How to replace LEFT outer join with INNER join in SQL? - sql

Question

I have a view on which I need to provide cluster Indexing the problem is in order to provide cluster indexing the it should not have any of the left or right outer joins , and I want to replace the LEFT outer join with INNER join , one of the ways which I can think of is to insert a dummy value with lets say -1 in the right table and by doing this even if all the Ids from the left table wont match Ids from the right table in INNER JOIN but since we have inserted -1 in the right table and we are using IsNULL(u.UserId,-1) it should return all the values from the left table but somehow this approach is not working.
create table Users(
UserId int,
UserName nvarchar(255)
)
insert into Users values(1,'sid429')
insert into Users values(2,'ru654')
insert into Users values(3,'dick231')
create table managers
(
caseId int,
CaseName nvarchar(255),
UserId int
)
insert into managers values (100,'Case1',1)
insert into managers values (101,'Case2',2)
insert into managers values (-1,NULL,-1)
select username from users u inner join managers m on m.UserId=IsNULL(u.UserId,-1)

Answer 1

Don't talk about indexes, but I think you could replace LEFT JOIN by INNER JOIN + UNION
select username from users u inner join managers m on m.UserId= u.UserId
UNION ALL
select username from users u WHERE NOT EXISTS (SELECT 1 FROM managers m WHERE m.UserId = u.UserId)

Answer 2

IsNull(u.UserId,-1) doesn't seem right - u.UserId is never null, since the absence of data is in the managers table - in this case, u.UserId will always have a value, but m.UserId might not, so IsNull(u.UserId, -1) won't work.
I'm intrigued to see a better answer, but I don't think you can do that - I think you eventually need to substitute the value conditionally to -1 if it doesn't exist in the other table, like this:
select username from users u
inner join managers m on m.UserId =
case when not exists(select * from managers where UserId = u.UserId)
then -1 else u.UserId end
This has the desired effect, but looking at the execution plan, won't help your performance issue.

Answer 3

You can replace a LEFT OUTER JOIN with an INNER JOIN if you add the missing values in the related table.
It has not worked for you because you have added a -1 value. But the not matching value on your INNER JOIN is a 3, not a null or a -1.
You can do so at runtime with an UNION, no need to permanently create those values as you have tried to do (inserting that -1 value) :
with expanded_managers as (
select CaseId, CaseName, UserId
from managers
union
select null, null, UserId
from users
where not exists (select * from managers where managers.UserId = users.UserId)
)
select UserName, CaseName
from users
inner join expanded_managers on expanded_managers.UserId = users.UserId

Answer 4

if you require only username it should be simple:
select distinct username from users u inner join managers m on m.UserId=u.UserId OR ( m.UserId=-1 AND u.userId = u.userId)

Answer 5

I have cleaned-up this part a bit. I had to guess the logical model, given that you did not specify any constraints.
create table Users (
UserId int not null
, UserName nvarchar(255) not null
, constraint pk_users primary key (UserId)
, constraint ak_users unique (UserName)
);
create table Cases (
CaseId int not null
, CaseName nvarchar(255) not null
, UserId int not null
, constraint pk_cases primary key (CaseId)
, constraint ak_cases unique (CaseName)
, constraint fk_cases foreign key (UserId)
references Users (UserId)
);
insert into Users values(1,'sid429') ;
insert into Users values(2,'ru654') ;
insert into Users values(3,'dick231');
insert into Cases values (100,'Case1',1);
insert into Cases values (101,'Case2',2);
This is mostly self-explanatory, but you have to understand that candidate keys (unique) for the result are: {UserID, CaseId}, {UserName, CaseName}, {UserID, CaseName}, {UserName, CaseId}. Not sure if you were expecting that.
with
R_00 as (
select UserId from Users
except
select UserId from Cases
)
select u.UserId
, u.UserName
, c.CaseId
, c.CaseName
from Users as u
join Cases as c on u.UserId = c.UserId
union
select u.UserId
, u.UserName
, (-1) as CaseId
, 'n/a'as CaseName
from Users as u
join R_00 as r on r.UserId = u.UserID
;
Another version of this, similar to other examples in the post.
select u.UserId
, u.UserName
, c.CaseId
, c.CaseName
from Users as u
join Cases as c on u.UserId = c.UserId
union
select u.UserId
, u.UserName
, (-1) as CaseId
, 'n/a' as CaseName
from Users as u
where not exists (select 1 from Cases as c where c.UserId = u.userId)
;