Help in joining two tables

Question

Hi ,
I need your help in the below scenario:
I have two tables.
One of my table 'table1' contains the below data:
Code:
Name,Value
12A,1
12B,1
12C,1
Table2 contains the below data:
Code:
value,result
1,12
1,24
1,56
1,423
1,32
1,3
I need to join based on value field.
My expected result is:
Code:
NAME,VALUE,RESULT
12A,1,12
12B,1,24
12C,1,56
12D,1,423
12E,1,32
12F,1,3
Depends on the number of records in the second table,we need to append A to Z at the end of the name field. The number of records will not exceed more than 26. How we can achieve this?
Thanks
Pandeeswaran

Answer 1

OK, now you give another useful information - that there always will be combinations of sums in table2 to match the values in table1. (It is difficult to help when you tell the specs one at a time :-) )
But it is not easy, because the code really will have to try and consider all possible combinations and then "choosing the right one" - this is easy for us humans, but not easy to encode in programming logic.
I have made an attempt:
SQL> set linesize 120
SQL> with table1 as (
  2     select 'A1' name, 123 id, 150 value from dual union all
  3     select 'A2' name, 123 id, 200 value from dual union all
  4     select 'A3' name, 123 id, 300 value from dual
  5  ), table2 as (
  6     select 123 id, 100 value from dual union all
  7     select 123 id, 100 value from dual union all
  8     select 123 id, 50  value from dual union all
  9     select 123 id, 100 value from dual union all
10     select 123 id, 100 value from dual union all
11     select 123 id, 100 value from dual union all
12     select 123 id, 100 value from dual
13  )
14  --
15  -- End of test data
16  --
17  select
18  t1.id, t1.name, t1.value, t2.value,
19  t1.rn, t1.minval, t1.maxval,
20  t2.rn, t2.sumval
21  from (
22     select
23     tab1.*,
24     nvl(sum(tab1.value) over (
25        partition by tab1.id
26        order by tab1.rn
27        rows between unbounded preceding and 1 preceding
28     ),0) minval,
29     sum(tab1.value) over (
30        partition by tab1.id
31        order by tab1.rn
32        rows between unbounded preceding and current row
33     ) maxval
34     from (
35        select
36        table1.*,
37        row_number() over (
38           partition by table1.id
39           order by table1.value desc
40        ) rn
41        from table1
42     ) tab1
43  ) t1
44  join (
45     select
46     tab2.*,
47     sum(tab2.value) over (
48        partition by tab2.id
49        order by tab2.rn
50     ) sumval
51     from (
52        select
53        table2.*,
54        row_number() over (
55           partition by table2.id
56           order by table2.value desc
57        ) rn
58        from table2
59     ) tab2
60  ) t2
61  on (t2.id = t1.id)
62  where t2.sumval > t1.minval
63  and t2.sumval <= t1.maxval
64  order by
65  t1.id,
66  t1.rn,
67  t2.rn
68  ;
        ID NA      VALUE      VALUE         RN     MINVAL     MAXVAL         RN     SUMVAL
       123 A3        300        100          1          0        300          1        100
       123 A3        300        100          1          0        300          2        200
       123 A3        300        100          1          0        300          3        300
       123 A2        200        100          2        300        500          4        400
       123 A2        200        100          2        300        500          5        500
       123 A1        150        100          3        500        650          6        600
       123 A1        150         50          3        500        650          7        650
7 rows selected.It does seem to work for your data sample, but it is much too simple a rule to work in general. My "rule" simply is to order the data by value descending and summing up until "enough" values have been added.
Consider this data sample instead:
SQL> with table1 as (
  2     select 'A1' name, 1 id, 100 value from dual union all
  3     select 'A2' name, 1 id, 200 value from dual union all
  4     select 'A3' name, 1 id, 300 value from dual union all
  5     select 'B1' name, 2 id, 100 value from dual union all
  6     select 'B2' name, 2 id, 200 value from dual
  7  ), table2 as (
  8     select 1 id, 25  value from dual union all
  9     select 1 id, 75  value from dual union all
10     select 1 id, 50  value from dual union all
11     select 1 id, 50  value from dual union all
12     select 1 id, 175 value from dual union all
13     select 1 id, 225 value from dual union all
14     select 2 id, 25  value from dual union all
15     select 2 id, 50  value from dual union all
16     select 2 id, 75  value from dual union all
17     select 2 id, 100 value from dual union all
18     select 2 id, 50  value from dual
19  )
20  --
21  -- End of test data
22  --
23  select
24  t1.id, t1.name, t1.value, t2.value,
25  t1.rn, t1.minval, t1.maxval,
26  t2.rn, t2.sumval
27  from (
28     select
29     tab1.*,
30     nvl(sum(tab1.value) over (
31        partition by tab1.id
32        order by tab1.rn
33        rows between unbounded preceding and 1 preceding
34     ),0) minval,
35     sum(tab1.value) over (
36        partition by tab1.id
37        order by tab1.rn
38        rows between unbounded preceding and current row
39     ) maxval
40     from (
41        select
42        table1.*,
43        row_number() over (
44           partition by table1.id
45           order by table1.value desc
46        ) rn
47        from table1
48     ) tab1
49  ) t1
50  join (
51     select
52     tab2.*,
53     sum(tab2.value) over (
54        partition by tab2.id
55        order by tab2.rn
56     ) sumval
57     from (
58        select
59        table2.*,
60        row_number() over (
61           partition by table2.id
62           order by table2.value desc
63        ) rn
64        from table2
65     ) tab2
66  ) t2
67  on (t2.id = t1.id)
68  where t2.sumval > t1.minval
69  and t2.sumval <= t1.maxval
70  order by
71  t1.id,
72  t1.rn,
73  t2.rn
74  ;
        ID NA      VALUE      VALUE         RN     MINVAL     MAXVAL         RN     SUMVAL
         1 A3        300        225          1          0        300          1        225
         1 A2        200        175          2        300        500          2        400
         1 A2        200         75          2        300        500          3        475
         1 A1        100         50          3        500        600          4        525
         1 A1        100         50          3        500        600          5        575
         1 A1        100         25          3        500        600          6        600
         2 B2        200        100          1          0        200          1        100
         2 B2        200         75          1          0        200          2        175
         2 B1        100         50          2        200        300          3        225
         2 B1        100         50          2        200        300          4        275
         2 B1        100         25          2        200        300          5        300
11 rows selected.In this dataset the simple ordering by value will not work - It should have been A3: (225,75), A2: (175,25) and A1: (50,50).
I cannot really think of a reasonably simple way to do it in SQL alone. Maybe using the MODEL clause would be possible, but not trivial. It is possible it would be easier to solve this in PL/SQL by iterating through a couple of arrays and intelligently trying the different combinations, rather than brute-force creating all combinations in a huge piece of SQL.
I am sorry, Pandeesh, but I can not think of a solution easily.
I might be able to do something if I fiddled with the problem for a couple of days, but that would be beyond the scope of this forum. That would be a consulting job rather than a bit of forum help :-)