Migrating legacy table to normalized data structure with foreign keys in Oracle SQL - sql

Question

I am having some trouble wrapping my head around remaking databases. I have a book database that includes only one table, where all of the authors data is included after each book. I'm trying to remake this database in order to have an author table and a book table.
I made the author table using :
CREATE TABLE AUTHORS
AS SELECT AUTHOR_NAME, AUTHOR_SURNAME, AUTHOR_BIRTHDATE
If I now want to remake the book table, how do I add the foreign key so that the author of each book will be the correct one? That is, if the first entry on the original book table was:
ISBN1 Title1 Author_Name1 Author_Surname1 Author_Birthdate1
How do I import this data into the new table so that the new author field, a foreign key, references the correct entry in the author table? Sorry if it's confusing.

Answer 1

You are looking to split the existing table into two tables, one to store the authors and the other for books. For this to work properly, you need to create a unique id for each author. Here is a step by step approach.
Assuming the following legacy data structure:
create table old_books (
isbn NUMBER(13, 0),
title VARCHAR2(200),
author_name VARCHAR2(200),
author_surname VARCHAR2(200),
author_birthdate DATE
);
And this sample data:
ISBN | TITLE | AUTHOR_NAME | AUTHOR_SURNAME | AUTHOR_BIRTHDATE
------------: | :----- | :---------- | :------------- | :---------------
1000000000001 | book 1 | name 1 | surname 1 | 01-MAR-90
1000000000002 | book 2 | name 2 | surname 2 | 01-MAR-95
1000000000003 | book 3 | name 1 | surname 1 | 01-MAR-90
First, let's create and feed the new data structure for authors (note that you don't want to use CREATE TABLE AS SELECT ... because this does not let you add constraints or other useful options).
To generate a unique author id, we use the IDENTITY feature (available starting Oracle 12c - without this feature, we would need to create a sequence and a trigger).
In legacy data, we assume that each author is uniquely identified by its name, surname and birthdate:
CREATE TABLE authors (
id NUMBER GENERATED ALWAYS AS IDENTITY,
name VARCHAR2(200),
surname VARCHAR2(200),
birthdate DATE,
PRIMARY KEY (id)
);
INSERT INTO AUTHORS (name, surname, birthdate)
SELECT DISTINCT author_name, author_surname, author_birthdate FROM old_books;
2 rows affected
SELECT * FROM authors;
ID | NAME | SURNAME | BIRTHDATE
-: | :----- | :-------- | :--------
1 | name 1 | surname 1 | 01-MAR-90
2 | name 2 | surname 2 | 01-MAR-95
With this first table in place, we can now create the books table. It contains a foreign key that references the primary key of the authors table. To feed the table, we need to join the legacy table with the new authors table to recover the id of each author:
CREATE TABLE books (
isbn NUMBER(13, 0),
title VARCHAR2(200),
author_id NUMBER,
CONSTRAINT book_author FOREIGN KEY(author_id) REFERENCES authors(id),
PRIMARY KEY (isbn)
);
INSERT INTO books(isbn, title, author_id)
SELECT ob.isbn, ob.title, a.id
FROM old_books ob
INNER JOIN authors a
ON a.name = ob.author_name
AND a.surname = ob.author_surname
AND a.birthdate = ob.author_birthdate;
3 rows affected
SELECT * FROM books;
ISBN | TITLE | AUTHOR_ID
------------: | :----- | --------:
1000000000001 | book 1 | 1
1000000000002 | book 2 | 2
1000000000003 | book 3 | 1
All set! Data is properly spread between the two tables, with the proper constraints in place. We can join both tables with a query like:
SELECT b.isbn, b.title, a.name, a.surname, a.birthdate
FROM authors a
INNER JOIN books b ON a.id = b.author_id;
ISBN | TITLE | NAME | SURNAME | BIRTHDATE
------------: | :----- | :----- | :-------- | :--------
1000000000001 | book 1 | name 1 | surname 1 | 01-MAR-90
1000000000002 | book 2 | name 2 | surname 2 | 01-MAR-95
1000000000003 | book 3 | name 1 | surname 1 | 01-MAR-90

Answer 2

You say that an author's first name plus surname are your author table's primary key. This is a valid approach. In case of two authors with the same name you'd have to find a solution like 'John' + 'Smith' and 'John R.' + 'Smith' or 'John' + 'Smith (the fantasy author)'. This is called a natural composite key, albeit not a perfect one as we may have to deal with duplicate names as mentioned. On the other hand there exist authors with the same name, so we may face this problem right away ;-)
Books are identified by their ISBN, which makes for an even better natural key, because there can be no duplicates. (Only if you wanted to add very old books or self-marketed books that have no ISBN, you'd have to create a fake ISBN.)
In order to have your book referring to an author, you must include the whole key, which is first and surname here. This is no redundancy, as this is the key needed to identify an author in your database.
CREATE TABLE books AS SELECT isbn, title, author_name, author_surname FROM old_table;
ALTER TABLE books ADD CONSTRAINT fk_book_author FOREIGN KEY (author_name, author_surname)
REFERENCES authors (author_name, author_surname);
An alternative would be to introduce surrogate (i.e. technical) keys. You would generate an ID (a number) for each book and each author and work with them. (That means the book table would contain an author_id.) But for a good database you should still think about what identifies a row naturally. This makes it easier for people who write the queries later. (E.g. someone asks to select a list of authors and the number of books they've written. How to write that query? Does it suffice to show first and surname or could we end up with two rows "John Smith | 5" and "John Smith | 2" and the enquirer saying they cannot use this ambiguous result?) Even when providing surrogate keys you should still have a unique constraint on the natural key, if there is one. For books with optional ISBNs this may be title + author_id and for authors it could be first name + surname + date of birth.
By the way: There exist books with more than one author ;-)