MySQL多个查询还是临时表？

Question

I need help from a SQL guru here...

I have data in the following format:

Key  Value
100  A
100  B
300  A
400  B
500  A
500  B
500  C
600  B
600  W

If I want to find all keys with A or B and A and B, but not AB with anything else...how would I do this? Would I need a temp table or can I somehow join the table to itself?

EDIT: The desired output would be Key 100 since it has A & B, Key 300 since it has A, Key 400 since it has B. Not Key 500 because it also contains a C, and not key 600 because it also contains a W

Answer 1

This should work for you:

SELECT DISTINCT t1.`key`
FROM table1 t1
WHERE t1.value in('A', 'B')
GROUP BY t1.`key`
HAVING COUNT(t1.value) = (SELECT COUNT(t2.value) 
                          FROM Table1 t2 
                          WHERE t1.`key` = t2.`key`)

SQL Fiddle Demo

This returns only the keys:

• Has A and B only, and nothing else so key = 500 shouldn't be returned. But key = 100 should be included.
• Has only A . As in key = 300 .
• Has only B so key = 600 shouldn't be returned since it contains one more value than B which is w . But key = 400 should be included.

Update: How is this working

If any key has a value IN('A', 'B') then it could also contains other values as well.

Thats why I added the HAVING clause:

HAVING COUNT(t1.value) = (SELECT COUNT(t2.value) 
                          FROM Table1 t2 
                          WHERE t1.`key` = t2.`key`)

The COUNT(t1.value) is compared to the total COUNT of the values for the same key, by using a correlated subquery SELECT COUNT(t2.value) FROM Table1 t2 WHERE t1.key = t2.key , for each value t1.value . So if the current key contains values other than 'A', 'B' then the COUNT(t1.value) won't equal to the COUNT of all values for the same key. For example the key = 500 has a count = 3 but using:

SELECT DISTINCT t1.`key`
FROM table1 t1
WHERE t1.value in('A', 'B')
GROUP BY t1.`key`

Without the HAVING clause will include it as well, since it has either A' or 'B'. But the HAVING clause eliminates this key since it has a count = 3 not equal to 2.

Answer 2

You should explain your question better, if i can understand correctly what you need you could use something like this:

SELECT Distinct K
FROM tab
WHERE
  value IN ('A', 'B') AND
  NOT EXISTS (SELECT Null from tab tab_1
              WHERE tab_1.K = tab.K and tab_1.value not in ('A', 'B'))