获取查询结果总摘要行的最佳方法

Question

我正在执行一个查询，该查询将获得数千行结果，而客户需要一行显示某些数字列的总和。 我通过使用分组分组来实现此目的，但是此功能最多支持32个不在聚合功能中的列。 我的问题是我必须返回将近45列，由于聚合函数，我只剩下10列。

原始查询如下所示：

    select
    o.Name, 
    ci.Id,
    ci.OriginId,
    ci.Varchar1,
    ci.Varchar2,
    ci.Varchar3,
    ci.Varchar4,
    ci.Varchar5,
    ci.Varchar6,
    ci.Varchar7,
    ci.Varchar8,
    ci.Varchar9,
    ci.Varchar10,
    ci.Varchar11,
    ci.Varchar12,
    ci.Varchar13,
    ci.Varchar14,
    ci.Varchar15,
    ci.Varchar16,
    ci.Varchar17,
    ci.Varchar18,
    ci.Varchar19,
    ci.Varchar20,       
    sum(ci.Decimal1) as Decimal1,
    sum(ci.Decimal1) as Decimal2,
    sum(ci.Decimal1) as Decimal3,
    sum(ci.Decimal1) as Decimal4,
    sum(ci.Decimal1) as Decimal5,
    sum(ci.Decimal1) as Decimal6,
    sum(ci.Decimal1) as Decimal7,
    sum(ci.Decimal1) as Decimal8,
    sum(ci.Decimal1) as Decimal9,
    sum(ci.Decimal1) as Decimal10,
    ci.Date1,
    ci.Date2,
    ci.Date3,
    ci.Date4,
    ci.Date5,
    ci.Date6,
    ci.Date7,
    ci.Date8,
    ci.Date9,
    ci.Date10

from 
    Items ci
    inner join Origins o 
    on ci.OriginId = o.Id 

group by grouping sets((
    o.Name,
    ci.Id,
    ci.OriginId,
    ci.Varchar1,
    ci.Varchar2,
    ci.Varchar3,
    ci.Varchar4,
    ci.Varchar5,
    ci.Varchar6,
    ci.Varchar7,
    ci.Varchar8,
    ci.Varchar9,
    ci.Varchar10,
    ci.Varchar11,
    ci.Varchar12,
    ci.Varchar13,
    ci.Varchar14,
    ci.Varchar15,
    ci.Varchar16,
    ci.Varchar17,
    ci.Varchar18,
    ci.Varchar19,
    ci.Varchar20,
    ci.Date1,
    ci.Date2,
    ci.Date3,
    ci.Date4,
    ci.Date5,
    ci.Date6,
    ci.Date7,
    ci.Date8,
    ci.Date9,
    ci.Date10), ())

我试图将查询一分为二，以使group by中的列数不达到可用的最大值。 如果我将每个查询分开执行，则会得到所需的结果，但是如果我将它们合并，则会出错（无法将nvarchar转换为数值）。

结果是这样的：

    select      
    o.name
    ci.Id,
    ci.OriginId,
    sum(ci.Decimal1) as Decimal1,
    sum(ci.Decimal1) as Decimal2,
    sum(ci.Decimal1) as Decimal3,
    sum(ci.Decimal1) as Decimal4,
    sum(ci.Decimal1) as Decimal5,
    sum(ci.Decimal1) as Decimal6,
    sum(ci.Decimal1) as Decimal7,
    sum(ci.Decimal1) as Decimal8,
    sum(ci.Decimal1) as Decimal9,
    sum(ci.Decimal1) as Decimal10,
    ci.Date1,
    ci.Date2,
    ci.Date3,
    ci.Date4,
    ci.Date5,
    ci.Date6,
    ci.Date7,
    ci.Date8,
    ci.Date9,
    ci.Date10

from 
    Items ci
    inner join Origins o 
    on ci.OriginId = o.Id 


group by grouping sets((
    o.Name,
    ci.Id,
    ci.OriginId,        
    ci.Date1,
    ci.Date2,
    ci.Date3,
    ci.Date4,
    ci.Date5,
    ci.Date6,
    ci.Date7,
    ci.Date8,
    ci.Date9,
    ci.Date10), ())

union

select  
    o.Name,
    ci.Id,
    ci.OriginId,
    ci.Varchar1,
    ci.Varchar2,
    ci.Varchar3,
    ci.Varchar4,
    ci.Varchar5,
    ci.Varchar6,
    ci.Varchar7,
    ci.Varchar8,
    ci.Varchar9,
    ci.Varchar10,
    ci.Varchar11,
    ci.Varchar12,
    ci.Varchar13,
    ci.Varchar14,
    ci.Varchar15,
    ci.Varchar16,
    ci.Varchar17,
    ci.Varchar18,
    ci.Varchar19,
    ci.Varchar20

from 
    Items ci
    inner join Origins o 
    on ci.OriginId = o.Id 

group by grouping sets((
    o.name,
    ci.Id,
    ci.OriginId,

    ci.Varchar1,
    ci.Varchar2,
    ci.Varchar3,
    ci.Varchar4,
    ci.Varchar5,
    ci.Varchar6,
    ci.Varchar7,
    ci.Varchar8,
    ci.Varchar9,
    ci.Varchar10,
    ci.Varchar11,
    ci.Varchar12,
    ci.Varchar13,
    ci.Varchar14,
    ci.Varchar15,
    ci.Varchar16,
    ci.Varchar17,
    ci.Varchar18,
    ci.Varchar19,
    ci.Varchar20), ())

另一种方法（如果可能的话）是通过对SQL中的集合进行分组来删除组并使用C＃生成一行，因为查询的结果是由IEnumerable接收的，但是我不知道SUM函数是否可用。

任何建议将被认真考虑。

提前致谢。

Answer 1

如果您要执行的操作基本上是所有数据加上总行，请考虑以下方法。 不要按包括所有未聚合列的分组集进行分组，而应按行ID分组（现有的ID在所有数据行中应该是唯一的，或者是用row_number()函数创建的人为的）。 还应考虑在计算总计后加入辅助表。

示例如下。

设置样本数据：

declare @origs table (id int, name varchar(20));
insert into @origs values (1, 'orig1'), (2, 'orig2');

declare @items table (
    id int, orig_id int,
    column1 varchar(20), column2 varchar(20),
    value1 float, value2 float);

insert into @items values
(1, 1, 'c1.1', 'c2.1', 100, 10)
,(2, 1, 'c1.2', 'c2.2', 200, 20)
,(3, 2, 'c1.3', 'c2.3', 300, 30);

下面的查询以您尝试的方式返回所有数据和总行：

select i.id, o.name as orig, i.column1, i.column2, sum(i.value1) val1, sum(i.value2) val2
from @items i
    join @origs o on o.id = i.orig_id
group by grouping sets ((i.id, o.name, i.column1, i.column2), ());

输出为：

id    orig  column1  column2  val1  val2
----- ----- -------- -------- ----- -----
1     orig1 c1.1     c2.1     100   10
2     orig1 c1.2     c2.2     200   20
3     orig2 c1.3     c2.3     300   30
NULL  NULL  NULL     NULL     600   60

将其与下一个查询进行比较，该查询将数据按单个列进行分组。 分组数据后， @origs连接辅助表@origs 。

;with items as (
    select
        case grouping(id) when 0 then max(id) else NULL end id,
        case grouping(id) when 0 then max(orig_id) else NULL end orig_id,
        case grouping(id) when 0 then max(column1) else NULL end column1,
        case grouping(id) when 0 then max(column2) else NULL end column2,
        val1 = sum(value1),
        val2 = sum(value2)
    from @items
    group by rollup (id)
)
select i.id, o.name as orig, i.column1, i.column2, i.val1, i.val2
from items i
    left join @origs o on o.id = i.orig_id;

输出是相同的 ：

id    orig  column1  column2  val1  val2
----- ----- -------- -------- ----- -----
1     orig1 c1.1     c2.1     100   10
2     orig1 c1.2     c2.2     200   20
3     orig2 c1.3     c2.3     300   30
NULL  NULL  NULL     NULL     600   60

Answer 2

仅给出几千行，我将使用存储过程将没有总计的结果存储到临时表或表值变量中，然后将结果作为该表的UNION ALL加上顶部的总计返回它的。