用 EXPLAIN 查看 JOIN 查询的执行计划

SQL 查询中可能会使用 JOIN 进行表连接，可以通过 EXPLAIN 语句来查看 JOIN 查询的执行计划。

示例

我们这里准备一个简单的示例，帮助你理解使用 EXPLAIN 解读 JOIN 查询的执行计划。

  1. > drop table if exists t1;
  2. > create table t1 (id int,ti tinyint unsigned,si smallint,bi bigint unsigned,fl float,dl double,de decimal,ch char(20),vch varchar(20),dd date,dt datetime);
  3. > insert into t1 values(1,1,4,3,1113.32,111332,1113.32,'hello','subquery','2022-04-28','2022-04-28 22:40:11');
  4. > insert into t1 values(2,2,5,2,2252.05,225205,2252.05,'bye','sub query','2022-04-28','2022-04-28 22:40:11');
  5. > insert into t1 values(3,6,6,3,3663.21,366321,3663.21,'hi','subquery','2022-04-28','2022-04-28 22:40:11');
  6. > insert into t1 values(4,7,1,5,4715.22,471522,4715.22,'good morning','my subquery','2022-04-28','2022-04-28 22:40:11');
  7. > insert into t1 values(5,1,2,6,51.26,5126,51.26,'byebye',' is subquery?','2022-04-28','2022-04-28 22:40:11');
  8. > insert into t1 values(6,3,2,1,632.1,6321,632.11,'good night','maybe subquery','2022-04-28','2022-04-28 22:40:11');
  9. > insert into t1 values(7,4,4,3,7443.11,744311,7443.11,'yes','subquery','2022-04-28','2022-04-28 22:40:11');
  10. > insert into t1 values(8,7,5,8,8758.00,875800,8758.11,'nice to meet','just subquery','2022-04-28','2022-04-28 22:40:11');
  11. > insert into t1 values(9,8,4,9,9849.312,9849312,9849.312,'see you','subquery','2022-04-28','2022-04-28 22:40:11');
  13. > drop table if exists t2;
  14. > create table t2 (id int,ti tinyint unsigned,si smallint,bi bigint unsigned,fl float,dl double,de decimal,ch char(20),vch varchar(20),dd date,dt datetime);
  15. > insert into t2 values(1,1,4,3,1113.32,111332,1113.32,'hello','subquery','2022-04-28','2022-04-28 22:40:11');
  16. > insert into t2 values(2,2,5,2,2252.05,225205,2252.05,'bye','sub query','2022-04-28','2022-04-28 22:40:11');
  17. > insert into t2 values(3,6,6,3,3663.21,366321,3663.21,'hi','subquery','2022-04-28','2022-04-28 22:40:11');
  18. > insert into t2 values(4,7,1,5,4715.22,471522,4715.22,'good morning','my subquery','2022-04-28','2022-04-28 22:40:11');
  19. > insert into t2 values(5,1,2,6,51.26,5126,51.26,'byebye',' is subquery?','2022-04-28','2022-04-28 22:40:11');
  20. > insert into t2 values(6,3,2,1,632.1,6321,632.11,'good night','maybe subquery','2022-04-28','2022-04-28 22:40:11');
  21. > insert into t2 values(7,4,4,3,7443.11,744311,7443.11,'yes','subquery','2022-04-28','2022-04-28 22:40:11');
  22. > insert into t2 values(8,7,5,8,8758.00,875800,8758.11,'nice to meet','just subquery','2022-04-28','2022-04-28 22:40:11');
  23. > insert into t2 values(9,8,4,9,9849.312,9849312,9849.312,'see you','subquery','2022-04-28','2022-04-28 22:40:11');

Hash Join

在 Hash Join 操作中，MatrixOne 首先读取表 t1 与 t2 中相对较小的一个表对其中每个需要被连接的值使用哈希函数，得到一个哈希表；然后对另一个表的每一行进行扫描并计算哈希值，与上一步生成的哈希表进行比对，如果有符合连接标准的值，则根据连接谓词生成一个新的连接表。

MatrixOne 中的 Hash Join 算子是多线程的，并且可以并发执行。

下面是一个 Hash Join 示例：

  1. > SELECT /*+ HASH_JOIN(t1, t2) */ * FROM t1, t2 WHERE t1.id = t2.id;
  2. +------+------+------+------+----------+---------+------+--------------+----------------+------------+---------------------+------+------+------+------+----------+---------+------+--------------+----------------+------------+---------------------+
  3. | id   | ti   | si   | bi   | fl       | dl      | de   | ch           | vch            | dd         | dt                  | id   | ti   | si   | bi   | fl       | dl      | de   | ch           | vch            | dd         | dt                  |
  4. +------+------+------+------+----------+---------+------+--------------+----------------+------------+---------------------+------+------+------+------+----------+---------+------+--------------+----------------+------------+---------------------+
  5. |    1 |    1 |    4 |    3 |  1113.32 |  111332 | 1113 | hello        | subquery       | 2022-04-28 | 2022-04-28 22:40:11 |    1 |    1 |    4 |    3 |  1113.32 |  111332 | 1113 | hello        | subquery       | 2022-04-28 | 2022-04-28 22:40:11 |
  6. |    2 |    2 |    5 |    2 |  2252.05 |  225205 | 2252 | bye          | sub query      | 2022-04-28 | 2022-04-28 22:40:11 |    2 |    2 |    5 |    2 |  2252.05 |  225205 | 2252 | bye          | sub query      | 2022-04-28 | 2022-04-28 22:40:11 |
  7. |    3 |    6 |    6 |    3 |  3663.21 |  366321 | 3663 | hi           | subquery       | 2022-04-28 | 2022-04-28 22:40:11 |    3 |    6 |    6 |    3 |  3663.21 |  366321 | 3663 | hi           | subquery       | 2022-04-28 | 2022-04-28 22:40:11 |
  8. |    4 |    7 |    1 |    5 |  4715.22 |  471522 | 4715 | good morning | my subquery    | 2022-04-28 | 2022-04-28 22:40:11 |    4 |    7 |    1 |    5 |  4715.22 |  471522 | 4715 | good morning | my subquery    | 2022-04-28 | 2022-04-28 22:40:11 |
  9. |    5 |    1 |    2 |    6 |    51.26 |    5126 |   51 | byebye       |  is subquery?  | 2022-04-28 | 2022-04-28 22:40:11 |    5 |    1 |    2 |    6 |    51.26 |    5126 |   51 | byebye       |  is subquery?  | 2022-04-28 | 2022-04-28 22:40:11 |
  10. |    6 |    3 |    2 |    1 |    632.1 |    6321 |  632 | good night   | maybe subquery | 2022-04-28 | 2022-04-28 22:40:11 |    6 |    3 |    2 |    1 |    632.1 |    6321 |  632 | good night   | maybe subquery | 2022-04-28 | 2022-04-28 22:40:11 |
  11. |    7 |    4 |    4 |    3 |  7443.11 |  744311 | 7443 | yes          | subquery       | 2022-04-28 | 2022-04-28 22:40:11 |    7 |    4 |    4 |    3 |  7443.11 |  744311 | 7443 | yes          | subquery       | 2022-04-28 | 2022-04-28 22:40:11 |
  12. |    8 |    7 |    5 |    8 |     8758 |  875800 | 8758 | nice to meet | just subquery  | 2022-04-28 | 2022-04-28 22:40:11 |    8 |    7 |    5 |    8 |     8758 |  875800 | 8758 | nice to meet | just subquery  | 2022-04-28 | 2022-04-28 22:40:11 |
  13. |    9 |    8 |    4 |    9 | 9849.312 | 9849312 | 9849 | see you      | subquery       | 2022-04-28 | 2022-04-28 22:40:11 |    9 |    8 |    4 |    9 | 9849.312 | 9849312 | 9849 | see you      | subquery       | 2022-04-28 | 2022-04-28 22:40:11 |
  14. +------+------+------+------+----------+---------+------+--------------+----------------+------------+---------------------+------+------+------+------+----------+---------+------+--------------+----------------+------------+---------------------+
  15. 9 rows in set (0.00 sec)
  17. mysql> EXPLAIN SELECT /*+ HASH_JOIN(t1, t2) */ * FROM t1, t2 WHERE t1.id = t2.id;
  18. +------------------------------------+
  19. | QUERY PLAN                         |
  20. +------------------------------------+
  21. | Project                            |
  22. |   ->  Join                         |
  23. |         Join Type: INNER           |
  24. |         Join Cond: (t1.id = t2.id) |
  25. |         ->  Table Scan on db1.t1   |
  26. |         ->  Table Scan on db1.t2   |
  27. +------------------------------------+
  28. 6 rows in set (0.01 sec)

MatrixOne 会按照以下顺序执行 Hash Join 算子：

  1. 分别并行扫描表 t2 和 t1。
  2. 执行 JOIN 的过滤查询：(t1.id = t2.id)。
  3. 执行 INNER JOIN。