CODEWITHSUNDEEP
sqloracle-database
finiteloop
finiteloop

Reputation: 4496

Counting Pairs in SQL

I have a database that contains 2 tables:

Users (user_id primary key), and Friends

The friends table is organized into 2 columns friend1, friend2, both of which contain user_ids as (foreign keys referencing Users). In each friend pair, friend1's user id is less than friend 2's user id.

I am trying to find a list of users who are not friends but who share the greatest amount of friends.

I have managed to do this in two separate queries:

Number of Shared Friends for users u1 and u2

select count(*)
from 
    ((select friend1 from friends where friend2 = u1 UNION 
    select friend2 from friends where friend1 = u1)
    INTERSECT
    (select friend1 from friends where friend2 = u2 UNION 
    select friend2 from friends where friend1 = u2))
;

Set of all user_id -> user_id pairs who are not friends:

select distinct
    u1.user_id as friend1,
    u2.user_id as friend2
from
    users u1,
    users u2
where
    u1.user_id < u2.user_id
minus
select friend1, friend2
from friends order by friend1;

However my ultimate goal is to get a result that is

user1  user2  shared_friends

such that user1 < user2, and user1 and user2 are not friends, and shared_friends is the count of how many friends these users have in common, which I have thus far been unsuccessful in achieving.

Upvotes: 1

Views: 387

Answers (2)

Thomas
Thomas

Reputation: 64674

The CTE's are just to provide some sample data.

With Users As
    (
    Select 1 As UserId, 'Alice' As Name
    Union All Select 2, 'Bob'
    Union All Select 3, 'Caroline'
    Union All Select 4, 'Doug'
    )
    , Friends As
    (
    Select 1 As Friend1, 2 As Friend2
    Union All Select 2, 1
    Union All Select 2, 3
    Union All Select 2, 4
    Union All Select 3, 1
    Union All Select 3, 4
    )
    , UserFriends As
    (
    Select U1.UserId
        , Case
            When F1.Friend1 = U1.UserId Then F1.Friend2
            Else F1.Friend1
            End As Friend
    From Users As U1
        Join Friends As F1
            On U1.UserId In(F1.Friend1,F1.Friend2)
    Group By U1.UserId
        , Case
            When F1.Friend1 = U1.UserId Then F1.Friend2
            Else F1.Friend1
            End     
    )
Select U1.Name, U2.Name
    , Count(*) As MutualFriendCount
    , Group_Concat(F.Name) As SharedFriends
From UserFriends As UF1
    Join UserFriends As UF2
        On UF2.Friend = UF1.Friend
    Join Users As U1
        On U1.UserId = UF1.UserId
    Join Users As U2
        On U2.UserId = UF2.UserId
    Join Users As F
        On F.UserId = UF1.Friend
            And F.UserId = UF2.Friend
Where UF1.UserId <> UF2.UserId
    And Not Exists  (
                    Select 1
                    From UserFriends As F3
                    Where F3.UserId = UF1.UserId
                        And F3.Friend = UF2.UserId
                    )
Group By U1.Name, U2.Name
Order By U1.Name, U2.Name

Upvotes: 0

Sparky
Sparky

Reputation: 15105

This should get you started, but it is Microsoft SQL (although pretty generic)

select a1.friend1 as User1,
       a2.friend1 as user2,
       count( distinct a1.friend2) as Shared_Friends
from friends a1 
join (select distinct friend1,friend2 from friends a2) a2 
       on a1.friend2=a2.friend2
left join friends a3 on a3.friend1=a1.friend1 and a3.friend2 = a2.friend1
where (a1.friend1=1 and a2.friend1=8) and a3.friend1 is null
group by a1.friend1,a2.friend1

Upvotes: 0

Related Questions