Azure SQL: Delete User records with redundant nested references - sql

Question

I have a quite complex requirement to delete UserSubscriptions records with redundant options, meaning if a UserId is associated with two or more subscriptions with same options I need to keep the first subscription and then delete the rest of the subscriptions, below are some of the scenarions.
Scenario 1: UserId 1 has three subscriptions (SubscriptionId 1, 2 and 3) and all of the subscriptions have same options (Email, Call, Fax) so for UserId 1 so the SubscriptionOptionIds 4,5,6,7,8,9 have to be deleted from UserSubscriptions
Scenario 2: UserId 2 has two subscriptions (SubscriptionId 1 and 2) and both the subscriptions DOES NOT have same options in this case nothing needs to be deleted
Scenarion 3: UserId 3 has two subscriptions (SubscriptionId 1 and 2) and both the subscriptions have same option (Email) so for UserId 3 so the SubscriptionOptionId 3 has to be deleted from UserSubscriptions
Below is my Table's DDL+DML
CREATE TABLE [Options](
[Id] [int] IDENTITY(1,1) NOT NULL,
[Name] [nvarchar](255) NOT NULL
) ON [PRIMARY]
GO
ALTER TABLE [Options] ADD CONSTRAINT [PK_Options] PRIMARY KEY CLUSTERED
(
[Id] ASC
)WITH (STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ONLINE = OFF, OPTIMIZE_FOR_SEQUENTIAL_KEY = OFF) ON [PRIMARY]
GO
CREATE TABLE [Subscriptions](
[Id] [int] IDENTITY(1,1) NOT NULL,
[Name] [nvarchar](255) NOT NULL
) ON [PRIMARY]
GO
ALTER TABLE [Subscriptions] ADD CONSTRAINT [PK_Subscriptions] PRIMARY KEY CLUSTERED
(
[Id] ASC
)WITH (STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ONLINE = OFF, OPTIMIZE_FOR_SEQUENTIAL_KEY = OFF) ON [PRIMARY]
GO
CREATE TABLE [SubscriptionsOptions](
[Id] [int] IDENTITY(1,1) NOT NULL,
[OptionId] [int] NOT NULL,
[SubscriptionId] [int] NOT NULL
) ON [PRIMARY]
GO
ALTER TABLE [SubscriptionsOptions] ADD CONSTRAINT [PK_SubscriptionsOptions] PRIMARY KEY CLUSTERED
(
[Id] ASC
)WITH (STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ONLINE = OFF, OPTIMIZE_FOR_SEQUENTIAL_KEY = OFF) ON [PRIMARY]
GO
ALTER TABLE [SubscriptionsOptions] WITH CHECK ADD CONSTRAINT [FK_SubscriptionsOptions.SubscriptionId_Subscriptions.Id] FOREIGN KEY([SubscriptionId])
REFERENCES [Subscriptions] ([Id])
GO
ALTER TABLE [SubscriptionsOptions] CHECK CONSTRAINT [FK_SubscriptionsOptions.SubscriptionId_Subscriptions.Id]
GO
ALTER TABLE [SubscriptionsOptions] WITH CHECK ADD CONSTRAINT [FK_SubscriptionsOptions.OptionId_Options.Id] FOREIGN KEY([OptionId])
REFERENCES [Options] ([Id])
GO
ALTER TABLE [SubscriptionsOptions] CHECK CONSTRAINT [FK_SubscriptionsOptions.OptionId_Options.Id]
GO
CREATE TABLE [UserSubscriptions](
[Id] [int] IDENTITY(1,1) NOT NULL,
[SubscriptionsOptionsId] [int] NOT NULL,
[userid] [int] NOT NULL
) ON [PRIMARY]
GO
ALTER TABLE [UserSubscriptions] ADD CONSTRAINT [PK_UserSubscriptions] PRIMARY KEY CLUSTERED
(
[Id] ASC
)WITH (STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ONLINE = OFF, OPTIMIZE_FOR_SEQUENTIAL_KEY = OFF) ON [PRIMARY]
GO
ALTER TABLE [UserSubscriptions] WITH CHECK ADD CONSTRAINT [FK_UserSubscriptions.SubscriptionsOptionsId_SubscriptionsOptions.Id] FOREIGN KEY([SubscriptionsOptionsId])
REFERENCES [SubscriptionsOptions] ([Id])
GO
ALTER TABLE [UserSubscriptions] CHECK CONSTRAINT [FK_UserSubscriptions.SubscriptionsOptionsId_SubscriptionsOptions.Id]
GO
INSERT INTO Options VALUES ('E-mail');
INSERT INTO Options VALUES ('Call');
INSERT INTO Options VALUES ('Fax');
INSERT INTO Subscriptions VALUES ('Promo1');
INSERT INTO Subscriptions VALUES ('Promo2');
INSERT INTO Subscriptions VALUES ('Promo3');
INSERT INTO SubscriptionsOptions VALUES (1,1);
INSERT INTO SubscriptionsOptions VALUES (1,2);
INSERT INTO SubscriptionsOptions VALUES (1,3);
INSERT INTO SubscriptionsOptions VALUES (2,1);
INSERT INTO SubscriptionsOptions VALUES (2,2);
INSERT INTO SubscriptionsOptions VALUES (2,3);
INSERT INTO SubscriptionsOptions VALUES (3,1);
INSERT INTO SubscriptionsOptions VALUES (3,2);
INSERT INTO SubscriptionsOptions VALUES (3,3);
INSERT INTO UserSubscriptions VALUES (1,1);
INSERT INTO UserSubscriptions VALUES (1,2);
INSERT INTO UserSubscriptions VALUES (1,3);
INSERT INTO UserSubscriptions VALUES (1,4);
INSERT INTO UserSubscriptions VALUES (1,5);
INSERT INTO UserSubscriptions VALUES (1,6);
INSERT INTO UserSubscriptions VALUES (1,7);
INSERT INTO UserSubscriptions VALUES (1,8);
INSERT INTO UserSubscriptions VALUES (1,9);
INSERT INTO UserSubscriptions VALUES (2,1);
INSERT INTO UserSubscriptions VALUES (2,2);
INSERT INTO UserSubscriptions VALUES (2,3);
INSERT INTO UserSubscriptions VALUES (2,4);
INSERT INTO UserSubscriptions VALUES (2,5);
INSERT INTO UserSubscriptions VALUES (3,1);
INSERT INTO UserSubscriptions VALUES (3,3);
Options
1 E-mail
2 Call
3 Fax
Subscriptions
1 Promo1
2 Promo2
3 Promo3
SubscriptionsOptions
id SubscriptionId OptionId
1 1 1
2 1 2
3 1 3
4 2 1
5 2 2
6 2 3
7 3 1
8 3 2
9 3 3
UserSubscriptions
userid SubscriptionOptionId
1 1
1 2
1 3
1 4
1 5
1 6
1 7
1 8
1 9
2 1
2 2
2 3
2 4
2 5
3 1
3 3
The final output of UserSubscriptions has to be like below, am having hard time for the deletion script for the above mentioned scenarios, I really appreciate if someone can help me with the deletion script
userid SubscriptionOptionId
1 1
1 2
1 3
2 1
2 2
2 3
2 4
2 5
3 1

Answer 1

Your requirement does not seem to match your expected output.
Your requirement says "delete UserSubscriptions records with redundant options, meaning if a UserId is associated with two or more subscriptions with same options I need to keep the first subscription and then delete the rest of the subscriptions". But in your expected output that is not satisfied.
For example, in your expected output you have deleted subscriptionOptionId 4 from userId 1, but you have not deleted it from userId 2. Why not?
Conversely, for userId 3 you have deleted subscriptionOptionId 3 from userSubscrptions, but subscriptionOptionId 1 has optionId 1, whereas subscriptionOptionId 3 has optionId 3. These are different options, so according to your requirement neither should have been deleted.
A query which actually deletes rows according to your stated requirement (but which will not produce your expected result) would be:
delete sub
from userSubscriptions sub
join SubscriptionsOptions opt on opt.id = sub.SubscriptionsOptionsId
where exists
(
select *
from userSubscriptions prevSub
join SubscriptionsOptions prevOpt on prevOpt.Id = prevSub.SubscriptionsOptionsId
where prevSub.userId = sub.userId
and prevSub.id < sub.id
)
Having said that, if I think about what this is modelling a little bit, I think the semantics of your requirements as described aren't really what you want. If I, as a user, am subscribed to two different promotions, it seems sensible that I could have both of those promotions use the same delivery option.
However, what would be redundant would be a situation where I am subscribed to a promotion, and the database says that I want that promo sent to me in a particular way, and it says it twice. For example, "allmhuran is subscribed to promo1 and wants it delivered by email", and "allmhuran is subscribed to promo1 and wants it delivered by email".
But there are no such redundancies in your sample data, because there are no redundancies in your subscriptionsOptions table. Every row is a unique combination of {subscriptionId, optionId}.
On the off chance that this is what you actually wanted to do, the delete statement only changes slightly:
delete sub
from userSubscriptions sub
join SubscriptionsOptions opt on opt.id = sub.SubscriptionsOptionsId
where exists
(
select *
from userSubscriptions prevSub
join SubscriptionsOptions prevOpt on prevOpt.Id = prevSub.SubscriptionsOptionsId
where prevSub.userId = sub.userId
and prevOpt.subscriptionId = opt.subscriptionId -- this condition was not in the previous statement
and prevSub.id < sub.id
)
If that is the intention, then I would also say that the existing schema is "overnormalized". All you really need is a UserSubscriptionOptions table (userId int, subscriptionId int, optionId int, primary key (userId, subscriptionId, optionId)

Answer 2

This is simpler than it seems:
Create a view over UserSubscriptions joined to SubscriptionOptions
Order by user, than as desired e.g. from "earliest" to "latest"
Add a hash across user id and option id
For each duplicate hash delete the corresponding SubscriptionOptions row
Once done, delete any Subscriptions rows with no corresponding SubscriptionOptions rows
This is doable in SQL or the imperative language of your choice.