Generate rows base on complex logic

Question

hi, i have a complex scenario that i am trying to solve. i will appreciate any help out there.
consider the following data.
WITH table1 AS
  SELECT 111 cid,  200 amt, 'P' ind FROM dual UNION all
  SELECT 111 cid,  200 amt, 'S' ind FROM dual UNION ALL
  SELECT 111 cid,  20  amt, 'AC' ind FROM dual UNION ALL
  SELECT 111 cid,  80 amt, 'S' ind FROM dual UNION ALL
  SELECT 222 cid,  100 amt, 'P' ind FROM dual UNION all
  SELECT 222 cid,  100 amt, 'S' ind FROM dual UNION ALL
  SELECT 222 cid,  20  amt, 'AC' ind FROM dual UNION ALL
  SELECT 333 cid,  300 amt, 'P' ind FROM dual UNION all
  SELECT 333 cid,  300 amt, 'S' ind FROM dual UNION ALL
  SELECT 333 cid,  30  amt, 'AC' ind FROM dual UNION ALL
  SELECT 333 cid,  90 amt, 'AC' ind FROM dual UNION all
  SELECT 333 cid,  20 amt, 'AC' ind FROM dual UNION all
)i need to break this down from column to rows which is something i know how to do.
however, the hard part is as follow. as you can see, for one particular cid, you could have
two rows with the same ind value. ex. cid=111 has two rows of ind='s', 333 has two rows with inc='ac'
what i need to do is as follow: lets take the simple scenario first
1. for cid=222 we can see that every rows has a different ind value. in this case, all i want to do is to
display columns to row. the output for 222 should be something like this
          cid       p_amt    s_amt   ac_amt
====      ====     =====   ======
222       100      100      20
   the only thing here is converting columns to rows.
2. the second step is the complex one. lets take cid=111. if you notice, there are two rows for ind=s
in this case what i want to do is to display this in two rows but the amount will differ.
the output should be something like this
         cid       p_amt    s_amt   ac_amt
====      ====     =====   ======
111       142.8    200      20
111       57.14    80       20
   what the output above shows is that columns got transform into rows but since there were two or more rows with same ind for a specific cid(ex 111)
i created two rows instead of 1 . also if you notice, the p_amount is not 200 but a percentage amount
i got the value of p_amt with the following formula: s_amount for each row/sum(ind='s') .
ex. (200 / (200+80 - sum of s ind)) * p_amt(ind=P) for the second rows same logic (80/(200+80) )* p_amt (ind=P)
same formula applay for ac amount (200 / (200+80 - sum of s ind)) * c_amt(ind=AC)
3. for cid=333 same logic as per step 2 but 3 ROWS needs to be display since there are 3 rows with ind=ac
in summary, when the ind for a specific cid is unique (no duplicate value, ex cid 222) then convert from columns to rows (1 row per specific cid)
when a particular cid has rows with same ind such as cid 111 and 333, then i want to convert columns to rows but instead of 1 row i want to convert into multiple
row and apply % to p_amt using formula above.
final output should be
         cid       p_amt    s_amt   ac_amt
====      ====     =====   ======
222       100      100      20
111       142.8    200      14.28
111       57.14    80       5.72
333       64.26    64.26    30
333       186.84   186.84   90
333       42.85    42.85    20
   can some one help write a query for such logic that produce result as above?
i am looking into having a query for this. but all solutions are welcome if it cannot be done in single query
thanks

Answer 1

Hi Frank, thanks for the query once again. i will run it later with the sample data. but first let me explain what i am trying to do here briefly.
here are answer to your comments and explanation follow
When there is only 1 row with ind='P' for a given cid, it looks like you want 1 row of output for each row where ind='AC', or for each row where ind='S', whichever is greater.
----yes when there is one P i want one row of output for each rows in AC or S only if S or AC has TWO or MORE rows whichever is greater
When there is more than 1 row with ind='P' for a given cid, it looks like you want each of those rows with ind='P' joined to each of the rows you would get otherwise.
--- for every P row same logic as point above applies for each P row. so if there are 1 P and two S we will see two rows. if there are two P and two S then we would see 4 rows
since for every P we need to apply the logic in in first question above
When all the rows for a cid have ind='P', you want them, so the query above has a outer-join between the "table" with the 'P' data and the other "tables". (Most of the "tables" are actually sub-query result sets.)
--- when all rows are P just display the data columns to rows for the same number of P rows. if there are two P, then two rows, if 3 P then 3 rows in row format.
also, if there is no P but you have S and AC. the number of rows display should be whichever is greater and of course the p_amt column will be null since there is no P. so in the case of cid=555, there should be one row for cid=555 with p amt null and s amount populated and c amount populated
so cid = 555 should be
cid     p_amt   s_amt    c_amt
===    =====    ======    =======
555    null     142.86  200
555    null     57.14    80
555    null     142.86  200
555    null     57.14    80In this case there is no P so take AC and apply formulat to S. also the same situation could happen where there is only S rows or only AC rows like cidi=666 and cid=777. these data should be diplay as same logic as cid=444
Below is an explanation
again consider the following data.
WITH table1 AS
  SELECT 111 cid,  200 amt, 'P' ind FROM dual UNION ALL
  SELECT 111 cid,  200 amt, 'P' ind FROM dual UNION all      --added another P
  SELECT 111 cid,  200 amt, 'S' ind FROM dual UNION ALL
  SELECT 111 cid,  20  amt, 'AC' ind FROM dual UNION ALL
  SELECT 111 cid,  80 amt, 'S' ind FROM dual UNION ALL
  SELECT 222 cid,  100 amt, 'P' ind FROM dual UNION all
  SELECT 222 cid,   80 amt, 'P' ind FROM dual UNION all    --added another P
  SELECT 222 cid,  100 amt, 'S' ind FROM dual UNION ALL
  SELECT 222 cid,  20  amt, 'AC' ind FROM dual UNION ALL
  SELECT 333 cid,  300 amt, 'P' ind FROM dual UNION all
  SELECT 333 cid,  300 amt, 'S' ind FROM dual UNION ALL
  SELECT 333 cid,  30  amt, 'AC' ind FROM dual UNION ALL
  SELECT 333 cid,  90 amt, 'AC' ind FROM dual UNION all
  SELECT 333 cid,  20 amt, 'AC' ind FROM dual union all
  SELECT 444 cid,  900 amt, 'P' ind FROM dual UNION all    --added these two
  SELECT 444 cid,  1000 amt, 'P' ind FROM dual UNION all
   SELECT 555 cid,  200 amt, 'S' ind FROM dual UNION ALL
   SELECT 555 cid,  200 amt, 'S' ind FROM dual UNION ALL
   SELECT 555 cid,  200 amt, 'AC' ind FROM dual union all
   SELECT 555 cid,  80 amt, 'AC' ind FROM dual union all
  SELECT 666 cid,  100amt, 'S' ind FROM dual UNION ALL
   SELECT 666 cid,  100amt, 'S' ind FROM dual UNION ALL
SELECT 777 cid,  50amt, 'AC' ind FROM dual UNION ALL
   SELECT 777 cid,  50amt, 'AC' ind FROM dual UNION ALL
this dataset represents amount. think of P indicator as a loan of $200 for cid=111. think of S as some type of interest. AC is along the same line as S.
one loan (P) could have multiple S and one or no AC.
one loan(P) could have just one row or more rows with P such as cid=444 and no S or AC
one loan P could have multiple AC and so on
each row in the output represent the percentage amount of the Loan base on the interest S or AC.
for example. lets take cid=111 in data above. we can see there are two Ps Two S and 1 AC.
for a particular cid, i want to check if there are one or more S or one or more AC.
in this case there are two S and i should display the data in row format in two rows. if there were 3 S then that would have been 3 rows display for cid=111.
if you look at the output below, you will see first two rows and the value for S remain the same as the data above(200 and 80)
if you look at the p_amt, the amt is NOT the same as the original data. the original data has 200. when you apply the formula mentioned in previous post, p amt gets split into different amount for the two rows and they add up to 200. if there were 3 S in the original data, then you will see 3 rows and the p amt as well as the c_amt will be less
since all need to add up to the original amount in the dataset. this scenario happen when there are two or more S or two or more AC.
the second set of data after the blank line, is the same as the first set. this is because there are two P with same amount, and the same logic should be apply for all the P present.
cid     p_amt   s_amt    c_amt
=====================
111  142.86  200.00   14.29
111   57.14   80.00    5.71
111  142.86  200.00   14.29
111   57.14   80.00    5.71lets take another example, cid=444 only has two P and no S or AC. in this case just display the data in row format. so there should be two rows only, s_amt and c_amt should be null since there is none.
another example is cid=222, there are two P, one S and one AC. since there is not two or more S or AC, then we need to convert all rows from columns to rows and display one rows. however, we can see that there is two P, so for the second P same logic apply. this is why you see duplicate rows because we need to apply same logic for every P row even though data might look the same.
222  100.00  100.00   20.00
222  80.00  100.00   20.00if there was only one P for cid=200 then we would have seen this output
222  100.00  100.00   20.00so in summary, we are converting from columns to rows for each set of cid and depending on the number of S rows or AC rows, that is how many rows should be display for a specific cid. then like in the case of cid=111, p_amt and c_amt in all rows should added up to the original amount. this is done by the RATION function that determine the percentage to apply base on the formula provided.
same with cid=333. in this case there are 3 AC so 3 rows should show and the p_amt and s_amt should be derive base on the percentage and add up to original value.
as i mentioned, there could be multiple P and they might have different amount, the same logic should apply for every P row therefore you could see duplication of rows such as output for cid=111
i know this is long but i hope this helps. feel free to ask any other questions and thanks for your help. i will try the query. if you feel you need to add something else to the query base on explanation above, feel free and let me know.
Edited by: elmasduro on Feb 16, 2011 7:07 AM
Edited by: elmasduro on Feb 16, 2011 7:14 AM
Edited by: elmasduro on Feb 16, 2011 7:42 AM
Edited by: elmasduro on Feb 16, 2011 8:48 AM
Edited by: elmasduro on Feb 16, 2011 10:37 AM