fthomas/refined

Simple refinement types for Scala

Oron Port

@soronpo

@coffius just a tip. From my experience it's better to have classes unconstrained with private construcor and constrained with a public definition in a companion.

Aleksei Shamenev

@coffius

@soronpo, Hmm... Interesting idea. But maybe you can elaborate it a little bit more. I am not sure if I understand it good enough.

Oron Port

@soronpo

@coffius
With your example I would have done:

class GenericStringValidator private [Min <: Int, Max <: Int](minValue : Min, maxValue : Max) {
}
object GenericStringValidator  {
  def apply[Min <: Int with Singleton, Max <: Int with Singleton](min : Min, max : Max)(implicit  r : Require[Min <= Max]) = new GenericStringValidator(min, max)
}

If you have further questions it's better to ask in the singleton-ops gitter

Aleksei Shamenev

@coffius

@soronpo, Ok, now I got it. No further questions. Thanks 🙇‍♂️

matrixbot

@matrixbot

@jarek_rozanski:matrix.org I was wondering if it possible to apply Refined types to key and value of a Map

Jarek Rozanski

@jrozanski

Apologies for clunky Matrix bot earlier. In essence I am trying to apply Refined to a generic Map, and defining refined type to Keys and Values. Imagine something along the lines of:

type ValidMap = Map[String Refined NonEmpty, Long Refined Closed[W.`0`.T, W.`100`,T]

Jens Grassel

@jan0sch

@jrozanski It should work, I've something like this:

type FormField = String Refined (Trimmed And NonEmpty)
type ErrorMessage = String Refined NonEmpty
type FormErrors = Map[FormField, List[ErrorMessage]]

However you'll need to construct our instances via Map or write an object ValidMap with the appropriate helper functions.

David Strawn

@isomarcte

Is there current work underway for supporting something analogous to RefineMacro in Scala 3.x.x?

fthomas

@fthomas:matrix.org

[m]

@isomarcte: I'm not working on a RefineMacro equivalent for Scala 3 right now. My current understanding is that it is not possible to have something similar in Scala 3. But I've not played a lot with Scala 3 yet, so maybe there is a way to have this functionality and I'm not aware of it.

Denis Novac

@DenisNovac

Hi everyone. I am trying to make this string restrictions:

prefix: String Refined PrefixConstraints
type PrefixConstraints = NonEmpty And Not[Contains['.']] And Forall[LowerCase] And Head[Letter]

But i am getting errors with strings like "t-m" because of dash:

[error] java.lang.Error: ConvertFailure(CannotConvert("t-m",eu.timepit.refined.api.Refined[String,eu.timepit.refined.boolean.And[eu.timepit.refined.boolean.And[eu.timepit.refined.boolean.And[eu.timepit.refined.boolean.Not[eu.timepit.refined.collection.Empty],eu.timepit.refined.boolean.Not[eu.timepit.refined.boolean.Not[eu.timepit.refined.collection.Forall[eu.timepit.refined.boolean.Not[eu.timepit.refined.generic.Equal[Char('.')]]]]]],eu.timepit.refined.collection.Forall[eu.timepit.refined.char.LowerCase]],eu.timepit.refined.collection.Head[eu.timepit.refined.char.Letter]]],Left predicate of (((!isEmpty(t-m) && !!(!(t == .) && !(- == .) && !(m == .))) && (isLower('t') && isLower('-') && isLower('m'))) && isLetter('t')) failed: Right predicate of ((!isEmpty(t-m) && !!(!(t == .) && !(- == .) && !(m == .))) && (isLower('t') && isLower('-') && isLower('m'))) failed: Predicate failed: (isLower('t') && isLower('-') && isLower('m')).),

This is happens because of Forall[LowerCase]. It seems that it is also restrict my string to only chars. Is there a workaround for this? I want to have numbers or dashes in my string with only lowercased chars.

Zhenhao Li

@Zhen-hao

hi, is there a version of applyRef that gives a ValidatedNel instead of Either

Zhenhao Li

@Zhen-hao

so for type A = String Refined P1 AND P2 applyRef will give Invalid(NonEmptyList(errorOnP1, errorOnP2)) instead of Left(combinedError)

Damien Favre

@benderpremier

Hi, I was wondering if someone had already developed something around typeclass derivation for refined types. e.g. given I have an Ordering[Int] in scope derive automatically an Ordering[Int Refined Positive] ?

Frank Thomas

@fthomas

@benderpremier Yes, see https://github.com/fthomas/refined/blob/master/modules/cats/shared/src/main/scala/eu/timepit/refined/cats/derivation.scala for the derivation mechanism and https://github.com/fthomas/refined/blob/7df4ff623dec9609848fd75d605ca4a8b9f6733f/modules/cats/shared/src/main/scala/eu/timepit/refined/cats/package.scala#L12-L31 for examples

1 reply

Frank Thomas

@fthomas

@Zhen-hao internally refined creates a Result[And[Result[P1], Result[P2]]] before this whole Result is translated to a String but there is no API that exposes the Result. A Result is either Passed or Failed and would allow in your example to inspect which of the predicates P1 and P2 failed.

Zhenhao Li

@Zhen-hao

thanks! I will give it a try when I get more time

Damien Favre

@benderpremier

Say I have positive integers on which I map _ % 3. Is there an alternative to using refineV[Interval.Closed[0,2]](x) that doesn’t wraps the result in an Either (Knowing that since I do a modulo, all possible positive integers will satisfy the predicate ? Should I do something with singleton-ops ?

Damien Favre

@benderpremier

for the moment I ended up using unsafeFrom(_ % 3) but I assume kind of an anti pattern

Piyush Purang

@ppurang

A very general question here - does any one know the limitations of refined under scala 3? Seems like val u : NonEmptyString = "u" doesn't work as intended and in this case what would be an okish work around?

[error] 85 |  val prefix: NonEmptyString = "u"
[error]    |                               ^^^
[error]    |    Found:    ("u" : String)
[error]    |    Required: eu.timepit.refined.types.all.NonEmptyString
[error]    |
[error]    |    The following import might make progress towards fixing the problem:
[error]    |
[error]    |      import eu.timepit.refined.auto.autoUnwrap

^that import is enabled

Dmytro Mitin

@DmytroMitin

@ppurang
https://scalacenter.github.io/scala-3-migration-guide/docs/macros/macro-libraries.html
alt

And implicit conversions in eu.timepit.refined.auto are macro-based.

Dominic Egger

@GrafBlutwurst

Hey. I was wondering if it's possible to summon a runtime representation of a Predicate namely for logging puproses, I can't log the values because of sensitivity concerns but want to log the predicate it fails. But I can't find anything in the code

Dominic Egger

@GrafBlutwurst

Right now I just have my own encoding using

refineDescribed[V, P](implicit validate:Validate[V,P], description: DescribeRefinement[P]): Either[RefinementError[P], V Refined P]

but I want to avoid reinventing the wheel

Brian P. Holt

@bpholt

I was surprised to learn today that

val x: PosInt = 5 << 20

compiles fine during the compile sbt task, but when compile:doc runs, it blows up:

[error] /…code.scala: compile-time refinement only works with literals
[error]   private val x: PosInt = 5 << 20
[error]                             ^
[info] Int(5242880) <: eu.timepit.refined.types.all.PosInt?
[info] false
[info] No documentation generated with unsuccessful compiler run

I'm curious why it's inconsistent. What is doc doing differently that would cause this? (Scala 2.12.12, if it matters)

Brian P. Holt

@bpholt

Is it possible to use capture groups in a MatchesRegex predicate, and then extract the captured value(s) somehow?

Aleksei Shamenev

@coffius

Hi there 👋
I wonder if it is possible to validate a size of a string in compile time?

In my case I'd like to check currency codes and make sure that all code's values are strings with 3 chars in length.
I have defined the corresponding type alias as type _Code = String Refined Size[W.3.T]
But when I try to use it like this

private val test: _Code = "123"

I get this compilation error

type mismatch;
 found   : String("123")
 required: _Code

And it is unclear for me either it is impossible to validate this rule during compilation or I am missing some imports.

Aleksei Shamenev

@coffius

UPD: found the answer here: https://gitter.im/fthomas/refined/archives/2016/09/13?at=57d838a7b8d4c138408de7bc

Alan Devlin

@spf3000

Hi everyone? Wondering if there is an idiomatic solution for the double definition error with refined types as parameters?
` ``
def getTrees(a: TreeID): List[Tree]
def getTrees(a: PlantID): List[Tree]
====>
double definition:
...
have same type after erasure: (a: String):
```

Oron Port

@soronpo

@spf3000 have you tried adding a dummy implicit argument?

def getTrees(a: PlantID)(implicit di : DummyImplicit): List[Tree]

Alan Devlin

@spf3000

@soronpo I didn't know about that trick - it's very neat - thanks :)

Jakub Kozłowski

@kubukoz:matrix.org

[m]

hey, are there plans to support refined macros in dotty?

Frank Thomas

@fthomas

@kubukoz:matrix.org I'm not sure yet if they can be ported to Scala 3. The Scala 2 macros use eval which is not available in Scala 3. Maybe if all Validate instances are declared as inline, it would be possible to do the validations at compile-time

Jakub Kozłowski

@kubukoz:matrix.org

[m]

I see

Frank Thomas

@fthomas

I experimented a little bit with implementing the refine macro in Scala 3 here: fthomas/refined#921 . It seems to be possible for predicates that are implemented in terms of functions like >, !, ==, etc. that can be evaluated at compile-time but it would break down for more complex predicates like MatchesRegex, Size, etc.

Oron Port

@soronpo

@fthomas did you try using compiletime.ops?

And I also think there is more to add there by PR-ing to dotty with what is required.

Both for refined and singleton-ops.

I haven't had a chance to look at this yet, but once I finish my Phd, I will get on that.

Hopefully during May

Frank Thomas

@fthomas

For that experiment I didn't looked at compiletime.ops since I was only looking at what functions can be called at compile-time on inline values and not types.

Oron Port

@soronpo

But inline values constraint can rely on type constraints

Endre Galaczi

@chiller

Hello, I have a question about refined-scalacheck . I am trying to produce an Arbitrary instance for String Refined Size[Less[20]] but it will not always succeed. I am puzzled by this because it can make one for Int Refined Less[20] so supposedly collection.buildableSizeArbitrary could use that instance to generate valid collection sizes, but I often get negative size values. Am I using the wrong implicits, or is this a bug in the library?

import eu.timepit.refined.api.Refined
import eu.timepit.refined.collection.Size
import eu.timepit.refined.numeric._
import eu.timepit.refined.scalacheck.all.lessArbitrary
import eu.timepit.refined.scalacheck.string.stringSizeArbitrary
import org.scalacheck.Gen

object Example extends App {

  type T1 = String Refined Size[Less[5]]

  case class CC(value: T1)

  println(implicitly(Gen.resultOf(CC.apply _)).sample)
}

Frank Thomas

@fthomas

@chiller If the Arbitrary for Int Refined Less[5] generates a negative value, Gen.buildableOfN cannot generate a value. Try using a predicate that excludes negative values, e.g. String Refined Size[Interval.Closed[0, 5]].

Endre Galaczi

@chiller

Thank you

Pelle Krøgholt

@pellekrogholt

Hi,

I'm using refined to refine a float value that I'm getting from db. I have a similar usage for int that works (tm)... But struggling with Integral[Float] implicit not in scope.

Context

type ValidValue           = Float Refined Divisible[0.25]

case class MyType(value: ValidValue)

  implicit val myTypeGet: Get[MyType] = Get[Float].temap { raw =>
    refineV[Divisible[0.25]](raw).map(MyType(_))
  }

raises error on compile

could not find implicit value for parameter v: eu.timepit.refined.api.Validate[Float, Divisible[0.25]]
    refineV[Divisible[0.25]](raw).map(MyType(_))

any pointers ?

Frank Thomas

@fthomas

Try using a Float literal: 0.25F

Pelle Krøgholt

@pellekrogholt

Thx for your pointer - I tried with ^^ same error :thought_balloon:

Where communities thrive

People

Repo info

Activity