SQL Server : FOR XML PATH - nesting / grouping

Question

I have data that looks like:

OrderID CustomerID  ItemID  ItemName
10000   1234        111111  Product A
10000   1234        222222  Product B
10000   1234        333333  Product C
20000   5678        111111  Product A
20000   5678        222222  Product B
20000   5678        333333  Product C

I want to write a T-SQL query in SQL Server to return the data like this:

<Root>
  <Order>
    <OrderID>10000</OrderID>
    <CustomerID>1234</CustomerID>
    <LineItem>
      <ItemID>11111</ItemId>
      <ItemName>Product A</ItemName>
    </LineItem>
    <LineItem>
      <ItemID>22222</ItemId>
      <ItemName>Product B</ItemName>
    </LineItem>
    <LineItem>
      <ItemID>33333</ItemId>
      <ItemName>Product B</ItemName>
    </LineItem>
  </Order>
  <Order>
    <OrderID>20000</OrderID>
    <CustomerID>5678</CustomerID>
    <LineItem>
      <ItemID>11111</ItemId>
      <ItemName>Product A</ItemName>
    </LineItem>
    <LineItem>
      <ItemID>22222</ItemId>
      <ItemName>Product B</ItemName>
    </LineItem>
    <LineItem>
      <ItemID>33333</ItemId>
      <ItemName>Product B</ItemName>
    </LineItem>
  </Order>
</Root>

I've tried returning the query in XML using:

FOR XML PATH ('Order'), root ('Root')

But that gives me an Order node for each row (6 in total) vs. just an order node for each orderId (2 in total).

Any ideas?

Oft

Answer 1

select  
    OrderID,
    CustomerID,
    (
        select 
        ItemID,
        ItemName
        from @Orders rsLineItem
        where rsLineItem.OrderID = rsOrders.OrderID
        for xml path('LineItem'), type
    )
from (select distinct OrderID, CustomerID from @Orders) rsOrders
FOR XML PATH ('Order'), root ('Root')

Anjaanjali

Thanks Bert. What does 'type' do in the subquery 'for xml path' statement? – Oft

@Oft It means, 'return this as the XML data type.' So in the query above, it's just returning the subquery as a small xml fragment. – Anjaanjali

Answer 2

For completion: here is a solution without subselect, that should perform faster for big tables. Instead it groups the table as many times as there are levels in the XML and identifies the level with GROUPING_ID (see https://technet.microsoft.com/en-us/library/bb522495(v=sql.105).aspx and https://learn.microsoft.com/en-us/sql/relational-databases/xml/use-explicit-mode-with-for-xml) :

with rsOrders as (
  select '10000' OrderID, '1234' CustomerID, '111111' ItemID, 'Product A' ItemName union
  select '10000' orderId, '1234' customerID, '222222' itemID, 'Product B' ItemName union
  select '10000' orderId, '1234' customerID, '333333' itemID, 'Product C' ItemName union
  select '20000' orderId, '5678' customerID, '111111' itemID, 'Product A' ItemName union
  select '20000' orderId, '5678' customerID, '222222' itemID, 'Product B' ItemName union
  select '20000' orderId, '5678' customerID, '333333' itemID, 'Product C' ItemName 
)
select case 
         when GROUPING_ID(ItemID) = 0 then 3 
         when GROUPING_ID(OrderID) = 0 then 2 
         else 1 
         end as tag,
       case 
           when GROUPING_ID(ItemID) = 0 then 2 
         when GROUPING_ID(OrderID) = 0 then 1 
         else null
       end as parent,
       null       as 'Root!1',
       OrderID    as 'Order!2!OrderID!element', 
       CustomerID as 'Order!2!CustomerID!element', 
       ItemID     as 'LineItem!3!ItemID!element', 
       ItemName   as 'LineItem!3!ItemName!element'
  from rsOrders
 group by grouping sets ((), (OrderID, CustomerID), (OrderID, CustomerID, ItemID, ItemName))
 order by OrderID, CustomerID, ItemID, ItemName
   for xml explicit, type

Unremitting