tsql - In SQL Server, what does "SET ANSI_NULLS ON" mean?

Question

The definition says:

When SET ANSI_NULLS is ON, a SELECT statement that uses WHERE column_name = NULL returns zero rows even if there are null values in column_name. A SELECT statement that uses WHERE column_name <> NULL returns zero rows even if there are non-null values in column_name.

Does this mean that no nulls will be included in this query?

SELECT Region
FROM employees
WHERE Region = @region

Or do ANSI_NULLs concern only queries like this one (where the WHERE includes the specific word NULL)?

SELECT Region
FROM employees
WHERE Region = NULL

Answer 1

It means that no rows will be returned if @region is NULL, when used in your first example, even if there are rows in the table where Region is NULL.

When ANSI_NULLS is on (which you should always set on anyway, since the option to not have it on is going to be removed in the future), any comparison operation where (at least) one of the operands is NULL produces the third logic value - UNKNOWN (as opposed to TRUE and FALSE).

UNKNOWN values propagate through any combining boolean operators if they're not already decided (e.g. AND with a FALSE operand or OR with a TRUE operand) or negations (NOT).

The WHERE clause is used to filter the result set produced by the FROM clause, such that the overall value of the WHERE clause must be TRUE for the row to not be filtered out. So, if an UNKNOWN is produced by any comparison, it will cause the row to be filtered out.

---

@user1227804's answer includes this quote:

If both sides of the comparison are columns or compound expressions, the setting does not affect the comparison.

from SET ANSI_NULLS^*

However, I'm not sure what point it's trying to make, since if two NULL columns are compared (e.g. in a JOIN), the comparison still fails:

create table #T1 (
    ID int not null,
    Val1 varchar(10) null
)
insert into #T1(ID,Val1) select 1,null

create table #T2 (
    ID int not null,
    Val1 varchar(10) null
)
insert into #T2(ID,Val1) select 1,null

select * from #T1 t1 inner join #T2 t2 on t1.ID = t2.ID and t1.Val1 = t2.Val1

The above query returns 0 rows, whereas:

SELECT * FROM #T1 t1 INNER JOIN #T2 t2 
   ON t1.ID = t2.ID 
  AND (   t1.Val1 = t2.Val1 
       OR t1.Val1 IS NULL 
      AND t2.Val1 IS NULL   )

Returns one row. So even when both operands are columns, NULL does not equal NULL. And the documentation for = doesn't have anything to say about the operands:

When you compare two NULL expressions, the result depends on the ANSI_NULLS setting:

If ANSI_NULLS is set to ON, the result is NULL¹, following the ANSI convention that a NULL (or unknown) value is not equal to another NULL or unknown value.

If ANSI_NULLS is set to OFF, the result of NULL compared to NULL is TRUE.

Comparing NULL to a non-NULL value always results in FALSE².

However, both ¹ and ² are incorrect - the result of both comparisons is UNKNOWN.

---

^*The cryptic meaning of this text was finally discovered years later. What it actually means is that, for those comparisons, the setting has no effect and it always acts as if the setting were ON. Would have been clearer if it had stated that SET ANSI_NULLS OFF was the setting that had no effect.