Parsing Null in SQL

NULL In the computer and programming world, it means unknown and uncertain. Although the Chinese translation is "empty", this void (null) is not that empty (empty). Null represents an unknown state, a future state. For example, I don’t know how much money Xiao Ming has in his pocket, but I can’t be sure it is 0. At this time, Null is used in computers to represent unknown and uncertain.

Although people who are proficient in SQL will not have any doubts about Null, it is still difficult to find an article that summarizes it comprehensively. I saw an English version and I feel good about it.

Tony Hoare invented the null Quote in 1965, and considered it a "billion-dollar mistake" he made. Even today, 50 years later, in SQL Null values ​​are also the culprit of many common errors.

Let’s take a look at some of the most shocking cases.

Null does not support size/equality judgment

The following two queries, no matter how many records there are in the users table, the returned records are 0 rows:

select * from users where deleted_at = null;
 
– result: 0 rows
 
select * from users where deleted_at != null;
 
– result: 0 rows

How could this be? All because null represents an "unknown" type. That is, it makes no sense to compare null with other values ​​using normal conditional operators. Null is not equal to Null (approximate understanding: an unknown value cannot be equal to an unknown value, and the relationship between the two is also unknown, otherwise mathematics and logic will be messed up). 　–

: The following SQL is suitable for MySQL. If it is Oracle, you need to add... from dual;

select null > 0;
 
– result: null
 
select null < 0;
 
– result: null
 
select null = 0;
 
– result: null
 
select null = null;
 
– result: null
 
select null != null;
 
– result: null

select * from users
 
where deleted_at is null;
 
– result: 所有被标记为删除的 users

If you want to determine whether the values ​​​​of two columns are different, you can use is distinct from:

select * from users
 
where has_address is distinct from has_photo
 
– result: 地址(address)或照片(photo)两者只有其一的用户

Subquery (subselect) is a very convenient way to

. For example, if you want to query users who do not have any packages, you can write the following query:

select * from users 
 
where id not in (select user_id from packages)

select * from users 
 
where id not in (1, 2, null)

This SQL statement will be converted to:

select * from users 
 
where id != 1 and id != 2 and id != null

We know that the result of id != null is an unknown value, null. And the result of the AND operation between any value and null is null, so it is equivalent to no other conditions. Then the reason for this result is null. The logical value is not true.

If the conditions are reversed, there will be no problem with the query results. Now we query users with packages.

select * from users 
 
where id in (select user_id from packages)

Similarly we can use a simple example:

select * from users
 
where id in (1, 2, null)

This SQL is converted to:

select * from users 
 
where id = 1 or id = 2 or id = null

Because the where clause is A series of or conditions, so it doesn't matter if one of them results in null. Non-true values ​​do not affect the calculation results of other parts of the clause, and are equivalent to being ignored.

When sorting, the null value is considered to be the largest. When sorting in descending order (descending), this will give you a headache, because null Value is ranked first.

The following query is to display user rankings based on scores, but it ranks users without scores to the top!

select name, points
 
from users
 
order by 2 desc;
 
– points 为 null 的记录排在所有记录之前!

There are two ideas for solving this type of problem. The simplest one is to use coalesce to eliminate the impact of null:

– 在输出时将 null 转换为 0 :
 
select name, coalesce(points, 0)
 
from users
 
order by 2 desc;
 
– 输出时保留 null, 但排序时转换为 0 :
 
select name, points
 
from users
 
order by coalesce(points, 0) desc;

There is also a way that requires the support of the database and specifies whether to put the null value first or last when sorting:

select name, coalesce(points, 0)
 
from users
 
order by 2 desc nulls last;

Of course, null can also be used to prevent errors from occurring, such as handling mathematical operation errors when dividing by zero.

Division by zero is a very egg-painfull error. SQL that was running fine yesterday suddenly went wrong when it was divided by 0. A common solution is to first use a case statement to determine whether the denominator is 0, and then perform the division operation.

select case when num_users = 0 then 0 
 
else total_sales/num_users end;

The way of ase statement is actually ugly, and the denominator is reused. If it's a simple situation, it's okay. If the denominator is a very complex

, then the tragedy will come: it is difficult to read, difficult to maintain and modify, and there will be a lot of bugs if you are not careful. At this time we can look at the benefits of null. Use nullif to make the denominator become null when it is 0. This will no longer report an error. When num_users = 0, the return result becomes null.

select total_sales/nullif(num_users, 0);
 
nullif 是将其他值转为 null, 而Oracle的 nvl 是将 null 转换为其他值。

If you don’t want it null, but if you want to convert to 0 or other numbers, you can use the coalesce function based on the previous SQL:

select coalesce(total_sales/nullif(num_users, 0), 0);
 
null 再转换回0

Tony Hoare may regret his mistake, but at least the problem of null can be easily solved. So go and practice your new ultimate move, and stay away from the invalid pit (nullifying) dug by null!

The above is the detailed content of Parsing Null in SQL. For more information, please follow other related articles on the PHP Chinese website!