在 SQL 服务器中以行方式求和进行透视

Question

TABLE-A:

PID	BrandName
1	Brand1
2	Brand2
3	Brand3
4	Brand4
5	Brand5
6	Brand6
7	Brand7
8	Brand8

TABLE-B:

CustNo	Name	BrandName	Qty	Amt
1	C1	Brand1	3	300
1	C1	Brand2	2	400
1	C1	Brand4	1	300
1	C1	Brand5	2	100
2	C2	Brand1	2	200
2	C2	Brand3	1	200
3	C3	Brand2	1	300
3	C3	Brand7	3	150

预期结果：-

CustNo	Name	Brand1	Brand2	Brand3	Brand4	Brand5	Brand7	Amt
1	C1	3	2	0	1	2	0	1100
2	C2	2	0	1	0	0	0	400
3	C3	0	1	0	0	0	3	450

我试过的支点：-

DECLARE @cols AS NVARCHAR(MAX),
    @query  AS NVARCHAR(MAX)

select @cols = STUFF((SELECT ',' + QUOTENAME(BrandName) from [TABLE-A] order by PID FOR XML PATH(''), TYPE).value('.', 'NVARCHAR(MAX)') ,1,1,'')

set @query = 'SELECT CustNo,[Name],' + @cols + '
             from 
             (
select CustNo,[Name],Qty,SUM(cast([amt] as float)) as Amt,BrandName from [TABLE-B] group by CustNo,[Name],BrandName,Qty
            ) x
            pivot 
            (
                max(Qty)
                for brandname in (' + @cols + ')
            ) p '

execute(@query)

Answer 1

您的查询几乎没有错误。首先，您 selected 列 'Slab' 不在您的 table 中（可能是由于从另一个查询复制）相反，您需要 select custno 和名称。

那么您的查询将运行但是您将针对每个客户有三行，因为每个客户在数量字段中有三个不同的值。背后的原因是内部查询中的group by子句（group by CustNo,[Name],BrandName,Qty）。相反，我使用 window 函数为每个客户求和（amt）。

我还使用了两组动态列名来去除结果中的空值。一个按照您在代码中使用的方式进行转换 (@cols)，其他列表包含 coalesce(columnname,0) 以将 null 转换为 0。

如果您使用的是 SQL Server 2017 及更高版本，那么我建议使用 string_agg() 来连接列名，因为它更容易、性能更快。我在查询 #2 中使用过它。

架构和插入语句：

create table [Table-A](PID int, BrandName varchar(50));
insert into [Table-A] values(1  ,'Brand1');
insert into [Table-A] values(2  ,'Brand2');
insert into [Table-A] values(3  ,'Brand3');
insert into [Table-A] values(4  ,'Brand4');
insert into [Table-A] values(5  ,'Brand5');
insert into [Table-A] values(6  ,'Brand6');
insert into [Table-A] values(7  ,'Brand7');
insert into [Table-A] values(8  ,'Brand8');

create table [TABLE-B]( CustNo  int,Name varchar(10),BrandName varchar(50),Qty int, Amt int);
insert into [TABLE-B] values(1  ,'C1',  'Brand1',   3,  300);
insert into [TABLE-B] values(1  ,'C1',  'Brand2',   2,  400);
insert into [TABLE-B] values(1  ,'C1',  'Brand4',   1,  300);
insert into [TABLE-B] values(1  ,'C1',  'Brand5',   2,  100);
insert into [TABLE-B] values(2  ,'C2',  'Brand1',   2,  200);
insert into [TABLE-B] values(2  ,'C2',  'Brand3',   1,  200);
insert into [TABLE-B] values(3  ,'C3',  'Brand2',   1,  300);
insert into [TABLE-B] values(3  ,'C3',  'Brand7',   3,  150);

查询#1（使用 stuff() 和 xml 路径 for()）

DECLARE @cols AS NVARCHAR(MAX),
    @colsForSelect AS NVARCHAR(MAX),
    @query  AS NVARCHAR(MAX)
 
 
SET @colsForSelect = STUFF((SELECT  ',' + ' Coalesce('+quotename(BrandName)+',0) '+ quotename(BrandName)
            FROM [TABLE-A] order by pid
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')
        
select @cols = STUFF((SELECT ',' + QUOTENAME(BrandName) from [TABLE-A] order by PID FOR XML PATH(''), TYPE).value('.', 'NVARCHAR(MAX)') ,1,1,'')

set @query = 'SELECT custno,name,' + @colsForSelect + ',Amt
             from 
             (
select CustNo,[Name],Qty,SUM(cast([amt] as float))over(partition by custno) as Amt,BrandName from [TABLE-B]             ) x
            pivot 
            (
                max(Qty)
                for brandname in (' + @cols + ')
            ) p '

execute(@query)

输出：

custno	name	Brand1	Brand2	Brand3	Brand4	Brand5	Brand7	Amt
1	C1	3	2	0	1	2	0	1100
2	C2	2	0	1	0	0	0	400
3	C3	0	1	0	0	0	3	450

查询#2（使用 string_agg() 而不是 stuff() 和 xml path()）

DECLARE @cols AS NVARCHAR(MAX),
    @colsForSelect AS NVARCHAR(MAX),
    @query  AS NVARCHAR(MAX)


select @colsForSelect= string_agg('coalesce('+BrandName+',0) '+QUOTENAME(BrandName)+' ','  ,') from [TABLE-A]

select @cols = string_agg(QUOTENAME(BrandName),',') from [TABLE-A]


set @query = 'SELECT custno,name,' + @colsForSelect + ',Amt
             from 
             (
select CustNo,[Name],Qty,SUM(cast([amt] as float))over(partition by custno) as Amt,BrandName from [TABLE-B]             ) x
            pivot 
            (
                max(Qty)
                for brandname in (' + @cols + ')
            ) p'

execute(@query)

输出：

custno	name	Brand1	Brand2	Brand3	Brand4	Brand5	Brand7	Amt
1	C1	3	2	0	1	2	0	1100
2	C2	2	0	1	0	0	0	400
3	C3	0	1	0	0	0	3	450

db<>fiddle here

在 SQL 服务器中以行方式求和进行透视

Pivot with row wise Sum in SQL Server

sql

tsql

sql-server

pivot