Solved

Exists versus In

Posted on 2014-02-19
Medium Priority
269 Views
This was an interview question asked:

Is there any scenario where exists would be used but not  IN.

My answer: Exists returns boolean value and IN returns a value and exists is faster performance-wise. They donot want to know about the performance but they wanted to know a scenario where only Exists could be used but not IN
0
Question by:d27m11y
[X]
Welcome to Experts Exchange

Add your voice to the tech community where 5M+ people just like you are talking about what matters.

• Help others & share knowledge
• Earn cash & points
• 3
• 2
• 2
• +2

LVL 77

Expert Comment

ID: 39871753
>>and exists is faster performance-wise

Tom Kyte explains why this may not be true:

>>wanted to know a scenario where only Exists could be used but not IN

Of the top of my head I cannot think of one.  I'm sure some other Expert will have one if I cannot think of one later.
0

LVL 23

Expert Comment

ID: 39871887
Prior to 10g, EXISTS was preferred for a small result set.  The CBO has long since been improved.
0

Author Comment

ID: 39872026
How about NULL values, am not sure.. can anyone shed light ?
0

LVL 77

Expert Comment

ID: 39872067

There is the example I couldn't think of.

IN lists cannot process nulls where EXISTS can.

Take the following test case:
``````drop table tab1 purge;
create table tab1(col1 char(1), col2 char(1));

insert into tab1 values('a','a');
insert into tab1 values('1',null);
commit;

select t1.col1 from tab1 t1 where exists (
select null from dual
);

select t1.col1 from tab1 t1 where col2 in (
select null from dual
);
``````
0

LVL 8

Expert Comment

ID: 39872784
@slightwv I haven't tested this but lines 8-10 don't look like making any sense. It will return all rows since select anything from dual will always be a non-empty resultset i.e. exists evaluates to constant true.

I believe exists makes sense on a relatively small table if you can use a good where condition e.g. using an index of a huge table:

``````select * from smalltable where exists
(select null from hugetable where hugetable.indexedfield = smalltable.somefield);

select * from smalltable where somefield in (select indexedfield from hugetable);
``````

As far as I can tell Oracle (as well as MySQL or PostgreSQL) would do something like:
- for IN: perform full table scan on hugetable and join the resultset with smalltable
- for EXISTS: perform full table scan on smalltable and a relatively small number of index searches on hugetable.

But I can imagine that one or more of the above systems, also possibly depending on table statistics, software version and perf tuning config, would rewrite the IN query to EXISTS or vica versa.
0

LVL 23

Expert Comment

ID: 39873353
To the interview question, then, your answer seems to be "no".  Disregarding performance as directed, IN does not currently have any declarative restriction.
0

LVL 32

Expert Comment

ID: 39874185
I think there might be cases when exists can be used but not in. Something like the following where you want to get an audit report but only where there is an error in the error_log table for today:

select * from audit_table where exists
(select 1 from error_log where errdate > trunc(sysdate));
0

LVL 8

Expert Comment

ID: 39876239
This can also be rewritten as "IN":

``````select * from audit_table where 1 in
(select 1 from error_log where errdate > trunc(sysdate));
``````

And this is actually a proof of concept: the following two boolean expressions are equal:

``````exists (select 1 from ...)
1 in (select 1 from...)
``````
Note that you can replace 1 with any non-null constant of any type.

There is a nuance in the original question though; "where exists would be used but not  IN." The concept above is not about "would" more like "could". Just in case it matters.
0

LVL 32

Expert Comment

ID: 39876768
Surrano, good point. I didn't think about the use of a literal value in a select of that literal value.
0

LVL 77

Accepted Solution

slightwv (䄆 Netminder) earned 2000 total points
ID: 39877250
>>is not about "would" more like "could". Just in case it matters.

Even though you discount my over-simplified test case as not making any sense.  I stand by my comment.

When nulls are involved, EXISTS will work an IN will not.  I can use EXISTS to look for nulls and I cannot rewrite it to use a null in an IN list.

I cannot say:   ... where column in (a,b,c,null)
0

Featured Post

Question has a verified solution.

If you are experiencing a similar issue, please ask a related question

Background In several of the companies I have worked for, I noticed that corporate reporting is off loaded from the production database and done mainly on a clone database which needs to be kept up to date daily by various means, be it a logical…
When it comes to protecting Oracle Database servers and systems, there are a ton of myths out there. Here are the most common.
This video explains at a high level about the four available data types in Oracle and how dates can be manipulated by the user to get data into and out of the database.
This videos aims to give the viewer a basic demonstration of how a user can query current session information by using the SYS_CONTEXT function
Suggested Courses
Course of the Month12 days, 22 hours left to enroll