Complicated SQL

Question

Dears,
I 'd like to create a SQL to return group data. I guess I don't create a procedure for manage this result, but I don't manage to create a SQL for this.
My strutuct is alike this:
create table t(c1 varchar2(1), c2 date, c3 number);
insert into t (c1,c2,c3) values ('X', '01/02/2004',1.2);
insert into t (c1,c2,c3) values ('X', '01/02/2004', 1.8);
insert into t (c1,c2,c3) values ('Y', '02/02/2004', 1.1);
insert into t (c1,c2,c3) values ('Z', '03/02/2004', 2.2);
insert into t (c1,c2,c3) values ('Z', '03/02/2004', 1.2);
insert into t (c1,c2,c3) values ('A', '04/02/2004', 1.9);
insert into t (c1,c2,c3) values ('A', '04/02/2004', 2.3);
insert into t (c1,c2,c3) values ('A', '04/02/2004', 0.2);
insert into t (c1,c2,c3) values ('A', '04/02/2004', 0.2);
insert into t (c1,c2,c3) values ('T', '04/02/2004', 1.3);
insert into t (c1,c2,c3) values ('T', '05/02/2004', 1.7);
insert into t (c1,c2,c3) values ('A', '06/02/2004', 0.9);
insert into t (c1,c2,c3) values ('T', '07/02/2004', 0.9);
insert into t (c1,c2,c3) values ('A', '07/02/2004', 0.0);
insert into t (c1,c2,c3) values ('X', '07/02/2004', 2.2);
insert into t (c1,c2,c3) values ('W', '08/02/2004', 4.7);
insert into t (c1,c2,c3) values ('X', '09/02/2004', 1.5);
insert into t (c1,c2,c3) values ('X', '09/02/2004', 0.15);
insert into t (c1,c2,c3) values ('X', '09/02/2004', 0.7);
(The inserts are according with the groups)
My problem is group result by condition c3<2 and so always that break this condition I have to do the sum rows preceding until last condition c3<2.
In short, I need the follow result, grouping by condition "c3 < 2", according order by date c2 :
|Last Line c1 | Number of days | Number of Rows
|of the Group | | (COUNT)
| Y | 2 | 3
| Z | 0 | 1
| A | 1 | 2
| A | 0 | 1
| A | 4 | 7
| X | 0 | 1
| W | 0 | 1
| X | 0 | 3
Anyone can help me?
Thanks.

Answer 1

The first group you have defined starts 01/02/2004 and ends 02/02/2004 and you say this is two days. However the third group you define starts 03/02/2004 and ends 04/02/2004 yet you record this as one day. What is the rule for determining the total days?
Neither am I entirely clear on this '<2' rule. If groups are determined by whether c3 is greater or less than two (in order of c2) why do these two rows constitute two groups and not one?
insert into t (c1,c2,c3) values ('X', '07/02/2004', 2.2);
insert into t (c1,c2,c3) values ('W', '08/02/2004', 4.7);
To separate rows into groups based on events which occur as you pass through the result set in a given order I generally use the LAG or LEAD analytic functions to identify the boundaries with (for example) the value 1 and then SUM () OVER (RANGE UNBOUNDED PRECEDING) to assign an number to each group in turn.
I suspect you can get this done with something similar to the following. You will notice my results do not tally exactly because of the reasons I have given above.
Oracle9i Enterprise Edition Release 9.2.0.4.0 - 64bit Production
With the Partitioning, OLAP and Oracle Data Mining options
JServer Release 9.2.0.4.0 - Production
SQL> CREATE TABLE t(c1 VARCHAR2(1), c2 DATE, c3 NUMBER);
Table created.
SQL> ALTER SESSION SET cursor_sharing = FORCE;
Session altered.
SQL> INSERT INTO t (c1,c2,c3) VALUES ('X', '01/FEB/2004',1.2);
1 row created.
SQL> INSERT INTO t (c1,c2,c3) VALUES ('X', '01/FEB/2004', 1.8);
1 row created.
(snip)
SQL> INSERT INTO t (c1,c2,c3) VALUES ('X', '09/FEB/2004', 0.15);
1 row created.
SQL> INSERT INTO t (c1,c2,c3) VALUES ('X', '09/FEB/2004', 0.7);
1 row created.
SQL> SELECT c1,
  2         MAX (c2) - MIN (c2) total_days,
  3         COUNT (*) total_rows
  4  FROM  (SELECT LAST_VALUE (c1) OVER (
  5                  PARTITION BY grp) c1, c2, grp
  6         FROM  (SELECT c1, c2,
  7                SUM (
  8                  CASE
  9                    WHEN c3 < 2 AND prv_c3 >= 2
10                      OR c3 >= 2 AND prv_c3 < 2
11                    THEN 1
12                  END)
13                  OVER (
14                    ORDER BY c2
15                      ROWS UNBOUNDED PRECEDING) grp
16                FROM (SELECT c1, c2, c3,
17                             LAG (c3) OVER (ORDER BY c2)  prv_c3
18                      FROM   t)))
19  GROUP BY c1, grp
20  ORDER BY MIN (c2), MAX (c2);
C TOTAL_DAYS TOTAL_ROWS
Y          1          3
Z          0          1
A          1          2
A          0          1
A          3          7
W          1          2
X          0          3
7 rows selected.
SQL> Padders