dry-validation

Data validation library based on predicate logic and rule composition.

Overview

Unlike other, well known, validation solutions in Ruby, dry-validation takes a different approach and focuses a lot on explicitness, clarity and preciseness of validation logic. It is designed to work with any data input, whether it's a simple hash, an array or a complex object with deeply nested data.

It is based on a simple idea that each validation is encapsulated by a simple, stateless predicate, that receives some input and returns either true or false.

Those predicates are encapsulated by rules which can be composed together using predicate logic. This means you can use the common logic operators to build up a validation schema.

It's very explicit, powerful and extendible.

Validations can be described with great precision, dry-validation eliminates ambigious concepts like presence validation where we can't really say whether some attribute or key is missing or it's just that the value is nil.

There's also the concept of type-safety, completely missing in other validation libraries, which is quite important and useful. It means you can compose a validation that does rely on the type of a given value. In example it makes no sense to validate each element of an array when it turns out to be an empty string.

The DSL

The core of dry-validation is rules composition and predicate logic. The DSL is a simple front-end for that. It only allows you to define the rules by using predicate identifiers. There are no magical options, conditionals and custom validation blocks known from other libraries. The focus is on pure validation logic.

Examples

Basic

Here's a basic example where we validate following things:

The input must have a key called :email
- Provided the email key is present, its value must be filled
The input must have a key called :age
- Provided the age key is present, its value must be an integer and it must be greater than 18

This can be easily expressed through the DSL:

require 'dry-validation'

class Schema < Dry::Validation::Schema
  key(:email) { |email| email.filled? }

  key(:age) do |age|
    age.int? & age.gt?(18)
  end
end

schema = Schema.new

errors = schema.call(email: '[email protected]', age: 19).messages

puts errors.inspect
# []

errors = schema.call(email: nil, age: 19).messages

puts errors.inspect
# { :email => [["email must be filled", nil]] }

A couple of remarks:

key assumes that we want to use the :key? predicate to check the existance of that key
age.gt?(18) translates to calling a predicate like this: schema[:gt?].(18, age)
age.int? & age.gt?(18) is a conjunction, so we don't bother about gt? unless int? returns true
You can also use | for disjunction
Schema object does not carry the input as its state, nor does it know how to access the input values, we pass the input to call and get error set as the response

Optional Keys

You can define which keys are optional and define rules for their values:

require 'dry-validation'

class Schema < Dry::Validation::Schema
  key(:email) { |email| email.filled? }

  optional(:age) do |age|
    age.int? & age.gt?(18)
  end
end

schema = Schema.new

errors = schema.call(email: '[email protected]').messages

puts errors.inspect
# []

errors = schema.call(email: '[email protected]', age: 17).messages

puts errors.inspect
# { :age => [["age must be greater than 18"], 17] }

Optional Values

When it is valid for a given value to be nil you can use none? predicate:

require 'dry-validation'

class Schema < Dry::Validation::Schema
  key(:email) { |email| email.filled? }

  key(:age) do |age|
    age.none? | (age.int? & age.gt?(18))
  end
end

schema = Schema.new

errors = schema.call(email: '[email protected]', age: nil).messages

puts errors.inspect
# []

errors = schema.call(email: '[email protected]', age: 19).messages

puts errors.inspect
# []

errors = schema.call(email: '[email protected]', age: 17).messages

puts errors.inspect
# { :age => [["age must be greater than 18"], 17] }

Optional Key vs Value

We make a clear distinction between specifying an optional key and an optional value. This gives you a way of being very specific about validation rules. You can define a schema which can give you precise errors when a key was missing or key was present but the value was nil.

This also comes with the benefit of being explicit about the type expectation. In the example above we explicitly state that :age can be nil or it can be an integer and when it is an integer we specify that it must be greater than 18.

Another benefit is that we can infer specific coercion rules when types are specified. In example Schema::Form will use form.nil type from dry-data to coerce empty strings into nil for you whenever you specify value.none? | value.int?. Furthermore it will try to coerce to int since that is our type expectation.

Nested Hash

We are free to define validations for anything, including deeply nested structures:

require 'dry-validation'

class Schema < Dry::Validation::Schema
  key(:address) do |address|
    address.hash? do
      address.key(:city) do |city|
        city.min_size?(3)
      end

      address.key(:street) do |street|
        street.filled?
      end

      address.key(:country) do |country|
        country.key(:name, &:filled?)
        country.key(:code, &:filled?)
      end
    end
  end
end

schema = Schema.new

errors = schema.call({}).messages

puts errors.inspect
# { :address => ["address is missing"] }

errors = schema.call(address: { city: 'NYC' }).messages

puts errors.to_h.inspect
# {
#   :address => [
#     { :street => ["street is missing"] },
#     { :country => ["country is missing"] }
#   ]
# }

Array Elements

You can use each rule for validating each element in an array:

class Schema < Dry::Validation::Schema
  key(:phone_numbers) do |phone_numbers|
    phone_numbers.array? do
      phone_numbers.each(&:str?)
    end
  end
end

schema = Schema.new

errors = schema.call(phone_numbers: '').messages

puts errors.inspect
# { :phone_numbers => [["phone_numbers must be an array", ""]] }

errors = schema.call(phone_numbers: ['123456789', 123456789]).messages

puts errors.inspect
# {
#   :phone_numbers => [
#     {
#       :phone_numbers => [
#         ["phone_numbers must be a string", 123456789]
#       ]
#     }
#   ]
# }

Rules Depending On Other Rules

When a rule needs input from other rules and depends on their results you can define it using rule DSL. A common example of this is "confirmation validation":

class Schema < Dry::Validation::Schema
  key(:password, &:filled?)
  key(:password_confirmation, &:filled?)

  rule(:password_confirmation, eql?: [:password, :password_confirmation])
end

A short version of the same thing:

class Schema < Dry::Validation::Schema
  confirmation(:password)
end

Notice that you must add :password_confirmation error message configuration if you want to have the error converted to a message.

Form Validation With Coercions

Probably the most common use case is to validate form params. This is a special kind of a validation for a couple of reasons:

The input is a hash with stringified keys
The input include values that are strings, hashes or arrays
Prior validation, we need to coerce values and symbolize keys based on the information from rules

For that reason, dry-validation ships with Schema::Form class:

require 'dry-validation'
require 'dry/validation/schema/form'

class UserFormSchema < Dry::Validation::Schema::Form
  key(:email) { |value| value.str? & value.filled? }

  key(:age) { |value| value.int? & value.gt?(18) }
end

schema = UserFormSchema.new

errors = schema.call('email' => '', 'age' => '18').messages

puts errors.inspect
# {
#   :email => [["email must be filled", nil]],
#   :age => [["age must be greater than 18 (18 was given)", 18]]
# }

There are few major differences between how it works here and in ActiveModel:

We have type checking as predicates, ie gt?(18) will not be applied if the value is not an integer
Thus, error messages are provided only for the rules that failed
There's a planned feature for generating "validation hints" which lists information about all possible rules
Coercion is handled by dry-data coercible hash using its form.* types that are dedicated for this type of coercions
It's very easy to add your own types and coercions (more info/docs coming soon)

Defining Custom Predicates

You can simply define predicate methods on your schema object:

class Schema < Dry::Validation::Schema
  key(:email) { |value| value.str? & value.email? }

  def email?(value)
    ! /magical-regex-that-matches-emails/.match(value).nil?
  end
end

You can also re-use a predicate container across multiple schemas:

module MyPredicates
  include Dry::Validation::Predicates

  predicate(:email?) do |input|
    ! /magical-regex-that-matches-emails/.match(value).nil?
  end
end

class Schema < Dry::Validation::Schema
  configure do |config|
    config.predicates = MyPredicates
  end

  key(:email) { |value| value.str? & value.email? }
end

You need to provide error messages for your custom predicates if you want them to work with Schem#messages interface.

You can learn how to do that in the Error Messages section.

List of Built-In Predicates

Basic

none?
eql?
key?

Types

str?
int?
float?
decimal?
bool?
date?
date_time?
time?
array?
hash?

Number, String, Collection

empty?
filled?
gt?
gteq?
lt?
lteq?
max_size?
min_size?
size?(int)
size?(range)
format?
inclusion?
exclusion?

Error Messages

By default dry-validation comes with a set of pre-defined error messages for every built-in predicate. They are defined in a yaml file which is shipped with the gem. This file is compatible with I18n format.

You can provide your own messages and configure your schemas to use it like that:

class Schema < Dry::Validation::Schema
  configure { |config| config.messages_file = '/path/to/my/errors.yml' }
end

You can also provide a namespace per-schema that will be used by default:

class Schema < Dry::Validation::Schema
  configure { |config| config.namespace = :user }
end

Lookup rules:

en:
  errors:
    size?:
      arg:
        default: "%{name} size must be %{num}"
        range: "%{name} size must be within %{left} - %{right}"

      value:
        string:
          arg:
            default: "%{name} length must be %{num}"
            range: "%{name} length must be within %{left} - %{right}"

    filled?: "%{name} must be filled"

    rules:
      email:
        filled?: "the email is missing"

      user:
        filled?: "%{name} name cannot be blank"

        rules:
          address:
            filled?: "You gotta tell us where you live"

Given the yaml file above, messages lookup works as follows:

messages = Dry::Validation::Messages.load('/path/to/our/errors.yml')

# matching arg type for size? predicate
messages[:size?, rule: :name, arg_type: Fixnum] # => "%{name} size must be %{num}"
messages[:size?, rule: :name, arg_type: Range] # => "%{name} size must within %{left} - %{right}"

# matching val type for size? predicate
messages[:size?, rule: :name, val_type: String] # => "%{name} length must be %{num}"

# matching predicate
messages[:filled?, rule: :age] # => "%{name} must be filled"
messages[:filled?, rule: :address] # => "%{name} must be filled"

# matching predicate for a specific rule
messages[:filled?, rule: :email] # => "the email is missing"

# with namespaced messages
user_messages = messages.namespaced(:user)

user_messages[:filled?, rule: :age] # "%{name} cannot be blank"
user_messages[:filled?, rule: :address] # "You gotta tell us where you live"

By configuring messages_file and/or namespace in a schema, default messages are going to be automatically merged with your overrides and/or namespaced.

I18n Integration

If you are using i18n gem and load it before dry-validation then you'll be able to configure a schema to use i18n messages:

require 'i18n'
require 'dry-validation'

class Schema < Dry::Validation::Schema
  configure { config.messages = :i18n }

  key(:email, &:filled?)
end

schema = Schema.new

# return default translations
puts schema.call(email: '').messages
{ :email => ["email must be filled"] }

# return other translations (assuming you have it :))
puts schema.call(email: '').messages(locale: :pl)
{ :email => ["email musi być wypełniony"] }

Important: I18n must be initialized before using schema, dry-validation does not try to do it for you, it only sets its default error translations automatically.

Rule AST

Internally, dry-validation uses a simple AST representation of rules and errors to produce rule objects and error messages. If you would like to programatically generate rules, it is a very simple process:

ast = [
  [
    :and,
    [
      [:key, [:age, [:predicate, [:key?, []]]]],
      [
        :and,
        [
          [:val, [:age, [:predicate, [:filled?, []]]]],
          [:val, [:age, [:predicate, [:gt?, [18]]]]]
        ]
      ]
    ]
  ]
]

compiler = Dry::Validation::RuleCompiler.new(Dry::Validation::Predicates)

# compile an array of rule objects
rules = compiler.call(ast)

puts rules.inspect
# [
#   #<Dry::Validation::Rule::Conjunction
#     left=#<Dry::Validation::Rule::Key name=:age predicate=#<Dry::Validation::Predicate id=:key?>>
#     right=#<Dry::Validation::Rule::Conjunction
#       left=#<Dry::Validation::Rule::Value name=:age predicate=#<Dry::Validation::Predicate id=:filled?>>
#       right=#<Dry::Validation::Rule::Value name=:age predicate=#<Dry::Validation::Predicate id=:gt?>>>>
# ]

# dump it back to ast
puts rules.map(&:to_ary).inspect
# [[:and, [:key, [:age, [:predicate, [:key?, [:age]]]]], [[:and, [:val, [:age, [:predicate, [:filled?, []]]]], [[:val, [:age, [:predicate, [:gt?, [18]]]]]]]]]]

Complete docs for the AST format are coming soon, for now please refer to this spec.

Status and Roadmap

This library is in a very early stage of development but you are encauraged to try it out and provide feedback.

For planned features check out the issues.

License

See LICENSE file.