MINUS vs NOT in where子句

Question

I have a query which retrieves data from a single table. However the query uses the MINUS clause like this:

SELECT field1, field2
FROM tab
WHERE field3 = 'a'
AND field4 = 'b'
    MINUS
SELECT field1, field2
FROM tab
WHERE field5 = 'c'
AND field6 = 'd'

Since this is selecting from the same table I am trying to rewrite this to get rid of the MINUS clause. I thought that something like this should work:

SELECT DISTINCT field1, field2
FROM tab
WHERE field3 = 'a'
AND field4 = 'b'
AND NOT (field5 = 'c'
AND field6 = 'd')

My reasoning is that if the MINUS clause excludes records in the second query, then wrapping the WHERE clause in an AND NOT should get rid of the same records. The MINUS clause also gets rid of duplicates which is why I added DISTINCT to the selection. However the problem is that my query is returning more records than the original.

What am I missing here?

Answer 1

Consider these two rows:

1, 2, a, b, x, y

1, 2, u, v, c, d

The MINUS operation will not return the pair (1, 2) but your query will. The c, d values may appear with the same 1, 2 but in a different row from the a, b

The fundamental distinction is that MINUS operates at the set level, while your NOT condition only works on one row at a time (the same row with the "required" values in the other columns).

Now: You CAN make your query a bit more efficient (although you can't avoid reading the base table twice). Use a NOT IN condition:

select field1, field2 from tab where field3 = 'a' and field4 = 'b'
and    (field1, field2) not in 
           (select field1, field2 from tab where field5 = 'c' and field6 = 'd');

Note (see spencer7593's comment below): As in all cases when NULLs may be present, NOT IN is not a good solution. Rather, a NOT EXISTS condition should be used. I won't elaborate, since it seems out of scope for the question asked (which was why the "NOT" solution is different from the "MINUS" solution).

Answer 2

In this solutions the table is being referenced only once.

select      field1,field2

from        tab

where       (field3,field4) in (('a','b'))
        or  (field5,field6) in (('c','d'))

group by    field1,field2

having      max(case when (field5,field6) in (('c','d')) then 1 else 0 end) = 0
;

OR

select      field1,field2

from        tab

group by    field1,field2

having      max(case when (field3,field4) in (('a','b')) then 1 else 0 end) = 1
        and max(case when (field5,field6) in (('c','d')) then 1 else 0 end) = 0
;