SQL to idenfity gaps in ranges

Question

A friday SQL teaser! 11.2.03.
I have a table with indicator and start, end ranges. I need an SQL to return ranges that are not excluded.
create table k_test (indicator varchar2(10),start_val number, end_val number);
insert into k_test values('INCLUDE',000,999);
insert into k_test values('EXCLUDE',000,503);
insert into k_test values('EXCLUDE',507,511);
insert into k_test values('EXCLUDE',514,999);A select statement please to return the ranges that are "open", i.e. not excluded.
504
505
506
512
513I can manage this in PL/SQL, keen to know a clever way in SQL.

Answer 1

Ok, let's do some advertising, because i think this solution is simple to understand and readable and need less effort to solve the problem in my opinion. An all can be seen stepwise.
First imaging an empty excel sheet.
So we start from here and fill in the start data.
--select r from (
select r, rng, start_val, end_val, indicator from t
model
dimension by (r)
measures(1 rng, start_val, end_val, indicator)
rules iterate(0) until iteration_number > start_val[0] (
rng[for r from start_val[iteration_number+1] to end_val[iteration_number+1] increment 1]=
case
when cv(r) >= start_val[iteration_number+1]
and cv(r) <= end_val[iteration_number+1]
and indicator[iteration_number+1]='EXCLUDE'
then 0
when indicator[iteration_number+1]='INCLUDE'
then 1
else nvl(rng[cv()],0)
end
--)where rng = 1
R     RNG     START_VAL     END_VAL     INDICATOR
2     1     0     503     EXCLUDE
3     1     507     511     EXCLUDE
4     1     514     999     EXCLUDE
1     1     0     999     INCLUDE
0     1     4     -      - The last "artificial" row is only used to hold the value for the max iterations needed, that's the number of rows in t. This cell (Excel!) is adressed with start_val[0].
So now we look at the result of first iteration rules iterate(1):
R     RNG     START_VAL     END_VAL     INDICATOR
2     1     0     503     EXCLUDE
3     1     507     511     EXCLUDE
4     1     514     999     EXCLUDE
1     1     0     999     INCLUDE
0     1     4     -      -
5     1     -      -      -
6     1     -      -      -
7     1     -      -      -
8     1     -      -      -
9     1     -      -      -
10     1     -      -      -
11     1     - just a snipplet. But we can see, that a INCLUDED-row was processed, because rng wa set to 1 on serveral cells. In fact in all rows from 0 to 999. The rows that were not there before, were created on that iteration.
so what will happen in the next iteration rules iterate(2)?
R     RNG     START_VAL     END_VAL     INDICATOR
2     0     0     503     EXCLUDE
3     0     507     511     EXCLUDE
4     0     514     999     EXCLUDE
1     0     0     999     INCLUDE
0     0     4     -      -
5     0     -      -      -
6     0     -      -      -
7     0     -      -      -
8     0     -
995     1     -      -      -
996     1     -      -      -
997     1     -      -      -
998     1     -      -      -
999     1     -      -      Obviously an excluded row was processed, the one with (0     503) and therefor rng was set to 0 in the rows from 0 to 503.
So that row processing goes on until all rows form the origin table are processed, in this case 4, and everytime rng in the processed range will marked with 0 or 1 depend on the indicator and leaved untouched if the row is out of the range in process.
rules iterate(9999999) until iteration_number > start_val[0] is just a construct not to hardcode the number of iterations.
At the end, all we have to do is counting the rows with rng 1.
So be honest, is this way too complicated ?-)