SQLTeam.com | Weblogs | Forums

Why these two statements do not return the same results?

#1

Hello

How can this return different results?

WHERE [COL1] <> 'VALUE1'
AND CASE [COL1] IS NULL THEN [COL2] ELSE [COL1] END <> 'VALUE1'

WHERE CASE [COL1] IS NULL THEN [COL2] ELSE [COL1] END <> 'VALUE1'

To me, they should return the same results!

Any idea?

1 Like
#2

When Col1 is NULL, the first clause in your WHERE condition - COL1 <> 'VALUE1' evaluates to unknown. Hence the where clause returns UNKNOWN.

You can see this more clearly if you examine the following test snippet. It will tell you that @Col1 is equal to VALUE1, even though it clearly is not. This is because the WHEN condition does not evaluate to TRUE, it evaluates to UNKNOWN, and so the ELSE clause gets executed.

DECLARE @Col1 VARCHAR(32) = NULL;

SELECT 
    CASE
        WHEN @Col1 <> 'VALUE1' THEN
            '@Col1 is not equal to VALUE1'
        ELSE
            '@Col1 is equal to VALUE1'
    END;

Lookup articles on how NULL behaves in SQL Server, for example here.

#3

hi

i tried to do some testing ..

in the first SQLclause its ignoring nulls
in the second SQL NULL is showing up

please click arrow to the left for Drop Create Sample data 
drop table #explaining 

create table #explaining 
(
col1 varchar(10) null , 
col2 varchar(10) 
) 
go 

insert into #explaining select 'VALUE1' , 'xxxxx'
insert into #explaining select 'yyyyyy' , 'xxxxx'
insert into #explaining select  Null , 'xxxxx'
go 

select 'Sample Data',* from #explaining

select 
        'SQL using [COL1] <> ''VALUE1'' '
	 ,  * 
from 
    #explaining 
WHERE [COL1] <> 'VALUE1' 
         AND 
	  CASE WHEN [COL1] IS NULL THEN [COL2] ELSE [COL1] END <> 'VALUE1'

select 
      'SQL using                         '
	  ,   * 
from 
    #explaining 
WHERE                        
    CASE WHEN [COL1] IS NULL THEN [COL2] ELSE [COL1] END <> 'VALUE1'

image

1 Like
#4

That is crazy! How can your code return that @Col1=VALUE1 when we declared @Col1=NULL???
It is mind boggling!

#5

Even if the WHEN evaluates to UNKNOWN, it is STILL NOT 'VALUE1' so it should never resolve to ELSE!!!

This is extremely confusing and dangerous. What is the proper way to handle this? Is it only a CASE thing or it works with IF and other statements?

From what comes to mind immediately, we will need to always do WHEN @Col1 <> ('VALUE1' OR NULL) or something!

#6

So to understand, when we do [Col1]<>'VALUE1', it will NOT return NULLs by default because NULLs are not truly <>'VALUE1', but they are UNKNOWN?

Thanks!

#7

False. NULLs add a level of complexity to value checking.

A NULL value is never = or <> any other value. After all, an unknown value could be 'VALUE1', so SQL can't say that it's "<>".

Also, a WHERE condition(s) must be "TRUE" to include a row, not just "might be TRUE" or "not proven FALSE".

Interestingly, a CHECK condition must be "FALSE" to prevent a row from being INSERTed. If you run the code below, you'll see that the NULL value IS inserted into #t1, even though it seems to break the CHECK constraint.

CREATE TABLE #t1 ( col1 varchar(10) NULL CHECK(col1 <> 'VALUE1') );
INSERT INTO #t1 VALUES('VALUE1') --<<-- not INSERTed
GO
INSERT INTO #t1 VALUES('VALUE2')
INSERT INTO #t1 VALUES(NULL) --<<-- this row *DOES* INSERT
SELECT * FROM #t1
GO
DROP TABLE #t1;
1 Like
#8

And then there are cases where null is interpreted as though NULL = NULL, as in the example below.

------------------------------------------------------------------
CREATE TABLE #tmp(id INT NULL UNIQUE)

-- this is successful
INSERT INTO #tmp VALUES (NULL);
-- this fails, because there is a NULL already, suggesting NULL=NULL.
INSERT INTO #tmp VALUES (NULL);

------------------------------------------------------------------
-- and then there is this
INSERT INTO #tmp VALUES (1),(2), (3);
-- where only two rows are returned even though there are four rows in the table 
-- of which only one row has the value 1.
SELECT * FROM #tmp WHERE id NOT IN (1)

-----------------------------------------------------------------
-- and then even funnier, where the query returns nothing.
CREATE TABLE #tmp2( id INT NULL UNIQUE)
INSERT INTO #tmp2 VALUES (1),(NULL);

SELECT * FROM #tmp AS t 
WHERE id NOT IN 
    ( SELECT t2.id FROM #tmp2 AS t2 )

It is well-worth spending the time to learn the nuances of the 3-valued logic in SQL to avoid many associated pitfalls such as these. There are very rational explanations and exceptions to the explanations as to why the behaviors are the way they are.

1 Like
#9

one way NULLS are dealt with

..if column is null replace with some value
and then compare that value

ISNULL(Column,'xxxxx') = 'xxxx'

#10

ISNULL(Column,'xxxxx') = 'xxxx'

NO, NO, NO !!

A function on a column prevents SQL from doing a seek on an index that is keyed by that column.

Thus, NEVER use an ISNULL() in a WHERE or JOIN clause.

Instead, do the more straightforward and clearer:

WHERE (column IS NULL OR column = 'xxxx')

This can also perform VASTLY better when an index on that column is available.

4 Likes