如何使此MySQL查询（带有子查询）更快？

Question

我有以下查询，它获取订单数据， 可以正常工作，但需要13.63秒 ：

SELECT
( SELECT meta_value FROM `myprefix_postmeta` as postmeta WHERE postmeta.meta_key = '_sku' AND postmeta.post_id = woocommerce_order_itemmeta3.meta_value ) AS sku,
woocommerce_order_items.order_item_name AS title,
SUM( woocommerce_order_itemmeta.meta_value ) AS quantity,
SUM( woocommerce_order_itemmeta2.meta_value ) AS total
FROM `myprefix_woocommerce_order_items` AS woocommerce_order_items
LEFT JOIN `myprefix_posts` AS posts ON posts.ID = woocommerce_order_items.order_id
LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS woocommerce_order_itemmeta ON woocommerce_order_itemmeta.order_item_id = woocommerce_order_items.order_item_id
LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS woocommerce_order_itemmeta2 ON woocommerce_order_itemmeta2.order_item_id =   woocommerce_order_items.order_item_id
LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS woocommerce_order_itemmeta3 ON woocommerce_order_itemmeta3.order_item_id =   woocommerce_order_items.order_item_id
WHERE posts.post_type = 'shop_order'
AND posts.post_status IN( 'wc-completed', 'wc-processing', 'wc-on-hold' )
AND woocommerce_order_itemmeta.meta_key = '_qty'
AND woocommerce_order_itemmeta2.meta_key = '_line_total' 
AND woocommerce_order_itemmeta3.meta_key = '_product_id'
GROUP BY woocommerce_order_itemmeta3.meta_value
ORDER BY total DESC

我想优化此查询以使其运行更快 。

也许只是花了很长时间才能返回。 但是，无论如何我都尝试了删除上面的子查询，结果如下：

SELECT
postmeta.meta_value AS sku,
woocommerce_order_items.order_item_name AS title,
SUM( woocommerce_order_itemmeta.meta_value ) AS quantity,
SUM( woocommerce_order_itemmeta2.meta_value ) AS total
FROM `myprefix_woocommerce_order_items` AS woocommerce_order_items
LEFT JOIN `myprefix_posts` AS posts ON posts.ID = woocommerce_order_items.order_id
LEFT JOIN `myprefix_postmeta` AS postmeta ON posts.ID = woocommerce_order_items.order_id
LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS woocommerce_order_itemmeta ON woocommerce_order_itemmeta.order_item_id = woocommerce_order_items.order_item_id
LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS woocommerce_order_itemmeta2 ON woocommerce_order_itemmeta2.order_item_id =   woocommerce_order_items.order_item_id
LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS woocommerce_order_itemmeta3 ON woocommerce_order_itemmeta3.order_item_id =   woocommerce_order_items.order_item_id
WHERE posts.post_type = 'shop_order'
AND posts.post_status IN( 'wc-completed', 'wc-processing', 'wc-on-hold' )
AND postmeta.meta_key = '_sku'
AND postmeta.post_id = woocommerce_order_itemmeta3.meta_value
AND woocommerce_order_itemmeta.meta_key = '_qty'
AND woocommerce_order_itemmeta2.meta_key = '_line_total' 
AND woocommerce_order_itemmeta3.meta_key = '_product_id'
GROUP BY woocommerce_order_itemmeta3.meta_value
ORDER BY total DESC

这几毫秒快一点 ，但是有什么方法可以进一步优化吗？

非常感谢您可能需要了解这些表背后的数据，它们是标准的WordPress数据库表和WooCommerce数据库架构 ，我无法真正共享这些表背后的数据。

上面的第二个示例存在一个附带问题，即如果SKU为NULL，则不包含该行，而最上面的一个包含该行，因此当前仍然是首选查询。

更新＃1

根据答案： https : //stackoverflow.com/a/54202562/8369600

我最初从答案的第1部分中更新了我的代码格式，并且效果很好。

然后，我在步骤2中使用了代码，并对格式进行了一些小的更改，以便在phpMyAdmin中为我工作，这些更改是：

SELECT
    (SELECT meta_value
        FROM `myprefix_postmeta` as postmeta 
        WHERE 
            postmeta.meta_key = '_sku' 
            AND postmeta.post_id = ( -- Substitute meta_value with CASE-expressions 
                    CASE
                        WHEN IM.meta_key = '_product_id' THEN IM.meta_value
                        ELSE NULL
                    END 
                ) -- IM3.meta_value               
            ) AS sku,
            -- think above is okay ^^^^
    I.order_item_name AS title,
    SUM(
        CASE -- Substitute meta_value with CASE-expressions 
            WHEN IM.meta_key = '_qty' THEN IM.meta_value
            ELSE 0 
        END 
    ) AS quantity,
    -- IM2.meta_value ) AS total
SUM( -- Substitute meta_value with CASE-expressions 
    CASE
            WHEN IM.meta_key = '_line_total' THEN IM.meta_value
            ELSE 0
    END 
) AS total       

FROM 
    `myprefix_woocommerce_order_items` AS I
    LEFT JOIN `myprefix_posts` AS posts 
    ON posts.ID = I.order_id
    LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS IM -- Make readable aliases
        ON IM.order_item_id = I.order_item_id 
            AND IM.meta_key IN('_qty', '_line_total', '_product_id')
    --         AND IM.meta_key = '_qty'   -- Move predicates from WHERE
    -- LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS IM2 
    --     ON IM2.order_item_id =   I.order_item_id
    --         AND IM2.meta_key = '_line_total' -- Move predicates from WHERE
    -- LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS IM3 
    --     ON IM3.order_item_id =   I.order_item_id
    --         AND IM3.meta_key = '_product_id' -- Move predicates from WHERE
WHERE 
    posts.post_type = 'shop_order'
    AND posts.post_status IN
    ( 'wc-completed', 'wc-processing', 'wc-on-hold' )
GROUP BY 
    (
        CASE
            WHEN IM.meta_key = '_product_id' THEN IM.meta_value
            ELSE NULL
        END 
    ) -- IM3.meta_value    
ORDER BY 
    total DESC

但是，这返回：

在所有行都填充有数据之前，请怀疑案件是否没有按照原始规定进行处理。

注意我还按照前面的注释之一的要求添加了指向WooCommerce数据库文档的链接。

Answer 1

步骤1：创建可读的别名并将meta_key -predicates移到JOIN

为了清楚起见，我采取了自由格式化您的查询。

我重命名了别名，并将meta_key -predicates从WHERE移到JOIN ：

SELECT
    (SELECT meta_value
        FROM `myprefix_postmeta` as postmeta 
        WHERE 
            postmeta.meta_key = '_sku' 
            AND postmeta.post_id = IM3.meta_value ) AS sku,
    I.order_item_name AS title,
    SUM (IM.meta_value ) AS quantity,
    SUM (IM2.meta_value ) AS total
FROM 
    `myprefix_woocommerce_order_items` AS I
    LEFT JOIN `myprefix_posts` AS posts 
    ON posts.ID = I.order_id
    LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS IM -- Make readable aliases
        ON IM.order_item_id = I.order_item_id 
            AND IM.meta_key = '_qty'   -- Move predicates from WHERE
    LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS IM2 
        ON IM2.order_item_id =   I.order_item_id
            AND IM2.meta_key = '_line_total' -- Move predicates from WHERE
    LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS IM3 
        ON IM3.order_item_id =   I.order_item_id
            AND IM3.meta_key = '_product_id' -- Move predicates from WHERE
WHERE 
    posts.post_type = 'shop_order'
    AND posts.post_status IN
    ( 'wc-completed', 'wc-processing', 'wc-on-hold' )
GROUP BY 
    IM3.meta_value
ORDER BY 
    total DESC

现在我们有了绝对等效的查询。

让我们看看我们可以做的进一步。

1. 使用相同谓词IM.order_item_id = I.order_item_id向myprefix_woocommerce_order_itemmeta 3个LEFT JOIN看起来不必要的重复
2. 但是我们在meta_key上有不同的谓词

步骤2：仅用1个替换3个JOIN，用CASE表达式替换JOINed字段

所以我们可以：

一种。 用一个JOIN替换3个JOIN

湾 用CASE表达式替换IM / IM2 / IM3字段

SELECT
    (SELECT meta_value
        FROM `myprefix_postmeta` as postmeta 
        WHERE 
            postmeta.meta_key = '_sku' 
            AND postmeta.post_id = ( -- Substitute meta_value with CASE-expressions 
                    CASE
                        WHEN IM.meta_key = '_product_id' THEN IM.meta_value
                        ELSE NULL
                    END 
                ) -- IM3.meta_value               
            ) AS sku,
    I.order_item_name AS title,
    SUM (
        CASE -- Substitute meta_value with CASE-expressions 
            WHEN IM.meta_key = '_qty' THEN IM.meta_value
            ELSE 0 
        END 
    ) AS quantity
    -- IM2.meta_value ) AS total
SUM ( -- Substitute meta_value with CASE-expressions 
    CASE
            WHEN IM.meta_key = '_line_total' THEN IM.meta_value
            ELSE 0
    END 
) AS total       

FROM 
    `myprefix_woocommerce_order_items` AS I
    LEFT JOIN `myprefix_posts` AS posts 
    ON posts.ID = I.order_id
    LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS IM -- Make readable aliases
        ON IM.order_item_id = I.order_item_id 
            AND IM.meta_key IN ('_qty', '_line_total', '_product_id')
    --         AND IM.meta_key = '_qty'   -- Move predicates from WHERE
    -- LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS IM2 
    --     ON IM2.order_item_id =   I.order_item_id
    --         AND IM2.meta_key = '_line_total' -- Move predicates from WHERE
    -- LEFT JOIN `myprefix_woocommerce_order_itemmeta` AS IM3 
    --     ON IM3.order_item_id =   I.order_item_id
    --         AND IM3.meta_key = '_product_id' -- Move predicates from WHERE
WHERE 
    posts.post_type = 'shop_order'
    AND posts.post_status IN
    ( 'wc-completed', 'wc-processing', 'wc-on-hold' )
GROUP BY 
    (
        CASE
            WHEN IM.meta_key = '_product_id' THEN IM.meta_value
            ELSE NULL
        END 
    ) -- IM3.meta_value    
ORDER BY 
    total DESC

PS

编写没有数据的SQL并不是很舒服。

而且此查询可能不适用于每个版本的mysql。

因此，请尝试理解原理，而不只是复制粘贴我的代码。

通知我们有关结果或问题。

Answer 2

如果woocommerce是WordPress的分支，那么他们可能在索引myprefix_woocommerce_order_itemmeta 。 调整此处的提示以提高性能。

我同意别人的看法-EAV模式模式糟透了。 上述帮助我的一些指标的建议。

例如，“ order”始终具有数量，product_id，line_total； 那么，为什么不将它们作为真实的列而不是将它们隐藏在另一个表中，就好像它们是稀有属性一样呢？ 对于只有一些商品具有“衣着尺寸”或“ F-stop”或“传输类型”的购物网站，EAV有点必要。 （在这种情况下，JSON 可能比EAV更好。）

模式审查

不错，woocommerce比WP更糟糕（ 参考 ）：

CREATE TABLE {$wpdb->prefix}woocommerce_order_itemmeta (
  meta_id BIGINT UNSIGNED NOT NULL auto_increment,
  order_item_id BIGINT UNSIGNED NOT NULL,
  meta_key varchar(255) default NULL,
  meta_value longtext NULL,
  PRIMARY KEY  (meta_id),
  KEY order_item_id (order_item_id),
  KEY meta_key (meta_key(32))
) $collate;

前缀索引（ meta_key(32) ）很少有用，并且通常适得其反。 尽管如此，可能我所有的建议（请参见上面的链接）都适用，并且名称进行了适当的更改。

Answer 3

首先要做的是检查在LEFT JOIN和过滤位置的所有列上是否都有正确的索引。 通常，缺少索引是导致查询缓慢的原因。

如果您想走得更远，我将重建索引。 也许一些零碎的东西也放慢了速度。

尽管不是必需的，您也可以尝试将wc-completed，wc-processing之类的迁移键放入自己的表中，并通过INT进行联接。

您也可以尝试规范化SUM甚至为它们创建一个计算列。