【发布时间】：2013-11-08 12:47:48
【问题描述】：
我有一个查询需要超过 2 小时才能执行数百万条记录，我不确定如何优化它以使其运行速度更快。 Here 是表和查询的 sql fiddle。
WITH frm
AS (SELECT product_id AS PId,
Min(Cast(product_startdate AS DATETIME)) AS PStartDate
FROM products
WHERE product_status IN ( 'F', 'R', 'M' )
GROUP BY product_id),
firstcount
AS (SELECT pid,
pstartdate,
(SELECT Count(*)
FROM products
WHERE product_id IN ((SELECT product_id
FROM products
WHERE product_status IN ( 'OR', 'OP' )
AND product_comments LIKE
'%CANCELLED%'
EXCEPT
(SELECT product_id
FROM products
WHERE product_status = 'DE'
UNION
SELECT product_id
FROM products
WHERE product_status = 'OR'
AND product_comments NOT LIKE
'%CANCELLED%'))
EXCEPT
(SELECT product_id
FROM products
WHERE product_status IN (
'RE', 'C', 'S', 'D' )
))
AND product_id = pid) AS v_count
FROM frm),
secondcount
AS (SELECT pid,
pstartdate,
CASE
WHEN v_count = 0 THEN (SELECT Count(*)
FROM products
WHERE product_id IN
(
SELECT product_id
FROM [dbo].products
WHERE
product_status IN ( 'F',
'R', 'M' )
AND product_startdate !=
'.'
EXCEPT
(SELECT
product_id
FROM products
WHERE
product_status = 'DE'
UNION
SELECT
product_id
FROM products
WHERE
product_status = 'OR'
AND product_comments NOT
LIKE '%CANCELLED%')
EXCEPT
(SELECT
product_id
FROM products
WHERE
product_status IN ( 'OR',
'OP' )
AND product_comments LIKE
'%CANCELLED%'
EXCEPT
(SELECT
product_id
FROM products
WHERE
product_status =
'DE'
UNION
SELECT product_id
FROM products
WHERE product_status =
'OR'
AND product_comments NOT
LIKE '%CANCELLED%'))
EXCEPT
(SELECT
product_id
FROM products
WHERE
product_status IN ( 'RE',
'C', 'S', 'D' )))
AND product_id = pid)
ELSE v_count
END AS v_count
FROM firstcount)
INSERT INTO products_del
(product_id,
product_startdate,
productdel_status)
SELECT pid,
pstartdate,
CASE
WHEN v_count != 0 THEN 'UNKNOWN'
ELSE NULL
END
FROM secondcount
SELECT *
FROM products_del
【问题讨论】：
-
在那个表上放一些索引 - 看看执行计划，它是表扫描后的表扫描。
-
如果您解释您的表结构并解释您想要实现的目标，这可能会有所帮助。这取决于您，让我们了解您的目标，而无需对您的查询进行逆向工程。
-
通用表表达式非常好用，但它们会减慢复杂查询的速度——尤其是大型查询。尝试将查询选择到临时表（或表变量）中，而不是嵌套公用表表达式，看看是否有帮助。
-
即使表是凌乱的或临时的，索引也应该加快查询速度。如果花费两个小时，构建它们的时间可能是延长查询时间的一个很好的权衡。
-
考虑在您的例外中添加连接提示，以指导查询计划，如第一个答案中所述，在此 SO 链接上：stackoverflow.com/questions/16084350/sql-except-performance/…
标签： sql sql-server query-optimization