选择不在另一个表中的行

Question

SELECT DISTINCT a.value
FROM a LEFT JOIN b 
    ON a.value = b.value 
      AND (b.field IS NULL OR b.field != 'my_string');

SELECT a.value
FROM a
WHERE a.value NOT IN
    (SELECT value 
     FROM b
     WHERE b.field = 'my_string');

From what I've read, doing a LEFT JOIN is faster. But I've also read that DISTINCT is code smell for inefficient query. How do I go about determining which query performs better in a worst case scenario?

EDIT: sorry, id is not a primary key, it's just another field. I'll replace it with value.

EDIT2: Looks like everybody's hung up on my first query. Let's say it looks like this instead. Isn't the logic the same?

SELECT DISTINCT a.value
FROM a LEFT JOIN b ON a.value = b.value 
WHERE (b.field IS NULL OR b.field != 'my_string');

EDIT3: Sample fiddle. http://sqlfiddle.com/#!2/500ea/1

EDIT4: Accepted answer. http://sqlfiddle.com/#!2/500ea/8

Answer 1

Your first join is non-sensical. It returns all a.id values in a . Remember, left join keeps all rows in the first table and matching rows in the second. I think you intend:

SELECT a.id
FROM a LEFT JOIN
     b 
     ON a.id = b.id AND b.field = 'my_string'
WHERE b.field IS NULL;

The distinct should be unnecessary, assuming that a.id is, well, a unique id.

An alternative is to use not exists :

SELECT a.id
FROM a
WHERE NOT EXISTS (SELECT 1 FROM b WHERE a.id = b.id AND b.field = 'my_string');

For performance, create an index on b(id, field) .

Answer 2

In general, the NOT IN (SELECT ...) is more inneficient than the LEFT JOIN , because the SELECT in the IN() condition must be executed once for every row in order to perform the filter. For small datasets, that's not a problem, but for large data sets, that can be quite a big headache quite inefficient.