Complex structures

Validating complex structures can quickly become a challenge. Akkurate has been designed from scratch for this task and provides you everything you need for any complexity levels.

Structured programming

Despite using a declarative approach for validation, Akkurate doesn't enforce you to use it and even encourages you to switch to imperative code when needed.

Loops

When validating arrays, maps or iterables, you can use a normal loop to iterate over each item:

@Validate
data class Library(val books: Set<Book>)

@Validate
data class Book(val title: String)

Validator<Library> {
    for (book in books) {
        book.title.isNotEmpty()
    }
}

Since iterating over a collection to validate its items is a common task, Akkurate also provides the each helper:

Validator<Library> {
    books.each { // this: Validatable<Book>
        title.isNotEmpty()
    }
}

Conditions

You can use conditions to apply constraints only when needed. Let's say our Library has a maximum capacity:

@Validate
data class Library(val books: Set<Book>, val maximumCapacity: Int)

The maximum capacity is infinite when it's equal to 0, so we want to constrain the size of the book collection only if the maximum capacity is at least 1. To achieve this, we can unwrap the value, add a condition, and create a constraint only if its positive.

Validator<Library> {
    val (max) = maximumCapacity
    if (max > 0) {
        books.hasSizeLowerThanOrEqualTo(max) otherwise {
            "Too many books"
        }
    }
}

Conditional constraints

Sometimes, you need to apply a constraint depending on the result of a previous one, like Twitter has to do in its settings when a user wants to update its username.

When Twitter launched, you could pick a one-character username. However, this is no longer possible; you're now forced to use at least five characters.

Now imagine being Twitter's product owner for today's settings page. Someone tries to change its username to “a”. This is less than five characters, and it already exists. So, what errors do you want to display?

Two errors? One about character count, and another one about an already taken username?
Or only one error about character count?

The answer is the second one:

A single input contains the value "a" while an error message is displayed below. The error states "Your username must be longer than 4 characters."

A single input contains the value "johndoe" while an error message is displayed below. The error states "That username has been taken. Please choose another."

Since new users can't register with handles shorter than 5 characters anyway, Twitter chose to skip the database check when the username is too short. Avoiding an unnecessary database query and restraining error spamming for the user.

This is a typical use case for conditional constraints. When a constraint is applied, you can read its satisfied property to check if the constraint is satisfied or not.

interface UserDao {
    fun existsByUsername(username: String): Boolean
}

// For the sake of the example, let's assume we've got the following
// global variable, acting as a DAO (Data Access Object).
val userDao = object : UserDao {
    override fun existsByUsername(username: String): Boolean = TODO()
}

@Validate
data class UserUpdate(val username: String)

Validator<UserUpdate> {
    // Retrieve the satisfied status of the constraint.
    val (isValidUsername) = username.hasLengthGreaterThanOrEqualTo(5)

    // Run the database check only if the last constraint succeeded.
    if (isValidUsername) {
        username.constrain {
            !userDao.existsByUsername(it)
        } otherwise { "This username is already taken" }
    }
}

Composition

It is common to validate the same properties in different places. To avoid applying the same constraints multiple times, you can use composition to reuse a validator inside another one.

Here we have a Person class used by two different properties: author and reviewers. To avoid code repetition, we create a validator for the Person class and then call it from Validator<Book>.

@Validate
data class Book(val author: Person, val reviewers: Set<Person>)

@Validate
data class Person(val fullName: String)

val validatePerson = Validator<Person> {
    fullName.isNotEmpty()
}

val validateBook = Validator<Book> {
    author.validateWith(validatePerson)
    reviewers.each { validateWith(validatePerson) }
}

Comparing values

You might sometimes need to check if two values are equal:

Validator<Library> {
    val library = this
    for (book in books) {
        constrain {
            book.height.unwrap() == library.shelfHeight.unwrap()
        } otherwise { "Book height must be equal to shelf height" }
    }
}

However, writing .unwrap() everywhere can quickly become cumbersome. This is why Validatable<T> implements equals and hashCode as pass-through methods to the underlying value. Which means you can remove the calls to unwrap to check for equality:

book.height.unwrap() == library.shelfHeight.unwrap()

book.height == library.shelfHeight

Transformation

Before validating some data, you might need to normalize it.

Imagine you add support for hashtags on the Book class:

@Validate
data class Book(val title: String, val hashtags: Set<String>)

val validateBook = Validator<Book> {
    title.isNotEmpty()
    hashtags.each { isNotEmpty() }
}

However, your users might prefix a hashtag with a #, which means the book will be considered valid if a hashtag contains this single character:

// This validation will succeed despite the book containing
// a hashtag with a single `#` character.
validateBook(
    Book(
        title = "The Lord of the Rings",
        hashtags = setOf("heroic", "#fantasy", "#")
    )
)

To solve this, the solution would be to trim this character before validation, this can be done with the map function:

val validateBook = Validator<Book> {
    title.isNotEmpty()
    hashtags.each {
        map { it.trimStart('#') }.isNotEmpty()
    }
}

Last modified: 24 May 2024