DB2 SQL JOIN比较

Question

I have the following data:

TABLE1: 
ADDRESS               INTER1    INTER2
12345 E AVE STREET    44444     55555555
888 OtherStreet       44444     55555555


TABLE2: 
ADDRESS             INTER1  INTER2
12345 E AVE STREET  44444   55555555

I can probably resolve in code this way:

if(tblOneInterOne == tblTwoInterOne ) {
   // compare address                                   
}

, but hoping to resolve in SQL.

SELECT 
     A.ADDRESS
   , A.INTER1
   , A.INTER2
FROM TABLE1 AS A
    LEFT JOIN TABLE2 AS B
    ON A.INTER1 = B.INTER1
    AND A.INTER2 = B.INTER2
    AND A.ADDRESS <> B.ADDRESS

When comparing with the above SQL, I still get this address: 12345 E AVE STREET on my data result. I have also tried joining with a sub-query, i have also tried grouping, and ordering by asc. I am out of ideas.

Expected output : If selecting from TABLE1, match on first two integers, if match, then compare on address. If match, do not display. If no match, then display non-match address. Vice-versa, if selecting from TABLE2, match on first two integers. If match, then compare on address, if no-match then the display should be null (no data).

Any help would be appreciated.

Thank you.

Answer 1

This query:

SELECT A.ADDRESS, A.INTER1, A.INTER2
FROM TABLE1 A LEFT JOIN
     TABLE2 B
     ON A.INTER1 = B.INTER1 AND
        A.INTER2 = B.INTER2 AND
        A.ADDRESS <> B.ADDRESS;

is going to return all rows in A . That is what LEFT JOIN does. That is what it is supposed to do.

If you want to filter rows from the first table, then change the JOIN type:

SELECT A.ADDRESS, A.INTER1, A.INTER2
FROM TABLE1 A INNER JOIN
     TABLE2 B
     ON A.INTER1 = B.INTER1 AND
        A.INTER2 = B.INTER2 AND
        A.ADDRESS <> B.ADDRESS;

You can also then move the inequality condition to WHERE . But using a LEFT JOIN and then moving the inequality to WHERE is rather ridiculous -- why specify an outer join and then undo it in the WHERE clause?

Answer 2

The solution to this issue was the following:

SELECT 
     A.ADDRESS
   , A.INTER1
   , A.INTER2
FROM TABLE1 AS A
    INNER JOIN TABLE2 AS B
       ON A.INTER1 = B.INTER1
       AND A.INTER2 = B.INTER2
    LEFT JOIN TABLE2 AS B1
       ON B1.ADDRESS <> B.ADDRESS
WHERE A.ADDRESS <> B.ADDRESS
AND B1.ADDRESS IS NOT NULL

This allows to match on the first two integer values, then joins on the address. However, this causes records with multiple addresses to still display. This is eliminated by the WHERE clause compare.

This was an actual production problem with provider data. Once the solution is discovered, it seems extremely simple--but arriving to the solution is always the tricky part, right? I hope this helps someone else.

Edit 04/24/2016

After more quality testing, this solution needed additional clause, which was:

SELECT 
     A.ADDRESS
   , A.INTER1
   , A.INTER2
FROM TABLE1 AS A
    INNER JOIN TABLE2 AS B
       ON A.INTER1 = B.INTER1
       AND A.INTER2 = B.INTER2
    LEFT JOIN TABLE2 AS B1
       ON B1.ADDRESS <> B.ADDRESS
WHERE NOT IN (SELECT B2.ADDRESS FROM TABLE2 AS B2 
                   INNER JOIN TABLE1 AS A2 
                     ON B2.INTER1 = A2.INTER1
                      AND B2.INTER2 = A2.INTER2)

 A.ADDRESS <> B.ADDRESS
AND B1.ADDRESS IS NOT NULL

Of course the 'WHERE NOT IN' allows the sub-query to exclude anything found. This can also be done using a CTE then excluding in the NOT IN clause. I hope this helps someone else.

Answer 3

try this :

SELECT 
        ifnull(f1.INTER1, f2.INTER1) as INTER1, 
        ifnull(f1.INTER2, f2.INTER2) as INTER2, 
        f1.ADDRESS as ADDRESS_A, f2.ADDRESS as ADDRESS_B,
        case 
        when f1.INTER1 then 'ADDRESS NOT IN TABLE1'
        when f2.INTER1 then 'ADDRESS NOT IN TABLE2'
        else 'ADDRESS ARE DIFFERENT' end as DIAGNOSTIC
FROM TABLE1 f1 
FULL OUTER JOIN TABLE2 f2 ON (f1.INTER1, f1.INTER2)  = (f2.INTER1, f2.INTER2)
where 
        f1.INTER1 is null or 
        f2.INTER1 is null or 
        f1.ADDRESS <> f2.ADDRESS;

Answer 4

--other solution if you want your result into multiple rows

select * from 
(
select ADDRESS, INTER1, INTER2 from table1
except
select ADDRESS, INTER1, INTER2 from table2
) tmp1

union all 

select * from 
(
select ADDRESS, INTER1, INTER2 from table2
except
select ADDRESS, INTER1, INTER2 from table1
) tmp2