在优化查询方面需要帮助

Question

其实我是用下面的查询来获取结果的，但是我觉得查询有点长，我可以帮助我优化下面的查询吗？

SELECT
    o.orderid AS id,
    sd.datestart AS startdate,
    sd.dateend AS enddate,
    o.comment AS comment,
    o.productname AS name,
    o.suspended AS sus,
    o.voided AS void,
    o.invoice_number AS inumber,
    GROUP_CONCAT(js.title) AS sections,
    GROUP_CONCAT(js.sectionid) AS sectionid,
    js.is_science_education AS isse,
    o.amount AS amount,
    js.isbundle,
    o.addedbyuserid,
    (SELECT
            COUNT(*)
        FROM
            institutiontoorder AS ito2
        WHERE
            orderid = ito.orderid) AS multiple
FROM
    institutiontoorder AS ito
        LEFT JOIN
    subscribedates AS sd ON ito.orderid = sd.orderid
        LEFT JOIN
    ordertosection AS ots ON ots.orderid = ito.orderid
        LEFT JOIN
    orders AS o ON o.orderid = ito.orderid
        LEFT JOIN
    institutions AS i ON i.institutionid = ito.institutionid
        JOIN
    journalsections AS js ON js.sectionid = ots.sectionid
WHERE
    ito.institutionid = '1419'
        AND o.voided != 1
        AND o.suspended != 1
        AND (dateend >= NOW())
GROUP BY o.orderid
ORDER BY o.orderid DESC;

我在使用多个连接，是否可以优化依赖子查询？

Answer 1

o:    (orderid, dateend)
ito:  (institutionid, orderid)
ots:  (orderid, sectionid)
js:   (sectionid, isbundle, is_science_education, title)
i:    (institutionid)
sd:   (orderid, dateend, datestart)

列的顺序通常很重要。如果这些列中的任何一个太大（例如，如果title 是TEXT），则将其关闭。否则，它们中的大多数都是“覆盖”索引，从而提供了一点额外的性能提升。

ito 列顺序对于处理 DRapp 的 ito2 很重要

Answer 2

对查询的细微调整，例如对其自己的预查询/聚合的 select count()，但仅适用于所讨论的一个机构。当您将某些 LEFT JOIN 添加到 WHERE 子句时，它们实际上变成了 INNER JOINS。

我实际上将 AND 部分移动到了所述表的相应 JOIN 位置。

我会确保以下索引

table               index
InstitutionToOrder  (institutionid, orderid)
orders              (orderid, voided, suspended)
subscribeDates      (orderid, dateEnd )
OrderToSection      (orderid, sectionid )

此外，您不需要到处都是“AS”，例如列 AS resultColumn 或表 A​​S 别名。隐含“AS”，分别在项目后面加一个空格。

SELECT
        o.orderid id,
        sd.datestart startdate,
        sd.dateend enddate,
        o.comment,
        o.productname name,
        o.suspended sus,
        o.voided void,
        o.invoice_number inumber,
        GROUP_CONCAT(js.title) sections,
        GROUP_CONCAT(js.sectionid) sectionid,
        js.is_science_education isse,
        o.amount,
        js.isbundle,
        o.addedbyuserid,
        preCount.multiple
    FROM
        institutiontoorder ito
            JOIN orders o 
                ON ito.orderid = o.orderid
                AND o.voided = 0
                AND o.suspended = 0
            JOIN subscribedates sd 
                ON ito.orderid = sd.orderid
                AND sd.dateend >= NOW()
            JOIN
                -- pre-aggregate counts once for just the 
                -- institution in question and join to it
                ( SELECT
                        ito2.orderid,
                        COUNT(*) muliple
                    FROM
                        institutiontoorder ito2
                    WHERE
                        ito2.institutionid = '1419'
                    group by
                        ito2.orderid ) AS PreCount
                on ito.orderid = PreCount.orderid
            LEFT JOIN ordertosection ots 
                ON ito.orderid = ots.orderid
                JOIN journalsections js 
                    ON ots.sectionid = js.sectionid
            LEFT JOIN institutions i 
                ON ito.institutionid = i.institutionid
    WHERE
        ito.institutionid = '1419'
    GROUP BY
        -- via Transitive Property, since o.orderid = ito.orderid,
        -- use the primary table in the query's orderID for group/order by
        ito.orderid
    ORDER BY 
        ito.orderid DESC;

根据 O. Jones 的 cmets，我完全同意对您订单的 VOIDED 和 SUSPENDED 过滤器进行不等式测试。这些值的选择是什么……只有 0 或 1，其中 1 表示 VOIDED 或 SUSPENDED？还是他们有其他价值。我已调整为相等 = 0（表示未作废且未暂停）