新特性解读 | 数组范围遍历功能

作者：杨涛涛
资深数据库专家，专研 MySQL 十余年。擅长 MySQL、PostgreSQL、MongoDB 等开源数据库相关的备份恢复、SQL 调优、监控运维、高可用架构设计等。目前任职于爱可生，为各大运营商及银行金融企业提供 MySQL 相关技术支持、MySQL 相关课程培训等工作。
本文来源：原创投稿
*爱可生开源社区出品，原创内容未经授权不得随意使用，转载请联系小编并注明来源。

---

大家都知道，MySQL 从 5.7 开始，增加了新的数据类型：JSON。

JSON 类型的出现，弥补了 MySQL 长期以来相比其他关系型数据库的非标准化特性略势，比如可以用 JSON 来实现类似字典、数组等类型。以及之后大量针对 JSON 类型字串的相关遍历方法与函数。

比如对数组来说，简单的遍历路径为 “$[数组小标]"或者"$[*]” 来一次性遍历所有元素。MySQL 8.0 又新增了对于数组元素的范围遍历，比如 “$[m to n]”，表示遍历下标为 m 到 n 的元素。

举个例子，以下 JSON 字串（随便找张表 explain format=json 的结果）：

# javascript
set @json_str1 = '{
  "query_block": {
    "table": {
      "filtered": "100.00",
      "cost_info": {
        "eval_cost": "898504.10",
        "read_cost": "17457.25",
        "prefix_cost": "915961.35",
        "data_read_per_join": "137M"
      },
      "table_name": "t1",
      "access_type": "ALL",
      "used_columns": [
        "id",
        "r1",
        "r2",
        "r3",
        "r4",
        "r5",
        "r6",
        "r7",
        "r8"
      ],
      "rows_examined_per_scan": 8985041,
      "rows_produced_per_join": 8985041
    },
    "cost_info": {
      "query_cost": "915961.35"
    },
    "select_id": 1
  }
}';

其中 key 为 used_columns 对应的值就是一个数组，现在我把这个数据遍历出来，并且生成以下结果：

# javascript
["id", "r1", "r2", "r3", "r4", "r5", "r6", "r7", "r8"]

["id", "r1", "r2", "r3", "r4", "r5", "r6", "r7"]

["id", "r1", "r2", "r3", "r4", "r5", "r6"]

["id", "r1", "r2", "r3", "r4", "r5"]

["id", "r1", "r2", "r3", "r4"]

["id", "r1", "r2", "r3"]

["id", "r1", "r2"]

["id", "r1"]

["id"]

先把这个键对应的值赋给一个变量：

mysql> select json_extract(@json_str1,"$.query_block.table.used_columns") into @json_array1;
Query OK, 1 row affected (0.00 sec)

接下来写个传统 PATH 遍历数据的存储过程来实现以上这个结果：

DELIMITER $$

USE `ytt`$$

DROP PROCEDURE IF EXISTS `sp_print_json_array_sample`$$

CREATE DEFINER=`root`@`%` PROCEDURE `sp_print_json_array_sample`(
    IN f_str1 JSON
    )
BEGIN
    DECLARE v_tmp_length TINYINT UNSIGNED DEFAULT 0;
    DECLARE i TINYINT UNSIGNED DEFAULT 0;

    SET v_tmp_length = JSON_LENGTH(f_str1);
    SET i = v_tmp_length;
    SET @v_tmp_result = f_str1;

    DROP TEMPORARY TABLE IF EXISTS tmp;
    CREATE TEMPORARY TABLE IF NOT EXISTS tmp(str1 JSON);

    WHILE i > 0
    DO
      SET i = i - 1;
      SET @stmt = CONCAT('select json_remove(@v_tmp_result,"$[',i,']") into @v_tmp_result');
      INSERT INTO tmp VALUES (@v_tmp_result);
      PREPARE s1 FROM @stmt;
      EXECUTE s1;
      END WHILE;
      DROP PREPARE s1;

      SET @v_stmt = NULL;
      SET @v_tmp_result = NULL;
      SELECT * FROM tmp;
END$$

DELIMITER ;

调用下这个存储过程：

mysql> call sp_print_json_array_sample(@json_array1);
+--------------------------------------------------------+
| str1                                                   |
+--------------------------------------------------------+
| ["id", "r1", "r2", "r3", "r4", "r5", "r6", "r7", "r8"] |
| ["id", "r1", "r2", "r3", "r4", "r5", "r6", "r7"]       |
| ["id", "r1", "r2", "r3", "r4", "r5", "r6"]             |
| ["id", "r1", "r2", "r3", "r4", "r5"]                   |
| ["id", "r1", "r2", "r3", "r4"]                         |
| ["id", "r1", "r2", "r3"]                               |
| ["id", "r1", "r2"]                                     |
| ["id", "r1"]                                           |
| ["id"]                                                 |
+--------------------------------------------------------+
9 rows in set (0.01 sec)

Query OK, 0 rows affected (0.01 sec)

那在 MySQL 8.0 下，可以直接把这个存储过程给优化掉，简化的写法如下：

DELIMITER $$

USE `ytt`$$

DROP PROCEDURE IF EXISTS `sp_print_json_array_sample`$$

CREATE DEFINER=`root`@`%` PROCEDURE `sp_print_json_array_sample`(
    IN f_str1 JSON
    )
BEGIN
    DECLARE v_tmp_length TINYINT UNSIGNED DEFAULT 0;
    DECLARE i TINYINT UNSIGNED DEFAULT 0;

    SET v_tmp_length = JSON_LENGTH(f_str1);
    SET i = v_tmp_length;
    DROP TEMPORARY TABLE IF EXISTS tmp;
    CREATE TEMPORARY TABLE IF NOT EXISTS tmp(str1 JSON);

    WHILE i > 0
    DO
      SET i = i - 1;
      SET @stmt = CONCAT(' insert into tmp select json_extract(''',f_str1,''','' $[ 0 to     ',i,']'') ');
      PREPARE s1 FROM @stmt;
      EXECUTE s1;
    END WHILE;

    DROP PREPARE s1;
    SET @v_stmt = NULL;

    SELECT str1 AS array_result FROM tmp;

END$$

DELIMITER ;

调用结果一样：

mysql> call sp_print_json_array_sample(@json_array1);
+--------------------------------------------------------+
| array_result                                           |
+--------------------------------------------------------+
| ["id", "r1", "r2", "r3", "r4", "r5", "r6", "r7", "r8"] |
| ["id", "r1", "r2", "r3", "r4", "r5", "r6", "r7"]       |
| ["id", "r1", "r2", "r3", "r4", "r5", "r6"]             |
| ["id", "r1", "r2", "r3", "r4", "r5"]                   |
| ["id", "r1", "r2", "r3", "r4"]                         |
| ["id", "r1", "r2", "r3"]                               |
| ["id", "r1", "r2"]                                     |
| ["id", "r1"]                                           |
| ["id"]                                                 |
+--------------------------------------------------------+
9 rows in set (0.00 sec)

Query OK, 0 rows affected (0.00 sec)

其中路径 “$[m to n]” , n 也可以是保留字 last，代表最后一个下标，比如：

mysql> select json_extract(@json_array1,'$[0 to last-7]') as result;
+--------------+
| result       |
+--------------+
| ["id", "r1"] |
+--------------+
1 row in set (0.00 sec)

总结

这里简单介绍了 MySQL 8.0 对于 JSON 数组的范围遍历功能，希望大家在开发过程中更加得心应手。