查找缺失对 (Oracle SQL)

Question

Lets say that I have a table called customer_accounts that looks like this:

Table: CUSTOMER_ACCOUNTS

+------------------+--------------+
| CUSTOMER_ID (PK) |  ACCOUNT_ID  | 
+------------------+--------------+
|        1         |          5   |
|        1         |          6   | 
|        1         |          8   | 
+------------------+--------------+

And I have another table called distinct_account_pairs that looks like this:

Table: DISTINCT_ACCOUNT_PAIRS (before)

+------------------+---------------+---------------+
| CUSTOMER_ID (PK) |  ACCOUNT_ID1  | ACCOUNT_ID2   |
+------------------+---------------+---------------+
|        1         |          6    |       5       |
+------------------+---------------+---------------+

How do I populate the distinct_account_pairs table with the missing pairs for each CUSTOMER_ID? In this example 5 and 8, 6 and 8 are missing for PK CUSTOMER_ID 1 in distinct_account_pairs, so after it should look something like this. Note that distinct_account_pairs could have account_id1 < account_id2 or vice versa. The pairs just need to be distinct for each CUSTOMER_ID.

Table: DISTINCT_ACCOUNT_PAIRS (after)

+------------------+---------------+---------------+
| CUSTOMER_ID (PK) |  ACCOUNT_ID1  | ACCOUNT_ID2   |
+------------------+---------------+---------------+
|        1         |          6    |       5       |
|        1         |          5    |       8       |
|        1         |          6    |       8       |
+------------------+---------------+---------------+

Answer 1

You can do it with self join and not exists:

 create table CUSTOMER_ACCOUNTS ( CUSTOMER_ID int,  ACCOUNT_ID  int);

 insert into CUSTOMER_ACCOUNTS values(        1         ,          5   );
 insert into CUSTOMER_ACCOUNTS values(        1         ,          6   ); 
 insert into CUSTOMER_ACCOUNTS values(        1         ,          8   );

 create table DISTINCT_ACCOUNT_PAIRS ( CUSTOMER_ID int,  ACCOUNT_ID1  int, ACCOUNT_ID2   int);
 insert into DISTINCT_ACCOUNT_PAIRS values(        1         ,          6    ,       5       );

Insert Query:

 insert into DISTINCT_ACCOUNT_PAIRS 
 select a.CUSTOMER_ID,a.ACCOUNT_ID,b.ACCOUNT_ID from CUSTOMER_ACCOUNTS a inner join CUSTOMER_ACCOUNTS b
 on a.CUSTOMER_ID = b.CUSTOMER_ID and a.ACCOUNT_ID < b.ACCOUNT_ID
 and not exists 
               (select 1 from DISTINCT_ACCOUNT_PAIRS d where d.CUSTOMER_ID=a.CUSTOMER_ID
               and ((d.ACCOUNT_ID1=a.ACCOUNT_ID and d.ACCOUNT_ID2=b.ACCOUNT_ID)
               or (d.ACCOUNT_ID1=b.ACCOUNT_ID and d.ACCOUNT_ID2=a.ACCOUNT_ID))
               );

Query: to select from DISTINCT_ACCOUNT_PAIRS

 SELECT * FROM DISTINCT_ACCOUNT_PAIRS;

Output:

CUSTOMER_ID	ACCOUNT_ID1	ACCOUNT_ID2
1	6	5
1	6	8
1	5	8

db<fiddle here

Answer 2

You would generate all the pairs and then compare them:

select ca1.customer_id, ca1.account_id, ca2.account_id
from customer_accounts ca1 join
     customer_accounts ca2
     on ca1.customer_id = ca2.customer_id and
        ca1.account_id < ca2.account_id left join
     distinct_account_pairs dap
     on dap.customer_id = ca1.customer_id and
        dap.account_id1 in (ca1.account_id, ca2.account_id) and
        dap.account_id2 in (ca1.account_id, ca2.account_id) 
where dap.customer_id is null;