yakivy / dupin   0.6.1

MIT License GitHub

Minimal, idiomatic, customizable validation Scala library.

Scala versions: 3.x 2.13 2.12 2.11
Scala.js versions: 1.x
Scala Native versions: 0.4

Dupin

Maven Central Sonatype Nexus (Snapshots) License: MIT Cats friendly

Dupin is a minimal, idiomatic, customizable validation Scala library.

You may find Dupin useful if you...

  • want a transparent and composable validation approach
  • need to return something richer than String as validation message
  • use effectful logic inside validator (Future, IO, etc...)
  • like parse don't validate style
  • have cats dependency and like their API style
  • need Scala 3, Scala JS or Scala Native support

Library is built around two type classes:

  • Validator[F[_], E, A] - is a self-sufficient validator for type A, represents a function A => F[ValidatedNec[E, Unit]]
  • Parser[F[_], E, A, B] - is a parser from type A to type B, represents a function A => F[IorNec[E, B]]

Table of contents

  1. Quick start
    1. Validate
    2. Parse
  2. Predefined validators
  3. Message customization
  4. Effectful validation
  5. Custom validating package
  6. Complex example
  7. Roadmap
  8. Changelog

Quick start

Add cats and dupin dependencies to the build file, let's assume you are using sbt:

libraryDependencies += Seq(
    "org.typelevel" %% "cats-core" % "2.9.0",
    "com.github.yakivy" %% "dupin-core" % "0.6.1",
)

Describe the domain:

case class Name(value: String)
case class Member(name: Name, age: Int)
case class Team(name: Name, members: Seq[Member])

Validate

Define validators:

import cats._
import dupin.basic.all._

//validator for simple type or value class
implicit val nameValidator: BasicValidator[Name] = BasicValidator
    .root[Name](_.value.nonEmpty, c => s"${c.path} should be non empty")

//idiomatic validator for complex type
implicit val memberValidator: BasicValidator[Member] =
    nameValidator.comapP[Member](_.name) combine
    BasicValidator.root[Int](
        a => a > 18 && a < 40,
        c => s"${c.path} should be between 18 and 40"
    ).comapP[Member](_.age)

//same validator but with combination helpers for better type resolving
val alternativeMemberValidator: BasicValidator[Member] = BasicValidator
    .success[Member]
    .combineP(_.name)(nameValidator)
    .combinePR(_.age)(a => a > 18 && a < 40, c => s"${c.path} should be between 18 and 40")

//derived validator
implicit val teamValidator: BasicValidator[Team] = BasicValidator
    .derive[Team]
    .combineR(_.members.size <= 8, _ => "team should be fed with two pizzas!")

//two stage validator
val failingTeamValidator: BasicValidator[Team] = teamValidator
    .andThen(BasicValidator.failure[Team](_ => "validation error after heavy computations"))

Validate them all:

import dupin.basic.all._

val validTeam = Team(
    Name("Bears"),
    List(
        Member(Name("Yakiv"), 26),
        Member(Name("Myroslav"), 31),
        Member(Name("Andrii"), 25)
    )
)

val invalidTeam = Team(
    Name(""),
    Member(Name(""), 0) :: (1 to 10).map(_ => Member(Name("Valid name"), 20)).toList
)

assert(validTeam.isValid)
assert(invalidTeam.validate == Validated.invalid(NonEmptyChain(
    ".members.[0].name should be non empty",
    ".members.[0].age should be between 18 and 40",
    ".name should be non empty",
    "team should be fed with two pizzas!",
)))
assert(failingTeamValidator.validate(validTeam) == Validated.invalid(NonEmptyChain(
    "validation error after heavy computations",
)))
assert(failingTeamValidator.validate(invalidTeam) == Validated.invalid(NonEmptyChain(
    ".members.[0].name should be non empty",
    ".members.[0].age should be between 18 and 40",
    ".name should be non empty",
    "team should be fed with two pizzas!",
)))

Parse

Enrich the domain with raw models to parse:

case class RawMember(name: String, age: Int)
case class RawTeam(name: String, members: List[RawMember])

Define parsers:

import cats._
import cats.implicits._
import dupin.basic.all._

// parser for simple type or value class
implicit val nameParser: BasicParser[String, Name] = BasicParser.root[String, Name](
    Option(_).filter(_.nonEmpty).map(Name.apply),
    c => s"${c.path} should be non empty",
)

//idiomatic parser for complex type
implicit val memberParser: BasicParser[RawMember, Member] =
    (
        nameParser.comapP[RawMember](_.name),
        BasicParser.idRoot[Int](
            Option(_).filter(a => a > 18 && a < 40),
            c => s"${c.path} should be between 18 and 40",
        ).comapP[RawMember](_.age),
    )
        .parMapN(Member.apply)

implicit val teamParser: BasicParser[RawTeam, Team] =
    (
        nameParser.comapP[RawTeam](_.name),
        memberParser.liftToTraverseCombiningP[List].comapP[RawTeam](_.members),
    )
        .parMapN(Team.apply)
        .andThen(
            //if you need identity parser that filters out value by condition,
            //you can simply create a validator and convert it to parser
            BasicValidator
                .root[Team](_.members.size <= 8, _ => "team should be fed with two pizzas!")
                .toParser
        )

Parse them all:

val validTeam = RawTeam(
    "Bears",
    List(
        RawMember("Yakiv", 26),
        RawMember("Myroslav", 31),
        RawMember("Andrii", 25)
    )
)

val invalidTeam = RawTeam(
    "",
    RawMember("", 0) :: (1 to 10).map(_ => RawMember("Valid name", 20)).toList
)

assert(validTeam.parse == Ior.right(Team(
    Name("Bears"),
    List(
        Member(Name("Yakiv"), 26),
        Member(Name("Myroslav"), 31),
        Member(Name("Andrii"), 25)
    )
)))
assert(invalidTeam.parse == Ior.left(NonEmptyChain(
    ".name should be non empty",
    ".members.[0].name should be non empty",
    ".members.[0].age should be between 18 and 40",
)))

Predefined validators

It also might be useful to extract and reuse validators for common types. Let's define validators for minimum and maximum Int value:

import dupin.basic.all._

def min(value: Int) = BasicValidator.root[Int](_ > value, c => s"${c.path} should be greater than $value")
def max(value: Int) = BasicValidator.root[Int](_ < value, c => s"${c.path} should be less than $value")

And since validators can be combined, you can use them to create more complex validators:

import cats._
import dupin.basic.all._

implicit val memberValidator: BasicValidator[Member] = BasicValidator
    .success[Member]
    .combineP(_.age)(min(18) && max(40).failureAs(_ => "updated validation message"))

val invalidMember = Member(Name("Ada"), 0)
val result = invalidMember.validate

assert(result == Validated.invalidNec(".age should be greater than 18"))

Message customization

But not many real projects use strings as validation messages, for example you want to support internationalization:

case class I18nMessage(
    description: String,
    key: String,
    params: List[String]
)

BasicValidator[A] is simply a type alias for Validator[Id, String, A], so you can define own validator type with partially applied builder:

import dupin._

type I18nValidator[A] = Validator[cats.Id, I18nMessage, A]
val I18nValidator = Validator[cats.Id, I18nMessage]

And start creating validators with custom messages:

import cats._

implicit val nameValidator: I18nValidator[Name] = I18nValidator.root[Name](
    _.value.nonEmpty,
    c => I18nMessage(
        s"${c.path} should be non empty",
        "validator.name.empty",
        List(c.path.toString())
    )
)

implicit val memberValidator: I18nValidator[Member] = I18nValidator
    .success[Member]
    .combinePI(_.name)
    .combinePR(_.age)(a => a > 18 && a < 40, c => I18nMessage(
        s"${c.path} should be between 18 and 40",
        "validator.member.age",
        List(c.path.toString())
    ))

Validation messages will look like:

import dupin.syntax._

val invalidMember = Member(Name(""), 0)
val result = invalidMember.validate

assert(result == Validated.invalid(NonEmptyChain(
    I18nMessage(
        ".name should be non empty",
        "validator.name.empty",
        List(".name")
    ),
    I18nMessage(
        ".age should be between 18 and 40",
        "validator.member.age",
        List(".age")
    )
)))

Effectful validation

For example, you want to allow only a limited list of names and it is stored in the database:

import scala.concurrent.Future

class NameService {
    private val allowedNames = Set("Ada")
    def contains(name: String): Future[Boolean] =
        // Emulation of DB call
        Future.successful(allowedNames(name))
}

So to be able to handle checks that return Future[Boolean], you just need to define your own validator type with partially applied builder:

import dupin._
import scala.concurrent.Future

type FutureValidator[A] = Validator[Future, String, A]
val FutureValidator = Validator[Future, String]

Then you can create validators with generic DSL (don't forget to import required type classes, as minimum Functor[Future]):

import cats.implicits._
import scala.concurrent.Future

val nameService = new NameService

implicit val nameValidator: FutureValidator[Name] = FutureValidator.rootF[Name](
    n => nameService.contains(n.value),
    c => s"${c.path} should be non empty"
)

implicit val memberValidator: FutureValidator[Member] = FutureValidator
    .success[Member]
    .combinePI(_.name)
    .combinePR(_.age)(a => a > 18 && a < 40, c => s"${c.path} should be between 18 and 40")

Validation result will look like:

import dupin.syntax._

val invalidMember = Member(Name(""), 0)
val result: Future[ValidatedNec[String, Member]] = invalidMember.validate

result.map(r => assert(r == Validated.invalid(NonEmptyChain(
    ".name should be non empty",
    ".age should be between 18 and 40"
))))

Custom validating package

To avoid imports boilerplate and isolating all customizations, you can define your own dupin package:

package object custom extends DupinCoreDsl with DupinSyntax {
    type CustomValidator[A] = Validator[Future, I18nMessage, A]
    val CustomValidator = Validator[Future, I18nMessage]

    type CustomParser[A, B] = Parser[Future, I18nMessage, A, B]
    val CustomParser = Parser[Future, I18nMessage]
}

Then you can start using custom validator type with a single import:

import cats.implicits._
import dupin.custom._

val nameService = new NameService

implicit val nameValidator: CustomValidator[Name] = CustomValidator.rootF[Name](
    n => nameService.contains(n.value),
    c => I18nMessage(
        s"${c.path} should be non empty",
        "validator.name.empty",
        List(c.path.toString())
    )
)

val validName = Name("Ada")
val valid: Future[Boolean] = validName.isValid

valid.map(assert(_))

Complex example

Let's assume that you need to build a method that receives a list of raw term models (each model is a product of term itself and a list of mistakes that people often make when typing this term, for example: "calendar" -> ["calender", "celender"]) and parses them before saving to the database. Here are some requirements:

  • suggested raw model:
case class RawTermModel(
    term: String,
    mistakes: List[String],
)
  • term and mistake should be a single word
  • term and mistake should not exist in the database:
type R[A] = Either[String, A]
trait TermRepository {
    def contains(term: Term): R[Boolean] = ...
}
  • terms should be unique among other terms in the list
  • mistakes should be unique among other mistakes and terms in the list
  • parsed model should have as minimum one mistake
  • suggested final model:
case class Term(value: String)
case class TermModel(
    term: Term,
    mistakes: NonEmptyList[Term],
)
  • if validation error occurs in term - skip the model and continue parsing
  • if validation error occurs in mistake - skip the mistake only and continue parsing
  • all validation errors should be collected and returned after parsing

So the parser from RawTermModel to TermModel, considering the requirements above, will look like:

//validation types to handle repository effect `R`
type CustomValidator[A] = Validator[R, String, A]
val CustomValidator = Validator[R, String]
type CustomParser[A, B] = Parser[R, String, A, B]
val CustomParser = Parser[R, String]

//parsers per requirement:

//term and mistake should be a single word
val termParser = CustomParser
    .root[String, Term](
        Option(_).filter(_.matches("\\w+")).map(Term.apply),
        c => s"${c.path}: cannot parse string '${c.value}' to a term"
    )

//term and mistake should not exist in the database
val repositoryTermParser = CustomValidator
    .rootF[Term](
        TermRepository.contains(_).map(!_),
        c => s"${c.path}: term '${c.value}' already exists"
    )
    .toParser

//intermediate model to aggregate parsed terms
case class HalfParsedTermModel(
    term: Term,
    mistakes: List[Term],
)

//terms should be unique among other terms in the list
val uniqueTermsParser = CustomParser
    //define list level context where terms should be unique
    .idContext[List[HalfParsedTermModel]] { _ =>
        val validTerms = mutable.Set.empty[Term]
        CustomValidator
            .root[Term](validTerms.add, c => s"${c.path}: term '${c.value}' is duplicate")
            .comapP[HalfParsedTermModel](_.term)
            .toParser
            //lift parser to `List` accumulating errors
            .liftToTraverseCombiningP[List]
    }

//mistakes should be unique among other mistakes and terms in the list
val uniqueTermsMistakesParser = CustomParser
    .idContext[List[HalfParsedTermModel]] { ms =>
        val validTerms = mutable.Set.from(ms.view.map(_.term))
        CustomParser
            .idContext[HalfParsedTermModel] { m =>
                CustomValidator
                    .root[Term](validTerms.add, c => s"${c.path}: mistake '${c.value}' is duplicate")
                    .toParser
                    .liftToTraverseCombiningP[List]
                    .comapP[HalfParsedTermModel](_.mistakes)
                    //update model with unique mistakes
                    .map(a => m.copy(mistakes = a))
            }
            .liftToTraverseCombiningP[List]
    }

//parsed model should have as minimum one mistake
def nelParser[A] = CustomParser
    .root[List[A], NonEmptyList[A]](_.toNel, c => s"${c.path}: cannot be empty")
val halfToFullModelParser = CustomParser
    .context[HalfParsedTermModel, TermModel](m =>
        nelParser[Term]
            .comapP[HalfParsedTermModel](_.mistakes)
            .map(mistakes => TermModel(m.term, mistakes))
    )

//combine all parsers together
val modelsParser = (
    termParser
        .andThen(repositoryTermParser)
        .comapP[RawTermModel](_.term),
    termParser
        .andThen(repositoryTermParser)
        .liftToTraverseCombiningP[List]
        .comapP[RawTermModel](_.mistakes),
)
    .parMapN(HalfParsedTermModel.apply)
    .liftToTraverseCombiningP[List]
    .andThen(uniqueTermsParser)
    .andThen(uniqueTermsMistakesParser)
    .andThen(halfToFullModelParser.liftToTraverseCombiningP[List])

(full list of test cases can be found here)

Roadmap

  • add unzip from index for validator/parser
  • enrich parser tests with validator cases
  • optimize Parser.liftToTraverseCombiningP, combineK is often slow, for example for lists
  • add complex example without parser for comparison
  • rename comap to contramap

Changelog

0.6.x

  • add Parser type
  • replace implicit conversion (comapToP, ...) with explicit lift methods (liftToTraverseP, ...)
  • a couple of minor fixes

0.5.x

  • simplify internal validator function
  • expose validator contravariant monoidal instance ContravariantMonoidal[Validator[F, E, *]]

0.4.x

  • add Scala 3 support for Scala Native
  • update Scala JS version
  • optimize path concatenation
  • separate F Validator methods (like rootF)
  • add Validator methods with context (like combineC)

0.3.x:

  • rename dupin.Validator.compose to dupin.Validator.comap, similar to cats.Contravariant.contramap
  • rename dupin.Validator.combinePK to dupin.Validator.combinePL, where L stands for "lifted" to reflect method signature
  • optimize a naive implementation of ValidatorComapToP.validatorComapToPForTraverse that threw StackOverflowException for long lists
  • minor refactorings

0.2.x:

  • migrate to mill build tool
  • add Scala 3, Scala JS and Scala Native support
  • expose validator monoid instance MonoidK[Validator[F, E, *]]
  • rename dupin.base package to dupin.basic
  • various refactorings and cleanups