Question

所以我有一个相对简单的查询，它根据WHERE为一系列输入选择存储函数的结果。完整代码如下。

让我感到困惑的是，当我通过mysql cli调用查询时，需要662秒才能返回单列的841行。现在，如果我将完全相同的代码复制/粘贴到phpmyadmin中，相同的值将在28秒内返回。

我需要做什么才能让mysql像phpmyadmin一样快地执行此查询？

我做了什么：

已验证phpmyadmin和php cli都使用相同的EXPLAIN计划。
- EXPLAIN计划现在包含在下面。
已验证phpmyadmin和php cli正在以同一用户身份执行查询。
测试了mysql cli是否比php cli具有更好的性能。
- OUTCOME： mysql cli在661秒内返回值，这可能与php cli的速度相同。我刚刚意识到SHOW PROCESSLIST在进程进入睡眠状态后不断计算时间。 上述问题标题和说明已经修订。

当前探索：

编辑存储的函数，以便使用WHERE IN而不是WHERE EXISTS。
- 建议http://dev.mysql.com/doc/refman/5.5/en/subquery-optimization-with-exists.html

完整的查询（包括无问题的字段）

SELECT 
    EC_ITM_PULL.SKU_NUM, 
    EC_ITM_PULL.COLOR_CD, 
    GM_SKU.COLOR_DES, 
    GM_ITM.DES1, 
    GM_ITM.DES2, 
    CUSTOM.EC_GET_ONHAND(SUBSTRING(GM_SKU.SKU_NUM, 1, 9), GM_SKU.COLOR_CD) OH, 
    UPPER(EC_ITM_PULL.ITEM_PULLED_INIT) INIT, 
    CUSTOM.EC_MOST_RECENT_EVENT(SUBSTRING(GM_SKU.SKU_NUM, 1, 9), GM_SKU.COLOR_CD, 'E') EVENT_CD 
FROM 
    EC_ITM_PULL, 
    GM_MERCH.GM_SKU, 
    GM_INV.GM_ITM, 
    EC_ITM 
WHERE 
        EC_ITM.SKU_NUM = EC_ITM_PULL.SKU_NUM 
    AND EC_ITM_PULL.COLOR_CD = EC_ITM.COLOR_CD 
    AND EC_ITM_PULL.ITEM_PULLED IS NOT NULL 
    AND GM_ITM.ITM_CD = SUBSTRING(EC_ITM.SKU_NUM, 1, 9) 
    AND EC_ITM.SKU_NUM = GM_SKU.SKU_NUM 
    AND EC_ITM.TO_STUDIO IS NULL 
    AND GM_ITM.ITM_CD = SUBSTRING(EC_ITM_PULL.SKU_NUM, 1, 9);

QUERY EXPLAIN PLAN：

+----+-------------+-------------+--------+---------------+---------+---------+-----------------------------------------------------+------+-------------+
| id | select_type | table       | type   | possible_keys | key     | key_len | ref                                                 | rows | Extra       |
+----+-------------+-------------+--------+---------------+---------+---------+-----------------------------------------------------+------+-------------+
|  1 | SIMPLE      | EC_ITM_PULL | ALL    | SKU_NUM       | NULL    | NULL    | NULL                                                | 2100 | Using where |
|  1 | SIMPLE      | GM_ITM      | eq_ref | PRIMARY       | PRIMARY | 38      | func                                                |    1 | Using where |
|  1 | SIMPLE      | GM_SKU      | ref    | SKU_NUM       | SKU_NUM | 38      | CUSTOM.EC_ITM_PULL.SKU_NUM                          |    1 |             |
|  1 | SIMPLE      | EC_ITM      | eq_ref | PRIMARY       | PRIMARY | 58      | GM_MERCH.GM_SKU.SKU_NUM,CUSTOM.EC_ITM_PULL.COLOR_CD |    1 | Using where |
+----+-------------+-------------+--------+---------------+---------+---------+-----------------------------------------------------+------+-------------+

问题存储功能

CREATE DEFINER=`root`@`localhost` FUNCTION `EC_GET_ONHAND`(`N_ITM_CD` VARCHAR(12), `N_COLOR_CD` VARCHAR(6)) RETURNS smallint(5)
BEGIN
    DECLARE TOTALOH SMALLINT(5);

    IF N_COLOR_CD IS NULL THEN SELECT IFNULL(SUM(IFNULL(RESERVE_QTY,0)+IFNULL(AVAIL_QTY,0)),0)
    INTO TOTALOH FROM GM_INV.GM_INV_LOC
    WHERE SKU_NUM 
    IN (
        SELECT SKU_NUM 
        FROM GM_MERCH.GM_SKU 
        WHERE ITM_CD = N_ITM_CD AND COLOR_CD IS NULL
    ) 
    AND (
        (
            (
                STORE_CD='85' 
                OR STORE_CD='95'
            ) 
            AND LOC_CD='STG72'
        ) 
        OR (
           (
               STORE_CD='72'
           ) 
           AND LOC_CD='RCV'
        )
    );

    ELSE SELECT IFNULL(SUM(IFNULL(RESERVE_QTY,0)+IFNULL(AVAIL_QTY,0)),0)
    INTO TOTALOH 
    FROM GM_INV.GM_INV_LOC
    WHERE SKU_NUM
    IN (
        SELECT SKU_NUM 
        FROM GM_MERCH.GM_SKU 
        WHERE ITM_CD = N_ITM_CD 
        AND COLOR_CD = N_COLOR_CD)
        AND (
            (
                (
                    STORE_CD='85' 
                    OR STORE_CD='95'
                ) 
                AND LOC_CD='STG72'
            ) 
            OR (
                (
                    STORE_CD='72'
                ) 
                AND LOC_CD='RCV'
            )
        );

    END IF;

    RETURN TOTALOH;
END

功能解释计划：

+----+--------------------+------------+----------------+---------------+---------------+---------+------------------+--------+--------------------------+
| id | select_type        | table      | type           | possible_keys | key           | key_len | ref              | rows   | Extra                    |
+----+--------------------+------------+----------------+---------------+---------------+---------+------------------+--------+--------------------------+
|  1 | PRIMARY            | GM_INV_LOC | ALL            | NULL          | NULL          | NULL    | NULL             | 509791 | Using where              |
|  2 | DEPENDENT SUBQUERY | GM_SKU     | index_subquery | SKU_ITM_COLOR | SKU_ITM_COLOR | 97      | func,const,const |      1 | Using index; Using where |
+----+--------------------+------------+----------------+---------------+---------------+---------+------------------+--------+--------------------------+

Answer 1

我通过重写我的存储函数解决了这个问题。这实际上将查询时间一直缩短到1.19秒，所以我实际上比phpmyadmin还要快得多=）

首先，我接受了问题中链接文章的建议并编辑了这一部分：

WHERE SKU_NUM 
    IN (
        SELECT SKU_NUM 
        FROM GM_MERCH.GM_SKU 
        WHERE ITM_CD = N_ITM_CD AND COLOR_CD IS NULL
    )

到此：

WHERE EXISTS 
    (
        SELECT 1 
        FROM GM_MERCH.GM_SKU, GM_INV.GM_INV_LOC
        WHERE 
            ITM_CD = N_ITM_CD 
        AND COLOR_CD IS NULL
        AND GM_INV_LOC.SKU_NUM = GM_MERCH.SKU_NUM
    )

这将查询的执行时间减少到不到180秒，但导致函数返回不正确的值，因为它正在为子查询中的每一行重复添加。

那么我觉得也许我根本不应该使用子查询。我重写了函数中的查找，如下所示：

SELECT IFNULL(SUM(IFNULL(RESERVE_QTY,0)+IFNULL(AVAIL_QTY,0)),0)
    INTO TOTALOH FROM GM_INV.GM_INV_LOC, GM_MERCH.SKU_NUM
    WHERE 
        ITM_CD = N_ITM_CD 
    AND COLOR_CD IS NULL
    AND GM_INV_LOC.SKU_NUM = GM_SKU.SKU_NUM
    AND (
        (
            (
                STORE_CD='85' 
                OR STORE_CD='95'
            ) 
            AND LOC_CD='STG72'
        ) 
        OR (
           (
               STORE_CD='72'
           ) 
           AND LOC_CD='RCV'
        )
    );

整个大查询现在在1.19秒内返回其841行，并具有准确的值。

Answer 2

我很高兴你找到了解决方案。但是phpmyadmin，另一个php应用程序或者cli之间有不同的时间通常意味着设置在各处都不一样。根据您的时间差异，但不知道所处理数据的大小，可能是因为您的CLI缺少内存并且需要SWAP ...要验证您的不同＆＃34; mysql访问环境＆＃34;我建议你运行这些查询并比较结果：

show variables like max_heap_table_size #chances are you will find diff here
show variables #shows all variables

希望它可以帮助那些偶然发现这种人物的人。

查询在phpmyadmin中需要28秒，但mysql命令行需要662秒

2 个答案: