idid
is a common interface for different Id types. It allows you to
define distinct types for each of your Id types, even though they might
have the same backing type (Int
, Long
, UUID
, etc).
Imagine you have a project with several entities, and they, in some cases,
share the same base type of Id. For instance, if you have a Customer
entity and Product
, both them might have Ids, and they might be an Int
,
Long
, or perhaps an UUID
. For example, they might look like this:
case class Customer(id: Int, /*...*/)
case class Product(id: Int, /*...*/)
And perhaps, at some point, you might have a method that receives multiple Ids. Take for example:
def isProductInBasket(customerId: Int, productId: Int) = ???
And that's when things start to get confusing. Simply swapping your parameters might cause a lot of undesired headaches and debugging sessions. And that's when you start to wonder, "isn't Scala a typed language? How can I make it differently"? Wouldn't you like to have different types for different Ids, even though their backing type is the same, and therefore allow the compiler to figure out that something is wrong in your code?
There are many ways to do that, but we were looking for a way that would be generic enough that you could have a base type and easily parse different Id types. Consider, for instance, you are using Play, and you have REST endpoints for the different entities, each having a specific Id type on the path. You don't want to write a binder for each of them, but instead it would be much nicer to have a common one that can parse them all. The same applies for writing those Ids to the database with, let's say, Slick and custom type mappers.
The example above could be written as the following:
import com.unstablebuild.idid._
case class CustomerId(underlying: Int) extends TypedId[Int]
implicit val customerIdFactory = Id.factory[CustomerId]
case class ProductId(underlying: Int) extends TypedId[Int]
implicit val customerIdFactory = Id.factory[ProductId]
case class Customer(id: CustomerId, /*...*/)
case class Product(id: ProductId, /*...*/)
def isProductInBasket(customerId: CustomerId, productId: ProductId) = ???
So far, that doesn't look like much. The magic, though, starts when using
the Id
object directly.
// Create an Id
val customerId = Id.create[CustomerId](123)
// Create random Ids
val customerId = Id.random[CustomerId]
// Parse from a String
val customerId = Id.parse[CustomerId]("123")
// Create from a default value (empty), i.e. 0
val customerId = Id.empty[CustomerId]
// Get its value
val underlyingId = Id.value(customerId)
This way, when you need to create something like a Play binder, you could declare it this way:
def idBinder[T <: Id : IdFactory] = new PathBindable[T] { /*...*/ }
It's import to point out that you can still use your Id class normally. We also advice to put your Id types and their factories on a Package Object, so they can be accessed more easily.
Instead of declaring a companion variable for each of your Id classes, one can instead use the implicit factory generator provided by com.unstablebuild.idid.factory.AutoIdFactory
or the com.unstablebuild.idid.auto
package. For instance:
object MyIds extends AutoIdFactory {
case class MyId(underlying: Int) extends TypedId[Int]
}
import MyIds._
val id = Id.random[MyId]
or
case class MyId(underlying: Int) extends TypedId[Int]
import com.unstablebuild.idid.auto._
val id = Id.random[MyId]
Default values for the underlying Id types, how to parse them, or how
random values are generated, are specific by instances of IdSource
. A
source is select through an implicit binding. Default sources are defined for the following types:
UUID
Int
Long
BigInt
String
If you require a type that is not available, all you have to do is declare
your own implicit source. To give an example, the source for an Int
looks like the following
implicit val intSource: IdSource[Int] = new IdSource[Int] {
override def random: Int = Random.nextInt()
override def parse(str: String): Int = str.toInt
override def empty: Int = 0
}
implicit def idOrdering[T <: Id](implicit ordering: Ordering[T#UID]): Comparator[T] =
Ordering.by[T, T#UID](_.underlying)
implicit def idBinder[T <: Id : IdFactory] = new PathBindable[T] {
override def bind(key: String, value: String): Either[String, T] =
Try(Id.parse[T](value)).toOption.toRight(s"Could not convert $value into ID")
override def unbind(key: String, id: T): String =
id.toString
}
implicit def idFormat[T <: Id](implicit factory: IdFactory[T], format: Format[T#UID]): Format[T] = new Format[T] {
override def writes(id: T): JsValue = format.writes(id.underlying)
override def reads(json: JsValue): JsResult[T] = json.validate[T#UID].map(Id.create[T])
}
The following will create a mapper from any defined type who also has also has a valid BaseColumnType
in Slick:
implicit def uuidMapper[T <: Id : IdFactory : ClassTag](implicit baseColumnType: BaseColumnType[T#UID]) =
MappedColumnType.base[T, T#UID](_.underlying, Id.create[T])
To use it with SBT, add the following to your build.sbt
file:
resolvers += Resolver.sonatypeRepo("public")
libraryDependencies += "com.unstablebuild" %% "idid" % "0.2.0"
Special thanks to Christian Wilhelm for the ideas behind this project.
./sbt +test +macros/test
./sbt +publishSigned +macros/publishSigned
./sbt sonatypeReleaseAll