所有日期和所有的用戶數量和類別

Question

我使用SQL Server 2012，並具有以下表創建矩陣：

CREATE TABLE [dbo].[Users](
    [UserId] [bigint] NOT NULL 
) 
CREATE TABLE [dbo].[CategoryIds](
    [CategoryId] [int] NOT NULL 
) 
CREATE TABLE [dbo].[DimDate](
    [Year_Value] [int] NOT NULL, 
    [Month_Value] [int] NOT NULL 
) 
CREATE TABLE [dbo].[TransactionsData](
    [TransactionId] [int] NOT NULL, 
[UserId] [bigint] NOT NULL, 
[CategoryId] [int] NOT NULL, 
[Date] [datetime] NOT NULL 
) 

表中的數據是這樣的：

Insert into [dbo].[Users] 
select 1 union all 
select 2 union all 
select 3 


Insert into [dbo].[CategoryIds] 
select 1 union all 
select 2 union all 
select 5 

insert into [dbo].[DimDate] 
select 2015, 3 union all 
select 2015,4 

insert into [dbo].[TransactionsData] 
select 1,1,1,'2015-03-01' union all 
select 2,1,1,'2015-03-20' union all 
select 3,2,5,'2015-03-02' union all 
select 4,2,5,'2015-03-05' 

我想有一個T-SQL代碼生成這樣的表格：

CREATE TABLE [dbo].[Matrix](
[Year_Value] [int] NOT NULL, 
[Month_Value] [int] NOT NULL, 
[UserId] [bigint] NOT NULL, 
[TransactionCount] [bigint] NOT NULL, 
[1] [int] NOT NULL, -- category with id 1 
[2] [int] NOT NULL, -- category with id 2 
[5] [int] NOT NULL -- category with id 5 etc 
) 

而矩陣表包含一個記錄對於每個年份和月份的每個用戶，無論該月份是否有該用戶的任何交易。如果沒有事務，則Year和Month和UserID仍作爲記錄存在，但事務計數爲0.對類別同樣有效：每個CategoryId成爲Matrix表的一列，每列中都有一個計數每年，每月和每個用戶有多少交易屬於每個類別。如果沒有針對特定用戶和特定類別的交易，則相應列中的值爲0。

我已經試過什麼：

select d.*, u.UserId, isnull(t.TransactionId,0) 
from [dbo].[DimDate] d 
outer apply (select UserId from [dbo].[Users]) u 
full outer join [dbo].[TransactionsData] t on u.UserId = t.UserId 

的數據在矩陣表將這個樣子，因爲上面的表中的數據：

INSERT INTO [dbo].[Matrix] 
SELECT 2015, 3, 1, 2, 2, 0, 0 UNION ALL 
SELECT 2015, 3, 2, 2, 0, 0, 2 UNION ALL 
SELECT 2015, 3, 3, 0, 0, 0, 0 UNION ALL 
SELECT 2015, 4, 1, 0, 0, 0, 0 UNION ALL 
SELECT 2015, 4, 2, 0, 0, 0, 0 UNION ALL 
SELECT 2015, 4, 3, 0, 0, 0, 0 

Answer 1

您可以在此使用dynamic crosstab做：

SQL Fiddle

DECLARE @sql NVARCHAR(MAX) = '' 

SELECT @sql = 
'SELECT 
    a.Year_Value 
    , a.Month_Value 
    , a.UserId 
    , COUNT(td.TransactionId) AS TransactionCount' + CHAR(10) 

SELECT @sql = @sql + 
' , SUM(CASE WHEN CategoryId = ' + CONVERT(VARCHAR(10), CategoryId) + ' THEN 1 ELSE 0 END) AS ' + QUOTENAME(CategoryId) + CHAR(10) 
FROM CategoryIds 

SELECT @sql = @sql + 
'FROM (
    SELECT 
     u.UserId, dd.Year_Value, m.Month_Value 
    FROM(VALUES 
     (1),(2),(3),(4),(5),(6),(7),(8),(9),(10),(11),(12) 
    )m(Month_Value) 
    CROSS JOIN(
     SELECT DISTINCT Year_Value FROM DimDate 
    )dd 
    CROSS JOIN Users u 
) a 
LEFT JOIN TransactionsData td 
    ON td.UserId = a.UserId 
    AND MONTH(td.Date) = a.Month_Value 
    AND YEAR(td.Date) = a.Year_Value 
GROUP BY 
    a.Year_Value, a.Month_Value, a.UserId 
ORDER BY 
    a.UserId, a.Year_Value, a.Month_Value' 

PRINT @sql 
EXEC (@sql) 

---

的想法是生成所有可能的UserId - Year_Value - Month_Value組合。然後，使用生成的組合並在TransactionsData上執行LEFT JOIN。

現在，要將CategoryIds顯示爲列，您需要執行條件聚合。這是非動態解決方案。這也是PRINT @sql命令的輸出。

SELECT 
    a.Year_Value 
    , a.Month_Value 
    , a.UserId 
    , COUNT(td.TransactionId) AS TransactionCount 
    , SUM(CASE WHEN CategoryId = 1 THEN 1 ELSE 0 END) AS [1] 
    , SUM(CASE WHEN CategoryId = 2 THEN 1 ELSE 0 END) AS [2] 
    , SUM(CASE WHEN CategoryId = 5 THEN 1 ELSE 0 END) AS [5] 
FROM (
    SELECT 
     u.UserId, dd.Year_Value, m.Month_Value 
    FROM(VALUES 
     (1),(2),(3),(4),(5),(6),(7),(8),(9),(10),(11),(12) 
    )m(Month_Value) 
    CROSS JOIN(
     SELECT DISTINCT Year_Value FROM DimDate 
    )dd 
    CROSS JOIN Users u 
) AS a 
LEFT JOIN TransactionsData td 
    ON td.UserId = a.UserId 
    AND MONTH(td.Date) = a.Month_Value 
    AND YEAR(td.Date) = a.Year_Value 
GROUP BY 
    a.Year_Value, a.Month_Value, a.UserId 
ORDER BY 
    a.UserId, a.Year_Value, a.Month_Value