Collapsing rows - date/time ranges

Question

Hi,
I need to collapse date/time ranges (1 range per row) into the smallest continuous block for a particular item, instance (date/time).
For example,
CREATE TABLE "BLOCK_TEST"
   "ID" NUMBER,
   "START_BLK" DATE,
   "STOP_BLK" DATE
alter session set nls_date_format = "YYYY-MM-DD HH24:MI";
INSERT INTO BLOCK_TEST ("ID","START_BLK","STOP_BLK") VALUES (1, TO_DATE('2006-07-03 01:00'), TO_DATE('2006-07-03 01:30'));
INSERT INTO BLOCK_TEST ("ID","START_BLK","STOP_BLK") VALUES (1, TO_DATE('2006-07-03 01:30'), TO_DATE('2006-07-03 02:00'));
INSERT INTO BLOCK_TEST ("ID","START_BLK","STOP_BLK") VALUES (1, TO_DATE('2006-07-03 02:00'), TO_DATE('2006-07-03 02:30'));
INSERT INTO BLOCK_TEST ("ID","START_BLK","STOP_BLK") VALUES (1, TO_DATE('2006-07-03 02:30'), TO_DATE('2006-07-03 03:00'));
INSERT INTO BLOCK_TEST ("ID","START_BLK","STOP_BLK") VALUES (1, TO_DATE('2006-07-03 03:30'), TO_DATE('2006-07-03 04:00'));
INSERT INTO BLOCK_TEST ("ID","START_BLK","STOP_BLK") VALUES (1, TO_DATE('2006-07-03 04:00'), TO_DATE('2006-07-03 04:30'));
INSERT INTO BLOCK_TEST ("ID","START_BLK","STOP_BLK") VALUES (2, TO_DATE('2006-07-03 02:00'), TO_DATE('2006-07-03 02:30'));
INSERT INTO BLOCK_TEST ("ID","START_BLK","STOP_BLK") VALUES (2, TO_DATE('2006-07-03 02:30'), TO_DATE('2006-07-03 03:00'));
INSERT INTO BLOCK_TEST ("ID","START_BLK","STOP_BLK") VALUES (2, TO_DATE('2006-07-03 03:00'), TO_DATE('2006-07-03 03:30'));
select * from block_test order by id, start_blk;
ID      START_BLK               STOP_BLK
1     2006-07-03 01:00     2006-07-03 01:30
1     2006-07-03 01:30     2006-07-03 02:00
1     2006-07-03 02:00     2006-07-03 02:30
1     2006-07-03 02:30     2006-07-03 03:00
1     2006-07-03 03:30     2006-07-03 04:00
1     2006-07-03 04:00     2006-07-03 04:30
2     2006-07-03 02:00     2006-07-03 02:30
2     2006-07-03 02:30     2006-07-03 03:00
2     2006-07-03 03:00     2006-07-03 03:30If ID = 1 and my instance is 2006-07-03 01:45, I need to obtain 2006-07-03 01:00 as the start of the block and 2006-07-03 03:00 as the end - notice the gap from 3:00 to 3:30 so the end would not be 4:30.
I can do this in a procedure but I was wondering if this could be done with just SQL?
Any hints, suggestions or opinions would be welcome.
Thanks,
dfg
EDIT: After re-reading this, need to clarify that I don't need to actually change or update the rows, just query to get the start and end of the block in question...
Message was edited by:
Indy

Answer 1

There might be better ways but the following worked for me
SQL> select
  2        first_value(start_blk) over (partition by id ORDER BY lvl desc),
  3        first_value(stop_blk) over (partition by id ORDER BY lvl desc)
  4  from (select level lvl, id, connect_by_root start_blk start_blk, stop_blk
  5        from (select id , start_blk, stop_blk
  6              from   block_test
  7              where  id=1
  8             )
  9        connect by prior stop_blk=start_blk
10        order by level desc
11       )
12  where to_date('2006-07-03 01:45', 'YYYY-MM-DD HH24:MI') between start_blk and stop_blk
13  and rownum < 2
14  ;
FIRST_VALUE(STAR FIRST_VALUE(STOP
2006-07-03 01:00 2006-07-03 03:00
SQL> select
  2        first_value(start_blk) over (partition by id ORDER BY lvl desc),
  3        first_value(stop_blk) over (partition by id ORDER BY lvl desc)
  4  from (select level lvl, id, connect_by_root start_blk start_blk, stop_blk
  5        from (select id , start_blk, stop_blk
  6              from   block_test
  7              where  id=1
  8             )
  9        connect by prior stop_blk=start_blk
10        order by level desc
11       )
12  where to_date('2006-07-03 03:15', 'YYYY-MM-DD HH24:MI') between start_blk and stop_blk
13  and rownum < 2
14  ;
no rows selected
SQL> select
  2        first_value(start_blk) over (partition by id ORDER BY lvl desc),
  3        first_value(stop_blk) over (partition by id ORDER BY lvl desc)
  4  from (select level lvl, id, connect_by_root start_blk start_blk, stop_blk
  5        from (select id , start_blk, stop_blk
  6              from   block_test
  7              where  id=1
  8             )
  9        connect by prior stop_blk=start_blk
10        order by level desc
11       )
12  where to_date('2006-07-03 03:45', 'YYYY-MM-DD HH24:MI') between start_blk and stop_blk
13  and rownum < 2
14  ;
FIRST_VALUE(STAR FIRST_VALUE(STOP
2006-07-03 03:30 2006-07-03 04:30
SQL>