Reputation: 759
I have table in database with XML column. Now I need to select some rows by two attributes from XML. So far I've come up with this:
SELECT o.Id
FROM Objects o
WHERE o.SerializedObject.value('(/object/param[@id="111"]/@value)[1]', 'varchar(8)') = '-1'
AND o.SerializedObject.value('(/object/param[@id="222"]/@value)[1]', 'varchar(8)') = '8'
EDIT:
XML is like:
<object>
<param id="1" value="111"/>
<param id="2" value="222"/>
...
<param id="200" value="4545"/>
<object>
Each object has ~2k params.
I'm wondering if there is a better way to do that with single XML query.
Upvotes: 0
Views: 1325
Reputation: 67291
This depends on your XML (you did not show an example, but I assume this is kind of EAV).
You can try using XML's method .exist()
:
DECLARE @mockup TABLE(ID INT IDENTITY,Comment VARCHAR(100),SerializedObject XML);
INSERT INTO @mockup VALUES
('just one of them','<object><param id="111" value="-1"/></object>')
,('both, but wrong values','<object><param id="111" value="-1"/><param id="222" value="-1"/></object>')
,('both, should fit','<object><param id="111" value="-1"/><param id="222" value="8"/></object>')
SELECT o.Id,o.Comment,o.SerializedObject
FROM @mockup o
WHERE o.SerializedObject.exist('/object[param[@id="111" and @value="-1"] and param[@id="222" and @value="8"]]')=1;
.exist()
is the fastest here, because it does not return any value. It will just return 1
on the first occurance found. This is especially fast, when there are many occurances of a <param id="111" value="???">
Otherwise you'd have to shred the whole lot and place the filter on the whole resultset.
And - of course! - the necessary hint: As told in a comment by Jeroen Mostert dealing with bigger XMLs might turn out as a bottle neck. If you need this more often, you might think about a relational design instead of big XMLs...
Upvotes: 1