NxtSchema

Installation

Add this line to your application's Gemfile:

gem 'nxt_schema'

And then execute:

$ bundle

Or install it yourself as:

$ gem install nxt_schema

What is it for?

NxtSchema is a type coercion and validation framework that allows you to coerce and validate arbitrary nested structures of data. The original idea is taken from https://dry-rb.org/gems/dry-schema and https://dry-rb.org/gems/dry-validation from the amazing dry.rb eco system. In contrast to dry-schema, NxtSchema aims to be a simpler solution that hopefully is easier to understand and debug. It also ships with some handy features that dry-schema does not implement.

Usage

PERSON = NxtSchema.schema(:person) do
  node(:first_name, :String)
  node(:last_name, :String)
  node(:email, :String, optional: true).validate(:includes, '@')
end

input = {
  first_name: 'Ändy',
  last_name: 'Robecke',
  email: '[email protected]'
}

result = PERSON.apply(input: input)

result.valid? # => true
result.output # => input

Nodes

A schema consists of a number of nodes. Every node has a name and an associated type for casting it's input when the schema is applied. Schemas can consist of 4 different kinds of nodes: 

NxtSchema::Node::Schema # => Hash of values 
NxtSchema::Node::Collection # => Array of values
NxtSchema::Node::AnyOf # => Any of the defined schemas
NxtSchema::Node::Leaf # => Node without sub nodes

The kind of node dictates how the schema is applied to the input. On the root level the following methods are available to create schemas:

  NxtSchema.schema { ... } # => Creates a schema node 
  NxtSchema.collection { ... } # => Creates an array of nodes
  NxtSchema.any_of { ... } # => Creates a collection of allowed schemas

Node predicate aliases

Of course these nodes can be combined and nested in arbitrary manner. When defining nodes within a schema, nodes are always required per default. You can create nodes with the node method or several useful helper methods. 

NxtSchema.schema(:person) do
  required(:first_name, :String) # => same as node(:first_name, :String)
  optional(:last_name, :String) # => same as node(:first_name, :String, optional: true)
  omnipresent(:email, :String) # => same as node(:first_name, :String, omnipresent: true)
end

NOTE: The methods above only apply to the keys of your schema and do not make any assumptions about values!

In other word this means that making a node optional only makes your node optional. When your input contains the key but the value is nil, you will still get an error in case there is no default or maybe expression that applies. Omnipresent node also only inject the node into the schema but do not inject a default value. In order to inject a key with value into a schema you also have to combine the node predicates with default value method described below. For clarification check out the examples below:

# Optional node without default value

schema = NxtSchema.schema(:person) do
  optional(:email, :String)
end

result = schema.apply(input: { email: nil })
result.errors # => {"person.email"=>["nil violates constraints (type?(String, nil) failed)"]}
result.output # => {:email=>nil}

result = schema.apply(input: {})
result.errors # => {}
result.output # => {}

# Optional node with default value

schema = NxtSchema.schema(:person) do
  optional(:email, :String).default('[email protected]')
end

result = schema.apply(input: { email: nil })
result.errors # => {}
result.output # => {:email=>"[email protected]"}

result = schema.apply(input: {})
result.errors # => {}
result.output # => {}

# Omnipresent node without default value

schema = NxtSchema.schema(:person) do
  omnipresent(:email, :String)
end

result = schema.apply(input: {})
result.errors # => {}
result.output # => {:email=>NxtSchema::Undefined}

# Omnipresent node with default value and maybe expression to allow default value to break type contract.

schema = NxtSchema.schema(:person) do
  omnipresent(:email, :String).default(nil).maybe(:nil?)
end

result = schema.apply(input: {})
result.errors # => {}
result.output # => {:email=>nil}

result = schema.apply(input: { email: '[email protected]' })
result.errors # => {}
result.output # => {:email=>"[email protected]"}
Conditionally optional nodes

You can also pass a proc as the optional option. This is a shortcut for adding a validation to the parent node that will result in a validation error in case the optional condition does not apply and the parent node does not contain a sub node with that name (here contact schema not including an email node). 

schema = NxtSchema.schema(:contact) do
  required(:first_name, :String)
  required(:last_name, :String)
  node(:email, :String, optional: ->(node) { node.up[:last_name].input == 'Robecke' })
end

result = schema.apply(input: { first_name: 'Andy', last_name: 'Other' })
result.errors # => {"contact"=>["Required key :email is missing"]}

result = schema.apply(input: { first_name: 'Andy', last_name: 'Robecke' })
result.errors # => {}

Combining Schemas

You can also simply reuse a schema by passing it to the node method as the type of a node. When doing so the schema will be cloned with the same options and configuration as the schema passed in. 

ADDRESS = NxtSchema.schema(:address) do
  required(:street, :String)
  required(:town, :String)
  required(:zip_code, :String)
end 

PERSON = NxtSchema.schema(:person) do
  required(:first_name, :String)
  required(:last_name, :String)
  optional(:address, ADDRESS)
end

Types

The type system is built with dry-types from the amazing https://dry-rb.org eco system. Even though dry-types also offers features such as default values for types as well as maybe types, these features are built directly into NxtSchema.

In NxtSchema every node has a type and you can either provide a symbol that will be resolved through the type system of the schema or you can directly provide an instance of dry type and thus use your custom types. This means you can basically build any kind of objects such as structs and models from your data and you are not limited to just hashes arrays and primitives.

Default type system

You can tell your schema which default type system it should use. Dry-Types comes with a few built in type systems. Per default NxtSchema will use nominal types if not specified otherwise. If the type cannot be resolved from the default type system that was specified NxtSchema will always fallback to nominal types. In theory you can provide a separate type system per node if that's what you need. 

NxtSchema.schema do
  required(:test, :String) # The :String will resolve to NxtSchema::Types::Nominal::String
end

NxtSchema.schema(type_system: NxtSchema::Types::JSON) do
  required(:test, :Date) # The :Date will resolve to NxtSchema::Types::JSON::Date
  # When the type does not exist in the default type system (there is non JSON::String) we fallback to nominal types
  required(:test, :String) 
end

NxtSchema.params

NxtSchema.params will give you a schema as root node with NxtSchema::Types::Params as default type system. This is suitable to validate and coerce your query params. 

NxtSchema.params do
  required(:effective_at, :DateTime) # would resolve to Types::Params::DateTime 
  required(:test, :String) # The :String will resolve to NxtSchema::Types::Nominal::String
  required(:advanced, NxtSchema::Types::Registry::Bool) # long version of required(:advanced, :Bool)
end

NxtSchema::Registry

To make use of NxtSchema.params in your controller you can simply include the NxtSchema::Registry to easily register and apply schemas: 

class UsersController < ApplicationController
  include NxtSchema::Registry

  # register the schema for the :create action
  schemas.register(
    :create, 
    NxtSchema.params do
      required(:first_name, :String)
      required(:last_name, :String)
    end
  )

  def create
    User.create!(**create_params) 
  end

  private

  def create_params
    # apply the registered schema  
    schemas.apply!(:create, params.fetch(:user))
  end
end

Custom types

You can also register custom types. In order to check out all the cool things you can do with dry types you should check out dry-types on https://dry-rb.org. But here is how you can add a type to the NxtSchema::Types module. 

NxtSchema.register_type(
  :MyCustomStrippedString,
  NxtSchema::Types::Strict::String.constructor(->(string) { string&.strip })
)

# once registered you can use the type in your schema

NxtSchema.schema(:company) do
  required(:name, :MyCustomStrippedString)
end

Values

Default values

# Define default values with the default method
required(:test, :DateTime).default(nil)
required(:test, :DateTime).default(-> { Time.current })

Maybe values

With maybe expressions you can halt coercion and allow your values to break the type contract. Note: This means that your output will simply be set to the input without coercing the value!

# Define maybe values (values that do not match the type)
required(:test, :String).maybe(:nil?)

nodes(:tests).maybe(:empty?) do # will allow the collection to be empty and thus not contain strings
  required(:test, :String)
end

Validations

NxtSchema comes with a simple validation system and ships with a small set of useful validators. Every node in a schema implements the :validate method. Similar to ActiveModel::Validations it allows you to simply add errors to a node based on some condition. When the node is yielded to your validation proc you have access to the nodes input with node.input and node.index when the node is within a collection of nodes as well as node.name. Furthermore you have access to the context that was passed in when defining the schema or passed to the apply method later.

NOTE: Validations only run when no maybe expression applies and the node input could be coerced successfully

  # Simple custom validation
  required(:test, :String).validate(-> (node) { node.add_error("#{node.input} is not valid") if node.input == 'not allowed' })
  # Built in validations
  required(:test, :String).validate(:attribute, :size, ->(s) { s < 7 }) 
  required(:test, :String).validate(:equal_to, 'same') 
  required(:test, :String).validate(:excluded_in, %w[not_allowed]) # excluded in the target: %w[not_allowed]
  required(:test, :String).validate(:included_in, %w[allowed]) # included in the target: %w[allowed]
  required(:test, :Array).validate(:excludes, 'excluded') # array value itself must exclude 'excluded' 
  required(:test, :Array).validate(:includes, 'included') # array value itself must include 'included'
  required(:test, :Integer).validate(:greater_than, 1) 
  required(:test, :Integer).validate(:greater_than_or_equal, 1) 
  required(:test, :Integer).validate(:less_than, 1) 
  required(:test, :Integer).validate(:less_than_or_equal, 1) 
  required(:test, :String).validate(:pattern, /\A.*@.*\z/) 
  required(:test, :String).validate(:query, :present?)

Custom validators

You can also register your custom validators. Therefore you can simply implement a class that returns a lambda on build. This lambda will be called with the node the validations runs on and it's input value. Except this, you are free in how your validator can be used. Check out the specs for some examples. 

class MyCustomExclusionValidator
  def initialize(target)
    @target = target
  end

  attr_reader :target

  def build
    lambda do |node, value|
      if target.exclude?(value)
        true
      else
        node.add_error("#{target} should not contain #{value}")
        false # validators must return false in the bad case (add_error already does this as per default)
      end
    end
  end
end

# Register your validators  
NxtSchema.register_validator(MyCustomExclusionValidator, :my_custom_exclusion_validator)

# and then simply reference it with the key you've registered it
schema = NxtSchema.schema(:company) do
  requires(:name, :String).validate(:my_custom_exclusion_validator, %w[lemonade])
end

schema.apply(name: 'lemonade').valid? # => false

Validation messages

  • Allow to specify a path to translations
  • Add translated errors
  • Interpolate with actual vs. expected

Combining validators

node(:test, String).validate(...) basically adds a validator to the node. Of course you can add multiple validators. But that means that they will all be executed. If you want your validator to only run in case another was false, you can use :validat_with do ... end in order to combine validators based on custom logic. 

NxtSchema.schema do
  required(:test, :Integer).validate_with do
    validator(:greater_than, 5) &&
      validator(:greater_than, 6) ||
      validator(:greater_than, 7)
  end
end

Note that this will not run subsequent validators once one was valuated to false and thus might not contain all error messages of all validators that would have failed.

Schema options

Optional keys strategies

Schemas in NxtSchema only look at the keys that you have defined in your schema, others are ignored per default. You can change this behaviour by providing a strategy for the :additional_keys option. 

# This will simply ignore any other key except test 
NxtSchema.schema(additional_keys: :ignore) do
  required(:test, :String) 
end

# This would give you an error in case you apply anything other than { test: '...' }
NxtSchema.schema(additional_keys: :restrict) do
  required(:test, :String) 
end

# This will merge other keys into your output
schema = NxtSchema.schema(additional_keys: :allow) do
  required(:test, :String) 
end

schema.apply(input: {test: 'getsafe', other: 'Heidelberg'})
schema.valid? # => true
schema.value # => { test: 'getsafe', other: 'Heidelberg' }

Transform keys

You may want to transform the keys from your input. Therefore specify the transform_keys option. This might be useful when you want your schema to return only symbolized keys for example. 

schema = NxtSchema.schema(transform_keys: ->(key) { key.to_sym}) do
  required(:test, :String)
end

schema.apply(input: { 'test' => 'getsafe' }) # => {:test=>"getsafe"}
schema.apply(input: { test: 'getsafe' }) # => {:test=>"getsafe"}

Adding meta data to nodes

You want to give nodes an ID or some other meta data? You can use the meta method on nodes for adding additional information onto any node. 

schema = NxtSchema.schema do
  ERROR_MESSAGES = {
    test: 'This is always broken'
  }

  required(:test, :String).meta(ERROR_MESSAGES).validate ->(node) { node.add_error(node.meta.fetch(node.name)) }
end

schema.apply(input: { test: 'getsafe' }) 
schema.error #  {"root.test"=>["This is always broken"]}

Contexts

When defining a schema it is possible to pass in a context option. This can be anything that you would like to access during building your schema. A context could provide custom validators or default values depending of the name of your nodes for instance.

Build time
context = OpenStruct.new(email_validator: ->(node) { node.input && node.input.includes?('@') }) 

NxtSchema.schema(:developers, context: context) do
  required(:first_name, :String)
  required(:last_name, :String)
  required(:email, :String).validate(context.email_validator)
end
Apply time

You can also pass in a context at apply time. If you do not pass in a specific context at apply time you can still access the context passed in at build time. Basically passing in a context at apply time will overwrite the context from before. You can access it simply through the node. 

build_context = OpenStruct.new(email_validator: ->(node) { node.input.includes?('@') })
apply_context = OpenStruct.new(default_role: 'BOSS')

schema = NxtSchema.schema(:developers, context: build_context) do
  # context at build time
  required(:email, :String).validate(context.email_validator) # 
  # access the context at apply time through the node 
  required(:role, :String).default { |_, node| node.context.default_role }
end

schema.apply(input: input, context: apply_context)

Development

After checking out the repo, run bin/setup to install dependencies. Then, run rake spec to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

To install this gem onto your local machine, run bundle exec rake install. To release a new version, update the version number in version.rb, and then run bundle exec rake release, which will create a git tag for the version, push git commits and tags, and push the .gem file to rubygems.org.

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/getand.

License

The gem is available as open source under the terms of the MIT License.

TODO:

  • Explain node interface
  • Add apply! method to readme
  • Allow to disable validation when applying --> Are there attributes that should be moved to apply time?
  • Should we have a global and a local registry for validators? --> Would be cool to register things for the schema only --> Would be cool if this was extendable
  • Do we need all off in order to combine multiple schemas?
  • Allow custom errors
  • Spec inheritance of params