2

我正在努力优化检索某些产品及其先前记录的定价详细信息的查询。

因为我的 3 个子查询使用 desc 顺序和限制 1 来返回最后记录的价格，并且日志表包含数万行（从长远来看可能包含数百万行），所以它在5秒的区域产生查询结果（大约有3000个产品）。

SELECT products.*, 
    (SELECT value_new FROM log_actions 
     WHERE action='edit' AND target_entity='net_price_euro' 
     AND target_table='products' 
     AND target_id=products.product_id
     ORDER BY log_time DESC LIMIT 1
    ) AS previous_net_price_euro,
    (SELECT value_new FROM log_actions 
     WHERE action='edit' AND target_entity='retail_sterling' 
     AND target_table='products' AND target_id=products.product_id
     ORDER BY log_time DESC LIMIT 1
     ) AS previous_retail_sterling,
    (SELECT value_new FROM log_actions 
     WHERE action='edit' AND target_entity='gp' 
     AND target_table='products' 
     AND target_id=products.product_id
     ORDER BY log_time DESC LIMIT 1
    ) AS previous_gp
FROM products
WHERE products.deleted IS NULL ORDER BY products.status asc

据我所知，我已将所有相关字段编入索引。在上面运行 EXPLAIN 会返回以下内容：

id  select_type     table   type    possible_keys   key     key_len     ref     rows    Extra
1   PRIMARY     products    ALL     deleted     NULL    NULL    NULL    3192    Using where
4   DEPENDENT SUBQUERY  log_actions     ref     target_entity,target_id,action,target_table     target_id   4   products.product_id     4   Using where; Using filesort
3   DEPENDENT SUBQUERY  log_actions     ref     target_entity,target_id,action,target_table     target_id   4   products.product_id     4   Using where; Using filesort
2   DEPENDENT SUBQUERY  log_actions     ref     target_entity,target_id,action,target_table     target_id   4   products.product_id     4   Using where; Using filesort

我已经阅读了添加 desc 索引可以提高性能，但进一步阅读似乎 MySQL 目前不支持这一点。

4

2 回答 2

2

这个查询重新思考了一点问题。你说得对，mysql还不支持DESC索引，这就是为什么我试图做一个比你的订购次数少的查询（你订购了3次）：

SELECT products.*, 
    l1.value_new AS previous_net_price_euro,
    l2.value_new AS previous_retail_sterling,
    l3.value_new AS previous_gp
FROM products
    INNER JOIN (
            SELECT target_id, target_entity, MAX(log_time) AS mx FROM log_actions
            WHERE
                action='edit' 
            AND target_table='products'
            GROUP BY target_id, target_entity
        ) lastvalues ON lastvalues.target_id = products.product_id
    INNER JOIN (
            SELECT target_id, target_entity, log_time, value_new FROM log_actions
            WHERE
                target_entity = 'net_price_euro' 
            AND action='edit' 
            AND target_table='products'
        ) l1 ON l1.target_id            = lastvalues.target_id 
                AND l1.log_time         = lastvalues.mx 
                AND l1.target_entity    = lastvalues.target_entity
    INNER JOIN (
            SELECT target_id, target_entity, log_time, value_new FROM log_actions
            WHERE
                target_entity = 'retail_sterling' 
            AND action = 'edit' 
            AND target_table = 'products'
        ) l2 ON l2.target_id            = lastvalues.target_id 
                AND l2.log_time         = lastvalues.mx 
                AND l2.target_entity    = lastvalues.target_entity
    INNER JOIN (
            SELECT target_id, target_entity, log_time, value_new FROM log_actions
            WHERE
                target_entity = 'gp' 
            AND action = 'edit' 
            AND target_table = 'products'
        ) l3 ON l3.target_id            = lastvalues.target_id 
                AND l3.log_time         = lastvalues.mx 
                AND l3.target_entity    = lastvalues.target_entity
WHERE products.deleted IS NULL 
ORDER BY products.status ASC

在这种情况下，您可以使用另一个包含 log_time 的索引，它可能会加速内部连接。

于 2013-01-30T15:09:04.663 回答
1

三个内联子查询都是相似的：

(SELECT value_new FROM log_actions 
 WHERE action = 'edit' 
   AND target_entity = 'net_price_euro' 
   AND target_table = 'products' 
   AND target_id = products.product_id
 ORDER BY log_time DESC LIMIT 1
)

我会在以下位置添加一个索引：

 (action, target_table, target_entity
 ,target_id, log_time, value_new)

索引中列的顺序对于前 3 列无关紧要，但对于后三列却很重要：(..., target_id, log_time, value_new)

于 2013-01-30T14:53:58.553 回答