016：Explain

一. Explain

EXPLAIN 官方文档

1.explain说明

• explain是解释SQL语句的执行计划，即显示该SQL语句怎么执行的

  • 使用explain的时候，也可以使用desc

• 5.6 版本支持DML语句进行explain解释

• 5.6 版本开始支持JSON格式的输出

注意：EXPLAIN查看的是执行计划，做SQL解析，不会去真的执行；且到5.7以后子查询也不会去执行。

(gcdb@localhost) 14:30:50 [mytest]> explain SELECT
    ->     t.TABLE_SCHEMA,
    ->     t.TABLE_NAME,
    ->     s.INDEX_NAME,
    ->     CARDINALITY,
    ->     TABLE_ROWS,
    ->     CARDINALITY / TABLE_ROWS AS SELECTIVITY
    -> FROM
    ->     information_schema.TABLES t,
    ->     (SELECT
    ->         table_schema, table_name, index_name, cardinality
    ->     FROM
    ->         information_schema.STATISTICS
    ->     WHERE
    ->         (table_schema , table_name, index_name, seq_in_index) IN (SELECT
    ->                 table_schema, table_name, index_name, MAX(seq_in_index)
    ->             FROM
    ->                 information_schema.STATISTICS
    ->             GROUP BY table_schema , table_name , index_name)) s
    -> WHERE
    ->     t.table_schema = s.table_schema
    ->         AND t.table_schema = 'employees'
    ->         AND t.table_name = s.table_name
    -> ORDER BY SELECTIVITY;
+----+--------------------+------------+------------+------+---------------+--------------+---------+------+------+----------+--------------------------------------------------------------------------------------------+
| id | select_type        | table      | partitions | type | possible_keys | key          | key_len | ref  | rows | filtered | Extra                             |
+----+--------------------+------------+------------+------+---------------+--------------+---------+------+------+----------+--------------------------------------------------------------------------------------------+
|  1 | PRIMARY            | t          | NULL       | ALL  | NULL          | TABLE_SCHEMA | NULL    | NULL | NULL |     NULL | Using where; Open_full_table; Scanned 1 database; Using temporary; Using filesort          |
|  1 | PRIMARY            | STATISTICS | NULL       | ALL  | NULL          | NULL         | NULL    | NULL | NULL |     NULL | Using where; Open_full_table; Scanned all databases; Using join buffer (Block Nested Loop) |
|  3 | DEPENDENT SUBQUERY | STATISTICS | NULL       | ALL  | NULL          | NULL         | NULL    | NULL | NULL |     NULL | Open_frm_only; Scanned all databases; Using temporary; Using filesort                      |
+----+--------------------+------------+------------+------+---------------+--------------+---------+------+------+----------+--------------------------------------------------------------------------------------------+
3 rows in set, 1 warning (0.00 sec)  -- 有warnings，这里相当于提供一个信息返回

(gcdb@localhost) 14:31:37 [mytest]> show warnings G;   -- 即将被弃用
*************************** 1. row ***************************
  Level: Warning
   Code: 1681
Message: 'EXTENDED' is deprecated and will be removed in a future release.
*************************** 2. row ***************************    -- 显示真正的执行语句
  Level: Note
   Code: 1003
Message: /* select#1 */ select `t`.`TABLE_SCHEMA` AS `TABLE_SCHEMA`,`t`.`TABLE_NAME` AS `TABLE_NAME`,`information_schema`.`STATISTICS`.`INDEX_NAME` AS `index_name`,`information_schema`.`STATISTICS`.`CARDINALITY` AS `cardinality`,`t`.`TABLE_ROWS` AS `TABLE_ROWS`,(`information_schema`.`STATISTICS`.`CARDINALITY` / `t`.`TABLE_ROWS`) AS `SELECTIVITY` from `information_schema`.`TABLES` `t` join `information_schema`.`STATISTICS` where ((`information_schema`.`STATISTICS`.`TABLE_NAME` = `t`.`TABLE_NAME`) and (`information_schema`.`STATISTICS`.`TABLE_SCHEMA` = `t`.`TABLE_SCHEMA`) and (`t`.`TABLE_SCHEMA` = 'employees') and <in_optimizer>((`information_schema`.`STATISTICS`.`TABLE_SCHEMA`,`information_schema`.`STATISTICS`.`TABLE_NAME`,`information_schema`.`STATISTICS`.`INDEX_NAME`,`information_schema`.`STATISTICS`.`SEQ_IN_INDEX`),<exists>(/* select#3 */ select 1,1,1,1 from `information_schema`.`STATISTICS` group by `information_schema`.`STATISTICS`.`TABLE_SCHEMA`,`information_schema`.`STATISTICS`.`TABLE_NAME`,`information_schema`.`STATISTICS`.`INDEX_NAME` having (((<cache>(`information_schema`.`STATISTICS`.`TABLE_SCHEMA`) = `information_schema`.`STATISTICS`.`TABLE_SCHEMA`) or <cache>(isnull(`information_schema`.`STATISTICS`.`TABLE_SCHEMA`))) and ((<cache>(`information_schema`.`STATISTICS`.`TABLE_NAME`) = `information_schema`.`STATISTICS`.`TABLE_NAME`) or <cache>(isnull(`information_schema`.`STATISTICS`.`TABLE_NAME`))) and ((<cache>(`information_schema`.`STATISTICS`.`INDEX_NAME`) = `information_schema`.`STATISTICS`.`INDEX_NAME`) or <cache>(isnull(`information_schema`.`STATISTICS`.`INDEX_NAME`))) and ((<cache>(`information_schema`.`STATISTICS`.`SEQ_IN_INDEX`) = max(`information_schema`.`STATISTICS`.`SEQ_IN_INDEX`)) or isnull(max(`information_schema`.`STATISTICS`.`SEQ_IN_INDEX`))) and <is_not_null_test>(`information_schema`.`STATISTICS`.`TABLE_SCHEMA`) and <is_not_null_test>(`information_schema`.`STATISTICS`.`TABLE_NAME`) and <is_not_null_test>(`information_schema`.`STATISTICS`.`INDEX_NAME`) and <is_not_null_test>(max(`information_schema`.`STATISTICS`.`SEQ_IN_INDEX`)))))) order by `SELECTIVITY`
2 rows in set (0.00 sec)

• 参数FORMAT
  • 使用format=json不仅仅是为了格式化输出效果，还有其他有用的显示信息。
  • 且当5.6版本后，使用MySQL Workbench，可以使用visual Explain方式显示详细的图示信息。

(gcdb@localhost) 16:24:02 [information_schema]> explain format=json  SELECT table_schema, table_name, index_name, MAX(seq_in_index) FROM information_schema.STATISTICS  GROUP BY table_schema , table_name , index_name G;
*************************** 1. row ***************************
EXPLAIN: {
  "query_block": {
    "select_id": 1,
    "cost_info": {
      "query_cost": "12.50"
    },
    "grouping_operation": {
      "using_temporary_table": true,
      "using_filesort": true,
      "cost_info": {
        "sort_cost": "2.00"
      },
      "table": {
        "table_name": "STATISTICS",
        "access_type": "ALL",
        "open_frm_only": true,
        "scanned_databases": "all",
        "used_columns": [
          "TABLE_CATALOG",
          "TABLE_SCHEMA",
          "TABLE_NAME",
          "NON_UNIQUE",
          "INDEX_SCHEMA",
          "INDEX_NAME",
          "SEQ_IN_INDEX",
          "COLUMN_NAME",
          "COLLATION",
          "CARDINALITY",
          "SUB_PART",
          "PACKED",
          "NULLABLE",
          "INDEX_TYPE",
          "COMMENT",
          "INDEX_COMMENT"
        ]
      }
    }
  }
}
1 row in set, 1 warning (0.00 sec)

ERROR:
No query specified

(gcdb@localhost) 16:25:29 [information_schema]>

2.Explain输出介绍

列	含义
id	执行计划的id标志
select_type	SELECT的类型
table	输出记录的表
partitions	符合的分区，[PARTITIONS]
type	JOIN的类型
possible_keys	优化器可能使用到的索引
key	优化器实际选择的索引
key_len	使用索引的字节长度
ref	进行比较的索引列
rows	优化器预估的记录数量
filtered	根据条件过滤得到的记录的百分比[EXTENDED]
extra	额外的显示选项

2.1. id

是指包含一组数字，表示查询中执行select子句或操作表的顺序。

口诀： id相等的从上往下看，id不等的从下往上看。但是在某些场合也不一定适用

2.2. select_type

select_type	含义
SIMPLE	简单SELECT(不使用UNION或子查询等)
PRIMARY	最外层的select
UNION	UNION中的第二个或后面的SELECT语句
DEPENDENT UNION	UNION中的第二个或后面的SELECT语句，依赖于外面的查询
UNION RESULT	UNION的结果
SUBQUERY	子查询中的第一个SELECT
DEPENDENT SUBQUERY	子查询中的第一个SELECT，依赖于外面的查询
DERIVED	派生表的SELECT(FROM子句的子查询)
MATERIALIZED	物化子查询
UNCACHEABLE SUBQUERY	不会被缓存的并且对于外部查询的每行都要重新计算的子查询
UNCACHEABLE UNION	属于不能被缓存的 UNION中的第二个或后面的SELECT语句

• MATERIALIZED
  • 产生中间临时表（实体）
  • 临时表自动创建索引并和其他表进行关联，提高性能
  • 和子查询的区别是，优化器将可以进行MATERIALIZED的语句自动改写成join，并自动创建索引

2.3. table

• 通常是用户操作的用户表
• <unionM, N> UNION得到的结果表
• <derivedN> 排生表，由id=N的语句产生
• <subqueryN> 由子查询物化产生的表，由id=N的语句产生

2.4. type

按照图上箭头的顺序来看，成本（cost）是从小到大

2.5. extra

• Using filesort：可以使用复合索引将filesort进行优化。提高性能
• Using index：比如使用覆盖索引
• Using where: 使用where过滤条件

Extra的信息是可以作为优化的提示，但是更多的是优化器优化的一种说明