README

A library for validating generic data in a functional fashion.

Basic idea

The idea is pretty simple. All goes around the following interface

interface Validation
{
    public function validate($data): ValidationResult;
}

where some $data comes in and a ValidationResult comes out.

A ValidationResult is a sum type which could be either valid, containing some valid $data, or invalid, containing some error messages.

This means that a validation could either succeed, and it that case you have the valid result at your disposition, or if fails, and you have the error messages to handle.

Immutability

Everything is immutable, so once you have created a validator you can not modify it, you can just create a new one.

On the other hand, immutability implies statelessness, and therefore you can reuse safely the same validator multiple times with different data.

Compositionality

The library provides two different mechanisms, which could be combined, which allow creating complex validators just putting simple ones together.

It goes without saying that you can create new validators to be used together with the existing ones.

Basic validators

The library provides several basic validators to validate native data structures. They all implement the Validation interface described above.

Context

How can you use a runtime value to validate some data, while you are creating your validators at build time? Well, that's what the context is there for.

Let's provide a motivating example for this: suppose we want to write a validator to check whether we are violating a uniqueness condition while updating a member in a collection. We will need to check it the data already exist in our collection, excluding the record we are currently updating. So we will need, beyond the data themselves, an identifier of the record which we are currently updating. We can pass this information in the context. Then the validator could look like

class CheckDuplicateExceptCurrentRecord
{
    private $recordRepository;
    
    public function __construct($recordRepository)
    {
        $this->recordRepository = $recordRepository;
    }

    public function validate($data, $context)
    {
        if ($recordRepository->containsDataExcludingRecord($data, $context['id'])) {
            return ValidationResult::errors(['DUPLICATE RECORD']);
        }
        
        return ValidationResult::valid($data);
    }
}

Then we can use our validator as follows

$validator = new CheckDuplicateExceptCurrentRecord($recordRepository);

$validator->validate($data, ['id' => $currentRecordId]);

Custom error formatters

The library itself does not want to impose how error messages should be structured and formatted. Therefore it allows the user to define his own error messages and his own error messages structure.

To do this every validator included in the library may be built using a custom error formatter.

The error formatter is nothing else but a callable which receives as input all the data known by the validator. Often this arguments will be just the data which need to be validated, but sometimes the error formatter could receive also the configuration parameters of the validator.

For example, the IsInstanceOf validator has a named constructor

public static function withClassNameAndFormatter(
    string $className,
    callable $errorFormatter
):

where the $errorFormatter needs to be a callable receiving as parameters the $className and the validation $data. For example we could use it as follows

$myValidator = IsInstanceOf::withClassNameAndFormatter(
    Foo::class,
    function ($className, $data) {
        return [
            sprintf(
                'The data %s is not an instance of %s',
                json_encode($data),
                $className
            )
        ];
    }
);

Translators

One specific usage of custom error formatters it translating the error messages.

Every validator has also a named constructor receiving a Translator to translate the library-defined error messages.

If you don't want to specify the translator for every single validator you define, there is also a TranslateErrors combinators which translates all the strings present in the return message.

Creating more complex validators

We mentioned above that compositionality is one one the main ideas beyond this library. In fact, we provide not one but two mechanisms to create complex validators starting from basic ones:

using combinators; a combinator is simply a function which creates a new validator starting from simpler ones in a very well defined way. The combinators this library provides are

using the lift, the sdo and the fdo functions provided in Result/functions.php to mimic applicative and monadic validation used in functional programming.

In the following sections we provide examples on how to use these two approaches.

Using combinators

Suppose you want to validate some data with the following format

[
    'name' => ... // non empty string
    'age' => ... // non-negative integer
]

To describe what we would like to check in plain text, we need to verify that:

the data we receive are an array
we have a name field and it should be a non-empty string
we have an age field and it should be a positive integer

The validator which does this check should look like:

Sequence::validations([
    new IsArray(),
    All::validations([
        Sequence::validations([
            HasKey::withKey('name'),
            Focus::on(
                function ($data) {
                    return $data['name'];
                },
                Sequence::validations([
                    new IsString(),
                    new NonEmpty()
                ])
            )
        ]),
        Sequence::validations([
            HasKey::withKey('age'),
            Focus::on(
                function ($data) {
                    return $data['age'];
                },
                Sequence::validations([
                    new IsInteger(),
                    IsGreaterThan::withBound(0)
                ])
            )
        ])
    ])
]);

Let's go through it step by step to understand every single piece.

We start with a Sequence validator. Its semantic is that it is going to perform a series of validation sequentially, one after the other, returning an error as soon as one fails.

In our case we start by checking that our data are an array, with the IsArray validator. Once we know we have an array, we can check if the two required fields exist. We can do these two operations independently, and, if they both fail, we want both the error messages. We use the All validator exactly for this, to say that all the listed conditions need to be verified and that we want all the error messages.

At this point we have the validations for name and the validations for age. For the former, we first check that the name key is present, with the HasKey validator. Then we want to validate the value of the name key; to do this we need to focus our attention not on the whole data structure, but just on the single value. We use the Focus validator to specify a callable which allows to inspect the specific value. At this point we use Sequence again to check that the value is a string, using the IsString validator, and to assert that it is not empty, with the NonEmpty validator.

For the age field, we do something similar. The only difference is that we check if it is an integer with the IsInteger validator and we use the IsAsAsserted validator, built with a user defined callable, to check that the value is a positive integer.

This example explains how to use several combinators to create complex validators. Actually, the specific validator we are considering here could be written more simply as

Associative::validations([
    'name' => Sequence::validations([
        new IsString(),
        new NonEmpty()
    ]),
    'age' => Sequence::validations([
        new IsInteger(),
        IsGreaterThan::withBound(0)
    ])
]);

removing a lot of boilerplate form the client code and allowing to concentrate on what is specific for the validation at hand.

Applicative and monadic validation

Applicative and monadic validations are common techniques in functional programming. Both consists in having several validators, applying them, and combining their results. The main difference between them is how failure is handled.

With the applicative style, all validators are computed independently and, in case of failure, all errors are returned. This is useful when you need to validate independent data and retrieve all the errors encountered.

With the monadic style, validators are applied sequentially, and the result of a validator is provided as input for the next one. This is needed whenever a validation depends on the result of a previous validation.

Probably in this case some code says more that a lot of vague words. You can check yourself how applicative and monadic validation work looking at the ApplicativeMonadicSpec tests.

How to use this library

This library is not build to be a ready-made artifact that you can install and start using immediately. On the contrary it provides just some basic elements which you can use to easily build your own validators.

The idea is that everyone could create his own library of validators, specific for his own domain and use case, composing the basic validators and custom ones with the help of the provided combinators.

marcosh / php-validation-dsl

Maintainers

Details