来自 3 个左连接的极慢选择查询

Question

我有一个非常慢的选择查询，我尝试了很多方法来加快它，但它仍然需要 分钟 来执行。我将解释这些表格以及我尝试过的内容。我想知道是否有更有效的方法来组织这个相当简单的查询。

大约有 18000 行。

查询解释

基本上头寸由经纪人拥有的客户所有，因此要将头寸与经纪人联系起来，我们通过clients 表。我正在显示与相应经纪人的头寸，以及它们是否通过covered_positions 链接到另一个头寸，以及关联的头寸（如果有）。

表格说明

• positions 包含客户完成的所有交易（股票/期权）。它有一个 PK positionsID (AI)，但 MySQL 似乎没有将它识别为 可能的索引。在下面解释结果。
• clients 包含所有客户信息（客户）。它有 PK clientsID (AI)，这是 positions 中的外键。
• login_users 包含经纪人（员工）。它的 PK user_id (AI) 是 clients 中的外键 broker_id。
• covered_positions 包含 2-5 个位置之间的链接（不是超链接）。它有 PK covered_id (AI)。 linked_id* 列是链接在一起的positionsIDs。

查询

delimiter $$

CREATE VIEW `main_join3` AS 
select `positions`.`id` AS `positionsID`,
`positions`.`order_list` AS `order_list`,
`positions`.`client_id` AS `client_id`,
`positions`.`client_name` AS `client_name`,
`positions`.`broker_name` AS `pos_broker_name`,
`positions`.`account_status` AS `pos_account_status`,
`positions`.`cost_basis` AS `cost_basis`,
`positions`.`cost_per_share` AS `cost_per_share`,
`positions`.`security` AS `security`,
`positions`.`security_type` AS `security_type`,
`positions`.`strike_price` AS `strike_price`,
`positions`.`option_type` AS `option_type`,
`positions`.`exp_month` AS `exp_month`,
`positions`.`exp_year` AS `exp_year`,
`positions`.`exp_date` AS `exp_date`,
`positions`.`buy_shares` AS `buy_shares`,
`positions`.`buy_date` AS `buy_date`,
`positions`.`buy_price` AS `buy_price`,
`positions`.`buy_commission` AS `buy_commission`,
`positions`.`buy_misc` AS `buy_misc`,
`positions`.`buy_cost` AS `buy_cost`,
`positions`.`sell_shares` AS `sell_shares`,
`positions`.`sell_date` AS `sell_date`,
`positions`.`sell_price` AS `sell_price`,
`positions`.`sell_commission` AS `sell_commission`,
`positions`.`sell_misc` AS `sell_misc`,
`positions`.`sell_cost` AS `sell_cost`,
`positions`.`loss` AS `loss`,
`positions`.`profit` AS `profit`,
`positions`.`funds_in_out` AS `funds_in_out`,
`positions`.`funds_in` AS `funds_in`,
`positions`.`funds_out` AS `funds_out`,
`positions`.`stop_order` AS `stop_order`,
`positions`.`notes` AS `notes`,
`positions`.`orange_highlight` AS `orange_highlight`,
`positions`.`date_created` AS `pos_date_created`,
`positions`.`date_updated` AS `pos_date_updated`,
`positions`.`created_by_id` AS `pos_cb_id`,
`positions`.`created_by_name` AS `pos_cb_name`,
`clients`.`id` AS `clientsID`,
`clients`.`broker_id` AS `broker_id`,
`clients`.`account_status` AS `clients_account_status`,
`login_users`.`user_id` AS `user_id`,
`login_users`.`user_level` AS `user_level`,
`login_users`.`username` AS `username`,
`login_users`.`name` AS `name`,
`covered_positions`.`id` AS `covered_id`,
`covered_positions`.`linked_id1` AS `linked_id1`,
`covered_positions`.`linked_id2` AS `linked_id2`,
`covered_positions`.`linked_id3` AS `linked_id3`,
`covered_positions`.`linked_id4` AS `linked_id4`,
`covered_positions`.`linked_id5` AS `linked_id5` 
from (((`positions` 
left join `clients` 
on((`positions`.`client_id` = `clients`.`id`))) 
left join `login_users` 
on((`clients`.`broker_id` = `login_users`.`user_id`))) 
left join `covered_positions` 
on(((`positions`.`id` = `covered_positions`.`linked_id1`) 
or (`positions`.`id` = `covered_positions`.`linked_id2`) 
or (`positions`.`id` = `covered_positions`.`linked_id3`) 
or (`positions`.`id` = `covered_positions`.`linked_id4`) 
or (`positions`.`id` = `covered_positions`.`linked_id5`))))$$

解释结果

id  select_type table   type    possible_keys   key key_len ref rows    Extra
1   SIMPLE  positions   ALL NULL    NULL    NULL    NULL    18070   
1   SIMPLE  clients eq_ref  PRIMARY PRIMARY 4   blackri_posting.positions.client_id 1   
1   SIMPLE  login_users eq_ref  PRIMARY,user_id PRIMARY 4   blackri_posting.clients.broker_id   1   
1   SIMPLE  covered_positions   ALL NULL    NULL    NULL    NULL    214

我尝试过的

• 从查询中删除所有未使用的列
• 将查询创建为视图
• 分析和优化数据库
• 更换服务器

非常感谢任何支持。

Answer 1

我设法使用触发器解决了这个问题。我没有使用那些昂贵的 OR，而是简单地在 positions 中添加了一个名为 link_id 的列，其中来自 covered_positions 的 id 是通过以下方式添加的：

DELIMITER $$
CREATE TRIGGER covereds_after_insert
    AFTER INSERT ON covered_positions
    FOR EACH ROW
BEGIN
    UPDATE positions
    SET link_id = NEW.id 
    WHERE positions.id 
    IN (NEW.linked_id1, NEW.linked_id2, NEW.linked_id3, NEW.linked_id4, NEW.linked_id5);
END
$$

与

DELIMITER $$
CREATE TRIGGER covereds_after_delete
    AFTER DELETE ON covered_positions
    FOR EACH ROW
BEGIN
    UPDATE positions
    SET link_id = NULL
    WHERE positions.id 
    IN (OLD.linked_id1, OLD.linked_id2, OLD.linked_id3, OLD.linked_id4, OLD.linked_id5);
END
$$

这样查询不会将 18,000 行变成大约 400 万行。

Answer 2

虽然次要，但你的括号用得太多了...这里是从/join修改的

from 
   positions
      left join clients
         ON positions.client_id = clients.id 
         left join login_users
            ON clients.broker_id = login_users.user_id
      left join covered_positions
         ON  positions.id = covered_positions.linked_id1
         or  positions.id = covered_positions.linked_id2  
         or  positions.id = covered_positions.linked_id3 
         or  positions.id = covered_positions.linked_id4
         or  positions.id = covered_positions.linked_id5

其次，只是为了笑，尝试将 MySQL 关键字“STRAIGHT_JOIN”添加到您的查询中......

SELECT STRAIGHT_JOIN（查询的其余部分）

第三，您的表“Covered_positions” 您是否对每个linked_id 列都有索引。如果无法在连接上使用索引，这将使您的查询爬网。

最后，如果您的覆盖头寸表有客户 ID，我绝对会将其添加为索引中的第一个字段，并在连接中包含客户 ID...这样，您至少会加入客户首先，然后是该客户中的职位。