Question

RGN - ＆gt;区域名称，NPR - ＆gt;需要针对每个区域计算的事件，RGF - >区域过滤器。所以在NPR我有汽车板块的事件，我需要为每个地区计算它们，并将其分成一周的7天。我的解决方案如下：

SELECT DISTINCT a.NAME,
COUNT(CASE WHEN DATEPART(dw,b.TM) = 1 THEN 1 ELSE null END) as MON,
COUNT(CASE WHEN DATEPART(dw,b.TM) = 2 THEN 1 ELSE null END) as TUE, 
COUNT(CASE WHEN DATEPART(dw,b.TM) = 3 THEN 1 ELSE null END) as WED, 
COUNT(CASE WHEN DATEPART(dw,b.TM) = 4 THEN 1 ELSE null END) as THU, 
COUNT(CASE WHEN DATEPART(dw,b.TM) = 5 THEN 1 ELSE null END) as FRI, 
COUNT(CASE WHEN DATEPART(dw,b.TM) = 6 THEN 1 ELSE null END) as SAT, 
COUNT(CASE WHEN DATEPART(dw,b.TM) = 7 THEN 1 ELSE null END) as SUN 
FROM RGN a 
LEFT JOIN NPR b 
ON a.ID IN  (
SELECT c.RGID
FROM RGF c
WHERE RTRIM(b.NUM) like RTRIM((REPLACE(c.RGFLT,'?','_')))
)
Group by a.NAME;

在过滤器中，我需要将?替换为_以获得like()表达式。

这个查询给了我很好的结果，但是在数据库中有4密耳的记录，就像20分钟就可以得到它。您是否知道如何做到完全相同但效率更高？

P.S。我尝试使用SUM（）而不是计数，但它不会改变任何东西。

@Edit我的下一次尝试是SELECT FROM NPR和join RGN，它有点快但不够。

Answer 1

SELECT a.NAME,
COUNT(CASE WHEN DATEPART(dw,b.TM) = 1 THEN 1 ELSE null END) as MON,
COUNT(CASE WHEN DATEPART(dw,b.TM) = 2 THEN 1 ELSE null END) as TUE, 
COUNT(CASE WHEN DATEPART(dw,b.TM) = 3 THEN 1 ELSE null END) as WED, 
COUNT(CASE WHEN DATEPART(dw,b.TM) = 4 THEN 1 ELSE null END) as THU, 
COUNT(CASE WHEN DATEPART(dw,b.TM) = 5 THEN 1 ELSE null END) as FRI, 
COUNT(CASE WHEN DATEPART(dw,b.TM) = 6 THEN 1 ELSE null END) as SAT, 
COUNT(CASE WHEN DATEPART(dw,b.TM) = 7 THEN 1 ELSE null END) as SUN
FROM RGN a, RGF c
left JOIN NPR b
ON RTRIM(b.NUM) like RTRIM((REPLACE(c.RGFLT,'?','_')))
WHERE a.ID = c.RGID AND b.TM BETWEEN
Group by a.RGNAME

执行时间为100k行~4s，足够好，因为它是~12s

更有效的SQL查询方式

1 个答案: