PolarDB这个问题怎么解决？ max没有排序语义 sql语义变了，改成这种窗口函数实现 可以走列存 也非常慢 直接把ap节点cpu拉满了，SELECT ep.product_id, p.products_name AS products_name_m, p.products_url_name, ep.enterprise_id, ent.enterprise_name_en AS enterprise_name
, ent.url_name AS enterprise_url_name, ep.specification, DATE_FORMAT(ep.price_date, ‘%Y-%m-%d’) AS price_date
, ep.price_unit, ep.price AS current_price
, (
SELECT tmp.price
FROM (
SELECT ep2.price, ROW_NUMBER() OVER (PARTITION BY ep2.product_id ORDER BY ep2.price_date DESC) AS rank_no
FROM echemi_db_v0_3.t_zy_enterprise_price ep2
LEFT JOIN t_zy_enterprise_price ep
ON ep2.product_id = ep.product_id
AND ep2.enterprise_id = ep.enterprise_id
WHERE IFNULL(ep2.delflag, ”) ‘1’
AND ep2.price_date < ep.price_date
AND IFNULL(ep2.price_date, ”) != ”
AND IFNULL(ep2.price, ”) != ”
AND IFNULL(ep2.delflag, ”) != ‘1’
) tmp
WHERE tmp.rank_no = 1
) AS previous_price, 0 AS subscribe_flag
FROM t_zy_enterprise_price ep
INNER JOIN t_e_products p
ON p.products_id = ep.product_id
AND p.information_flag = ‘1’
AND IFNULL(p.delflag, ‘0’) != ‘1’
INNER JOIN t_zy_enterprise ent
ON ent.id = ep.enterprise_id
AND IFNULL(ent.delflag, ‘0’) != ‘1’
INNER JOIN (
SELECT ep3.product_id, ep3.enterprise_id, MAX(ep3.price_date) AS max_price_date
FROM t_zy_enterprise_price ep3
GROUP BY ep3.product_id, ep3.enterprise_id
) pimax
ON pimax.product_id = ep.product_id
AND pimax.enterprise_id = ep.enterprise_id
AND pimax.max_price_date = ep.price_date
WHERE ep.delflag = ‘0’
ORDER BY subscribe_flag DESC, ep.price_date DESC, ep.update_date DESC, ep.id
LIMIT 5
以下为热心网友提供的参考意见
那得用window改，SELECT
tmp1.*,
(
SELECT
tmp.price
FROM
(
SELECT
ep2.price,
ROW_NUMBER() OVER (
PARTITION BY ep2.product_id
ORDER BY
ep2.price_date DESC
) AS rank_no
FROM
echemi_db_v0_3.t_zy_enterprise_price ep2
LEFT JOIN t_zy_enterprise_price ep ON ep2.product_id = ep.product_id
AND ep2.enterprise_id = ep.enterprise_id
WHERE
IFNULL(ep2.delflag, ”) ‘1’
AND ep2.price_date < ep.price_date
AND IFNULL(ep2.price_date, ”) != ”
AND IFNULL(ep2.price, ”) != ”
AND IFNULL(ep2.delflag, ”) != ‘1’
) tmp
WHERE
tmp.rank_no = 1
) AS previous_price,
0 AS subscribe_flag
FROM
(
SELECT
ep.product_id,
p.products_name AS products_name_m,
p.products_url_name,
ep.enterprise_id,
ent.enterprise_name_en AS enterprise_name,
ent.url_name AS enterprise_url_name,
ep.specification,
DATE_FORMAT(ep.price_date, ‘%Y-%m-%d’) AS price_date,
ep.price_unit,
ep.price AS current_price,
ep.price_date as pd,
ep.update_date as ud,
ep.id as id
FROM
t_zy_enterprise_price ep
INNER JOIN t_e_products p ON p.products_id = ep.product_id
AND p.information_flag = ‘1’
AND IFNULL(p.delflag, ‘0’) != ‘1’
INNER JOIN t_zy_enterprise ent ON ent.id = ep.enterprise_id
AND IFNULL(ent.delflag, ‘0’) != ‘1’
INNER JOIN (
SELECT
ep3.product_id,
ep3.enterprise_id,
MAX(ep3.price_date) AS max_price_date
FROM
t_zy_enterprise_price ep3
GROUP BY
ep3.product_id,
ep3.enterprise_id
) pimax ON pimax.product_id = ep.product_id
AND pimax.enterprise_id = ep.enterprise_id
AND pimax.max_price_date = ep.price_date
WHERE
ep.delflag = ‘0’
ORDER BY
subscribe_flag DESC,
ep.price_date DESC,
ep.update_date DESC,
ep.id
LIMIT
5
) tmp1
ORDER BY
subscribe_flag DESC,
tmp1.pd DESC,
tmp1.ud DESC,
tmp1.id
LIMIT
5
试试这个，先做LIMIT 5减少下window的计算量，也可以试试把
SELECT ep2.price, ep2.price_date FROM t_zy_enterprise_price ep2 ORDER BY ep2.price_date DESC, ep2.price desc limit 1;
改为
select max(ep2.price) FROM t_zy_enterprise_price ep2 where ep2.price_date = (SELECT max(price_date) from t_zy_enterprise_price);此回答整理自钉群“PolarDB专家面对面 – HTAP（列存索引）”