Vapor’s validation API

The very very first thing I might like to point out you is a matter that I’ve with the present validation API for the Vapor framework. I all the time needed to make use of it, as a result of I actually just like the validator features however sadly the API lacks numerous options which might be essential for my wants.

If we check out our beforehand created Todo instance code, you would possibly do not forget that we have solely put some validation on the create API endpoint. That is not very protected, we must always repair this. I’ll present you the right way to validate endpoints utilizing the built-in API, to see what is the challenge with it. 🥲

As a way to show the issues, we’ll add a brand new Tag mannequin to our Todo gadgets.

import Vapor
import Fluent

closing class TagModel: Mannequin {

    static let schema = "tags"
    static let idParamKey = "tagId"
   
    struct FieldKeys {
        static let title: FieldKey = "title"
        static let todoId: FieldKey = "todo_id"
    }
    
    @ID(key: .id) var id: UUID?
    @Subject(key: FieldKeys.title) var title: String
    @Mum or dad(key: FieldKeys.todoId) var todo: TodoModel
    
    init() { }
    
    init(id: UUID? = nil, title: String, todoId: UUID) {
        self.id = id
        self.title = title
        self.$todo.id = todoId
    }
}

So the principle thought is that we’re going to have the ability to tag our todo gadgets and save the todoId reference for every tag. This isn’t going to be a world tagging answer, however extra like a easy tag system for demo functions. The relation will probably be mechanically validated on the database degree (if the db driver helps it), since we’ll put a international key constraint on the todoId area within the migration.

import Fluent

struct TagMigration: Migration {

    func put together(on db: Database) -> EventLoopFuture<Void> {
        db.schema(TagModel.schema)
            .id()
            .area(TagModel.FieldKeys.title, .string, .required)
            .area(TagModel.FieldKeys.todoId, .uuid, .required)
            .foreignKey(TagModel.FieldKeys.todoId, references: TodoModel.schema, .id)
            .create()
    }

    func revert(on db: Database) -> EventLoopFuture<Void> {
        db.schema(TagModel.schema).delete()
    }
}

You will need to point out this once more: NOT each single database helps international key validation out of the field. This is the reason it will likely be extraordinarily vital to validate our enter knowledge. If we let customers to place random todoId values into the database that may result in knowledge corruption and different issues.

Now that now we have our database mannequin & migration, here is how the API objects will appear like. You’ll be able to put these into the TodoApi goal, since these DTOs may very well be shared with a consumer facet library. 📲

import Basis

public struct TagListObject: Codable {
    
    public let id: UUID
    public let title: String

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

public struct TagGetObject: Codable {
    
    public let id: UUID
    public let title: String
    public let todoId: UUID
    
    public init(id: UUID, title: String, todoId: UUID) {
        self.id = id
        self.title = title
        self.todoId = todoId
        
    }
}

public struct TagCreateObject: Codable {

    public let title: String
    public let todoId: UUID
    
    public init(title: String, todoId: UUID) {
        self.title = title
        self.todoId = todoId
    }
}

public struct TagUpdateObject: Codable {
    
    public let title: String
    public let todoId: UUID
    
    public init(title: String, todoId: UUID) {
        self.title = title
        self.todoId = todoId
    }
}

public struct TagPatchObject: Codable {

    public let title: String?
    public let todoId: UUID?
    
    public init(title: String?, todoId: UUID?) {
        self.title = title
        self.todoId = todoId
    }
}

Subsequent we prolong our TagModel to help CRUD operations, in case you adopted my first tutorial about the right way to construct a REST API utilizing Vapor, this needs to be very acquainted, if not please learn it first. 🙏

import Vapor
import TodoApi

extension TagListObject: Content material {}
extension TagGetObject: Content material {}
extension TagCreateObject: Content material {}
extension TagUpdateObject: Content material {}
extension TagPatchObject: Content material {}

extension TagModel {
    
    func mapList() -> TagListObject {
        .init(id: id!, title: title)
    }

    func mapGet() -> TagGetObject {
        .init(id: id!, title: title, todoId: $todo.id)
    }
    
    func create(_ enter: TagCreateObject) {
        title = enter.title
        $todo.id = enter.todoId
    }
        
    func replace(_ enter: TagUpdateObject) {
        title = enter.title
        $todo.id = enter.todoId
    }
    
    func patch(_ enter: TagPatchObject) {
        title = enter.title ?? title
        $todo.id = enter.todoId ?? $todo.id
    }
}

The tag controller goes to look similar to the todo controller, for now we cannot validate something, the next snippet is all about having a pattern code that we will fantastic tune in a while.

import Vapor
import Fluent
import TodoApi

struct TagController {

    personal func getTagIdParam(_ req: Request) throws -> UUID {
        guard let rawId = req.parameters.get(TagModel.idParamKey), let id = UUID(rawId) else {
            throw Abort(.badRequest, motive: "Invalid parameter `(TagModel.idParamKey)`")
        }
        return id
    }

    personal func findTagByIdParam(_ req: Request) throws -> EventLoopFuture<TagModel> {
        TagModel
            .discover(strive getTagIdParam(req), on: req.db)
            .unwrap(or: Abort(.notFound))
    }

    
    
    func record(req: Request) throws -> EventLoopFuture<Web page<TagListObject>> {
        TagModel.question(on: req.db).paginate(for: req).map { $0.map { $0.mapList() } }
    }
    
    func get(req: Request) throws -> EventLoopFuture<TagGetObject> {
        strive findTagByIdParam(req).map { $0.mapGet() }
    }

    func create(req: Request) throws -> EventLoopFuture<Response> {
        let enter = strive req.content material.decode(TagCreateObject.self)

        let tag = TagModel()
        tag.create(enter)
        return tag
            .create(on: req.db)
            .map { tag.mapGet() }
            .encodeResponse(standing: .created, for: req)
    }
    
    func replace(req: Request) throws -> EventLoopFuture<TagGetObject> {
        let enter = strive req.content material.decode(TagUpdateObject.self)

        return strive findTagByIdParam(req)
            .flatMap { tag in
                tag.replace(enter)
                return tag.replace(on: req.db).map { tag.mapGet() }
            }
    }
    
    func patch(req: Request) throws -> EventLoopFuture<TagGetObject> {
        let enter = strive req.content material.decode(TagPatchObject.self)

        return strive findTagByIdParam(req)
            .flatMap { tag in
                tag.patch(enter)
                return tag.replace(on: req.db).map { tag.mapGet() }
            }
    }

    func delete(req: Request) throws -> EventLoopFuture<HTTPStatus> {
        strive findTagByIdParam(req)
            .flatMap { $0.delete(on: req.db) }
            .map { .okay }
    }
}

In fact we may use a generic CRUD controller class that might extremely cut back the quantity of code required to create related controllers, however that is a distinct matter. So we simply need to register these newly created features utilizing a router.

import Vapor

struct TagRouter: RouteCollection {

    func boot(routes: RoutesBuilder) throws {

        let tagController = TagController()
        
        let id = PathComponent(stringLiteral: ":" + TagModel.idParamKey)
        let tagRoutes = routes.grouped("tags")
        
        tagRoutes.get(use: tagController.record)
        tagRoutes.put up(use: tagController.create)
        
        tagRoutes.get(id, use: tagController.get)
        tagRoutes.put(id, use: tagController.replace)
        tagRoutes.patch(id, use: tagController.patch)
        tagRoutes.delete(id, use: tagController.delete)
    }
}

Additionally just a few extra adjustments within the configure.swift file, since we would prefer to make the most of the Tag performance now we have to register the migration and the brand new routes utilizing the TagRouter.

import Vapor
import Fluent
import FluentSQLiteDriver

public func configure(_ app: Utility) throws {

    if app.surroundings == .testing {
        app.databases.use(.sqlite(.reminiscence), as: .sqlite, isDefault: true)
    }
    else {
        app.databases.use(.sqlite(.file("Assets/db.sqlite")), as: .sqlite)
    }

    app.http.server.configuration.hostname = "192.168.8.103"
    app.migrations.add(TodoMigration())
    app.migrations.add(TagMigration())
    strive app.autoMigrate().wait()

    strive TodoRouter().boot(routes: app.routes)
    strive TagRouter().boot(routes: app.routes)
}

Another factor, earlier than we begin validating our tags, now we have to place a brand new @Kids(for: .$todo) var tags: [TagModel] property into our TodoModel, so it should be far more simple to fetch tags.

When you run the server and attempt to create a brand new tag utilizing cURL and a pretend UUID, the database question will fail if the db helps international keys.

curl -X POST "http://127.0.0.1:8080/tags/" 
    -H 'Content material-Kind: utility/json' 
    -d '{"title": "take a look at", "todoId": "94234a4a-b749-4a2a-97d0-3ebd1046dbac"}'

This isn’t preferrred, we must always defend our database from invalid knowledge. Properly, to begin with we do not need to permit empty or too lengthy names, so we must always validate this area as effectively, this may be achieved utilizing the validation API from the Vapor framework, let me present you the way.

extension TagCreateObject: Validatable {
    public static func validations(_ validations: inout Validations) {
        validations.add("title", as: String.self, is: !.empty)
        validations.add("title", as: String.self, is: .rely(...100) && .alphanumeric)
    }
}

func create(req: Request) throws -> EventLoopFuture<Response> {
    strive TagCreateObject.validate(content material: req)
    let enter = strive req.content material.decode(TagCreateObject.self)

    let tag = TagModel()
    tag.create(enter)
    return tag
        .create(on: req.db)
        .map { tag.mapGet() }
        .encodeResponse(standing: .created, for: req)
}

Okay, it seems nice, however this answer lacks just a few issues:

• You’ll be able to’t present customized error messages
• The element is all the time a concatenated outcome string (if there are a number of errors)
• You’ll be able to’t get the error message for a given key (e.g. “title”: “Title is required”)
• Validation occurs synchronously (you possibly can’t validate primarily based on a db question)

That is very unlucky, as a result of Vapor has very nice validator features. You’ll be able to validate characters (.ascii, .alphanumeric, .characterSet(_:)), numerous size and vary necessities (.empty, .rely(_:), .vary(_)), collections (.in(_:)), test null inputs, validate emails and URLs. We should always attempt to validate the todo identifier primarily based on the out there todos within the database.

It’s potential to validate todoId’s by operating a question with the enter id and see if there’s an current report in our database. If there isn’t any such todo, we cannot permit the creation (or replace / patch) operation. The issue is that now we have to place this logic into the controller. 😕

func create(req: Request) throws -> EventLoopFuture<Response> {
    strive TagCreateObject.validate(content material: req)
    let enter = strive req.content material.decode(TagCreateObject.self)
    return TodoModel.discover(enter.todoId, on: req.db)
        .unwrap(or: Abort(.badRequest, motive: "Invalid todo identifier"))
        .flatMap { _ in
            let tag = TagModel()
            tag.create(enter)
            return tag
                .create(on: req.db)
                .map { tag.mapGet() }
                .encodeResponse(standing: .created, for: req)
        }
}

It will do the job, however is not it unusual that we’re doing validation in two separate locations?

My different drawback is that utilizing the validatable protocol means that you would be able to’t actually go parameters for these validators, so even in case you asynchronously fetch some required knowledge and someway you progress the logic contained in the validator, the entire course of goes to really feel like a really hacky answer. 🤐

Actually, am I lacking one thing right here? Is that this actually how the validation system works in the preferred net framework? It is fairly unbelievable. There should be a greater means… 🤔

Async enter validation This methodology that I’ll present you is already out there in Feather CMS, I consider it is fairly a complicated system in comparison with Vapor’s validation API. I am going to present you the way I created it, first we begin with a protocol that’ll include the fundamental stuff wanted for validation & outcome administration.

import Vapor

public protocol AsyncValidator {
    
    var key: String { get }
    var message: String { get }

    func validate(_ req: Request) -> EventLoopFuture<ValidationErrorDetail?>
}

public extension AsyncValidator {

    var error: ValidationErrorDetail {
        .init(key: key, message: message)
    }
}

It is a fairly easy protocol that we’ll be the bottom of our asynchronous validation movement. The important thing will probably be used to identical to the identical means as Vapor makes use of validation keys, it is principally an enter key for a given knowledge object and we’ll use this key with an acceptable error message to show detailed validation errors (as an output content material).

import Vapor

public struct ValidationErrorDetail: Codable {

    public var key: String
    public var message: String
    
    public init(key: String, message: String) {
        self.key = key
        self.message = message
    }
}

extension ValidationErrorDetail: Content material {}

So the concept is that we’ll create a number of validation handlers primarily based on this AsyncValidator protocol and get the ultimate outcome primarily based on the evaluated validators. The validation methodology can appear like magic at first sight, but it surely’s simply calling the async validator strategies if a given secret is already invalidated then it will skip different validations for that (for apparent causes), and primarily based on the person validator outcomes we create a closing array together with the validation error element objects. 🤓

import Vapor

public struct RequestValidator {

    public var validators: [AsyncValidator]
    
    public init(_ validators: [AsyncValidator] = []) {
        self.validators = validators
    }
    
    
    public func validate(_ req: Request, message: String? = nil) -> EventLoopFuture<Void> {
        let preliminary: EventLoopFuture<[ValidationErrorDetail]> = req.eventLoop.future([])
        return validators.cut back(preliminary) { res, subsequent -> EventLoopFuture<[ValidationErrorDetail]> in
            return res.flatMap { arr -> EventLoopFuture<[ValidationErrorDetail]> in
                if arr.comprises(the place: { $0.key == subsequent.key }) {
                    return req.eventLoop.future(arr)
                }
                return subsequent.validate(req).map { outcome in
                    if let outcome = outcome {
                        return arr + [result]
                    }
                    return arr
                }
            }
        }
        .flatMapThrowing { particulars in
            guard particulars.isEmpty else {
                throw Abort(.badRequest, motive: particulars.map(.message).joined(separator: ", "))
            }
        }
    }

    public func isValid(_ req: Request) -> EventLoopFuture<Bool> {
        return validate(req).map { true }.get well { _ in false }
    }
}

Do not wrap your head an excessive amount of about this code, I am going to present you the right way to use it straight away, however earlier than we may carry out a validation utilizing our new instruments, we want one thing that implements the AsyncValidator protocol and we will truly initialize. I’ve one thing that I actually like in Feather, as a result of it may well carry out each sync & async validations, in fact you possibly can give you extra easy validators, however this can be a good generic answer for a lot of the instances.

import Vapor

public struct KeyedContentValidator<T: Codable>: AsyncValidator {

    public let key: String
    public let message: String
    public let optionally available: Bool

    public let validation: ((T) -> Bool)?
    public let asyncValidation: ((T, Request) -> EventLoopFuture<Bool>)?
    
    public init(_ key: String,
                _ message: String,
                optionally available: Bool = false,
                _ validation: ((T) -> Bool)? = nil,
                _ asyncValidation: ((T, Request) -> EventLoopFuture<Bool>)? = nil) {
        self.key = key
        self.message = message
        self.optionally available = optionally available
        self.validation = validation
        self.asyncValidation = asyncValidation
    }
    
    public func validate(_ req: Request) -> EventLoopFuture<ValidationErrorDetail?> {
        let optionalValue = strive? req.content material.get(T.self, at: key)

        if let worth = optionalValue {
            if let validation = validation {
                return req.eventLoop.future(validation(worth) ? nil : error)
            }
            if let asyncValidation = asyncValidation {
                return asyncValidation(worth, req).map { $0 ? nil : error }
            }
            return req.eventLoop.future(nil)
        }
        else {
            if optionally available {
                return req.eventLoop.future(nil)
            }
            return req.eventLoop.future(error)
        }
    }
}

The primary thought right here is that we will go both a sync or an async validation block alongside the important thing, message and optionally available arguments and we carry out our validation primarily based on these inputs.

First we attempt to decode the generic Codable worth, if the worth was optionally available and it’s lacking we will merely ignore the validators and return, in any other case we must always attempt to name the sync validator or the async validator. Please word that the sync validator is only a comfort device, as a result of in case you do not want async calls it is less difficult to return with a bool worth as a substitute of an EventLoopFuture<Bool>.

So, that is how one can validate something utilizing these new server facet Swift validator parts.

func create(req: Request) throws -> EventLoopFuture<Response> {
    let validator = RequestValidator.init([
        KeyedContentValidator<String>.init("name", "Name is required") { !$0.isEmpty },
        KeyedContentValidator<UUID>.init("todoId", "Todo identifier must be valid", nil) { value, req in
            TodoModel.query(on: req.db).filter(.$id == value).count().map {
                $0 == 1
            }
        },
    ])
    return validator.validate(req).flatMap {
        do {
            let enter = strive req.content material.decode(TagCreateObject.self)
            let tag = TagModel()
            tag.create(enter)
            return tag
                .create(on: req.db)
                .map { tag.mapGet() }
                .encodeResponse(standing: .created, for: req)
        }
        catch {
            return req.eventLoop.future(error: Abort(.badRequest, motive: error.localizedDescription))
        }
    }
}

This looks as if a bit extra code at first sight, however do not forget that beforehand we moved out our validator right into a separate methodology. We will do the very same factor right here and return an array of AsyncValidator objects. Additionally a “actual throwing flatMap EventLoopFuture” extension methodology may assist us enormously to take away pointless do-try-catch statements from our code.

Anyway, I am going to go away this up for you, but it surely’s simple to reuse the identical validation for all of the CRUD endpoints, for patch requests you possibly can set the optionally available flag to true and that is it. 💡

I nonetheless need to present you yet another factor, as a result of I do not like the present JSON output of the invalid calls. We will construct a customized error middleware with a customized context object to show extra particulars about what went improper throughout the request. We want a validation error content material for this.

import Vapor

public struct ValidationError: Codable {

    public let message: String?
    public let particulars: [ValidationErrorDetail]
    
    public init(message: String?, particulars: [ValidationErrorDetail]) {
        self.message = message
        self.particulars = particulars
    }
}

extension ValidationError: Content material {}

That is the format that we would like to make use of when one thing goes improper. Now it might be good to help customized error codes whereas retaining the throwing nature of errors, so for that reason we’ll outline a brand new ValidationAbort that is going to include all the pieces we’ll want for the brand new error middleware.

import Vapor

public struct ValidationAbort: AbortError {

    public var abort: Abort
    public var message: String?
    public var particulars: [ValidationErrorDetail]

    public var motive: String { abort.motive }
    public var standing: HTTPStatus { abort.standing }
    
    public init(abort: Abort, message: String? = nil, particulars: [ValidationErrorDetail]) {
        self.abort = abort
        self.message = message
        self.particulars = particulars
    }
}

It will permit us to throw ValidationAbort objects with a customized Abort & detailed error description. The Abort object is used to set the right HTTP response code and headers when constructing the response object contained in the middleware. The middleware is similar to the built-in error middleware, besides that it may well return extra particulars in regards to the given validation points.

import Vapor

public struct ValidationErrorMiddleware: Middleware {

    public let surroundings: Setting
    
    public init(surroundings: Setting) {
        self.surroundings = surroundings
    }

    public func reply(to request: Request, chainingTo subsequent: Responder) -> EventLoopFuture<Response> {
        return subsequent.reply(to: request).flatMapErrorThrowing { error in
            let standing: HTTPResponseStatus
            let headers: HTTPHeaders
            let message: String?
            let particulars: [ValidationErrorDetail]

            change error {
            case let abort as ValidationAbort:
                standing = abort.abort.standing
                headers = abort.abort.headers
                message = abort.message ?? abort.motive
                particulars = abort.particulars
            case let abort as Abort:
                standing = abort.standing
                headers = abort.headers
                message = abort.motive
                particulars = []
            default:
                standing = .internalServerError
                headers = [:]
                message = surroundings.isRelease ? "One thing went improper." : error.localizedDescription
                particulars = []
            }

            request.logger.report(error: error)

            let response = Response(standing: standing, headers: headers)

            do {
                response.physique = strive .init(knowledge: JSONEncoder().encode(ValidationError(message: message, particulars: particulars)))
                response.headers.replaceOrAdd(title: .contentType, worth: "utility/json; charset=utf-8")
            }
            catch {
                response.physique = .init(string: "Oops: (error)")
                response.headers.replaceOrAdd(title: .contentType, worth: "textual content/plain; charset=utf-8")
            }
            return response
        }
    }
}

Primarily based on the given surroundings we will report the main points or disguise the interior points, that is completely up-to-you, for me this strategy works the perfect, as a result of I can all the time parse the problematic keys and show error messages contained in the consumer apps primarily based on this response.

We simply have to change one line within the RequestValidator & register our newly created middleware for higher error reporting. Here is the up to date request validator:

.flatMapThrowing { particulars in
    guard particulars.isEmpty else {
        throw ValidationAbort(abort: Abort(.badRequest, motive: message), particulars: particulars)
    }
}


app.middleware.use(ValidationErrorMiddleware(surroundings: app.surroundings))

Now in case you run the identical invalid cURL request, you must get a means higher error response.

curl -i -X POST "http://192.168.8.103:8080/tags/" 
    -H 'Content material-Kind: utility/json' 
    -d '{"title": "eee", "todoId": "94234a4a-b749-4a2a-97d0-3ebd1046dbac"}'

# HTTP/1.1 400 Dangerous Request
# content-length: 72
# content-type: utility/json; charset=utf-8
# connection: keep-alive
# date: Wed, 12 Might 2021 14:52:47 GMT
#
# {"particulars":[{"key":"todoId","message":"Todo identifier must be valid"}]}

You’ll be able to even add a customized message for the request validator once you name the validate operate, that’ll be out there below the message key contained in the output.

As you possibly can see that is fairly a pleasant approach to take care of errors and unify the movement of all the validation chain. I am not saying that Vapor did a foul job with the official validation APIs, however there’s undoubtedly room for enhancements. I actually love the big variety of the out there validators, however however I freakin’ miss this async validation logic from the core framework. ❤️💩

One other good factor about this strategy is that you would be able to outline validator extensions and enormously simplify the quantity of Swift code required to carry out server facet validation.

I do know I am not the one one with these points, and I actually hope that this little tutorial will aid you create higher (and extra protected) backend apps utilizing Vapor. I can solely say that be happy to enhance the validation associated code for this Todo venture, that is a very good observe for certain. Hopefully it will not be too laborious so as to add extra validation logic primarily based on the supplied examples. 😉