Problems with splitting result into different rows. Group by-issue?

Question

Hi all!
I have a table that looks a bit like this:
Pallet          Status          Number          Time
A          MoveIn          48          11:11
A          Pick          -1          11:11
A          Pick          -1          11:12
A          Pick          -1          11:12
A          MoveOut     45          11:13
A          MoveIn          45          18:20
A          Pick          -1          18:23
A          Pick          -1          18:23
A          Pick          -1          18:23
A          Pick          -1          18:24
A          MoveOut     41          18:25
A          MoveIn          41          21:31
A          Pick          -1          21:55
.....This is logging pallets that is moved to one station and then number of boxes are picked off. (For making it more simple, I only show one pallet in above example.)
I would like to present the result with one row for each time the pallet have been at the station, like:
Pallet          Picked          Time
A          3          11:11
A          5          18:20
A          ....          .....This means that the same pallet might be showned on several rows, as it has been to the station several times. I have tried several different "GROUP BY" but I cannot make it work.
For example I have tried to group by hour, but this fails if a pallet is moved twice per hour or if boxes are picket at different hours (eg, first box picked 09:59, last box 10:01)
Does anyone have a suggestion how to solve this?
Thanks in advance and Kind regards / Anders from Sweden

Answer 1

Hej, Anders,
Can you explain what defines a group?
It looks like you're saying that a group is a sub-set of a pallet. If we arrange all the rows for a pallet in order by time, then a new group will start every time we have a row where status='MoveIn'. That is, a group consists of a row with status='MoveIn', and all the rows immediately following it, in order by time, up to, but not including, the next row with status=-'MoveIn'. Is that right?
If so, you can use the analytic COUNT (or SUM) function to find out how many rows with status='MoveIn' have already occurred, in order by time, like this:
WITH     got_grp          AS
     SELECT     pallet, status, time
     ,     COUNT ( CASE
                         WHEN  status = 'MoveIn'
                   THEN  1
                     END
                ) OVER ( PARTITION BY  pallet
                            ORDER BY          time
                    )           AS grp
     FROM    table_x
SELECT       pallet
,       COUNT ( CASE
                  WHEN  status = 'Pick'
                THEN  1
              END
            )          AS picked
,       MIN (time)          AS start_time
FROM       got_grp
GROUP BY  pallet
,            grp
ORDER BY  pallet
,            grp
;If you'd care to post CREATE TABLE and INSERT statements for your sample data, then I could test this.
This assumes that the combination (pallet, time) is unique. Your sample data includes:
Pallet          Status          Number          Time
A          MoveIn          48          11:11
A          Pick          -1          11:11I assume that you're just not displaying the full time, and your data is really something like:
Pallet          Status          Number          Time
A          MoveIn          48          23-Sep-2011 11:11:01
A          Pick          -1          23-Sep-2011 11:11:08Otherwise, how do you know if these two rows are in te same group or not? You may need to add a tie-breaker to the analytic ORDER BY clause. For example, if you say "ORDER BY time, *status* ", then, if the two rows above really did have exactly the same time, down to the second, then the one with status='MoveIn' would be considered to be earlier than the row with status='Pick'.