SQL Server 2012 - A Little Guidance - sql

Question

I have searched the net but I am certain I must not be phrasing my keywords correctly because I am not finding possible solutions for my problem. think it might be recursion but I'm not quite certain.
I have a table that has the following categories:
ID, Author, Customer, Group
A sample dataset would be like:
ID | Author | Customer | Group
------------------------------------------
1 | Paula Hawkins | John Doe | NULL
2 | Harlan Coben | John Doe | NULL
3 | James Patterson| John Doe | NULL
4 | Paula Hawkins | Jane Doe | NULL
5 | James Patterson| Jane Doe | NULL
6 | James Patterson| Steven Doe| NULL
7 | Harlan Coben | Steven Doe| NULL
8 | Paula Hawkins | Harry Doe | NULL
9 | James Patterson| Harry Doe | NULL
Its possible a customer may have one ore more then one author checked out so what I am trying to do is group them with a unique id based on what total are checked out (regardless of the customer name):
ID | Author | Customer | Group
--------------------------------------------
1 | Paula Hawkins | John Doe | 1
2 | Harlan Coben | John Doe | 1
3 | James Patterson| John Doe | 1
4 | Paula Hawkins | Jane Doe | 2
5 | James Patterson| Jane Doe | 2
6 | James Patterson| Steven Doe | 3
7 | Harlan Coben | Steven Doe | 3
8 | Paula Hawkins | Harry Doe | 2
9 | James Patterson| Harry Doe | 2
its very possible the same customer could be found hundreds of times for multiple books so the final group category would represent the unique value for that customer (other customers would have the same value only if everything they have checked out also matches everything the other customer has checked out).
Using the above data, Harry and Jane have the exact same authors checked out so they are in the same group but John and Steven have different combinations so they have their own unique group.
Hopefully this makes sense. Is this what is called recursion? If so then I will look towards a cte solution that uses some sort of ranking for the unique id value. Thanks for any help you give.

Answer 1

Not sure how to get your exact group order, but to just group customers together you can combine their authors with FOR XML and group the customers based on exact matches.
WITH cte AS (
SELECT
*,
RANK() OVER (ORDER BY Authors) [Group]
FROM (
SELECT
[Customer],
STUFF((SELECT ',' + [Author]
FROM myTable WHERE Customer = mt.Customer
ORDER BY Author
FOR XML PATH('')), 1, 1, '') AS Authors
FROM
myTable mt
GROUP BY [Customer] ) t
)
SELECT
mt.[ID],
mt.[Author],
mt.[Customer],
cte.[Group]
FROM
cte
JOIN myTable mt ON mt.Customer = cte.Customer
ORDER BY mt.[ID]
SQL FIDDLE DEMO

Answer 2

Try using cursors... Cursors are slow, but they're also easier to understand..
Here's a sample implementation...
DECLARE #GroupExists Bit
DECLARE #CurrGroup Int
DECLARE #NextGroup Int
DECLARE #Customer VARCHAR(250)
SET #NextGroup = 1
DECLARE customer_cursor CURSOR FAST_FORWARD
FOR SELECT distinct Customer FROM dbo.TableName
OPEN customer_cursor
FETCH NEXT FROM customer_cursor
INTO #Customer
WHILE ##FETCH_STATUS = 0
BEGIN
SET #GroupExists = 0
--Test condition to check if group of authors in in use
IF #GroupExists = 1 Then
BEGIN
UPDATE dbo.TableName
SET Group = #CurrGroup
WHERE Customer = #Customer
END
ELSE
BEGIN
UPDATE dbo.TableName
SET Group = #NextGroup
WHERE Customer = #Customer
SET #NextGroup= #NextGroup+ 1
END
FETCH NEXT FROM customer_cursor
INTO #Customer
END

Answer 3

You should be able to generate groups using standard SQL. The following query should do the job; I make no promises of its performance though.
WITH
CTE_CheckOutBookCount AS
(
SELECT [ID]
,[Author]
,[Customer]
,COUNT([Author]) OVER (PARTITION BY [Customer]) AS [CheckOutBooks] -- Count the number of books checked out by each customer. This will be used for our initial compare between customers.
FROM CheckedOutBooks
),
CTE_AuthorAndCountCompare AS
(
SELECT CB.[ID]
,CBC.[Customer] AS MatchedCustomers
FROM CTE_CheckOutBookCount CB
INNER JOIN CTE_CheckOutBookCount CBC ON CB.[Author] = CBC.[Author] AND CB.[CheckOutBooks] = CBC.[CheckOutBooks] --Join customer information on number of books checked out and author name of books checked out.
)
,CTE_MatchedCustomers
AS
(
SELECT
[ID]
,[Author]
,[Customer]
--Get the minimum record id of customers which match exactly on count and authors checked out. This will be used to help generate group ID.
,(
SELECT MIN(ID)
FROM CTE_AuthorAndCountCompare
WHERE CheckedOutBooks.[Customer] = CTE_AuthorAndCountCompare.MatchedCustomers
) MinID
FROM CheckedOutBooks
)
SELECT
[ID]
,[Author]
,[Customer]
,DENSE_RANK() OVER (ORDER BY MinID) AS [Group] -- Generate new group id
FROM CTE_MatchedCustomers
ORDER BY ID