A better way to get current and previous status?

Question

I'm working on a system where status messages are read into a table for items of equipment. My requirement is to find all the messages arriving within a particular time window, together with the status value and the status value for the previous message sent for that equipment item, if any. This previous message can (and probably will) have a date outside the window I'm looking at, or might not exist at all for a brand new equipment item.
The table looks something like this:
create table eq_state
(eq_id      NUMBER,   -- ID of the equipment
event_date DATE,     -- Date/Time of the message
status     NUMBER)   -- Status passed in the messageThere are indexes on (event_date,eq_id) and on eq_id. It'll get very big, so it's important that indexes are used wherever possible.
Here's what I have so far:
WITH eq_state AS
SELECT 1 eq_id, TO_DATE('01/05/2013 12:01:00','DD/MM/YYYY HH:MI:SS') event_date, 1 status FROM DUAL
UNION ALL
SELECT 2 eq_id, TO_DATE('01/05/2013 01:01:00','DD/MM/YYYY HH:MI:SS') event_date, 2 status FROM DUAL
UNION ALL
SELECT 2 eq_id, TO_DATE('01/05/2013 12:02:00','DD/MM/YYYY HH:MI:SS') event_date, 3 status FROM DUAL
UNION ALL
SELECT 3 eq_id, TO_DATE('01/05/2013 10:00:00','DD/MM/YYYY HH:MI:SS') event_date, 4 status FROM DUAL
UNION ALL
SELECT 3 eq_id, TO_DATE('01/05/2013 12:03:00','DD/MM/YYYY HH:MI:SS') event_date, 5 status FROM DUAL
UNION ALL
SELECT 3 eq_id, TO_DATE('01/05/2013 12:04:00','DD/MM/YYYY HH:MI:SS') event_date, 6 status FROM DUAL
SELECT eq_id,TO_CHAR(event_date,'DD/MM/YYYY HH:MI:SS') event_date,curr_state,prev_state
FROM  (SELECT Es.Eq_Id,
              Es.Event_Date,
              Es.Status AS Curr_State,
              Es2.Status AS Prev_State,
              RANK() OVER (PARTITION BY es.eq_id,es.event_date ORDER BY es2.event_date DESC) the_rank
       FROM   Eq_State Es
       LEFT OUTER JOIN Eq_State Es2
       ON     Es2.Eq_Id = Es.Eq_Id
       AND    Es2.Event_Date < Es.Event_Date
       WHERE  Es.Event_Date BETWEEN TO_DATE('01/05/2013 12:00:00','DD/MM/YYYY HH:MI:SS')
                                AND TO_DATE('01/05/2013 12:05:00','DD/MM/YYYY HH:MI:SS')) x
WHERE  the_rank = 1
ORDER BY 1,2This gives me the desired results and seems to use sensible indexes (I'll spare you the explain plan):
     EQ_ID      EVENT_DATE          CURR_STATE PREV_STATE
         1 01/05/2013 12:01:00          1          
         2 01/05/2013 12:02:00          3          2
         3 01/05/2013 12:03:00          5          4
         3 01/05/2013 12:04:00          6          5So what's the problem? Just curiosity really - the technique of finding the RANK() in the inline view, and then selecting rows where it's equal to 1 seems like a bit of a kludge. Is there a more elegant way to do it? Maybe something using LAG() or KEEP DENSE_RANK LAST?
I'm still not really up-to-speed with analytic functions, so I'm hoping to learn something from those of you that are!
ORACLE 11.2.0.1.0 by the way

Answer 1

I would use another way, firstly sorting by any unique criteria and using joins
WITH eq_state AS
SELECT 1 eq_id, TO_DATE('01/05/2013 12:01:00','DD/MM/YYYY HH:MI:SS') event_date, 1 status FROM DUAL
UNION ALL
SELECT 2 eq_id, TO_DATE('01/05/2013 01:01:00','DD/MM/YYYY HH:MI:SS') event_date, 2 status FROM DUAL
UNION ALL
SELECT 2 eq_id, TO_DATE('01/05/2013 12:02:00','DD/MM/YYYY HH:MI:SS') event_date, 3 status FROM DUAL
UNION ALL
SELECT 3 eq_id, TO_DATE('01/05/2013 10:00:00','DD/MM/YYYY HH:MI:SS') event_date, 4 status FROM DUAL
UNION ALL
SELECT 3 eq_id, TO_DATE('01/05/2013 12:03:00','DD/MM/YYYY HH:MI:SS') event_date, 5 status FROM DUAL
UNION ALL
SELECT 3 eq_id, TO_DATE('01/05/2013 12:04:00','DD/MM/YYYY HH:MI:SS') event_date, 6 status FROM DUAL
), sorted as (
  select row_number() over (order by a.event_date) as rn, a.* from eq_state a
select * from sorted s1
left join sorted s2 on (s2.rn = s1.rn -1)
order by s1.rn
        RN      EQ_ID EVENT_DATE                STATUS         RN      EQ_ID EVENT_DATE                STATUS
         1          2 01.05.2013 01:01:00            2                                                       
         2          3 01.05.2013 10:00:00            4          1          2 01.05.2013 01:01:00            2
         3          1 01.05.2013 12:01:00            1          2          3 01.05.2013 10:00:00            4
         4          2 01.05.2013 12:02:00            3          3          1 01.05.2013 12:01:00            1
         5          3 01.05.2013 12:03:00            5          4          2 01.05.2013 12:02:00            3
         6          3 01.05.2013 12:04:00            6          5          3 01.05.2013 12:03:00            5