I have an XML hierarchy like this in an XML type column of a table with 10,000 records-
<Root>
<Elem1>
<Parent1>
<Separator>
<Child1/>
</Separator>
</Parent1>
</Elem1>
</Root>
I have a query like this -
DECLARE @Root VARCHAR(50)
DECLARE @Entity VARCHAR(50)
DECLARE @ParentNode VARCHAR(50)
DECLARE @Separator VARCHAR(50)
DECLARE @ChildNode VARCHAR(50)
SET @Root = 'Root'
SET @Entity = 'Elem1'
SET @ParentNode = 'Parent1'
SET @Separator = 'separator'
SET @ChildNode = 'Child1'
select Parent.P.value('.', 'varchar(max)') as MyValue,
T.uniqueId, T.XMLCol
from [XMLTable] as T
cross apply
(SELECT
XMLTable.XMLCol.query('(/*[local-name()=sql:variable("@Root")]/*[local-name(.)=sql:variable("@Entity")]/*[local-name(.)=sql:variable("@ParentNode")]/*[local-name(.)=sql:variable("@Separator")]/*[local-name(.)=sql:variable("@ChildNode")])[1]'
) as Parent(P)
How can I further Optimize this query. Currently, it is taking 2 secs and if I do further INNER JOINS, it adds up to the time. I tried creating a PRIMARY Index on the XML Column, but it takes more time !
EDIT- If I hardcode the path instead of using variables, then it takes less than a second. But, I want it in a Table-Valued function and cannot hardcode path ?
EDIT - Solution
select x.value('(Parent1/Separator1/Child1)[1]', 'varchar(max)') as Col1,
x.value('(Parent2/Separator2/Child2)[1]', 'varchar(max)') as Col2,
x.value('(Parent3)[1]', 'varchar(max)') as Col3
from [XMLTable] T
cross apply T.XMLCOL.nodes('/Root/Elem1') a(x)
The above query takes just about one sec. It seems to be the fastest of all. So, instead of going for a Table valued function with params, the above query can be dynamically prepared and executed on the fly using ADO.NET capabilities
Correct me please ... ?
If you only want one value out of every row there is no need to use
cross apply
.Another way to get the same is to use FLWOR. In my limited tests this will run a bit faster.