将Access Crosstab / PIVOT查询转换为T-SQL

Question

Access Query that I need to convert to work under SQL Server:

TRANSFORM Sum([T_Leads]![OrderType]='New Order')-1 & " / " & Sum([T_Leads]![OrderType]='Change Order')-1 
AS [New / Change]   
SELECT Employees.EmployeeName as Name, Count(T_Leads.OrderType) AS Total
FROM Employees INNER JOIN T_Leads ON Employees.EmployeeID = T_Leads.EmployeeID
WHERE (((T_Leads.Date)>Date()-7))
and [Employees.LeadRotation] <> "Inactive"
GROUP BY Employees.EmployeeName
ORDER BY T_Leads.Date
PIVOT T_Leads.Date;

The output displays a list of employees currently taking leads (who are not "inactive"). For the column headers, the date is shown for the previous seven days (if a lead was submitted on that day), and two totals are displayed under each date. One for the total number of New Orders received, and another for total number of Change Orders. I've not been able to find any examples that generate the date columns and display two values under each column.

The Access query currently produces output like this in a GridView:

+-------------+-------+----------+----------+----------+----------+-----------+
| Name        | Total | 4/5/2016 | 4/6/2016 | 4/7/2016 | 4/8/2016 | 4/11/2016 |  
+-------------+-------+----------+----------+----------+----------+-----------+
| Doe, Jane   |  9    | 0/1      | 0/2      | 0/3      | /        | 0/3       |
+-------------+-------+----------+----------+----------+----------+-----------+
| Guy, Some   | 4     | 0/1      | 0/1      | /        | /        | 0/2       | 
+-------------+-------+----------+----------+----------+----------+-----------+
| Doe, John   | 10    | 0/1      | 1/1      | 2/1      | 0/3      | 0/1       |

Sample Data:

| EmployeeID  | Customer         | Date       | OrderType|
+-------------+------------------+------------+----------+
| 1           | Fake Customer    | 2016-05-14 | New      | 
+-------------+------------------+------------+----------+
| 2           | Some Company     | 2016-05-13 | Change   | 
+-------------+------------------+------------+----------+
| 3           | Stuff Inc.       | 2016-05-14 | New      | 
+-------------+------------------+------------+----------+
| 3           | Cool Things      | 2016-05-12 | Change   |

Answer 1

You'd need to dynamically produce your pivot columns then do case statements for each of them. The following is an example of how you could do it:

IF OBJECT_ID('tmpEmployees_Test', 'U') IS NOT NULL DROP TABLE tmpEmployees_Test;
CREATE TABLE tmpEmployees_Test (EmployeeID INT, EmployeeName VARCHAR(255));
INSERT tmpEmployees_Test (EmployeeID, EmployeeName)
VALUES (1, 'Doe, Jane'), (2, 'Doe, John'), (3, 'Guy, Some');

IF OBJECT_ID('tmpOrders_Test', 'U') IS NOT NULL DROP TABLE tmpOrders_Test;
CREATE TABLE tmpOrders_Test (EmployeeID INT, Customer VARCHAR(255), Date DATE, OrderType VARCHAR(255));
INSERT tmpOrders_Test (EmployeeID, Customer, Date, OrderType)
VALUES (1, 'Fake Customer', '2016-05-14', 'New')
        , (2, 'Some Company', '2016-05-13', 'Change')
        , (3, 'Stuff Inc.', '2016-05-14', 'New')
        , (3, 'Cool Things', '2016-05-12', 'Change')
        , (3, 'Amazing Things', '2016-05-12', 'Change');

DECLARE @startDate DATE = '2016-05-14', @cols VARCHAR(MAX) = '', @cols2 VARCHAR(MAX) = '';

SELECT @cols += ', CONVERT(VARCHAR(255), SUM(CASE WHEN O.Date = ''' + CONVERT(VARCHAR(255), DATEADD(DD, X.Y, @startDate)) + ''' AND O.OrderType = ''New'' THEN 1 ELSE 0 END)) + ''/'' + CONVERT(VARCHAR(255), SUM(CASE WHEN O.Date = ''' + CONVERT(VARCHAR(255), DATEADD(DD, X.Y, @startDate)) + ''' AND O.OrderType = ''Change'' THEN 1 ELSE 0 END)) ' + QUOTENAME(CONVERT(VARCHAR(255), DATEADD(DD, X.Y, @startDate), 103)) + CHAR(10) + CHAR(9) + CHAR(9)
        , @cols2 += ', CASE WHEN ' + QUOTENAME(CONVERT(VARCHAR(255), DATEADD(DD, X.Y, @startDate), 103)) + ' = ''0/0'' THEN ''/'' ELSE ' + QUOTENAME(CONVERT(VARCHAR(255), DATEADD(DD, X.Y, @startDate), 103))  + ' END ' + QUOTENAME(CONVERT(VARCHAR(255), DATEADD(DD, X.Y, @startDate), 103)) + CHAR(10) + CHAR(9)
FROM (VALUES (0),(-1),(-2),(-3),(-4),(-5),(-6)) X(Y)
JOIN tmpOrders_Test O ON O.Date = DATEADD(DD, X.Y, @startDate)
GROUP BY X.Y
ORDER BY X.Y;

DECLARE @SQL VARCHAR(MAX) = '
WITH T AS (
    SELECT E.EmployeeID
        , COUNT(*) Total
        ' + @cols + '
    FROM tmpEmployees_Test E
    JOIN tmpOrders_Test O ON O.EmployeeID = E.EmployeeID
    WHERE O.Date BETWEEN ''' + CONVERT(VARCHAR(255), DATEADD(dd, -6, @startDate)) + ''' AND ''' + CONVERT(VARCHAR(255), @startDate) + '''
    GROUP BY E.EmployeeID)
SELECT E.EmployeeName
    , Total
    ' + @cols2 + '
FROM T
JOIN tmpEmployees_Test E ON E.EmployeeID = T.EmployeeID;'

--PRINT @SQL;
EXEC(@SQL);

This mirrors the output you're expecting (as far as I can tell), even if it looks a bit messy. I don't think you can produce your desired output without it being a bit messy, though.

Note: The CTE in the dynamic SQL is just to get rid of all the '0/0' and make them '/' and it seems the easiest way to do it.

Answer 2

Currently, your crosstab is not straightforward for easy pivot translation as you have two conditional aggregates which are then concatenated into a string value. Plus, you have an aggregate group count.

In TSQL, this query will be a little intricate. Consider the below approach using two derived table pivot aggregate queries joined together with use of CAST .

Source (temp table -thanks to @ZLK)

CREATE TABLE #tmpOrders (EmployeeID INT, Customer VARCHAR(255), 
                         Date DATE, OrderType VARCHAR(255));
INSERT #tmpOrders (EmployeeID, Customer, Date, OrderType)
VALUES (1, 'Fake Customer', '2016-05-14', 'New'),
       (2, 'Some Company', '2016-05-13', 'Change'),
       (3, 'Stuff Inc.', '2016-05-14', 'New'),
       (3, 'Cool Things', '2016-05-12', 'Change'),
       (3, 'Amazing Things', '2016-05-12', 'Change');

Query (inner joins of two pivots)

SELECT pvt1.Name, pvt1.Total,
       CAST(pvt1.[2016-05-12] AS VARCHAR(2)) + '/' 
            +  CAST(pvt2.[2016-05-12] AS VARCHAR(2)) As [05/12/2016], 
       CAST(pvt1.[2016-05-13] AS VARCHAR(2)) + '/' 
            +  CAST(pvt2.[2016-05-13] AS VARCHAR(2)) As [05/13/2016], 
       CAST(pvt1.[2016-05-14] AS VARCHAR(2)) + '/' 
            +  CAST(pvt2.[2016-05-14] AS VARCHAR(2)) As [05/14/2016]        
FROM     
    (SELECT t.Customer as Name, t.[Date],
            COUNT(*) As Total, 
            SUM(CASE WHEN t.OrderType = 'New' THEN 1 ELSE NULL END) AS NewOrders

      FROM #tmpOrders t
      WHERE (t.[Date] > GETDATE()-7)    
      GROUP BY t.Customer, t.[Date]) AS src1     
      PIVOT 
          (
            COUNT(src1.[NewOrders]) 
            FOR src1.[Date] IN ([2016-05-12], [2016-05-13], [2016-05-14])
          ) AS pvt1

INNER JOIN 
    (SELECT t.Customer as Name, t.[Date],
            COUNT(*) As Total, 
            SUM(CASE WHEN t.OrderType = 'Change' THEN 1 ELSE NULL END) AS ChangeOrders
      FROM #tmpOrders t
      WHERE (t.[Date] > GETDATE()-7)    
      GROUP BY t.Customer, t.[Date]) AS src1     
      PIVOT 
          (
            COUNT(src1.[ChangeOrders]) 
            FOR src1.[Date] IN ([2016-05-12], [2016-05-13], [2016-05-14])
          ) AS pvt2        
ON pvt1.Name = pvt2.Name

Output

--Name          Total   05/12/2016  05/13/2016  05/14/2016
--Amazing Things    1         0/1          0/0         0/0
--Cool Things       1         0/1          0/0         0/0
--Fake Customer     1         0/0          0/0         1/0
--Some Company      1         0/0          0/1         0/0
--Stuff Inc.        1         0/0          0/0         1/0

Do note: one thing you lose in translation is the dynamic quality of Access' crosstab query which will output all existing values (up until the 255 column limit). In SQL Server, for regular pivots you need to declare such values in advance or use stored procs/functions to return all values dynamically.

Answer 3

IF OBJECT_ID('tmpEmployees_Test', 'U') IS NOT NULL DROP TABLE tmpEmployees_Test;
CREATE TABLE tmpEmployees_Test (EmployeeID INT, EmployeeName VARCHAR(255));
INSERT tmpEmployees_Test (EmployeeID, EmployeeName)
VALUES (1, 'Doe, Jane'), (2, 'Doe, John'), (3, 'Guy, Some');


IF OBJECT_ID('tmpOrders_Test', 'U') IS NOT NULL DROP TABLE tmpOrders_Test;
CREATE TABLE tmpOrders_Test (EmployeeID INT, Customer VARCHAR(255), Date DATE, OrderType VARCHAR(255));
INSERT tmpOrders_Test (EmployeeID, Customer, Date, OrderType)
VALUES (1, 'Fake Customer', '2016-05-14', 'New')
        , (2, 'Some Company', '2016-05-13', 'Change')
        , (3, 'Stuff Inc.', '2016-05-14', 'New')
        , (3, 'Cool Things', '2016-05-12', 'Change')
        , (3, 'Amazing Things', '2016-05-12', 'Change');


DECLARE @columns NVARCHAR(MAX), @sql NVARCHAR(MAX);
SET @columns = N'';
SELECT @columns += N', p.' + QUOTENAME(Name) 
  FROM (SELECT distinct CONVERT(nvarchar(30) , p.Date , 101) as Name FROM dbo.tmpOrders_Test AS p where [Date] > GETDATE()-7
  ) AS x;
-- Kept it for formatting Purpose
DECLARE @columns1 NVARCHAR(MAX)
SET @columns1 = N'';
SELECT @columns1 += N', ISNULL(p.' + QUOTENAME(Name) + ',''/'') AS ' + QUOTENAME(Name) 
  FROM (SELECT distinct CONVERT(nvarchar(30) , p.Date , 101) as Name FROM dbo.tmpOrders_Test AS p where [Date] > GETDATE()-7
  ) AS x;


SET @sql = N'
SELECT EmployeeName, Count(*) as Total  ' +  @columns1  + '
FROM
(
    SELECT  EmployeeID, EmployeeName' + ''+ @columns1 + '' + '
    FROM
    (
      SELECT    o.employeeID,EmployeeName, CAST(COUNT(case WHEN OrderType = ''New'' then 1  end) as varchar(5)) + ''/'' + 
                CAST(COUNT(case WHEN OrderType = ''Change'' then 1  end) as varchar(5)) as OrderType, CONVERT(nvarchar(30) , p.Date , 101) as Date
       FROM     dbo.tmpOrders_Test AS p
                    INNER JOIN dbo.tmpEmployees_Test AS o
                     ON p.EmployeeID = o.EmployeeID
       GROUP BY EmployeeName, Date, o.employeeID
    ) AS j
    PIVOT
    (
      Max(OrderType) FOR Date IN ('
      + STUFF(REPLACE(@columns, ', p.[', ',['), 1, 1, '')
      + ')
    ) AS p) as p JOIN tmpOrders_Test as m on p.employeeID = m.employeeID
where [Date] > GETDATE()-7
GROUP BY EmployeeName ' + @columns + '
';

PRINT @sql;
EXEC sp_executesql @sql;

This one is using dynamic Pivot. You might want to do this business logic on Application or Reporting Side instead of complex sql.