Finding Missing Numbers

Question

I want to find missing numbers from a column which is alpha numeric, its last five positions are numeric. At first three positions there are different combinations of alphas i.e. *'BA ','BAW','SA ','SAA' ..........* if alphas are two then 3rd position is blank. I also need it to generalised this to cover all existing alpha numeric series.
SELECT SUBSTR (a.claim_no, 4, 5) + 1
  FROM core_business.cb_pensioner a
WHERE SUBSTR (a.claim_no, 1, 3) = 'BA '
   AND NOT EXISTS (
                 SELECT NULL
                   FROM core_business.cb_pensioner b
                  WHERE to_number(SUBSTR (b.claim_no, 4, 5)) =  to_number(SUBSTR (a.claim_no, 4, 5)) + 1)
   order by SUBSTR (a.claim_no, 4, 5);I am getting ORA-01722 error

Answer 1

You have two issues with such a task.
Problem 1 is how to convert the trailing part into a number so that you can compare values.
to_number(SUBSTR (b.claim_no, 4, 5)) works, but only if the value in b.claim_no is always a number regardless of the other column criteria.
A statement like
WHERE SUBSTR (a.claim_no, 1, 3) = 'BA ' /* filter 1 */
and to_number(SUBSTR (b.claim_no, 4, 5)) > 1 /* filter 2 */
...might fail because your table might have a value like
b.claimno = 'XYZ1234b'
The way you wrote the query the database will decide which filter criteria it executes first. It might be filter1, but could also be filter2. If filter2 is executed first then you will get the error message.
Problem 2 how to order and compare different rows, so that you can find the missing values.
Here the analytic LEAD(LAG) function is very helpful. it will give you access to the next (or previous row).
Both problems can be overcome.
Here is my try, but I'm not sure if it works in your case. You might need to adapt a little.
This should work if all the rows are numeric that belong to the specific groups. if you can have wrong data in there too, then a different approach is needed.
with testdata as (select 'XYZ1234b' claim_no from dual union all
            select 'BA 12345' claim_no from dual union all
            select 'BA 12346' claim_no from dual union all
            select 'BA 12350' claim_no from dual union all
            select 'BAW11111' claim_no from dual union all
            select 'BAW11113' claim_no from dual union all
            select 'XYZ1234b' claim_no from dual )
/* end of test data creation */
select claim_no, substr(claim_no, 1,3) part1, substr(claim_no, 4) part2,
       to_number(substr(claim_no, 4)) current_num,
       lead(to_number(substr(claim_no, 4))) over (partition by substr(claim_no, 1,3) order by to_number(substr(claim_no, 4))) next_num,
       lead(to_number(substr(claim_no, 4))) over (partition by substr(claim_no, 1,3) order by to_number(substr(claim_no, 4)))
        - 1 - to_number(substr(claim_no, 4))
        as "missing_values"
from testdata
where substr(claim_no, 1,3) in ('BAW','BA ')
order by claim_no;
CLAIM_NO PAR PART2 CURRENT_NUM   NEXT_NUM missing_values
BA 12345 BA  12345       12345      12346              0
BA 12346 BA  12346       12346      12350              3
BA 12350 BA  12350       12350
BAW11111 BAW 11111       11111      11113              1
BAW11113 BAW 11113       11113Problem 1 is solved because the where condition is always executed before the function in the select clause, especially in connection with an analytic function.
Problem 2 is also solved by using the analytic function LEAD function. And this is fast, because you access the same table only one time. Other solution might do a subquery to fetch the value from the previous/next row. This means several accesses to the same table and is usually much slower.
Edited by: Sven W. on Sep 20, 2010 2:32 PM