Question

I have a table (time_period)

The table lists several time periods like so.

unique_id         type_id          sequence_number     billing_id   ...
    13              LJK1             1                  4356
    14              CNS3             2                  4356
    15              LJK1             3                  4356
    16              AOX4             4                  4356
    17              CNS3             1                  9827
    18              POD2             2                  9827

It gets tricky. I need to return all the type_id pairs from one period to the next period based on incrementing sequence numbers. For example, for billing_id=’4356′, I need to retrieve this resultset

-----------------------------------
LJK1       | CNS3                 -- type_ids pair from seq 1 to seq 2
CNS3       | LJK1                 -- type_ids pair from seq 2 to seq 3
LJK1       | AOX4                 -- type_ids pair from seq 3 to seq 4

The count of sequence_numbers per order is variable (as high as 10, as low as 2). I’ll be doing this for X number of orders. From this, I’ll be doing further processing, but getting these pairings itself is puzzling me. I don’t need order_id or sequence numbers in the resultset.

Are recursive CTEs needed here? Perhaps cursors? Or maybe using pivot? I’ve just been brainstorming and researching, but I’m not too experienced with any of those and could use some direction.

Using SQL Server 2008

Answer 1

I don’t think you need a recursive CTE nor cursors. Try this:

    SELECT A.type_id, B.type_id
    FROM time_period A
    LEFT JOIN time_period B
    ON A.billing_id = B.billing_id
    AND A.sequence_number = B.sequence_number - 1
    WHERE A.billing_id = 4356