Postgres excludes NULL in WHERE id != int query

问题:

I came up against a strange problem in Postgres yesterday when trying to filter out user ids from a stats table. When we did, for example, user_id != 24, postgres excluded the rows where user_id is NULL as well.

I created the following test code which shows the same results.

CREATE TEMPORARY TABLE test1 (
    id int DEFAULT NULL
);

INSERT INTO test1 (id) VALUES (1), (2), (3), (4), (5), (2), (4), (6), 
    (4), (7), (5), (9), (5), (3), (6), (4), (3), (7), 
    (NULL), (NULL), (NULL), (NULL), (NULL), (NULL), (NULL);     

SELECT COUNT(*) FROM test1;

SELECT id, COUNT(*) as count
FROM test1
GROUP BY id;

SELECT id, COUNT(*) as count
FROM test1
WHERE id != 1
GROUP BY id;

SELECT id, COUNT(*) as count
FROM test1
WHERE (id != 1 OR id IS NULL)
GROUP BY id;

The first query just counts all the rows. The second counts the number of each value, including nulls. The third excludes the value 1 and also all the nulls. The fourth is a work around to exclude value 1 and still include the nulls.

For what I'm trying to use this query for, null values should always be included.

Is the work around the only way to do this? Is this expected Postgres behaviour?

回答1:

Your "work around" is the usual way to do it. Everything is behaving as expected.

The reason is simple: nulls are neither equal, nor not equal, to anything. This makes sense when you consider that null means "unknown", and the truth of a comparison to an unknown value is also unknown.

The corollary is that:

• null = null is not true
• null = some_value is not true
• null != some_value is not true

The two special comparisons IS NULL and IS NOT NULL exist to deal with testing if a column is, or is not, null. No other comparisons to null can be true.

回答2:

The IS DISTINCT FROM predicate exists for this purpose. It's described as:

not equal, treating null like an ordinary value

So just doing id IS DISTINCT FROM 1 should work.

Reference: https://www.postgresql.org/docs/11/functions-comparison.html