使用Oracle，如何查看不匹配的结果？

Question

I have two oracle tables with similar structures, the main difference is that one is a subset of the other. What I want to do is perform a join that will return all rows that are not in the subset table.

The issue is that the unique identifier is about 6-7 columns. So I can not just do a WHERE NOT IN clause. Does anyone know which type of join or how I can do this?

Answer 1

Assuming your primary key is (COL1,COL2,...) :

The easy answer is to use the MINUS set operator :

SELECT COL1,COL2 FROM T
MINUS SELECT COL1,COL2 FROM U;

If you feel the need, or have more complex logic, you might prefer using an outer join :

SELECT COL1,COL2 
FROM T LEFT JOIN U
USING(COL1,COL2)
WHERE U.ROWID IS NULL;

Please notice the WHERE clause that filters out the result set to only keep PK of rows not present in the U table.

In addition, all comparisons are only performed on the primary key , not checking the other columns. See http://sqlfiddle.com/#!4/6e42f/13 . According to execution plan, you will see that in my very simple example, the OUTER JOIN performs better.

---

That being said, the subquery syntax might perform even better, as the Oracle optimizer build almost the same execution plan than for the outer join. But , the key difference is the use of an anti nested loop , removing the need for the extra NOT NULL filter:

SELECT COL1,COL2 FROM T
WHERE (COL1,COL2) NOT IN (SELECT COL1,COL2 FROM U);

As far as I know, the NOT IN operator is the only way to perform the anti join you need.

Answer 2

You can write a MINUS query. Make sure you mention exact column names, ie exact number of columns with same data type in both SQLs in the MINUS query.

The result set thus returned is the difference in the rows between the two tables.

Answer 3

where not in is the answer.

the query will be something like this:

select blam, hoot, kapow
from schmarr
where blam not in
( select blam from smashy )

The column blam must have the following properties:

1. identifies a row in both the schmarr and the smashy tables.
2. is neither the naughty column 6 nor the naughty column 7 you mentioned.

the schmarr table is the superset table. the smashy table is the subset table.

Answer 4

let's assume Table-B is subset of Table-A.

    select * from Table-A left outer join Table-b on 
      Table-A.column1 = Table-b.column2 and 
      Table-A.column2 = Table-b.column2 and ....

where the columns1,column2,... are the unique identifiers