Question

I have a SQL Server table AccountAction which is denormalised. It is a flattened version of the Account and Action tables, which I’m hoping should be a lot quicker for reporting queries over millions of rows. One Account can have many Actions, so the table looks similar to:

Account     Action
account1    action1
account1    action2
account1    action10
account2    action5

However I’m having some trouble getting the information back for a restricted subset in a simple stored procedure.

select Account, Action
from AccountAction
where ???

What I’m looking for is to get the first X accounts, with all their actions. So this will be a dynamic number of rows. So using the example table above if I passed in 1, I would get 3 rows (i.e. give me all rows for the first account).

(I don’t mind that the account name will be in each row – it is pivoted elsewhere)

Do I need to use a ROWNUM or similar to restrict the rows? I’m sure this must be a simpler issue than I’ve found so far.

EDIT

The answers using TOP won’t work, in the example I’d be wanting 3 rows returned if I said ‘give me one (the first) account’. But how do I know there will be 3? Its dynamic. Also they may not be sequential, what if account1’s action99 was at position 55 million in the results.

Answer 1

WITH
  SequencedData
AS
(
  SELECT
    DENSE_RANK() OVER (ORDER BY Account) AS account_sequence_id,
    *
  FROM
    AccountAction
)
SELECT
  *
FROM
  SequenceData
WHERE
  account_sequence_id = ???

Or, for multiples…

WHERE
  account_sequence_id BETWEEN 3 AND 5    -- For the 3rd, 4th and 5th accounts.