as is a no-macro, no-reflection, opinionated type refinement library for Scala 3. It is powered by e to handle invalid value errors.
| Latest Version | Java Version | Scala Version |
|---|---|---|
| 1.0.1 | 21 | 3.4.2 |
If you use SBT, add following to your build.sbt:
libraryDependencies += "dev.akif" %% "as" % "1.0.1"If you use Maven, add following to your pom.xml:
<dependencies>
<dependency>
<groupId>dev.akif</groupId>
<artifactId>as_3</artifactId>
<version>1.0.1</version>
</dependency>
</dependencies>If you use Gradle, add following to your project's build.gradle:
dependencies
{
implementation('dev.akif:as_3:1.0.1')
}Assume we have following Author and Book types.
case class Author(id: Long, name: String)
case class Book(id: Long, authorId: Long, name: String)Long and String are primitives and may not contain valid values for ids or names. Even if you implement ids and names with their own types to guard against invalid values, author's id is different from a book's id, and they should not be of same type.
To overcome this, we can have separate refined types.
Here is how you can create one, refining Long as AuthorId:
- Make an
opaque typealiasAuthorId = Longin a separate file- Optionally add a subtype constraint
<: Longif you want to be able to makeAuthorIdtype a subtype ofLong(in other words, to be able to assign anAuthorIdvalue to aLongreference)
- Optionally add a subtype constraint
- Create a companion object extending
UnderlyingType as RefinedType(Long as AuthorIdin this case) - Implement
undefinedValueandvalidationmembers
// AuthorId.scala
import dev.akif.as.*
opaque type AuthorId <: Long = Long
// Parenthesis are required because `as` is used as an infix type, which would otherwise be `as[Long, AuthorId]`
object AuthorId extends (Long as AuthorId):
override val undefinedValue: Long = 0L
override val validation: Validation[Long] =
Validation.min(lowerLimit = 0L, inclusive = false)This incurs no additional allocation because AuthorId is defined as an opaque type alias. Within AuthorId.scala file, it can be treated as a Long. However, in any other file, AuthorId and Long are completely different types.
Let's create other types too.
// AuthorName.scala
import dev.akif.as.*
opaque type AuthorName <: String = String
object AuthorName extends (String as AuthorName):
override val undefinedValue: String = ""
override val validation: Validation[String] =
Validation.all(
Validation.make(predicate = !_.isBlank, failureMessage = _ => "Value is blank"),
Validation.make(predicate = _.length <= 64, failureMessage = _ => "Value has more than 64 characters")
)
// BookId.scala
import dev.akif.as.*
opaque type BookId <: Long = Long
object BookId extends (Long as BookId):
override val undefinedValue: Long = 0L
override val validation: Validation[Long] =
Validation.min(lowerLimit = 0L, inclusive = false)
// BookName.scala
import dev.akif.as.*
opaque type AuthorName <: String = String
object AuthorName extends (String as AuthorName):
override val undefinedValue: String = ""
override val validation: Validation[String] =
Validation.all(
Validation.make(predicate = !_.isBlank, failureMessage = _ => "Value is blank"),
Validation.make(predicate = _.length <= 64, failureMessage = _ => "Value has more than 64 characters")
)Now let's change the original example.
case class Book(id: AuthorId, name: AuthorName)
case class Book(id: BookId, authorId: AuthorId, name: BookName)Now our types are fine-tuned so that they will have correct values, and we are forced to check everything at compile time. We need to use apply method, as extension method or their variants to create a refined value. Creating a refined type by applying an A to an A as B returns an E or B. In other words, object construction can fail and this is reflected to types. Since we have ors in the construction, it's easy to combine them, even in for comprehensions.
import dev.akif.as.*
import e.scala.*
// Failure({"name":"invalid-data","causes":[{"message":"Value is less than or equal to 0"}],"data":{"type":"AuthorId","value":"-1"}})
val authorId1: E or AuthorId = AuthorId(-1L)
// Success(1)
val authorId2: E or AuthorId = AuthorId(1L)
// Can also be written with the extension method
// val authorId2: E or AuthorId = 1L.as[AuthorId]
// Success(Author(1, Mehmet Akif Tütüncü))
val maybeAuthor: E or Author =
for
id <- AuthorId(1L)
name <- AuthorName("Mehmet Akif Tütüncü")
yield
Author(id, name)In cases where you need to have the refined AuthorId directly instead of E or AuthorId, you can use:
AuthorId.applyOrUndefined(-1L)(or-1L.asOrUndefined[AuthorId]) method that will return you the undefined value when validation failsAuthorId.unsafe(-1L)(or-1L.asUnsafe[AuthorId]) method that will throw the validation error as an exception, hence the name unsafe
// Book(0, 0, Type Refinement in Scala)
val book: Book = Book(
BookId.undefined,
AuthorId.applyOrEmpty(-1L),
BookName.unsafe("Type Refinement in Scala")
)Accessing the unrefined value is possible via the conveniently named extension method value. If defined as a subtype, one can also directly assign a refined value to a reference of the unrefined type.
// 0L
val bookId: Long = book.id.value
// "Type Refinement in Scala"
val bookName: String = book.name // since BookName alias is defined as a subtype of Stringas is built with SBT. You can use clean, compile, test tasks for development and testing.
as packages are published to Maven Central, and they are versioned according to semantic versioning. Release process is managed by sbt-release.
All contributions are welcome, including requests to feature your project utilizing as. Please feel free to send a pull request. Thank you.
as is licensed with MIT License.