优化用于在两个表之间查找匹配的 MySQL 查询答案

【问题标题】：Optimize MySQL query used to find matches between two tables优化用于在两个表之间查找匹配的 MySQL 查询
【发布时间】：2019-03-13 05:10:39
【问题描述】：

在我正在进行的一个项目中，我有两个表：

consumption：包含来自客户的历史订单，其字段指定了他们购买的产品的功能（每行一个产品）
product：包含当前产品库存

数据库引擎是 InnoDB。

目标：

应用程序必须显示双方的匹配，我的意思是：
当我列出当前产品库存时，我想显示一列显示有多少历史订单与特定产品匹配
当我列出历史订单时，我想查看有多少产品与特定历史订单匹配

consumption 和 product 表以及其他相关表的数据库结构：

CREATE TABLE `consumption` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `created_by_id` INT(11) NULL DEFAULT NULL,
    `client_id` INT(11) NOT NULL,
    `data_import_id` INT(11) NULL DEFAULT NULL,
    `tmp_consumption_id` INT(11) NULL DEFAULT NULL,
    `material_id` INT(11) NULL DEFAULT NULL,
    `quality_id` INT(11) NULL DEFAULT NULL,
    `thick` DECIMAL(10,3) NULL DEFAULT NULL,
    `thick_max` DECIMAL(10,3) NULL DEFAULT NULL,
    `width` DECIMAL(10,2) NULL DEFAULT NULL,
    `width_max` DECIMAL(10,2) NULL DEFAULT NULL,
    `long` INT(11) NULL DEFAULT NULL,
    `long_max` INT(11) NULL DEFAULT NULL,
    `purchase_price` DECIMAL(10,2) NULL DEFAULT NULL,
    `sale_price` DECIMAL(10,2) NULL DEFAULT NULL,
    `comments` VARCHAR(255) NULL DEFAULT NULL,
    `annual_consumption` DECIMAL(10,3) NULL DEFAULT NULL,
    `type` ENUM('consumption','request') NULL DEFAULT 'consumption',
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    `covering_grammage` VARCHAR(64) NULL DEFAULT NULL,
    `asp_sup_acab` VARCHAR(64) NULL DEFAULT NULL,
    PRIMARY KEY (`id`),
    INDEX `fk_consumption_client1` (`client_id`),
    INDEX `created_by_id` (`created_by_id`),
    INDEX `material_id` (`material_id`),
    INDEX `quality_id` (`quality_id`),
    CONSTRAINT `consumption_ibfk_1` FOREIGN KEY (`material_id`) REFERENCES `material` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION,
    CONSTRAINT `consumption_ibfk_2` FOREIGN KEY (`quality_id`) REFERENCES `quality` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION,
    CONSTRAINT `fk_consumption_client1` FOREIGN KEY (`client_id`) REFERENCES `client` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=30673
;

CREATE TABLE `product` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `warehouse_id` INT(11) NULL DEFAULT NULL,
    `created_by_id` INT(11) NULL DEFAULT NULL,
    `data_import_id` INT(11) NULL DEFAULT NULL,
    `tmp_product_id` INT(11) NULL DEFAULT NULL,
    `code` VARCHAR(32) NOT NULL,
    `material_id` INT(11) NULL DEFAULT NULL,
    `quality_id` INT(11) NULL DEFAULT NULL,
    `covering_id` INT(11) NULL DEFAULT NULL,
    `finish_id` INT(11) NULL DEFAULT NULL,
    `source` VARCHAR(128) NULL DEFAULT NULL,
    `thickness` DECIMAL(10,3) NULL DEFAULT NULL,
    `width` INT(11) NULL DEFAULT NULL,
    `tons` DECIMAL(10,3) NULL DEFAULT NULL,
    `re` INT(11) NULL DEFAULT NULL,
    `rm` INT(11) NULL DEFAULT NULL,
    `a_percent` INT(11) NULL DEFAULT NULL,
    `comments` VARCHAR(255) NULL DEFAULT NULL,
    `price` DECIMAL(10,2) NULL DEFAULT NULL,
    `deleted` TINYINT(1) NOT NULL DEFAULT '0',
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    PRIMARY KEY (`id`),
    INDEX `warehouse_id` (`warehouse_id`),
    INDEX `material_id` (`material_id`),
    INDEX `quality_id` (`quality_id`),
    INDEX `covering_id` (`covering_id`),
    INDEX `finish_id` (`finish_id`),
    CONSTRAINT `product_ibfk_1` FOREIGN KEY (`material_id`) REFERENCES `material` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION,
    CONSTRAINT `product_ibfk_2` FOREIGN KEY (`quality_id`) REFERENCES `quality` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION,
    CONSTRAINT `product_ibfk_3` FOREIGN KEY (`covering_id`) REFERENCES `covering` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION,
    CONSTRAINT `product_ibfk_4` FOREIGN KEY (`finish_id`) REFERENCES `finish` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION,
    CONSTRAINT `product_ibfk_5` FOREIGN KEY (`warehouse_id`) REFERENCES `warehouse` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=740
;
CREATE TABLE `client` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `zone_id` INT(11) NULL DEFAULT NULL,
    `zone2_id` INT(11) NULL DEFAULT NULL,
    `code` VARCHAR(64) NOT NULL,
    `business_name` VARCHAR(255) NULL DEFAULT NULL,
    `fiscal_name` VARCHAR(255) NULL DEFAULT NULL,
    `nif` VARCHAR(15) NULL DEFAULT NULL,
    `contact_short_name` VARCHAR(128) NULL DEFAULT NULL,
    `contact_full_name` VARCHAR(128) NULL DEFAULT NULL,
    `email` VARCHAR(255) NULL DEFAULT NULL,
    `group` VARCHAR(255) NULL DEFAULT NULL,
    `status` TINYINT(1) NOT NULL DEFAULT '1',
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    PRIMARY KEY (`id`),
    UNIQUE INDEX `code_UNIQUE` (`code`),
    INDEX `zone_id` (`zone_id`),
    INDEX `zone2_id` (`zone2_id`),
    CONSTRAINT `client_ibfk_1` FOREIGN KEY (`zone_id`) REFERENCES `zone` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=443
;
CREATE TABLE `client_group` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `code` VARCHAR(15) NOT NULL,
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    PRIMARY KEY (`id`),
    UNIQUE INDEX `code` (`code`)
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=49
;
CREATE TABLE `client_has_group` (
    `client_id` INT(11) NOT NULL,
    `group_id` INT(11) NOT NULL,
    INDEX `client_id` (`client_id`),
    INDEX `group_id` (`group_id`),
    CONSTRAINT `client_has_group_ibfk_1` FOREIGN KEY (`client_id`) REFERENCES `client` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION,
    CONSTRAINT `client_has_group_ibfk_2` FOREIGN KEY (`group_id`) REFERENCES `client_group` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
;
CREATE TABLE `covering` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `code` VARCHAR(128) NOT NULL,
    `group` VARCHAR(128) NULL DEFAULT NULL,
    `equivalence` VARCHAR(128) NULL DEFAULT NULL,
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    PRIMARY KEY (`id`),
    UNIQUE INDEX `code_UNIQUE` (`code`)
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=55
;
CREATE TABLE `finish` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `code` VARCHAR(128) NOT NULL,
    `group` VARCHAR(128) NULL DEFAULT NULL,
    `equivalence` VARCHAR(128) NULL DEFAULT NULL,
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    PRIMARY KEY (`id`),
    UNIQUE INDEX `code_UNIQUE` (`code`)
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=42
;
CREATE TABLE `material` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `code` VARCHAR(128) NOT NULL,
    `group` VARCHAR(128) NULL DEFAULT NULL,
    `equivalence` VARCHAR(128) NULL DEFAULT NULL,
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    PRIMARY KEY (`id`),
    UNIQUE INDEX `code_UNIQUE` (`code`),
    INDEX `group` (`group`),
    INDEX `equivalence` (`equivalence`)
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=46
;
CREATE TABLE `quality` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `code` VARCHAR(128) NOT NULL,
    `group` VARCHAR(128) NULL DEFAULT NULL,
    `equivalence` VARCHAR(128) NULL DEFAULT NULL,
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    PRIMARY KEY (`id`),
    UNIQUE INDEX `code_UNIQUE` (`code`),
    INDEX `group` (`group`),
    INDEX `equivalence` (`equivalence`)
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=980
;
CREATE TABLE `user_filter` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `user_id` INT(11) NOT NULL,
    `filter_type` ENUM('consumption','product') NOT NULL DEFAULT 'consumption',
    `name` VARCHAR(255) NOT NULL,
    `is_default` TINYINT(1) NOT NULL DEFAULT '0',
    `client_status` TINYINT(1) NULL DEFAULT NULL,
    `client_group` VARCHAR(45) NULL DEFAULT NULL,
    `material` VARCHAR(15) NULL DEFAULT NULL,
    `quality` VARCHAR(64) NULL DEFAULT NULL,
    `thickness` VARCHAR(45) NULL DEFAULT NULL,
    `width` VARCHAR(45) NULL DEFAULT NULL,
    `tons` VARCHAR(45) NULL DEFAULT NULL,
    `covering` VARCHAR(45) NULL DEFAULT NULL,
    `finish` VARCHAR(45) NULL DEFAULT NULL,
    `re` VARCHAR(45) NULL DEFAULT NULL,
    `rm` VARCHAR(45) NULL DEFAULT NULL,
    `a_percent` VARCHAR(45) NULL DEFAULT NULL,
    `comments` VARCHAR(255) NULL DEFAULT NULL,
    `price` VARCHAR(45) NULL DEFAULT NULL,
    `warehouse` VARCHAR(45) NULL DEFAULT NULL,
    `date` VARCHAR(45) NULL DEFAULT NULL,
    `type` ENUM('consumption','request') NULL DEFAULT NULL,
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    PRIMARY KEY (`id`),
    INDEX `fk_user_filter_user1` (`user_id`),
    INDEX `filter_type` (`filter_type`),
    CONSTRAINT `fk_user_filter_user1` FOREIGN KEY (`user_id`) REFERENCES `user` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=5
;
CREATE TABLE `warehouse` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `name` VARCHAR(128) NOT NULL,
    `zone_id` INT(11) NULL DEFAULT NULL,
    `zone2_id` INT(11) NULL DEFAULT NULL,
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    PRIMARY KEY (`id`),
    INDEX `zone_id` (`zone_id`),
    INDEX `zone2_id` (`zone2_id`),
    CONSTRAINT `warehouse_ibfk_1` FOREIGN KEY (`zone_id`) REFERENCES `zone` (`id`) ON UPDATE NO ACTION ON DELETE NO ACTION
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=37
;
CREATE TABLE `zone` (
    `id` INT(11) NOT NULL AUTO_INCREMENT,
    `zone2_id` INT(11) NULL DEFAULT NULL,
    `name` VARCHAR(128) NOT NULL,
    `date_add` DATETIME NOT NULL,
    `date_upd` DATETIME NOT NULL,
    PRIMARY KEY (`id`),
    INDEX `zone2_id` (`zone2_id`)
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=49
;

为了能够在两个表之间找到匹配项，我做了什么：

我在 consumption 和 product 表之间创建了一个 LEFT JOIN 查询（如果需要，它也可以与其他表连接）。

看起来像这样：

SELECT cons.`id` as `consumption_id`, cons.`client_id` as `consumption_client_id`, cons.`material_id` as `consumption_material_id`, cons.`quality_id` as `consumption_quality_id`, cons.`thick` as `consumption_thick`, cons.`thick_max` as `consumption_thick_max`, cons.`width` as `consumption_width`, cons.`width_max` as `consumption_width_max`, cons.`long` as `consumption_long`, cons.`long_max` as `consumption_long_max`, cons.`type` as `consumption_type`, cons.`date_add` as `consumption_date_add`, prod.`id` as `product_id`, prod.`warehouse_id` as `product_warehouse_id`, prod.`code` as `product_code`, prod.`material_id` as `product_material_id`, prod.`quality_id` as `product_quality_id`, prod.`covering_id` as `product_covering_id`, prod.`finish_id` as `product_finish_id`, prod.`thickness` as `product_thickness`, prod.`width` as `product_width`, prod.`tons` as `product_tons` 
      FROM consumption cons 
        INNER JOIN client cli
          ON cli.id=cons.client_id
        LEFT JOIN client_has_group cli_gr
          ON cli_gr.client_id=cons.client_id
        LEFT JOIN product prod
          ON 

          (
            (cons.material_id=prod.material_id)
              OR 
            prod.material_id IN (
              SELECT id FROM material WHERE `equivalence`=(
                  SELECT `equivalence` FROM material WHERE id=cons.material_id
                )
                AND `group`=(
                  SELECT `group` FROM material WHERE id=cons.material_id
                )
            )
          )


 AND 

          (
            (cons.quality_id=prod.quality_id)
              OR 
            prod.quality_id IN (
              SELECT id FROM quality WHERE `equivalence`=(
                  SELECT `equivalence` FROM quality WHERE id=cons.quality_id
                )
                AND `group`=(
                  SELECT `group` FROM quality WHERE id=cons.quality_id
                )
            )
          )


 AND (prod.thickness  >= (cons.thick - 0.1) AND prod.thickness <= (cons.thick_max + 0.1))
 AND (prod.width  >= (cons.width - 1000) AND prod.width <= (cons.width_max + 1000)) 

             WHERE 1 > 0 AND prod.deleted=0 AND cli.status=1 AND cons.date_add >= '2017-10-08 00:00:00' 
        GROUP BY cons.id, prod.id

当我想列出产品并显示每个产品的消费匹配时，我有一个简单列出产品的主查询，然后我将该查询与上面的前一个查询连接起来，并按产品 ID 计算匹配分组。

SELECT t.*,
       count(f.consumption_id) AS matchesCount
FROM `product` t
LEFT JOIN (...previous query here...) f ON f.product_id=t.id
GROUP BY t.id

其他注意事项/注意事项：

应用程序使用两个表中具有相同名称的几个字段，以便在JOIN 中使用ON 查找匹配项
该应用程序还使用更复杂的业务逻辑，例如，产品材料可以相等，也可以位于等价表或组内
用户可以保存个人过滤器，这就是为什么使用user_filter 表的原因，因此作为用户，我可以保存多个“搜索”并快速从一个切换到另一个
匹配必须实时显示，我的意思是，动态计算，而不是通过任何 cronjob，因为用户过滤器总是会改变
应用程序现在将处理的数据量将是消耗表中的大约 35k 行和产品表中的大约 1.5k 行
托管应用程序的服务器是运行 MySQL 的专用服务器（64GB RAM）

我有 3k 行消费和 100 个产品的良好性能，现在有 10k+ 消费和 600 个产品，开始从 nginx 获取网关超时。猜测查询耗时过长。

我已经知道如果ON原因有很多条件它会更快，因为结果集更小，但如果条件很宽，它会超时，我猜结果行也是许多。也许连接会产生数百万行。

我想问的是：

为了在两个表之间进行数据“实时匹配”，我是否走在正确的道路上？使用 JOIN 是一个很好的解决方案吗？我想不出其他方法。
除了尝试优化查询和索引之外，我是否可以对服务器进行任何调整以充分利用服务器硬件？
在另一个项目中做过类似事情的人有什么其他技巧或技巧吗？

更新 1：在此处添加完整查询以列出具有消费匹配的产品：

SELECT t.*,
       count(f.consumption_id) AS matchesCount
FROM `product` t
LEFT JOIN (


SELECT cons.`id` as `consumption_id`, cons.`client_id` as `consumption_client_id`, cons.`material_id` as `consumption_material_id`, cons.`quality_id` as `consumption_quality_id`, cons.`thick` as `consumption_thick`, cons.`thick_max` as `consumption_thick_max`, cons.`width` as `consumption_width`, cons.`width_max` as `consumption_width_max`, cons.`long` as `consumption_long`, cons.`long_max` as `consumption_long_max`, cons.`type` as `consumption_type`, cons.`date_add` as `consumption_date_add`, prod.`id` as `product_id`, prod.`warehouse_id` as `product_warehouse_id`, prod.`code` as `product_code`, prod.`material_id` as `product_material_id`, prod.`quality_id` as `product_quality_id`, prod.`covering_id` as `product_covering_id`, prod.`finish_id` as `product_finish_id`, prod.`thickness` as `product_thickness`, prod.`width` as `product_width`, prod.`tons` as `product_tons` 
      FROM consumption cons 
        INNER JOIN client cli
          ON cli.id=cons.client_id
        LEFT JOIN client_has_group cli_gr
          ON cli_gr.client_id=cons.client_id
        LEFT JOIN product prod
          ON 

          (
            (cons.material_id=prod.material_id)
              OR 
            prod.material_id IN (
              SELECT id FROM material WHERE `equivalence`=(
                  SELECT `equivalence` FROM material WHERE id=cons.material_id
                )
                AND `group`=(
                  SELECT `group` FROM material WHERE id=cons.material_id
                )
            )
          )



             WHERE 1 > 0 AND prod.deleted=0 AND cli.status=1 AND cons.date_add >= '2017-10-08 00:00:00' 
        GROUP BY cons.id, prod.id


) f ON f.product_id=t.id
GROUP BY t.id

查询时间：00:02:41（+ 0,078 秒网络）。

注意：子查询 JOIN 单独运行会产生 600k 行。我正在考虑尝试以某种方式对其进行分组以使其更小。

更新 2：通过在子查询中进行计数来实现重大改进，从而减少用于 JOIN 的结果集

基本上是子查询，而不是返回 600k+ 行，它只返回与产品或消费一样多的行，具体取决于您要查找的内容。为此，matchesCount 已移至子查询内部而不是外部，并且 group by 已更改，具体取决于您要显示的列表。

这是最终查询现在的样子：

列出消费并统计每次消费匹配的产品：

SELECT SQL_NO_CACHE `t`.*, 
          IFNULL(f.matchesCount, 0) AS matchesCount
   FROM `consumption` `t`
   LEFT JOIN
     (SELECT cons.`id` AS `consumption_id`,
             cons.`client_id` AS `consumption_client_id`,
             cons.`material_id` AS `consumption_material_id`,
             cons.`quality_id` AS `consumption_quality_id`,
             cons.`thick` AS `consumption_thick`,
             cons.`thick_max` AS `consumption_thick_max`,
             cons.`width` AS `consumption_width`,
             cons.`width_max` AS `consumption_width_max`,
             cons.`long` AS `consumption_long`,
             cons.`long_max` AS `consumption_long_max`,
             cons.`type` AS `consumption_type`,
             cons.`date_add` AS `consumption_date_add`,
             prod.`id` AS `product_id`,
             prod.`warehouse_id` AS `product_warehouse_id`,
             prod.`code` AS `product_code`,
             prod.`material_id` AS `product_material_id`,
             prod.`quality_id` AS `product_quality_id`,
             prod.`covering_id` AS `product_covering_id`,
             prod.`finish_id` AS `product_finish_id`,
             prod.`thickness` AS `product_thickness`,
             prod.`width` AS `product_width`,
             prod.`tons` AS `product_tons`,
             count(prod.`id`) AS matchesCount
      FROM consumption cons
      INNER JOIN client cli ON cli.id=cons.client_id
      LEFT JOIN product prod ON ((cons.material_id=prod.material_id)
                                 OR prod.material_id IN
                                   (SELECT id
                                    FROM material
                                    WHERE `equivalence`=
                                        (SELECT `equivalence`
                                         FROM material
                                         WHERE id=cons.material_id )
                                      AND `group`=
                                        (SELECT `group`
                                         FROM material
                                         WHERE id=cons.material_id ) ))
      AND ((cons.quality_id=prod.quality_id)
           OR prod.quality_id IN
             (SELECT id
              FROM quality
              WHERE `equivalence`=
                  (SELECT `equivalence`
                   FROM quality
                   WHERE id=cons.quality_id )
                AND `group`=
                  (SELECT `group`
                   FROM quality
                   WHERE id=cons.quality_id ) ))
      AND (prod.thickness >= (cons.thick - 0.1)
           AND prod.thickness <= (cons.thick_max + 0.1))
      AND (prod.width >= (cons.width - 1000)
           AND prod.width <= (cons.width_max + 1000))
      WHERE 1 > 0
        AND prod.deleted=0
        AND cli.status=1
        AND cons.date_add >= '2017-10-08 00:00:00'
      GROUP BY cons.id) f ON f.consumption_id=t.id

   GROUP BY t.id

列出产品并计算与每个产品匹配的消费：

SELECT SQL_NO_CACHE t.*,
          IFNULL(f.matchesCount, 0) AS matchesCount
   FROM `product` `t`
   LEFT JOIN
     (SELECT cons.`id` AS `consumption_id`,
             cons.`client_id` AS `consumption_client_id`,
             cons.`material_id` AS `consumption_material_id`,
             cons.`quality_id` AS `consumption_quality_id`,
             cons.`thick` AS `consumption_thick`,
             cons.`thick_max` AS `consumption_thick_max`,
             cons.`width` AS `consumption_width`,
             cons.`width_max` AS `consumption_width_max`,
             cons.`long` AS `consumption_long`,
             cons.`long_max` AS `consumption_long_max`,
             cons.`type` AS `consumption_type`,
             cons.`date_add` AS `consumption_date_add`,
             prod.`id` AS `product_id`,
             prod.`warehouse_id` AS `product_warehouse_id`,
             prod.`code` AS `product_code`,
             prod.`material_id` AS `product_material_id`,
             prod.`quality_id` AS `product_quality_id`,
             prod.`covering_id` AS `product_covering_id`,
             prod.`finish_id` AS `product_finish_id`,
             prod.`thickness` AS `product_thickness`,
             prod.`width` AS `product_width`,
             prod.`tons` AS `product_tons`,
             count(cons.`id`) AS matchesCount
      FROM consumption cons
      INNER JOIN client cli ON cli.id=cons.client_id
      LEFT JOIN product prod ON cons.material_id=prod.material_id
      AND cons.quality_id=prod.quality_id
      WHERE 1 > 0
        AND prod.deleted=0
        AND cli.status=1
      GROUP BY prod.id) f ON f.product_id=t.id
   WHERE deleted=0
   GROUP BY t.id

两个查询的执行时间都不到 1 秒（每个）。

注意：我仍然在我的应用程序中使用之前的查询，例如，当我想要细分匹配单次消费的产品列表时，或者相反。在这种情况下，我已经为每个消费 id 或产品 id 添加了一个过滤器，这大大减少了结果集的大小。

【问题讨论】：

您消费了您的产品表中不存在的产品？我会先修复数据。

标签： mysql performance join match

【解决方案1】：

为什么需要左加入 client_has_group cli_gr ON cli_gr.client_id=cons.client_id 它从未使用过
为什么需要 GROUP BY cons.id, prod.id 如果您选择所有字段可能只选择您需要的内容

试试这个选择，我觉得会更快

SELECT count(*), prod.*    
FROM consumption cons 
INNER JOIN client cli ON cli.id=cons.client_id   
INNER JOIN   material m ON  m.id=cons.material_id    
INNER JOIN   quality q ON  q.id=cons.quality_id
    LEFT JOIN product prod
    ON    
      (
        (cons.material_id=prod.material_id)
          OR 
        prod.material_id IN (
          SELECT id FROM material WHERE `equivalence`=m.equivalence
            AND `group`=m.group
        )
      )
AND 
      (
        (cons.quality_id=prod.quality_id)
          OR 
        prod.quality_id IN (
          SELECT id FROM quality WHERE `equivalence`=q.equivalence
            AND `group`=q.group
        )
      )
 AND (prod.thickness  >= (cons.thick - 0.1) AND prod.thickness <= (cons.thick_max + 0.1))
 AND (prod.width  >= (cons.width - 1000) AND prod.width <= (cons.width_max + 1000)) 

 WHERE 1 > 0 AND prod.deleted=0 AND cli.status=1 AND cons.date_add >= '2017-10-08 00:00:00'
group by  prod.id

最好在后台做计算，并在产品和消费表中添加这个字段。

【讨论】：

【解决方案2】：

如果client_has_group 是“many:1”，那是错误的做法。你不需要额外的桌子。

INT 始终为 4 个字节。考虑较小的数据类型。最终，数据库的大小可能会增加您的问题。

您真的需要date_add 和date_upd。它们看起来像你永远不会使用的杂物。

在可行的情况下避免使用IN ( SELECT ... )。切换到JOIN 或EXISTS。

为什么有这么多带有代码 + 组 + 等价的表？他们可以是一个单一的群体吗？您需要所有 3 列吗？你需要id，因为code 是UNIQUE？有时模式会“过度规范化”，并且性能会受到影响，而不会占用太多空间。

OR 在某些情况下是性能杀手。

“相关子查询”在某些情况下很有用，但最好通过JOIN 来完成这个：

AND `group` = ( SELECT `group` FROM quality WHERE id=cons.quality_id )

注意与JOIN 的聚合（例如COUNT）；你可能会得到一个夸大的价值。这是因为JOIN 先发生。

【讨论】：