Add data to many-to-many relation with one SQL command - sql

Question

I have a basic understanding of SQL databases and I might overlooked something, but I cannot figure out the following problem: there is a many-to-many relationship (for example: users - user_roles - roles). Is it possible to add (new) role to a (new) user with one SQL command (atomic operation)? Currently I use Sqlite.
I am aware of the SELECT last_insert_rowid(); command and with this and several SQL commands I can achieve what I want. But I want to incorporate it into one command (so the server, in this case Sqlite, can optimize the query, etc.). I have no idea, how it is done in real life (one command vs. several one in one transaction), that´s the root cause of this question.
So far this is what I was able to do:
pragma foreign_keys = on;
CREATE TABLE users (
user_id integer primary key autoincrement,
user_name text not null unique
);
CREATE TABLE roles (
role_id integer primary key autoincrement,
role_name text not null unique
);
CREATE TABLE user_roles (
user_id integer not null,
role_id integer not null,
foreign key (user_id) references users(user_id),
foreign key (role_id) references roles(role_id),
primary key (user_id, role_id)
);
insert into users (user_name) values ('Joe');
insert into roles (role_name) values ('admin');
insert into user_roles (user_id, role_id) values (
(select user_id from users where user_name = 'Joe'),
(select role_id from roles where role_name = 'admin')
);
If both user and role exists (Joe and admin), then it works fine.
But I cannot figure out, how to achieve "add-if-missing-then-return-id" behavior if Joe or admin is mission from database.
Example (both user and role are missing):
insert into user_roles (user_id, role_id) values (
(select user_id from users where user_name = 'Bill'),
(select role_id from roles where role_name = 'user')
);
Result:
Execution finished with errors.
Result: NOT NULL constraint failed: user_roles.user_id

Answer 1

You could create view from user_roles table:
CREATE VIEW user_roles_view AS
SELECT
U.user_name, R.role_name
FROM user_roles AS UR
INNER JOIN users AS U ON u.user_id = UR.user_id
INNER JOIN roles AS R ON r.role_id = UR.role_id;
Views in SQLite are read-only unless you create an INSTEAD OF trigger on it. This way you can specify a command or sequence of commands that are executed when the view is modified using INSERT, UPDATE or DELETE statement. For INSERT it could go like this:
CREATE TRIGGER user_roles_view_insert INSTEAD OF INSERT ON user_roles_view
BEGIN
INSERT OR IGNORE INTO users (user_name) VALUES (NEW.user_name);
INSERT OR IGNORE INTO roles (role_name) VALUES (NEW.role_name);
INSERT OR IGNORE INTO user_roles (user_id, role_id) VALUES (
(SELECT user_id FROM users WHERE user_name = NEW.user_name),
(SELECT role_id FROM roles WHERE role_name = NEW.role_name)
);
END;
Note the usage of INSERT OR IGNORE to prevent inserting duplicate values into all of the three tables. Here's how you would insert values via the view:
INSERT INTO user_roles_view VALUES ('Joe', 'admin');
-- The above statement creates:
-- a row (1, 'Joe') in table users,
-- a row (1, 'admin) in table roles,
-- a row (1, 1) in table user_roles.
INSERT INTO user_roles_view VALUES ('Joe', 'admin');
-- The above statement doesn't add any additional records, because all appropriate records
-- already exist.
INSERT INTO user_roles_view VALUES ('Joe', 'system');
-- The above statement creates:
-- a row (2, 'system') in table roles,
-- a row (1, 2) in table user_roles.
INSERT INTO user_roles_view VALUES ('Alice', 'admin'), ('Bob', 'system');
-- The above statement creates:
-- a row (2, 'Alice') in table users,
-- a row (3, 'Bob') in table users,
-- a row (2, 1) in table user_roles,
-- a row (3, 2) in table user_roles
All of the above statements produce the following output from user_roles_view (SELECT * FROM user_roles_view):
user_name
role_name
Joe
admin
Joe
system
Alice
admin
Bob
system