sql server - Combine single ids into range

Question

I have a table that contains lots of integers. This table gets queried and the results end up being turned into xml. If the table contains for example the following items:

SELECT itemId FROM items WHERE enabled = true
1
2
3
5

The my final xml output after some processing would be:

 <item id="1" />
 <item id="2" />
 <item id="3" />
 <item id="5" />

The xml ends up being fairly large and alot of the items are actually ranges. What I would like to do is update my query to combine ranges (alot of these items are 'neighbours' so the xml generated would be quite a bit smaller). I'm trying to get the procedures results to be more like this:

1-3
5

So that the final XML looks something like this (if I can just change the procedure, the XML processing can stay the same):

<item id="1-3"/>
<item id="5"/>

I was thinking my best route may be to use a self join where table1.itemId = table2.itemId - 1 but I haven't been able to get it working. Does anyone have any suggestions on how I can go about this?

Answer 1

Would this help?

SELECT
     MIN(ItemID)
    ,MAX(ItemID)
FROM
    (
        SELECT ItemID, RANK() OVER (ORDER BY ItemID) R FROM Items
    ) Tmp
GROUP BY
    ItemID - R

Answer 2

I'd think this should do the trick: 1) order by itemID 2) use OVER...PARTITION to get row number 3) use it in a recursive Common Table Expression that joins a number to all others where anchor + row number equals the ItemID, thereby finding all sequential numbers 4) group by the anchor in an outer query and then use MIN and MAX to get the range.