oracle - SQL joining question

Question

Here is my situation:

I have one table that contains a list of drugs sold containing the NDC (an identifier), quantity sold, and whether the drug is a brand name or generic. I have another table that contains the prescription numbers, dates, and NDCs.

I need to generate a list of the most recent 4 prescription numbers for the top 50 generic drugs and the top 50 Brand name drugs.

Simplified example:

Drug_list:
NDC   QTY      Type
123   50       Generic
125   47       Brand
128   34       Generic
...
549   1        Brand
294   1        Generic

Claims_list:
NDC  RX_num  Date
123  1234    20081027
123  4194    20090517
594  12598   20091012

How would I write a join to generate a list of

NDC RX1, RX2, RX3, RX4

where NDC are the 50 most common 'Brand' NDCs, and the following RXs are the RX numbers of the most recent claims?

~~~~~~~

So far I've got this:

select t.ndc, cl.rx, cl.date from (
select * from (
select * from (
select * from drug_list where brand = 'Generic')
order by qty)
where rownum < 51) t
join claims_list cl on cl.ndc = t.ndc
order by t.ndc, cl.date;

Which gets me part of the way there. From there, how do I trim it down to only 4 results per NDC? And, is it possible to get it in the following from:

NDC, RX1, RX2, RX3, RX4

If I have to report it as:

NDC1, RX1
NDC1, RX2
NDC1, RX3
NDC1, RX4
NDC2, RX1
NDC2, RX2
NDC2, RX3
NDC2, RX4
NDC3, RX1
... etc

but I would prefer to have it on one line.

~~~~ (as requested by a comment: create table statements for example tables):

create table drug_list
(NDC varchar2(15), QTY number, type varchar2(10));

create table claims_list
(NDC varchar2(15), RX_num varchar2(20), "date" date);

Answer 1

You can use a combination of Analytics (if you are on a recent enough version of Oracle) and a Pivot Table to do it. This should work with your dataset.

select ndc,
       max(decode(rn, 1, rx_num, null)) rx1,
       max(decode(rn, 2, rx_num, null)) rx2,
       max(decode(rn, 3, rx_num, null)) rx3,
       max(decode(rn, 4, rx_num, null)) rx4
  from (select *
          from (select claims_list.ndc,
                       claims_list.rx_num,
                       row_number() over (partition by claims_list.ndc order by claims_list.date desc) rn
                  from claims_list,
                       (select * 
                          from (select *
                                  from drug_list
                                 where type = 'Generic'
                                order by qty desc
                               )
                         where rownum < 51
                       ) drug_list
                 where drug_list.ndc = claims_list.ndc
               )
         where rn < 5
        order by ndc, rn
       )
group by ndc;

The inner query uses analytics to pull the most recent 4 rx numbers for each drug based on the claim date. Then we use a pivot to take it from 4 lines per drug to one line with 4 columns.