使用自连接和子查询优化更新 SQL 查询
Optimizing an UPDATE SQL Query with self joins and subqueries
我的 SQL 查询更新了数据库中的所有股票，但它的工作效率不高，有时我会收到 504 超时错误。该代码工作正常。我怎样才能让它更好地工作。
P.S: 缺少prepared statements请忽略，稍后补上
关于 tables 的一些信息（Wordpress Woocommerce 插件默认 tables）：
wp_posts: 这 table 包括帖子。 （帖子可以是产品或产品变体。例如产品是蝴蝶 T 恤，产品变体是蝴蝶 T 恤红色大号）。
wp_postmeta：此 table 包含有关帖子的元信息。例如，如果产品变体有库存，或者它是什么颜色，或者它是什么尺寸。
//This array gives, which products are there, and their respective categories.
$allProducts = array("Fermuarlı Kapşonlu Sweatshirt" => "'2653','2659'","Kapşonlu Sweatshirt" => "'2646','2651'","Sweatshirt" => "'2644','2650'","Kadın Tişört" => "'2654','2656'","Atlet" => "'2655','2657'","Tişört" => "'2643','2304'");
//Below arrays gives information about, which product variations are out of stock.
$tisort_OutOfStock =array();
$atlet_OutOfStock =array("all_colors"=>"'3xl','4xl','5xl'");
$kadin_tisort_OutOfStock =array("all_colors"=>"'xxl','3xl','4xl','5xl'");
$sweatshirt_OutOfStock =array("beyaz"=>"'xxl','3xl','4xl','5xl'","kirmizi"=>"'xxl','3xl','4xl','5xl'","bordo"=>"'5xl'","antrasit"=>"'5xl'");
$kapsonlu_sweatshirt_OutOfStock =array("gri-kircilli"=>"'5xl'");
$fermuarli_kapsonlu_sweatshirt_OutOfStock =array("gri-kircilli"=>"'5xl'","siyah"=>"'5xl'");
//Reset stocks before updating.
$resetStocks = "UPDATE wp_postmeta set meta_value = 'instock' where meta_key = '_stock_status'";
$wpdb->query($resetStocks);
echo "Stoklar are reseted<br>";
//Foreach product, foreach color, update if product doesn't have stock.
foreach( $allProducts as $key => $urun ){
switch ($key) {
case "Kadın Tişört": $tempArray = $kadin_tisort_OutOfStock; break;
case "Fermuarlı Kapşonlu Sweatshirt": $tempArray = $fermuarli_kapsonlu_sweatshirt_OutOfStock; break;
case "Kapşonlu Sweatshirt": $tempArray = $kapsonlu_sweatshirt_OutOfStock; break;
case "Sweatshirt": $tempArray = $sweatshirt_OutOfStock; break;
case "Atlet": $tempArray = $atlet_OutOfStock; break;
case "Tişört": $tempArray = $tisort_OutOfStock; break;
}
foreach( $tempArray as $color => $size ){
$query = "UPDATE wp_postmeta set meta_value = 'outofstock' where meta_key = '_stock_status' and post_id in
(
select post_id from (select * from wp_postmeta) AS X where meta_key = 'attribute_pa_beden' and meta_value in (".$size.")
and post_id in (select post_id from (select * from wp_postmeta) AS Y where meta_key = 'attribute_pa_renk' and ((meta_value = '".$color."') OR ('".$color."' = 'all_colors')))
and post_id in (select id from wp_posts where post_type = 'product_variation' and post_parent in (select object_id FROM wp_term_relationships where term_taxonomy_id in (".$urun.")))
)";
global $wpdb;
$updatedRowCount = $wpdb->query($query);
}
}
SELECT post_id
from
(
SELECT *
from wp_postmeta
) AS X
where ...
-->
SELECT post_id
FROM wp_postmeta
WHERE ...
（子查询只会减慢速度。）
and post_id in (
SELECT post_id
而不是 IN ( SELECT ... )，使用 JOIN。
除了这些技巧之外，请参阅我的 tips 关于改进 postmeta 架构的内容。
首先处理密钥 SELECT:
SELECT post_id
from
(
SELECT *
from wp_postmeta
) AS X
where meta_key = 'attribute_pa_beden'
and meta_value in (".$size.")
and post_id in (
SELECT post_id
from
(
SELECT *
from wp_postmeta
) AS Y
where meta_key = 'attribute_pa_renk'
and ((meta_value = '".$color."')
OR ('".$color."' = 'all_colors'))
)
and post_id in (
SELECT id
from wp_posts
where post_type = 'product_variation'
and post_parent in (
SELECT object_id
FROM wp_term_relationships
where term_taxonomy_id in (".$urun.")))
)";
是的，我知道您需要从 UPDATE wp_postmeta "hide" wp_postmeta，但我们可以重新安排事情以提高效率。请注意，在 过滤之前，您有两种情况如何获取整个 wp_postmeta ？这使得无法使用任何索引，因此很慢。
SELECT m1.post_id
FROM wp_postmeta AS m1
JOIN wp_postmeta AS m2 USING(post_id)
JOIN wp_posts AS p2 USING(post_id)
JOIN wp_term_relationships AS tr ON p2.post_parent = tr.object_id
WHERE m1.meta_key = 'attribute_pa_beden' AND m1.meta_value in ("$size")
AND m2.meta_key = 'attribute_pa_renk' AND ( m1.meta_value = '$color'
OR '$color' = 'all_colors' )
AND p2.post_type = 'product_variation'
AND tr.term_taxonomy_id IN ($urun)
忘记 UPDATE，直到你调试了这个 SELECT。 （我可能犯了一些错误，但它看起来不是更简单吗？它会 运行 快很多，尤其是使用我推荐的索引。）
带颜色的OR可能会被优化掉，所以我不担心。
我无法预测优化器将从 4 个 table 中的哪一个开始，因此需要这些索引来为其提供选择：
tr: (term_taxonomy_id, object_id) -- in this order
posts: (post_type, post_id) -- in this order
postmeta: (meta_key, meta_value) -- see note below
Optimizer 选择从哪个 table 开始后，它将依次转到其他每个 table；订单对我们来说无关紧要。这些附加索引可能有用：
posts: (post_parent, post_id) -- in this order
postmeta: (post_id, meta_key, meta_value) -- see note below
如果 meta_value 是 LONGTEXT，那么它 不能 在索引中，所以把它去掉。 （不，不要为 "prefix" 索引而烦恼。）
如果您使用的是 MySQL 5.5 或 5.6，meta_key 太长，无法作为索引；查看我的 link 了解多种解决方法。
EAV 架构很糟糕，您正在找出原因。
回到 UPDATE 的拼凑，添加包装器：
UPDATE wp_postmeta AS m
JOIN ( SELECT post_id
FROM ( the above query )
) AS kludge USING (post_id)
SET m.meta_value = 'outofstock'
WHERE m.meta_key = '_stock_status'
我的 SQL 查询更新了数据库中的所有股票，但它的工作效率不高，有时我会收到 504 超时错误。该代码工作正常。我怎样才能让它更好地工作。
P.S: 缺少prepared statements请忽略，稍后补上
关于 tables 的一些信息（Wordpress Woocommerce 插件默认 tables）：
wp_posts: 这 table 包括帖子。 （帖子可以是产品或产品变体。例如产品是蝴蝶 T 恤，产品变体是蝴蝶 T 恤红色大号）。
wp_postmeta：此 table 包含有关帖子的元信息。例如，如果产品变体有库存，或者它是什么颜色，或者它是什么尺寸。
//This array gives, which products are there, and their respective categories.
$allProducts = array("Fermuarlı Kapşonlu Sweatshirt" => "'2653','2659'","Kapşonlu Sweatshirt" => "'2646','2651'","Sweatshirt" => "'2644','2650'","Kadın Tişört" => "'2654','2656'","Atlet" => "'2655','2657'","Tişört" => "'2643','2304'");
//Below arrays gives information about, which product variations are out of stock.
$tisort_OutOfStock =array();
$atlet_OutOfStock =array("all_colors"=>"'3xl','4xl','5xl'");
$kadin_tisort_OutOfStock =array("all_colors"=>"'xxl','3xl','4xl','5xl'");
$sweatshirt_OutOfStock =array("beyaz"=>"'xxl','3xl','4xl','5xl'","kirmizi"=>"'xxl','3xl','4xl','5xl'","bordo"=>"'5xl'","antrasit"=>"'5xl'");
$kapsonlu_sweatshirt_OutOfStock =array("gri-kircilli"=>"'5xl'");
$fermuarli_kapsonlu_sweatshirt_OutOfStock =array("gri-kircilli"=>"'5xl'","siyah"=>"'5xl'");
//Reset stocks before updating.
$resetStocks = "UPDATE wp_postmeta set meta_value = 'instock' where meta_key = '_stock_status'";
$wpdb->query($resetStocks);
echo "Stoklar are reseted<br>";
//Foreach product, foreach color, update if product doesn't have stock.
foreach( $allProducts as $key => $urun ){
switch ($key) {
case "Kadın Tişört": $tempArray = $kadin_tisort_OutOfStock; break;
case "Fermuarlı Kapşonlu Sweatshirt": $tempArray = $fermuarli_kapsonlu_sweatshirt_OutOfStock; break;
case "Kapşonlu Sweatshirt": $tempArray = $kapsonlu_sweatshirt_OutOfStock; break;
case "Sweatshirt": $tempArray = $sweatshirt_OutOfStock; break;
case "Atlet": $tempArray = $atlet_OutOfStock; break;
case "Tişört": $tempArray = $tisort_OutOfStock; break;
}
foreach( $tempArray as $color => $size ){
$query = "UPDATE wp_postmeta set meta_value = 'outofstock' where meta_key = '_stock_status' and post_id in
(
select post_id from (select * from wp_postmeta) AS X where meta_key = 'attribute_pa_beden' and meta_value in (".$size.")
and post_id in (select post_id from (select * from wp_postmeta) AS Y where meta_key = 'attribute_pa_renk' and ((meta_value = '".$color."') OR ('".$color."' = 'all_colors')))
and post_id in (select id from wp_posts where post_type = 'product_variation' and post_parent in (select object_id FROM wp_term_relationships where term_taxonomy_id in (".$urun.")))
)";
global $wpdb;
$updatedRowCount = $wpdb->query($query);
}
}
SELECT post_id
from
(
SELECT *
from wp_postmeta
) AS X
where ...
-->
SELECT post_id
FROM wp_postmeta
WHERE ...
（子查询只会减慢速度。）
and post_id in (
SELECT post_id
而不是 IN ( SELECT ... )，使用 JOIN。
除了这些技巧之外，请参阅我的 tips 关于改进 postmeta 架构的内容。
首先处理密钥 SELECT:
SELECT post_id
from
(
SELECT *
from wp_postmeta
) AS X
where meta_key = 'attribute_pa_beden'
and meta_value in (".$size.")
and post_id in (
SELECT post_id
from
(
SELECT *
from wp_postmeta
) AS Y
where meta_key = 'attribute_pa_renk'
and ((meta_value = '".$color."')
OR ('".$color."' = 'all_colors'))
)
and post_id in (
SELECT id
from wp_posts
where post_type = 'product_variation'
and post_parent in (
SELECT object_id
FROM wp_term_relationships
where term_taxonomy_id in (".$urun.")))
)";
是的，我知道您需要从 UPDATE wp_postmeta "hide" wp_postmeta，但我们可以重新安排事情以提高效率。请注意，在 过滤之前，您有两种情况如何获取整个 wp_postmeta ？这使得无法使用任何索引，因此很慢。
SELECT m1.post_id
FROM wp_postmeta AS m1
JOIN wp_postmeta AS m2 USING(post_id)
JOIN wp_posts AS p2 USING(post_id)
JOIN wp_term_relationships AS tr ON p2.post_parent = tr.object_id
WHERE m1.meta_key = 'attribute_pa_beden' AND m1.meta_value in ("$size")
AND m2.meta_key = 'attribute_pa_renk' AND ( m1.meta_value = '$color'
OR '$color' = 'all_colors' )
AND p2.post_type = 'product_variation'
AND tr.term_taxonomy_id IN ($urun)
忘记 UPDATE，直到你调试了这个 SELECT。 （我可能犯了一些错误，但它看起来不是更简单吗？它会 运行 快很多，尤其是使用我推荐的索引。）
带颜色的OR可能会被优化掉，所以我不担心。
我无法预测优化器将从 4 个 table 中的哪一个开始，因此需要这些索引来为其提供选择：
tr: (term_taxonomy_id, object_id) -- in this order
posts: (post_type, post_id) -- in this order
postmeta: (meta_key, meta_value) -- see note below
Optimizer 选择从哪个 table 开始后，它将依次转到其他每个 table；订单对我们来说无关紧要。这些附加索引可能有用：
posts: (post_parent, post_id) -- in this order
postmeta: (post_id, meta_key, meta_value) -- see note below
如果 meta_value 是 LONGTEXT，那么它 不能 在索引中，所以把它去掉。 （不，不要为 "prefix" 索引而烦恼。）
如果您使用的是 MySQL 5.5 或 5.6，meta_key 太长，无法作为索引；查看我的 link 了解多种解决方法。
EAV 架构很糟糕，您正在找出原因。
回到 UPDATE 的拼凑，添加包装器：
UPDATE wp_postmeta AS m
JOIN ( SELECT post_id
FROM ( the above query )
) AS kludge USING (post_id)
SET m.meta_value = 'outofstock'
WHERE m.meta_key = '_stock_status'