34

I'm using Postgres with this query

select 
*
from Entity this_ 
where 
(this_.ID not in (null))

Why does this give me no results? I would expect to get all rows where id is not null

with

(this_.ID not in (1))

i get the expected results

wutzebaer
  • 14,365
  • 19
  • 99
  • 170

6 Answers6

39

The result of [not] in (null) will always be null. To compare to null you need is [not] null or is [not] distinct from null

select *
from Entity this_ 
where this_.ID is not null

If you want where (ID not in (1,null)) as in your comment you can do

where ID is not null and ID not in (1)
Clodoaldo Neto
  • 118,695
  • 26
  • 233
  • 260
14

PostgreSQL uses NULL as undefined value.
What you're asking is to return the items that are not in a list or an undefined value. Since undefined means that you do not know what's inside, PostgreSQL does not return any item because simply can not respond to the request.
While the request:

select * from Entity where id in (1, null)

can return the records, because if it finds an element with ID = 1 knows that is in the collection
the request:

select * from Entity where (ID not in (1, null))

can not be satisfied because the null value can be any value.

AndreaBoc
  • 3,077
  • 2
  • 17
  • 22
  • 4
    the explanation is good, but `NULL` is `NULL`, and not _any value_. It is a special value but **definitely not _ANY value_**. Actually, for me it was enough to remember: `where null = null` - returns 0 lines, `where null != null` - 0 lines, `null > null` - zero lines, and _almost all other operations with NULL as argument returns `false`_. all these work for the `IN`-operator as well (because it is basically almost the same as comparing `==` the item to every element in the `( parentheses - list )` ). – maxkoryukov Sep 01 '21 at 08:41
  • 1
    actually, most operations with `NULL` return `NULL` (not `false`) sorry for a misleading comment – maxkoryukov May 25 '23 at 20:15
12

answer

The solution was already posted another answer, but the same page, and my solution is almost the same:

where
  "field" is NOT NULL
  AND "field" not in (1)

don't forget, the inverted version of the condition (is null or in (list)) uses the OR-operator (instead of AND):

where
  "field" is NULL
  OR "field" in (1)

bonus: SQL and NULL

and here is a thing a good SQL-developer has somewhere in the subconscious zone (the request was tested in Postgres, but I'm pretty sure it is the behavior from the standard ANSI SQL):

select
    -- simple things
    1 = 1,              -- true
    1 = 2,              -- false
    -- "simple things" with null
    1  = null,          -- null (and it is not `false` if you expected this)
    1 != null,          -- null (and it is not `true` if you expected this)
    -- "strange" area:
    null  = null,       -- null (and it is not `true` and not `false`)
    null != null,       -- null (and it is not `true` and not `false`)
    1 > null,           -- null (yeah, after 4 previous examples it is exactly what you expected)
    1 < null,           -- null
    null < null,        -- null
    null > null,        -- null
    -- value IN ()
    1 in (1,2),         -- true
    1 in (1,null),      -- true
    1 in (2, 3),        -- false
    1 in (2, null),     -- null (!!! surprise?)
    3 between (1, NLL),-- null
    -- value NOT IN
    1 not in (1,2),     -- false
    1 not in (1,null),  -- false
    1 not in (2, 3),    -- true
    1 not in (2, null), -- null (!!! surprise?)
    -- NULL IN/NOT IN
    null in (1,2),      -- null
    null in (NULL),     -- null
    null not in (1,2),  -- null
    null not in (NULL), -- null
    -- and surprise:
    NOT NULL            -- NULL (!!, but most probably you foresaw/knew this)

as you can see - if the null is an operand then the result is null and in a boolean context (for example, in WHERE-clauses) - it is falsey. Though, falsey is not the same as false, and NOT (1=NULL) is NULL, but not truly, so both these requests return 0 lines:

-- 1
select 1 where (1 = null)

-- 2
select 1 where NOT (1 = null)

I hope it was useful

maxkoryukov
  • 4,205
  • 5
  • 33
  • 54
  • =/ this links exactly here? – wutzebaer Sep 02 '21 at 19:47
  • > link is here — yes, that man posted his answer first, so I added a link to his answer (which is actually the same as mine, or mine is the same as his), I just added some notes on how PG-operations handle NULL-operands – maxkoryukov Sep 02 '21 at 19:59
2

I have had similar in problem and eventually came up with the following solution;

select * from public."Employee_11" where (COALESCE("Name",'@'),"Surname") 
    in (
        ('@','dummy')
    )

It does return the records which Name column has null values. You can also use this for not in clause which will return not null records of Name;

  select * from public."Employee_11" where (COALESCE("Name",'@'),"Surname") 
        not in (
            ('@','dummy')
        )
Ktt
  • 469
  • 3
  • 8
  • 18
0

I had similar problem. My ego that I knew SQL well, got punctured. Here is the simplified example from scott/tiger tables.

select empno, ename from emp where deptno not in (10, 20, null);

Returned nothing. Although I use NOT IN condition very sparingly because it does not use index and is very slow. I prefer to use OUTER JOIN instead.

I tried this query in Postgres and Oracle both and results are the same. So, must be a standards compliant result. NULL behaves this way only in NOT IN condition.

BB23850
  • 109
  • 1
  • 11
  • 1
    All DBMS behave this way. `deptno not in (10, 20, null);` is the same as `detpno <> 10 AND detpno <> 20 AND deptno <> NULL` which is (e.g. for detpno=1) the same as `false AND false AND null` which in turn is `NULL` - which in turn is "not true" for the WHERE clause. The result is independent of the value of `deptno` and thus it's "not true" for all rows in the table and consequently it returns nothing. –  Sep 01 '21 at 09:41
-2

You can use <> ANY operator. Example on your code:

select 
  * 
from Entity this_ 
where 
   (this_.ID <> ANY (null))
Ahmet Erkan ÇELİK
  • 2,364
  • 1
  • 26
  • 28