Question

I have a parent table (Person) with a child table (Address). I need to send all person and address combinations to a service to be processed, but I am restricted in that I can only send up to 10,000 at a time. When I send data for a person I need to make sure I send all of the data for that person, so they can’t be split across batches.

I want to be able to pull the data using SQL only, no additional code (this is to be used in an SSIS package).

Example: I want to pull only max of 5 from the data below

---Person---  ---------Address----------
ID  Name      AddressID    Address
--- --------  ------------ -------------
1   Bill      1            1 The Street
1   Bill      2            3 The Road
2   Jeff      3            7 The Hill
3   Mike      4            9 The Avenue
4   Ross      5            11 Main Road     --- This is the 5th row, must be ignored
4   Ross      6            12 Down Under    --- This record would be orphaned if 
                                                we pulled top 5

Answer 1

I found this seemed to be the best way to do it:

WITH data AS
(
      select TOP 10001 ROW_NUMBER() OVER (ORDER BY p.PersonID) as [Rank], 
             p.PersonID, a.AddressID
      from Input.Person p
      JOIN Input.Address a on p.PersonID = a.PersonID
)

select d1.*
FROM data d1
LEFT JOIN data d2 on d1.PersonID = d2.PersonID
      AND d2.[RANK] > 10000 
WHERE d1.[RANK] <= 10000
AND d2.PersonID IS NULL
ORDER BY 1

---

The TOP 10001 allows me to limit the data searched as if the 10000th and 10001st record share a parent, then the 10000th should be ignored.