Prioritizing distinct counts from aggregate table

Question

I am a relatively new developer with limited PL/SQL experience trying to modify an existing query that extracts counts from an aggregate table. What I need to do is set it up so that it only counts the record once based on it's priority. This is currently set up in a DECODE Statement counting only totals (not distinct nor prioritized). What I’m trying to accomplish would be something like this:
If Service 'A' exists count it and stop. If more than one Service ‘A’ exists count only once.
If Service 'A' does not exist then check if 'B' exist, if 'B' exists count it and stop. If more than one Service ‘B’ exists count only once.
If Service 'B' does not exist then check if 'C' exist, if 'C' exists count it and stop. If more than one Service ‘C’ exists count only once. etc.
It may be possible to have multiple services in a given category and if so should be counted only once.
Here is a snipit of the existing code....
COUNT
(DECODE (service_id,
1, program_enrollment_id
) basic_readjustment,
COUNT
(DECODE (service_id,
2, program_enrollment_id
) occupation_skills,
COUNT (DECODE (service_id,
3, program_enrollment_id
) on_the_job,
COUNT
(DECODE (service_id,
4, program_enrollment_id
) placement_assist
Any help you can give would be appreciated!
Mark

Answer 1

Yes, although based on your last post, it is not quite a simple as my example. Your original query as posted will count each occurence of each service_id whether there is one or many for a group. The DISTINCT on the innermost query, against wia_bas_mv, will remove duplicates for a particular service_id, but would leave multiple records it there were different service_ids for the same group.
The trick is to to get only one record for each group, and make sure that that record is the "highest" priority record. Since your priority order does not match the sort order of the service_ids, you need to add an extra DECODE to get them to "sort" in the right order. Assuming that I have done the service_id mapping correctly, this should be close to what you are looking for:
SELECT region_id, region_name, officegroupname, officegroupid,
       SUM(basic_readjustment) basic_readjustment,
       SUM(occupation_skills) occupation_skills,
       SUM(on_the_job) on_the_job, SUM(placement_assist) placement_assist,
       SUM(other) other,
       SUM(basic_readjustment + occupation_skills + on_the_job +
           placement_assist + other) total
FROM (SELECT officegroupname, officegroupid, region_id, region_name,
             SUM(DECODE(service_id, 'C', 1, 0)) basic_readjustment,
             SUM(DECODE(service_id, 'A', 1, 0)) occupation_skills,
             SUM(DECODE(service_id, 'B', 1, 0)) on_the_job,
             SUM(DECODE(service_id, 'D', 1, 0)) placement_assist,
             SUM(DECODE(service_id, 'E', 1, 0) other
      FROM (SELECT program_enrollment_id, officegroupname,
                   officegroupid, region_id, region_name,
                   MIN(DECODE(service_id, 368, 'A', 369, 'B', 144, 'C',
                                          114, 'D', 'E')) service_id
            FROM wia_base_mv
            WHERE region_id < 13 AND
                  program_value = 'VAL04_TAA' AND
                  actual_completion_date IS NULL
            GROUP BY program_enrollment_id, officegroupname,
                     officegroupid, region_id, region_name)
      GROUP BY officegroupname, officegroupid, region_id, region_name
      UNION ALL
      SELECT r.region_id, r.region_name, og.officegroupname,
             og.officegroupid, 0 basic_readjustment, 0 occupation_skills,
             0 on_the_job, 0 placement_assist, 0 other
      FROM officegroup og, regions r, office_data od
      WHERE r.region_id = od.region_id AND
            og.officeid = od.office_id)
GROUP BY region_id, region_name, officegroupname, officegroupid
ORDER BY region_id HTH
John