Populating a distinct list from repetitive data in SQL Server

Question

I need to gather a list of distinct employees from an XML file that contains a log of sales made by each employee. Unfortunately, the data in the XML file isn't exactly "consistent". The file is structured like so:

<Sale EmployeeId="67890" EmployeeName="John" EmployeeManagerId="12345" 
      CustomerName="Bob" SaleNumber="..." />
<Sale EmployeeId="67890" EmployeeName="John" EmployeeManagerId="12345" 
      CustomerName="Pat" SaleNumber="..." />
<Sale EmployeeId="67890" EmployeeName=""     EmployeeManagerId="12345" 
      CustomerName="Sally" SaleNumber="..." />
<Sale EmployeeId="67890" EmployeeName=""     EmployeeManagerId="12345" 
      CustomerName="Sue" SaleNumber="..." />
<Sale EmployeeId="67890" EmployeeName="John" EmployeeManagerId=""      
      CustomerName="Jack" SaleNumber="..." />
<Sale EmployeeId="58203" EmployeeName="Fred" EmployeeManagerId=""      
      CustomerName="Bill" SaleNumber="..." />

This XML file is uploaded to a web application, which passes its contents (as XML) to a stored procedure in SQL Server for processing. Because of the size of this file (up to 30,000 elements), I would like to do as little processing in the web application as possible.

The best solution I have come up with so far is to create a temporary table with one row for each distinct EmployeeId and ManagerId value. Then, for each row in the table, loop through the XML elements that have a matching EmployeeId until I find an entry where the name is not null (then repeat for ManagerId).

So, for each unique employee ID, I would be iterating over the results twice to see if I can find their name and manager's ID.

Once the file is processed, I would expect the Employee table to look like this:

+---------+------+------------+
| Id (PK) | Name | ManagerId  |
+---------+------+------------+
| 12345   | NULL | NULL       |
| 67890   | John | 12345      |
| 58203   | Fred | NULL       |
+---------+------+------------+

Is there a more efficient (and less procedural) solution for this?

Answer 1

This gets the results, but may require some cleanup work if the sample data is different.

DECLARE @T TABLE ( x XML )
INSERT  INTO @T
        ( x )
VALUES  ( '<Sale EmployeeId="67890" EmployeeName="John" EmployeeManagerId="12345"        CustomerName="Bob" SaleNumber="..." />' )
    ,   ( '<Sale EmployeeId="67890" EmployeeName="John" EmployeeManagerId="12345"        CustomerName="Pat" SaleNumber="..." />' ),
        ( '<Sale EmployeeId="67890" EmployeeName=""     EmployeeManagerId="12345"        CustomerName="Sally" SaleNumber="..." />' )
     ,  ( '<Sale EmployeeId="67890" EmployeeName=""     EmployeeManagerId="12345"        CustomerName="Sue" SaleNumber="..." />' ),
        ( '<Sale EmployeeId="67890" EmployeeName="John" EmployeeManagerId=""             CustomerName="Jack" SaleNumber="..." />' ),
        ( '<Sale EmployeeId="58203" EmployeeName="Fred" EmployeeManagerId=""             CustomerName="Bill" SaleNumber="..." />' ) 

;WITH c 
AS (

SELECT DISTINCT ID = x.value('(/Sale/@EmployeeId)[1]', 'int')
      , NAME = x.value('(/Sale/@EmployeeName)[1]', 'varchar(4)')
      , ManagerID = x.value('(/Sale/@EmployeeManagerId)[1]', 'int')
FROM    @t
WHERE  x.value('(/Sale/@EmployeeName)[1]', 'varchar(4)') <> ''
)

SELECT ID, NAME, ManagerID =MIN( NULLIF(ManagerID, 0))
FROM c 
GROUP BY ID, Name
UNION 
SELECT ManagerID, NULL, NULL
FROM c
WHERE ManagerID NOT IN (SELECT DISTINCT ID FROM c)
    AND ManagerID <> 0

Answer 2

declare @xml xml = '
<Sale EmployeeId="67890" EmployeeName="John" EmployeeManagerId="12345" 
      CustomerName="Bob" SaleNumber="..." />
<Sale EmployeeId="67890" EmployeeName="John" EmployeeManagerId="12345" 
      CustomerName="Pat" SaleNumber="..." />
<Sale EmployeeId="67890" EmployeeName=""     EmployeeManagerId="12345" 
      CustomerName="Sally" SaleNumber="..." />
<Sale EmployeeId="67890" EmployeeName=""     EmployeeManagerId="12345" 
      CustomerName="Sue" SaleNumber="..." />
<Sale EmployeeId="67890" EmployeeName="John" EmployeeManagerId=""      
      CustomerName="Jack" SaleNumber="..." />
<Sale EmployeeId="58203" EmployeeName="Fred" EmployeeManagerId=""      
      CustomerName="Bill" SaleNumber="..." />'

-- "E1 is all employees"
;with E1 as      
(
  select T.N.value('@EmployeeId', 'int') as Id,
         T.N.value('@EmployeeName', 'nvarchar(100)') as Name,
         T.N.value('@EmployeeManagerId', 'int') as ManagerID
  from @xml.nodes('/Sale') as T(N)
),
-- E2 groups on id to get only one emp for each id
E2 as
(
  select Id, max(Name) as Name, nullif(max(ManagerID), 0) as ManagerID
  from E1 
  group by Id
),
-- "All manager id's"
M as
(
  select distinct T.N.value('@EmployeeManagerId', 'int') as Id
  from @xml.nodes('/Sale') as T(N)
  where T.N.value('@EmployeeManagerId', 'int') <> 0       
)
-- "All unique employees"
select Id, Name, ManagerID
from E2
union all
-- "Add managers with a lookup against emp for name and manager id"
select M.Id, E2.Name, E2.ManagerID
from M
  left outer join E2 
    on M.Id = E2.ID