DValidation is a little, opinionated domain object validation toolkit on top of scalaz. While scalaz already offers the great Validation abstractions for building, composing and transforming validations it is a very general tool.
DValidation tries to be more opinionated with the goal to offer tools for common validation situations including:
- A set of built-in specific and generic validators
- A custom error aggregation type (DomainErrors)
- An abstract, translation friendly error type (DomainError)
- A set of classes representing common error cases (IsEmptyStringError, IsNotGreaterThenError, ...)
- A Path abstraction for locating errors in a class hierarchy
- Converters from other error representations
- Tools for creating domain specific error types
- Utilities to support testing validations
DValidation is available via the Sonatype OSS repository:
resolvers += Resolver.sonatypeRepo("releases")
The current release targets the Scala 2.10.x and 2.11.x series together with scalaz 7.0.6 or 7.1.0.
libraryDependencies += "net.atinu" %% "dvalidation" % "0.2"
libraryDependencies += "org.scalaz" %% "scalaz-core" % "7.1.0"
Lets start with this simple domain model:
trait Classification
case object StringInstrument extends Classification
case object Keyboard extends Classification
abstract class Instrument(val classification: Classification)
case object BassGuitar extends Instrument(StringInstrument)
case object Guitar extends Instrument(StringInstrument)
case object Piano extends Instrument(Keyboard)
case class Musician(name: String, age: Int, instruments: Seq[Instrument])
A way to get started with DValidation is to define an ad-hoc validation using the predefined combinators:
val mikael = Musician("Mikael Åkerfeldt", 40, List(Guitar, BassGuitar))
val martin = Musician("Martin Mendez", 17, List(BassGuitar))
val res: DValidation[Musician] = mikael.validateWith(
notBlank(mikael.name) forAttribute 'name,
ensure(mikael.age)("error.dvalidation.legalage", 18)(_ > 18) forAttribute 'age,
hasElements(mikael.instruments) forAttribute 'instruments
)
The object mikael is valid and therefore the validation will result in a Success:
Success(User(Mikael Åkerfeldt,40))
martin on the other hand is too young which is also reflected in the validation result:
Failure(DomainError(path: /age, value: 17, msgKey: error.dvalidation.legalage, args: 18))
The definition of reusable validations can be achieved using a DValidator which can be defined as follows:
val musicianValidator: DValidator[Musician] = Validator.template[Musician] { musician =>
musician.validateWith(
notBlank(musician.name) forAttribute 'name,
ensure(musician.age)(key = "error.dvalidation.legalage", args = 18)(_ > 18) forAttribute 'age,
hasElements(musician.instruments) forAttribute 'instruments
)
}
musicianValidator(mikael)
musicianValidator(martin)
A common issue when validating business objects is the validation of sequences since errors
are either global or have to be mapped to a specific element. In the example the
hasElements
combinator is used to define a global validation rule. validSequence
is
a combinator which applies a given validation to all elements of a sequence. The
withValidations
combinator will include the resulting sequence of DValidations
in the
parent validation context.
val max = Musician("Max Mustermann", 29, List(Piano, Guitar))
val stringInstrumentValidator = Validator.template[Instrument](i =>
ensure(i)(
key = "error.dvalidation.stringinstrument",
args = i.classification)(_.classification == StringInstrument)
)
max.validateWith(
notBlank(max.name) forAttribute 'name,
ensure(max.age)("error.dvalidation.legalage", 18)(_ > 18) forAttribute 'age,
hasElements(max.instruments) forAttribute 'instruments
).withValidations(
validSequence(max.instruments, stringInstrumentValidator) forAttribute 'instruments
)
This validation will result in one DomainError
with a path indicating the position in
the sequence.
Failure(DomainError(path: /instruments/[0], value: Piano, msgKey: error.dvalidation.stringinstrument, args: Keyboard))
Validator | Syntax | Works for | Example |
---|---|---|---|
notBlank | String | notBlank("a") or notBlank(" ", trimWhitespace = true) | |
notZero | scalaz.Monoid | ||
hasElements | Traversable | hasElements(List(1,2,3)) | |
isSome | Option | isSome(Option(2)) | |
isTrySuccess | Try | isTrySuccess(Try{ "bla" }) | |
isEqual | is_== | scalaz.Order | 1 is_== 1 |
isEqualStrict | is_=== | scalaz.Order | 1 is_=== 1 |
isGreaterThan | is_> / is_>= | scalaz.Order | 2 is_> 1 |
isSmallerThan | is_< / is_< | scalaz.Order | 1 is_< 2 |
isInRange | scalaz.Order | isInRange(4, min = 1, max = 5) | |
hasSize | dvalidation.Sized | hasSize(List(1, 2, 3), min = 4) | |
hasLength | String | hasLength("1", min = 1) | |
italic types are required implicit views |
Validation often get applied to data which lives in a context. The state of the context now becomes a requirement
for a successful validation. If for example an Option[String]
is used to represent an optional academic title
a validation expects the String
to be non empty if the Option
is a Some
. To support those use cases higher
order validators have been introduces for Option
and Try
. With this tool the academic title example can be
solved like this:
val academicTitle: Option[String] = Some("Dr.")
validOptRequired(academicTitle)(title => notBlank(title))
The ensure combinator offers a simple approach how to define a custom validator.
def isEqual[T](valueCheck: T, valueExcept: T): DValidation[T] =
ensure(valueCheck)("error.dvalidation.isequal", valueExcept)(a => a == valueExcept)
DValidation offers generic validators which are only useful when applied to a concrete class
or use case. For example the isSmallerThan validator works for all instances with a defined scalaz.Order
type class. A validation for Java 8 java.time.LocalDateTime for example would look like this:
object DateValidation {
private implicit val lDtOrder = scalaz.Order.order[LocalDateTime]((a, b) =>
if (a.isBefore(b)) Ordering.LT else Ordering.GT)
val inPast = Validator.template[LocalDateTime](_ is_< LocalDateTime.now())
val inFuture = Validator.template[LocalDateTime](_ is_> LocalDateTime.now())
}
Since isSmallerThen
does not know for which class the validation was done, the validator can only
yield an IsNotLowerThenError
. This is not an optimal situation if specific translations
have to be presented to the user or if business logic should is bound to this error scenario.
DValidation offers so called ErrorMaps
to map the generic error value of a validator to a use case
specific error representation. In the context of the date validation this could look like this:
object DateValidation {
private implicit val lDtOrder = scalaz.Order.order[LocalDateTime]((a, b) =>
if (a.isBefore(b)) Ordering.LT else Ordering.GT)
private implicit val toInPastError = ErrorMap.mapKey[IsNotLowerThenError]("dvalidaiton.dateInPast")
private implicit val toInFutureError = ErrorMap.mapKey[IsNotGreaterThenError]("dvalidaiton.dateInFuture")
val inPast = Validator.template[LocalDateTime](_ is_< LocalDateTime.now())
val inFuture = Validator.template[LocalDateTime](_ is_> LocalDateTime.now())
}
scalaz Validation offers an applicative validation style with which validations can be composed
and mapped to a larger validation. In this aspect the applicative validation is similar to the
common validation style of DValidation, with the difference that the mapping from element validations
to a bigger validation is done explicitly via a mapping function. Because of this the applicative
validation style is potentially more type safe, if the apply
function of a case class is used. The cost
of this style is the fixed structure of the expected validations, which can lead to more complicated
validation code.
val musicianValidatorApplicative = Validator.template[Musician] { musician =>
val stringInstrument = validSequence(musician.instruments, stringInstrumentValidator).collapse
val atLeastOneString = stringInstrument.disjunction
.flatMap(value => hasElements(value).disjunction).validation
val hasLegalAge = ensure(musician.age)(key = "error.dvalidation.legalage", args = 18)(_ > 18)
((notBlank(musician.name) forAttribute 'name) |@|
(hasLegalAge forAttribute 'age) |@|
(atLeastOneString forAttribute 'instruments))(Musician.apply)
}
DValidation offers conversions between Scala standard library types and DValidation.
Some(1).asValidation
// => Success(1)
val opt: Option[Int] = None
opt.asValidation
// => Failure(DomainError(path: /, value: None, msgKey: error.dvalidation.isNone))
scala.util.Success(1).asValidation
// => Success(1)
val exception = new IllegalArgumentException
scala.util.Failure(exception).asValidation
// => Failure(DomainError(path: /, value: java.lang.IllegalArgumentException, msgKey: error.dvalidation.isTryFailue))
A DValidation
and DValidator
are defined as follows:
type DValidation[T] = scalaz.Validation[DomainErrors, T]
type DValidator[T] = T => DValidation[T]
In other words, in the failure case of a scalaz.Valdation
the DomainErrors
type is used,
which itself is a list of at least one DomainError. This type has attributes which
allow clients to handle errors appropriately. The path
attribute helps to map an error to
the domain object which is responsible for the error. The value
attribute will return the
value of the validated attribute. Together with the msgKey
and args
attributes clients
can build language specific error messages, which can be represented to users.
The library packages some predefined type of DomainErrors like IsEmptySeqError
which already
chooses appropriate value and msgKey values.