Unique constraint including field in referenced table - sql-server

Question

Is is possible to add a uniqueness constraint to a table that would include a field from a referenced table? If so, how can that be done?
(Note that I'm using Microsoft SQL Server.)
To explain why this is necessary, here's a simplified version of my problem.
Firstly, let's say I have a table of things, defining the fixed properties of each individual thing:
CREATE TABLE dbo.things(
id INT IDENTITY(1,1) NOT NULL
CONSTRAINT [things$PrimaryKey] PRIMARY KEY CLUSTERED,
thing_name NVARCHAR(20) NOT NULL
CONSTRAINT [things$thing_name] UNIQUE,
-- etc.
);
I also have different configurations, for a range of different elements, not just things, and these are captured in a table too:
CREATE TABLE dbo.configurations(
id INT IDENTITY(1,1) NOT NULL
CONSTRAINT [configurations$PrimaryKey] PRIMARY KEY CLUSTERED,
config_name NVARCHAR(20) NOT NULL
CONSTRAINT [configurations$config_name] UNIQUE,
-- etc.
);
I then want to have different configurations for each thing:
CREATE TABLE dbo.thing_configurations(
id INT IDENTITY(1,1) NOT NULL
CONSTRAINT [thing_configurations$PrimaryKey] PRIMARY KEY CLUSTERED,
thing_id INT NOT NULL
CONSTRAINT [thing_configurations$things_fk] REFERENCES dbo.things(id),
quantity INT NOT NULL
CONSTRAINT [thing_configurations$quantity] CHECK(quantity >= 0),
-- etc.
);
Next, I want to identify the configuration of each thing for each particular configuration in an association table. I can ensure that the configuration and associated thing configuration are unique, but I want to ensure that each thing has no more than one thing configuration for each configuration.
CREATE TABLE dbo.config_thing_configurations(
config_id INT NOT NULL
CONSTRAINT [config_thing_configurations$configurations_fk] REFERENCES dbo.configurations(id),
thing_config_id INT NOT NULL
CONSTRAINT [config_thing_configurations$thing_configurations_fk] REFERENCES dbo.thing_configurations(id)
-- How can I ensure that I do not have multiple configurations of a each thing in a configuration.
);
More specifically, say I have these things:
id
thing_name
1
Apple
2
Bicycle
...
...
And these configurations:
id
config_name
1
Base configuration.
2
Experiment #1
...
...
And these thing configurations:
id
thing_id
quantity
1
1 (Apple)
20
2
1 (Apple)
30
3
2 (Bicycle)
5
...
...
How can I prevent each configuration from having two or more different configurations of each individual thing in the thing_configurations table?
config_id
thing_config_id
1
1 (20 Apples)
1
2 (30 Apples)
...
...
For example, in the above table, I have two conflicting configurations for Apples in configuration 1.
Ideally, I'd like to be able to enforce a uniqueness constraint for config_id and the thing_id referenced through the thing_config_id. Is that possible?

Answer 1

So this is possible, through the creation of an indexed view in combination with a unique index.
In this specific example, this is achieved by first creating a view that associates the configuration and thing identifiers:
CREATE VIEW dbo.thing_config_view WITH SCHEMABINDING AS
SELECT
thing_configurations.thing_id,
config_thing_configurations.config_id
FROM dbo.config_thing_configurations
INNER JOIN dbo.thing_configurations ON config_thing_configurations.thing_config_id = thing_configurations.id;
Next, we create a unique index on those fields from the view:
CREATE UNIQUE CLUSTERED INDEX [thing_config_view$unique_thing_configurations] ON dbo.thing_config_view(thing_id, config_id);
and it works exactly as required.

Answer 2

This could be handled through Triggers. The trigger content would depend on what you want to occur when a query is called that would violate the condition. Something like this for insert (untested):
CREATE TRIGGER tr_config_thing_configurations ON config_thing_configurations
INSTEAD OF INSERT
AS
IF EXISTS(
SELECT 1
FROM inserted i
INNER JOIN thing_configurations tc ON tc.id = i.thing_config_id
INNER JOIN things t ON t.id = tc.thing_id
WHERE t.id IN (
SELECT it.id
FROM config_thing_configurations ictc
INNER JOIN thing_configurations itc ON itc.id = ictc.thing_config_id
INNER JOIN things it ON it.id = itc.thing_id
WHERE
ictc.config_id = i.config_id
)
BEGIN
RAISERROR ('Configuration already has a configuration for thing', 16, 1);
ROLLBACK TRANSACTION;
RETURN
END
INSERT INTO config_thing_configurations (config_id, thing_config_id) SELECT config_id, thing_config_id FROM inserted;
GO
You would need another for update or you could write a single to handle both cases.