Identifying records with inconsistent relationships - sql

Question

You can run all this SQL and see the results here.
Skip to the results ands the ensuing problems to get to the meat of the issue.
I have a table of Clubs (a club as in a group or organization of people, like a "swim club" or a "knitting club").
DECLARE #club TABLE (
Id INT
,Name NVARCHAR(255)
);
INSERT INTO #club VALUES
(1, 'Swim Club')
,(2, 'Knitting Club')
,(3, 'Bridge Club');
I have a table of Members.
DECLARE #member TABLE (
Id INT
,Name NVARCHAR(255)
);
INSERT INTO #member VALUES
(1, 'John Jones')
,(2, 'Sally Smith')
,(3, 'Rod Roosevelt')
,(4, 'Bobby Burns')
,(5, 'Megan Moore');
Members can belong to many Clubs, and so there is a Membership table that connects Clubs to Members (and also describes the membership dues price).
DECLARE #membership TABLE (
Id INT
,Member INT --FK to #member
,Club INT --FK to #club
,Dues INT --the cost of membership
);
INSERT INTO #membership VALUES
(1,1,1,10)
,(2,1,2,5)
,(3,2,1,10)
,(4,2,3,20)
,(5,3,1,10)
,(6,3,2,5)
,(7,4,2,5)
,(8,4,3,20)
,(9,5,1,10)
,(10,5,3,20);
Most Members just pay their associated Dues. Some Members, however, are Sponsored by other Members. And so those Sponsored members will have their Dues paid by another Member (the Sponsor). Therefore, we have a Sponsorship table. The Sponsorship table connects Sponsors (paying for the Dues) to a Sponsee (having their Dues paid by a Sponsor) for a particular Club. Because a Sponsorship is specific to a Club, a Sponsorship record connects two Membership records and NOT two Member records.
DECLARE #sponsorship TABLE (
Id INT
,Sponsee_Membership INT --FK to Sponsee's #membership record
,Sponsor_Membership INT --FK to Sponsor's #membership record
);
INSERT INTO #sponsorship VALUES
(1,5,1)
,(2,8,4)
,(3,9,3)
,(4,10,4);
To get a full view of our Clubs/Memberships/Sponsors, we have:
SELECT
mship.Id AS 'Mship'
,mem.Name AS 'Member'
,c.Name AS 'Club'
,mship.Dues
,spons_mem.Name AS 'Sponsor'
FROM
#membership AS mship
JOIN #member AS mem
ON mship.Member = mem.Id
JOIN #club AS c
ON mship.Club = c.Id
LEFT JOIN #sponsorship AS spons
ON spons.Sponsee_Membership = mship.Id
LEFT JOIN #membership AS spons_mship
ON spons_mship.Id = spons.Sponsor_Membership
LEFT JOIN #member AS spons_mem
ON spons_mem.Id = spons_mship.Member;
which gives us these Results:
Mship Member Club Dues Sponsor
1 John Jones Swim Club 10 NULL
2 John Jones Knitting Club 5 NULL
3 Sally Smith Swim Club 10 NULL
4 Sally Smith Bridge Club 20 NULL
5 Rod Roosevelt Swim Club 10 John Jones
6 Rod Roosevelt Knitting Club 5 NULL
7 Bobby Burns Knitting Club 5 NULL
8 Bobby Burns Bridge Club 20 Sally Smith
9 Megan Moore Swim Club 10 Sally Smith
10 Megan Moore Bridge Club 20 Sally Smith
Sponsorships SHOULD span all shared memberships.
That is, if Sally sponsors Bobby, any time they are both in the same club, Sally will be identified as Bobby's sponsor.
We can see this in lines Mship=7 and Mship=8.
Bobby and Sally are both in the Bridge Club, so Sally is identified as Bobby's sponsor for his Bridge Club membership.
Sally is NOT a member of the Knitting Club, and so Bobby's Knitting Club membership does not show Sally as a sponsor.
Sorry for the long setup. Here's my actual question:
How can I identify where a Sponsorship is missing?
From the example, we have lines Mship=5 and Mship=6.
John is Rod's sponsor.
We can see the sponsorship for Rod's Swim Club membership.
Rod and John are also both Knitting Club members,
but Rod does not show John as a sponsor for his Knitting Club membership.
This is incorrect and this is what I am after.
I want to query all such missing sponsorships.
I can accomplish this using cursors / WHILE loops, but I know that such solutions are usually not taking a proper set-based approach. What would a proper query for this look like?
Many thanks.

Answer 1

Here is a SQL query that might respond to your requirement.
The logic is to use a subquery to generate a mapping between sponsors and sponsees based on mapping member.id instead of memberships.id ; for this, we use aggregation. Then, the outer query searches for clubs where both the sponsor and the sponsee participate, but for which no relation is declared in the sponsorship table
The query returns one record for each offending membership, with the sponsee and sponsor names.
SELECT mship1.Id, m1.Name Member, m2.Name Sponsor, c.Name Club, mship1.Dues
FROM
#membership mship1
INNER JOIN #club c ON c.Id = mship1.Club
INNER JOIN (
SELECT ms1.Member Sponsee_Member , MAX(ms2.Member) Sponsor_Member
FROM #sponsorship ss
INNER JOIN #membership ms1 ON ms1.Id = Sponsee_Membership
INNER JOIN #membership ms2 ON ms2.Id = Sponsor_Membership
GROUP BY ms1.Member
) rels ON rels.Sponsee_Member = mship1.Member
INNER JOIN #membership mship2 ON mship2.Member = rels.Sponsor_Member AND mship2.Club = mship1.Club
INNER JOIN #member m1 ON m1.Id = mship1.Member
INNER JOIN #member m2 ON m2.Id = mship2.Member
LEFT JOIN #sponsorship sship ON sship.Sponsor_Membership = mship2.Id
WHERE sship.Id IS NULL
;
In the rextester that you provided, this returns :
Id | Member | Sponsor | Club | Dues
-----|-----------------|--------------|----------------|-----
6 | Rod Roosevelt | John Jones | Knitting Club | 5
Working on creating this query got me thinking that you could optimize your database design. The current model will make it hard to maintain consistency : your question by itself demonstrates that. In the future, what happens if a sponsor registers in a new club, where one of his sponsee already participate ? Once again you will need to detect the missing sponshorship relation, and somehow create it.
You actually have a 1-1 relationship between sponsors and sponsee, since you stated that sponsorships SHOULD span all shared memberships. It does not look like you would allow a sponsee to have several sponsors, even across different clubs.
I would suggest that you drop the sponshorship table and store a self-foreign key to the sponsor directly in the member table. Starting from there, it is easy to check which clubs both members have in common, and properly assign the dues using a SQL query.

Answer 2

I found that I can gather this in a single query by gathering all Sponsored Relationships with a CTE, finding all memberships that SHOULD have a sponsorship based on that, then removing all existing sponsorships with an EXCEPT. I'm left with sponsorships that should exist but do not.
WITH sponsored_relationships AS (
SELECT DISTINCT
sponsee_member.Id AS Sponsee
,sponsor_member.Id AS Sponsor
FROM
#sponsorship AS s
JOIN #membership AS sponsee_mship
ON s.Sponsee_Membership = sponsee_mship.Id
JOIN #member AS sponsee_member
ON sponsee_mship.Member = sponsee_member.Id
JOIN #membership AS sponsor_mship
ON s.Sponsor_Membership = sponsor_mship.Id
JOIN #member AS sponsor_member
ON sponsor_mship.Member = sponsor_member.Id
)
SELECT
see_mem.Name AS Sponsee
,sor_mem.Name AS Sponsor
,c.Name AS Club
FROM
sponsored_relationships AS sr
JOIN #member AS see_mem
ON sr.Sponsee = see_mem.Id
JOIN #membership AS see_mship
ON see_mship.Member = see_mem.Id
JOIN #member AS sor_mem
ON sr.Sponsor = sor_mem.Id
JOIN #membership AS sor_mship
ON sor_mship.Member = sor_mem.Id
JOIN #club AS c
ON (see_mship.Club = c.Id
AND sor_mship.Club = c.Id
)
EXCEPT
SELECT
see_mem.Name AS Sponsee
,sor_mem.Name AS Sponsor
,c.Name AS Club
FROM
sponsored_relationships AS sr
JOIN #member AS see_mem
ON sr.Sponsee = see_mem.Id
JOIN #membership AS see_mship
ON see_mship.Member = see_mem.Id
JOIN #member AS sor_mem
ON sr.Sponsor = sor_mem.Id
JOIN #membership AS sor_mship
ON sor_mship.Member = sor_mem.Id
JOIN #club AS c
ON (see_mship.Club = c.Id
AND sor_mship.Club = c.Id
)
JOIN #sponsorship AS sship
ON (sship.Sponsee_Membership = see_mship.Id
AND sship.Sponsor_Membership = sor_mship.Id
);