Question

I have 3 tables that join with together and i want select distinct rows on the B.val3. i use something like this :

SELECT A.val1,A.val2, B.val1, B.val2, B.val3, C.val1, C.val2
FROM A INNER JOIN
B ON A.val1 = B.val1 INNER JOIN
C ON A.val1 = C.val1

this statement get to me 24 rows with duplicate B.val3
now, when i use SELECT DISTINCT A.val1,A.val2, B.val1,..., all 24 records are retrieved, but i want only distinct rows on the B.val3 retrieved (8 rows)

when i use something like :

SELECT DISTINCT A.val1,A.val2, B.val1,
...
C ON A.val1 = C.val1 GROUP BY B.val3

i received an error :

Msg 8120, Level 16, State 1, Line 13
Column 'A.val1' is invalid in the select list because it is not contained in either an aggregate function or the GROUP BY clause.

EDIT : structure and data

table A as follow : val2 is foreign key of B and val3 is foreign key C

val1   val2    val3
-----  -----   -----
1       100     200
2       100     201
3       101     200
4       102     200
5       102     201

table B :

val1   val2    val3
-----  -----   -----
100     a2      aaa
101     b2      bbb
102     c2      ccc

table C :

val1   val2    
-----  -----   
200     a3      
201     b3      

and i want get something like :

A.val1  A.val2  A.val3  B.val1  B.val2  B.val3  C.val1  C.val2
-----   -----   -----   -----   -----   -----   -----   -----
1       100      200     100      a2     aaa      200    a3
3       101      200     101      b2     bbb      200    a3      
4       102      201     102      c2     ccc      201    b3    

Answer 1

You could try something like this:

    select * 
    from 
    (
    SELECT A.val1,A.val2, B.val1, B.val2, B.val3, C.val1, C.val2
            , row_number() over 
            (
                partition by b.val3 
                order by A.val1, A.val2, B.val1, B.val2, C.val1, C.val2
            ) r
    FROM A 
    INNER JOIN B ON A.val1 = B.val1 
    INNER JOIN C ON A.val1 = C.val1
    ) x
    where x.r = 1

or

    SELECT max(A.val1)
    ,max(A.val2)
    , max(B.val1)
    , max(B.val2)
    , B.val3
    , max(C.val1)
    , max(C.val2)
    FROM A 
    INNER JOIN B ON A.val1 = B.val1 
    INNER JOIN C ON A.val1 = C.val1
    group by b.val3

Depending on what you’re trying to achieve. If those don’t do what you’re after, please can you provide more info on what you’re hoping to do / example data?

The issue you have is when selecting a distinct b.val3 there may be multiple records associated:

• Are values in column V3 in table B unique?
• Are values in column V1 in tables A and/or C unique?

If the answers to either of the above questions are no, you need to give SQL a way to decide which of the multiple possible records/results to select when choosing what data to display for the other columns.

---

EDIT

Based on example data given above, please find a script to replicate the sample info & display the solution:

if object_id('a') is not null drop table a
if object_id('b') is not null drop table b
if object_id('c') is not null drop table c
go
create table b 
(
    val1 int not null identity(100,1) primary key clustered
    , val2 nvarchar(2) not null
    , val3 nvarchar(3) not null
)
go
create table c
(
    val1 int not null identity(200,1) primary key clustered
    , val2 nvarchar(2) not null
)
go
create table a
(
    val1 int not null identity(1,1) primary key clustered
    , val2 int not null constraint fk_a_b foreign key references b(val1)
    , val3 int not null constraint fk_a_c foreign key references c(val1)
)
go

--ids 100 - 105
insert b
select 'a2', 'aaa'
union all select 'b2', 'bbb'
union all select 'c2', 'ccc'
union all select 'c3', 'ccc' --val3 is not unique
union all select 'c4', 'ccc' --
union all select 'b3', 'bbb' --

--ids 200 - 204
insert c
select 'a3'
union all select 'b3'
union all select 'c3'
union all select 'd3'
union all select 'e3'

insert a
select 100, 200
union all select 100, 200
union all select 100, 201
union all select 101, 200
union all select 102, 200
union all select 102, 201
union all select 103, 201
union all select 104, 201
union all select 105, 201
union all select 105, 202
union all select 105, 203
union all select 105, 204

--what does the full result set look like?
SELECT  A.val1 aval1
,       A.val2 aval2
,       B.val1 bval1
,       B.val2 bval2
,       B.val3 bval3
,       C.val1 cval1
,       C.val2 cval2  
FROM A 
INNER JOIN B 
    ON A.val2 = B.val1 
INNER JOIN C 
    ON A.val3 = C.val1 

--now show unique B's
select Aval1, Aval2, Bval1, Bval2, Bval3, Cval1, Cval2     
from      
(     
    SELECT  A.val1 aval1
    ,       A.val2 aval2
    ,       B.val1 bval1
    ,       B.val2 bval2
    ,       B.val3 bval3
    ,       C.val1 cval1
    ,       C.val2 cval2       
    , row_number() over              
    (                 
        partition by b.val3                  
        order by b.val1, c.val1 --try playing with this to see how the results change / see what fits your requirements           
    ) r     
    FROM A 
    INNER JOIN B 
        ON A.val2 = B.val1 
    INNER JOIN C 
        ON A.val3 = C.val1    
) x     
where x.r = 1 

--what wasn't included in the unique B result set, but was in the full set?
select Aval1, Aval2, Bval1, Bval2, Bval3, Cval1, Cval2     
from      
(     
    SELECT  A.val1 aval1
    ,       A.val2 aval2
    ,       B.val1 bval1
    ,       B.val2 bval2
    ,       B.val3 bval3
    ,       C.val1 cval1
    ,       C.val2 cval2       
    , row_number() over              
    (                 
        partition by b.val3                  
        order by b.val1, c.val1 --try playing with this to see how the results change / see what fits your requirements           
    ) r     
    FROM A 
    INNER JOIN B 
        ON A.val2 = B.val1 
    INNER JOIN C 
        ON A.val3 = C.val1    
) x     
where x.r > 1