CODEWITHSUNDEEP
sqlsql-server
Gen Wan
Gen Wan

Reputation: 2009

Return matched-pair and the non-matched pair for each ID only once

Any help would be appreciated.

I have two sample tables here.

Table A:

ID |Name
123|REG
123|ERT
124|REG
124|ACR

Table B

ID |Name
123|REG
123|WWW
124|REG
124|ADR

Here is the simple join output and I will explain my question in the comments:

*Yes -- I want this row

*No -- I don't want this row 

AID|Aname|BID|Bname
123|REG  |123|REG  --Yes-- Matched-pair for id '123'
123|ERT  |123|REG  --No--'REG' already had one match. 'ERT' should pair with 'WWW' for id '123'
123|REG  |123|WWW  --No--The same reason as above
123|ERT  |123|WWW  --Yes--non-matched pair for id '123'
124|REG  |124|REG
124|ACR  |124|REG
124|REG  |124|ADR
124|ACR  |124|ADR

My desired result:

AID|Aname|BID|Bname
123|ERT  |123|WWW
123|REG  |123|REG
124|REG  |124|REG
124|ACR  |124|ADR

SQL server 2017.

Thank you in advance.

My approach (Inspired by the post from @The Impaler)

;with CTEall as(
select A.id as AID, A.NAME as Aname, b.id as BID,b.NAME as Bname  from A
inner join B on A.id = B.id),
match as (
select A.id as AID, A.NAME as Aname, b.id as BID,b.NAME as Bname  
from A inner join B on A.id = B.id and A.NAME = B.NAME)
select *
from CTEall 
where Aname not in (select Aname from match where AID = BID) 
and Bname not in (select Aname from match where BID = AID)
union all
select * from match
order by 1

Upvotes: 0

Views: 54

Answers (2)

The Impaler
The Impaler

Reputation: 48810

I would do:

with
m as ( -- matched rows
  select a.id as aid, a.name as aname, b.id as bid, b.name as bname
  from table_a a
  join table_b b on a.id = b.id and a.name = b.name
),
l as ( -- unmatched "left rows"
  select a.id, a.name,
    row_number() over(partition by id order by name) as rn
  from table_a a
  left join table_b b on a.id = b.id and a.name = b.name
  where b.id is null
),
r as ( -- unmatched "right rows"
  select b.id, b.name,
    row_number() over(partition by id order by name) as rn
  from table_b b
  left join table_a a on a.id = b.id and a.name = b.name
  where a.id is null
)
select aid, aname, bid, bname from m
union all
select l.id, l.name, r.id, r.name
from l
join r on r.id = l.id and r.rn = l.rn

Note: This solution may be a little bit overkill, since matches all unmatched rows when there are multiple ones per ID... something that is not necessary. Per OP comments there always be a single unmatched row per ID.

Upvotes: 1

Tab Alleman
Tab Alleman

Reputation: 31785

Often when you think about the logic you want in a different way, the answer (or at least AN answer) becomes obvious.

I am thinking of your logic this way:

JOIN Table A to Table B such that A.ID=B.ID (always) AND EITHER A.Name=B.Name OR A.Name doesn't have a Match in B, and B.Name doesn't have a match in A.

This logic is pretty easy to express in SQL

WHERE a.ID=b.ID 
AND (
  a.Name=b.Name OR (
    NOT EXISTS(SELECT * FROM TableB b2 WHERE b2.ID=a.ID AND b2.Name=a.Name)
    AND 
    NOT EXISTS(SELECT * FROM TableA a2 WHERE a2.ID=b.ID AND a2.Name=b.Name)
  )
)

Upvotes: 4

Related Questions