如何选择甚至不存在的值

Question

我使用的是 SQL-Server 2005

我有两个表，用户和订单。每个用户可以有多个订单。通过 userID 连接的表。订单有日期列（即下订单的时间）。用户有registrationSite栏（这是用户及其所有订单背后的附属网站）。

我想选择每天和网站的订单总和，所以 即使网站没有售出或用户注册，我也会有 0 作为总价值。

这是缺少突出显示部分的当前查询。

select sum(orderSum)*40/100-0.17,count(*),
registrationSite,
dateadd(dd,datediff(dd,0,cu.date),0)
from Users cu 
inner join Orders cp 
on cu.userID=cp.userID
group by dateadd(dd,datediff(dd,0,cu.date),0),registrationSite
order by dateadd(dd,datediff(dd,0,cu.date),0),registrationSite

Answer 1

使用LEFT OUTER JOIN 而不是INNER JOIN ..

from Users cu 
inner join Orders cp 

应该变成

from Users cu 
left outer join Orders cp 

这意味着所有用户及其附带的订单（如果存在）。（但无论订单如何，用户都将是所有用户）

---

更新

您需要创建一个 tally 表，其中包含日期范围（天）和left outer join ..

看这里：How to get a table of dates between x and y in sql server 2005

Answer 2

如果您使用的是 SQL 2005 或更高版本，则可以使用公用表表达式动态构建日历表。在您的原始帖子中，您正在对 cu.date 进行日期数学运算，但别名 cu 代表的是用户表而不是订单表。我假设这应该是cp.date？即，您的描述谈到了订单日期，但您的 SQL 使用了用户表中的日期。

With OrderDateBoundaries As
    (
    Select Cast(DateDiff(d,0,Min([Date]))) As MinDate
        , Cast(DateDiff(d,0,Max([Date]))) As MaxDate
    From Orders
    )
    , Calendar As
    (
    Select MinDate As [Date]
    From OrderDateBoundaries
    Union All
    Select DateAdd(d, 1, [Date])
    From Calendar
    Where [Date] <= DateAdd(d, 1, (
                                    Select MaxDate
                                    From OrderDateBoundaries
                                    ))
    )
Select Calendar.[Date]
    , Coalesce(Sum(O.ordersum) * 40 / 100 - 0.17,0) As OrderSum
    , Count(*)
    , RegistrationSite
From Calendar
        Left Join (Users As U
            Join Orders As O
                On O.userId = U.UserId)
             On Cast(DateDiff(d, 0, O.OrderDate) As datetime) =Calendar.Date
Group By Calendar.[Date], registrationsite
Option(MaxRecursion 0);

如果您确实想加入Users.Date，那么只需进行简单更改即可。另外，不清楚ordersum是从哪个表中存储的。

With DateBoundaries As
    (
    Select Cast(DateDiff(d,0,Min([Date]))) As MinDate
        , Cast(DateDiff(d,0,Max([Date]))) As MaxDate
    From Users
    )
    , Calendar As
    (
    Select MinDate As [Date]
    From DateBoundaries
    Union All
    Select DateAdd(d, 1, [Date])
    From Calendar
    Where [Date] <= DateAdd(d, 1, (
                                    Select MaxDate
                                    From DateBoundaries
                                    ))
    )
Select Calendar.[Date]
    , Coalesce(Sum(O.ordersum) * 40 / 100 - 0.17,0) As OrderSum
    , Count(*)
    , RegistrationSite
From Calendar
        Left Join (Users As U
            Join Orders As O
                On O.userId = U.UserId)
             On Cast(DateDiff(d,0,U.Date) As datetime) =Calendar.Date
Group By Calendar.[Date], RegistrationSite
Option(MaxRecursion 0);