Oracle, find differences in records between two "sets"

Question

Some scripts:

create table set_1
( id number(12));

create table set_2
( id number(12));

create table map_set
( id_set_1 number(12),
  id_set_2 number(12)
);
  
  
  insert into set_1 values(1);
  insert into set_1 values(2);
  insert into set_1 values(3);
  insert into set_1 values(4);
  
  insert into map_set values(1,5);
  insert into map_set values(2,6);
  insert into map_set values(3,7);
  insert into map_set values(4,8);
  
  insert into set_2 values(5);
  insert into set_2 values(6);
  insert into set_2 values(7);
  insert into set_2 values(8);

create table attrib_1(
id           number(12),
set_id       number(12),
attrib_id    number(2),
val          number(10,2));

create table attrib_2
(
id number(12),
set_id number(12),
attrib_id number(2),
val       number(10,2)
);

insert into attrib_1 values(1,1,1,10.2);
insert into attrib_1 values(2,1,2,2.5);
insert into attrib_1 values(3,1,3,5.23);

SELECT * FROM attrib_1;

insert into attrib_2 values(1,5,2,4.54);
insert into attrib_2 values(2,5,3,89.5);

There is a - let's call them sets. There is also a table which mapping this sets, so in this case fe set from set_1 table with id = 1 is a copy of set_2 with id =2. Sets can have attributes, one or more. Then we make some changes in this attrib sets.

What I want is to detect differences between this attributes between "same" sets. I've wrote this:

   SELECT  *
    FROM    (
              SELECT a1.id       ,
                     a1.attrib_id,
                     a1.val      ,
                     s1.id  AS set_id,
                     mp.id_set_2
              
              FROM   set_1   s1,
                     map_Set mp,
                     attrib_1 a1
              WHERE  mp.id_set_1 = s1.id
              AND    s1.id = 1
              and    a1.set_id = S1.id
            ) s1
    full    join
            (
              SELECT a2.id       ,
                     a2.attrib_id,
                     a2.val      ,
                     s2.id  AS set_id
              FROM   set_2   s2,
                     attrib_2 a2
              WHERE  s2.id = 5
              and    a2.set_id = S2.id
            ) s2
    ON      s1.id_set_2 = s2.set_id
    AND     s1.attrib_id = s2.attrib_id

I want to create for.. loop plsql code where it gets all records and:

• if attrib_id's are the same and val is different - update record
• if there is attrib in set_2 but missing in set_1 - delete
• if there is attrib in set_1 but missing in set_2 - add and here's the problem:

I need an id of set to do INSERT, but there is no id "empty attrib" records.

I want output to look like this:

• to have ALL data about sets, and left join attributes.

I tried left joins, this is closest I've got to my goal but now I'm stuck.

Any ideas? Thanks!

Edit: i prefer to do this in one SQL. Triggers on tables forbides operatioms on few id's in one insert.

Answer 1

You want to perform three operations. They can be done with three separate queries:

1. If attrib_id's are the same and val is diffrent - update record

update (
  select x.*, u.bval
  from attrib_1 x
  join (
    select
      m.*, a.attrib_id as aid, a.val as aval, b.val as bval
    from map_set m
    join attrib_1 a on a.set_id = m.id_set_1
    full join attrib_2 b on b.set_id = m.id_set_2 and b.attrib_id = a.attrib_id
  ) u on u.aid = x.attrib_id and u.aval <> u.bval
)
set val = bval

2. If there is attrib in set_2 but missing in set_1 - delete

delete from attrib_2
where (set_id, attrib_id) in (
  select b.set_id, b.attrib_id
  from map_set m
  join attrib_1 a on a.set_id = m.id_set_1
  right join attrib_2 b on b.set_id = m.id_set_2 and b.attrib_id = a.attrib_id
  where a.set_id is null
);

3. If there is attrib in set_1 but missing in set_2 - add

insert into attrib_2 (id, set_id, attrib_id, val)
select
  100, -- probably use an identity or sequence
  m.id_set_2, a.attrib_id, a.val
from map_set m
join attrib_1 a on a.set_id = m.id_set_1
left join attrib_2 b on b.set_id = m.id_set_2 and b.attrib_id = a.attrib_id
where b.set_id is null

Final Result:

 ID_SET_1  ID_SET_2  AID  AVAL  BVAL 
 --------- --------- ---- ----- ---- 
 1         5         1    10.2  10.2 
 1         5         2    4.54  4.54 
 1         5         3    89.5  89.5 

See running example at db<>fiddle . I added one extra row to your example to test case #3 (add).