ]
   }
  }
}';

第一级：

mysql> select json_keys(@json_str1) as 'first_object';
+-----------------+
| first_object |
+-----------------+
| ["query_block"] |
+-----------------+
1 row in set (0.00 sec)

第二级：

mysql> select json_keys(@json_str1,'$.query_block') as 'second_object';
+-------------------------------------+
| second_object |
+-------------------------------------+
| ["table", "cost_info", "select_id"] |
+-------------------------------------+
1 row in set (0.00 sec)

第三级：

mysql> select json_keys(@json_str1,'$.query_block.table') as 'third_object'\G
*************************** 1. row ***************************
third_object: 
[
"key",
"ref",
"filtered",
"cost_info",
"key_length",
"table_name",
"access_type",
"used_columns",
"possible_keys",
"used_key_parts",
"rows_examined_per_scan",
"rows_produced_per_join"
]
1 row in set (0.01 sec)

第四级：

mysql> select json_extract(@json_str1,'$.query_block.table.cost_info') as 'forth_object'\G
*************************** 1. row ***************************
forth_object: {
"eval_cost":"0.20",
"read_cost":"0.00",
"prefix_cost":"0.00",
"data_read_per_join":"176"
}
1 row in set (0.00 sec)

那我们把这个JSON 串转换为表。

SELECT * FROM JSON_TABLE(@json_str1,
         "$.query_block"
         COLUMNS(
           rowid FOR ORDINALITY,
       NESTED PATH '$.table' 
       COLUMNS (
           a1_1 varchar(100) PATH '$.key',
           a1_2 varchar(100) PATH '$.ref[0]',
           a1_3 varchar(100) PATH '$.filtered',
           nested path '$.cost_info' 
           columns (
                a2_1 varchar(100) PATH '$.eval_cost' ,
                a2_2 varchar(100) PATH '$.read_cost',
                a2_3 varchar(100) PATH '$.prefix_cost',
                a2_4 varchar(100) PATH '$.data_read_per_join'

       ),
       a3 varchar(100) PATH '$.key_length',
           a4 varchar(100) PATH '$.table_name',
           a5 varchar(100) PATH '$.access_type',
           a6 varchar(100) PATH '$.used_key_parts[0]',
           a7 varchar(100) PATH '$.rows_examined_per_scan',
           a8 varchar(100) PATH '$.rows_produced_per_join',
           a9 varchar(100) PATH '$.key'

       ),
         NESTED PATH '$.cost_info' 
      columns (
      b1_1 varchar(100) path '$.query_cost'
       ),
         c INT path "$.select_id"
      )
    ) AS tt;



+-------+------+-------+--------+------+------+------+------+------+----------+-------+------+------+------+------+------+------+
| rowid | a1_1 | a1_2 | a1_3 | a2_1 | a2_2 | a2_3 | a2_4 | a3 | a4 | a5 | a6 | a7 | a8 | a9 | b1_1 | c |
+-------+------+-------+--------+------+------+------+------+------+----------+-------+------+------+------+------+------+------+
| 1 | id | const | 100.00 | 0.20 | 0.00 | 0.00 | 176 | 8 | bigtable | const | id | 1 | 1 | id | NULL | 1 |
| 1 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | 1.00 | 1 |
+-------+------+-------+--------+------+------+------+------+------+----------+-------+------+------+------+------+------+------+
2 rows in set (0.00 sec)

当然，JSON_table 函数还有其他的用法，我这里不一一列举了，详细的参考手册。

转载于:https://my.oschina.net/actiontechoss/blog/3066374