Making combinations of attributes unique in PostgreSQL - sql

Question

There is an option in postgresql where we can have a constraint such that we can have multiple attributes of a table together as unique
UNIQUE (A, B, C)
Is it possible to take attributes from multiple tables and make their entire combination as unique in some way
Edit:
Table 1: List of Book
Attributes: ID, Title, Year, Publisher
Table 2: List of Author
Attributes: Name, ID
Table 3: Written By: Relation between Book and Author
Attributes: Book_ID, Author_ID
Now I have situation where I don't want (Title, Year, Publisher, Authors) get repeated in my entire database

Answer 1

There are 3 solutions to this problem:
You add a column "authorID" to the table "book", as a foreign key. You can then add the UNIQUE constraint to the table "book".
We can have a foreign key on the 2 columns (bookID, author ID) which references the table bookAuthor.
You create a Trigger on insert on the table "book" which checks whether the combination exist and does not insert if it does exist. You will find a working example of this option below.
Whilst working on this option I realised that the JOIN to WrittenBy must be done on Title and not ID. Otherwise we can record the same book as many times as we like just by using a new ID. The problem with using the title is that the slightest change in spelling or punctuation means that it is treated as a new title.
In the example the 3rd insert has failed because it already exists. In the 4th have left 2 spaces in "Tom Sawyer" and it is accepted as a different title.
Also as we use a join to find out the author the real effect of our rule is exactly the same as if we had a UNIQUE constraint on the table books on columns Title, Year and Publisher. This means that all that I have coded is a waste of time.
We thus decide, after coding it, that this option is not effective.
We could create a fourth table with the 4 columns and a UNIQUE constraint on all 4. This seems a heavy solution compared to option 1.
CREATE TABLE Book (
ID int primary key,
Title varchar(25),
Year int,
Publisher varchar(10) );
CREATE TABLE Author (
ID int primary key,
Name varchar(10)
);
CREATE TABLE WrittenBy(
Book_ID int primary key,
Titlew varchar(25),
Author_ID int
);
CREATE FUNCTION book_insert_trigger_function()
RETURNS TRIGGER
LANGUAGE PLPGSQL
AS $$
DECLARE
authID INTEGER;
coun INTEGER;
BEGIN
IF pg_trigger_depth() <> 1 THEN
RETURN NEW;
END IF;
SELECT MAX(Author_ID) into authID
FROM WrittenBy w
WHERE w.Titlew = NEW.Title;
SELECT COUNT(*) INTO coun FROM
Book b LEFT JOIN WrittenBy w ON
b.Title = w.Titlew
WHERE NEW.year = b.year
AND NEW.title=b.title
AND NEW.publisher=b.publisher
AND authID = COALESCE(w.Author_ID,authID);
IF coun > 0 THEN
RETURN null; -- this means that we do not insert
ELSE
RETURN NEW;
END IF;
END;
$$
;
CREATE TRIGGER book_insert_trigger
BEFORE INSERT
ON Book
FOR EACH ROW
EXECUTE PROCEDURE book_insert_trigger_function();
INSERT INTO WrittenBy VALUES
(1,'Tom Sawyer',1),
(2,'Huckleberry Finn',1);
INSERT INTO Book VALUES (1,'Tom Sawyer',1950,'Classics');
INSERT INTO Book VALUES (2,'Huckleberry Finn',1950,'Classics');
INSERT INTO Book VALUES (3,'Tom Sawyer',1950,'Classics');
INSERT INTO Book VALUES (3,'Tom Sawyer',1950,'Classics');
SELECT *
FROM Book b
LEFT JOIN WrittenBy w on w.Titlew = b.Title
LEFT JOIN Author a on w.author_ID = a.ID;
>
> id | title | year | publisher | book_id | titlew | author_id | id | name
> -: | :--------------- | ---: | :-------- | ------: | :--------------- | --------: | ---: | :---
> 1 | Tom Sawyer | 1950 | Classics | 1 | Tom Sawyer | 1 | null | null
> 2 | Huckleberry Finn | 1950 | Classics | 2 | Huckleberry Finn | 1 | null | null
> 3 | Tom Sawyer | 1950 | Classics | null | null | null | null | null
>
db<>fiddle here