Desk joins in Fluent 4

Database fashions

Fluent is a Swift ORM framework written for Vapor. You need to use fashions to characterize rows in a desk, migrations to create the construction for the tables and you’ll outline relations between the fashions utilizing Swift property wrappers. That is fairly a easy approach of representing father or mother, baby or sibling connections. You may “keen load” fashions by these predefined relation properties, which is nice, however typically you do not need to have static sorts for the relationships.

I am engaged on a modular CMS and I am unable to have hardcoded relationship properties contained in the fashions. Why? Effectively, I need to have the ability to load modules at runtime, so if module A relies upon from module B by a relation property then I am unable to compile module A independently. That is why I dropped many of the cross-module relations, nonetheless I’ve to jot down joined queries. 😅

Buyer mannequin

On this instance we’re going to mannequin a easy Buyer-Order-Product relation. Our buyer mannequin may have a fundamental identifier and a reputation. Think about the next:

last class CustomerModel: Mannequin, Content material {
    static let schema = "prospects"
    
    @ID(key: .id) var id: UUID?
    @Discipline(key: "title") var title: String

    init() { }

    init(id: UUID? = nil, title: String) {
        self.id = id
        self.title = title
    }
}

Nothing particular, only a fundamental Fluent mannequin.

Order mannequin

Clients may have a one-to-many relationship to the orders. Which means a buyer can have a number of orders, however an order will all the time have precisely one related buyer.

last class OrderModel: Mannequin, Content material {
    static let schema = "orders"
    
    @ID(key: .id) var id: UUID?
    @Discipline(key: "date") var date: Date
    @Discipline(key: "customer_id") var customerId: UUID

    init() { }

    init(id: UUID? = nil, date: Date, customerId: UUID) {
        self.id = id
        self.date = date
        self.customerId = customerId
    }
}

We may benefit from the @Dad or mum and @Little one property wrappers, however this time we’re going to retailer a customerId reference as a UUID kind. In a while we’re going to put a overseas key constraint on this relation to make sure that referenced objects are legitimate identifiers.

Product mannequin

The product mannequin, similar to the client mannequin, is completely impartial from anything. 📦

last class ProductModel: Mannequin, Content material {
    static let schema = "merchandise"
    
    @ID(key: .id) var id: UUID?
    @Discipline(key: "title") var title: String

    init() { }

    init(id: UUID? = nil, title: String) {
        self.id = id
        self.title = title
    }
}

We will create a property with a @Sibling wrapper to specific the connection between the orders and the merchandise, or use joins to question the required information. It actually does not matter which approach we go, we nonetheless want a cross desk to retailer the associated product and order identifiers.

OrderProductModel

We will describe a many-to-many relation between two tables utilizing a 3rd desk.

last class OrderProductModel: Mannequin, Content material {
    static let schema = "order_products"
    
    @ID(key: .id) var id: UUID?
    @Discipline(key: "order_id") var orderId: UUID
    @Discipline(key: "product_id") var productId: UUID
    @Discipline(key: "amount") var amount: Int

    init() { }

    init(id: UUID? = nil, orderId: UUID, productId: UUID, amount: Int) {
        self.id = id
        self.orderId = orderId
        self.productId = productId
        self.amount = amount
    }
}

As you possibly can see we will retailer additional data on the cross desk, in our case we’re going to affiliate portions to the merchandise on this relation proper subsequent to the product identifier.

Migrations

Happily, Fluent provides us a easy option to create the schema for the database tables.

struct InitialMigration: Migration {

    func put together(on db: Database) -> EventLoopFuture<Void> {
        db.eventLoop.flatten([
            db.schema(CustomerModel.schema)
                .id()
                .field("name", .string, .required)
                .create(),
            db.schema(OrderModel.schema)
                .id()
                .field("date", .date, .required)
                .field("customer_id", .uuid, .required)
                .foreignKey("customer_id", references: CustomerModel.schema, .id, onDelete: .cascade)
                .create(),
            db.schema(ProductModel.schema)
                .id()
                .field("name", .string, .required)
                .create(),
            db.schema(OrderProductModel.schema)
                .id()
                .field("order_id", .uuid, .required)
                .foreignKey("order_id", references: OrderModel.schema, .id, onDelete: .cascade)
                .field("product_id", .uuid, .required)
                .foreignKey("product_id", references: ProductModel.schema, .id, onDelete: .cascade)
                .field("quantity", .int, .required)
                .unique(on: "order_id", "product_id")
                .create(),
        ])
    }

    func revert(on db: Database) -> EventLoopFuture<Void> {
        db.eventLoop.flatten([
            db.schema(OrderProductModel.schema).delete(),
            db.schema(CustomerModel.schema).delete(),
            db.schema(OrderModel.schema).delete(),
            db.schema(ProductModel.schema).delete(),
        ])
    }
}

If you wish to keep away from invalid information within the tables, you must all the time use the overseas key and distinctive constraints. A overseas key can be utilized to examine if the referenced identifier exists within the associated desk and the distinctive constraint will ensure that just one row can exists from a given area.

Becoming a member of database tables utilizing Fluent 4

We now have to run the InitialMigration script earlier than we begin utilizing the database. This may be achieved by passing a command argument to the backend software or we will obtain the identical factor by calling the autoMigrate() methodology on the applying occasion.

For the sake of simplicity I’ll use the wait methodology as an alternative of async Futures & Guarantees, that is advantageous for demo functions, however in a real-world server software you must by no means block the present occasion loop with the wait methodology.

That is one doable setup of our dummy database utilizing an SQLite storage, however after all you need to use PostgreSQL, MySQL and even MariaDB by the obtainable Fluent SQL drivers. 🚙

public func configure(_ app: Utility) throws {

    app.databases.use(.sqlite(.file("db.sqlite")), as: .sqlite)

    app.migrations.add(InitialMigration())

    attempt app.autoMigrate().wait()

    let prospects = [
        CustomerModel(name: "Bender"),
        CustomerModel(name: "Fry"),
        CustomerModel(name: "Leela"),
        CustomerModel(name: "Hermes"),
        CustomerModel(name: "Zoidberg"),
    ]
    attempt prospects.create(on: app.db).wait()
    
    let merchandise = [
        ProductModel(name: "Hamburger"),
        ProductModel(name: "Fish"),
        ProductModel(name: "Pizza"),
        ProductModel(name: "Beer"),
    ]
    attempt merchandise.create(on: app.db).wait()

    
    let order = OrderModel(date: Date(), customerId: prospects[0].id!)
    attempt order.create(on: app.db).wait()

    let beerProduct = OrderProductModel(orderId: order.id!, productId: merchandise[3].id!, amount: 6)
    attempt beerProduct.create(on: app.db).wait()
    let pizzaProduct = OrderProductModel(orderId: order.id!, productId: merchandise[2].id!, amount: 1)
    attempt pizzaProduct.create(on: app.db).wait()
}

We now have created 5 prospects (Bender, Fry, Leela, Hermes, Zoidberg), 4 merchandise (Hamburger, Fish, Pizza, Beer) and one new order for Bender containing 2 merchandise (6 beers and 1 pizza). 🤖

Inside be part of utilizing one-to-many relations

Now the query is: how can we get the client information based mostly on the order?

let orders = attempt OrderModel
    .question(on: app.db)
    .be part of(CustomerModel.self, on: OrderModel.$customerId == CustomerModel.$id, methodology: .inside)
    .all()
    .wait()

for order in orders {
    let buyer = attempt order.joined(CustomerModel.self)
    print(buyer.title)
    print(order.date)
}

The reply is fairly easy. We will use an inside be part of to fetch the client mannequin by the order.customerId and buyer.id relation. Once we iterate by the fashions we will ask for the associated mannequin utilizing the joined methodology.

Joins and lots of to many relations

Having a buyer is nice, however how can I fetch the related merchandise for the order? We will begin the question with the OrderProductModel and use a be part of utilizing the ProductModel plus we will filter by the order id utilizing the present order.

for order in orders {
    

    let orderProducts = attempt OrderProductModel
        .question(on: app.db)
        .be part of(ProductModel.self, on: OrderProductModel.$productId == ProductModel.$id, methodology: .inside)
        .filter(.$orderId == order.id!)
        .all()
        .wait()

    for orderProduct in orderProducts {
        let product = attempt orderProduct.joined(ProductModel.self)
        print(product.title)
        print(orderProduct.amount)
    }
}

We will request the joined mannequin the identical approach as we did it for the client. Once more, the very first parameter is the mannequin illustration of the joined desk, subsequent you outline the relation between the tables utilizing the referenced identifiers. As a final parameter you possibly can specify the kind of the be part of.

Inside be part of vs left be part of

There’s a nice SQL tutorial about joins on w3schools.com, I extremely suggest studying it. The primary distinction between an inside be part of and a left be part of is that an inside be part of solely returns these data which have matching identifiers in each tables, however a left be part of will return all of the data from the bottom (left) desk even when there aren’t any matches within the joined (proper) desk.

There are lots of several types of SQL joins, however inside and left be part of are the most typical ones. If you wish to know extra in regards to the different sorts you must learn the linked article. 👍

Abstract

Desk joins are actually useful, however you need to watch out with them. It’s best to all the time use correct overseas key and distinctive constraints. Additionally think about using indexes on some rows if you work with joins, as a result of it will probably enhance the efficiency of your queries. Pace will be an essential issue, so by no means load extra information from the database than you really want.

There is a matter on GitHub in regards to the Fluent 4 API, and one other one about querying particular fields utilizing the .area methodology. Lengthy story quick, joins will be nice and we want higher docs. 🙉