Generating arbitrary (legal) Unicode character with scalacheck?

4
votes

I'm trying to create a generator that produces (non-zero-length) legal unicode strings, with scalacheck 1.6.6 and specs 1.7 (scala 2.8.1).

I hoped I could just create generators like:

object Generators {
  def unicodeChar: Gen[Char] = 
    choose(Math.MIN_CHAR, Math.MAX_CHAR).map(_.toChar).filter(
      c => Character.isDefined(c))
  def unicodeStr: Gen[String] = for(cs <- listOf1(unicodeChar)) yield cs.mkString
}

...then use them from specs like:

import org.specs.Specification
import org.specs.matcher.ScalaCheckMatchers

object CoreSpec extends Specification with ScalaCheckMatchers {        
  "The core" should {    
    "pass trivially" in {
      Generators.unicodeStr must pass((s: String) => s == s)
    }
  }
}

But, it seems that using filter in unicodeChar causes a problem:

Specification "CoreSpec"
  The core should
  x pass trivially
    Gave up after only 64 passed tests. 500 tests were discarded.

If I remove the filter from unicodeChar, my test passes, but I run into other problems later, since my strings are not always well-defined unicode.

Thanks in advance for any suggestions on how to achieve this.

4
scalaspecsscalacheck
It seems a bit easier if I don't use Specs, and provide a new implicit Params instance. But that only fixes the symptom, it doesn't really solve the underlying cause. – Eric Bowman - abstracto -

4 Answers

9
votes

Try filtering out the characters before creating the generator:

val unicodeChar: Gen[Char] = Gen.oneOf((Math.MIN_CHAR to Math.MAX_CHAR).filter(Character.isDefined(_)))

It will be more memory intensive, since the complete list of unicode characters will be allocated when the generator is created, but only one instance of that list will be used so it shouldn't be a big problem.

5
votes

I don't know how it was in 2010, but nowadays you can use Arbitrary: 

  import org.scalacheck.Arbitrary
  import org.scalacheck.Gen

  val unicodeChar: Gen[Char] = Arbitrary.arbChar.arbitrary
  val unicodeString: Gen[String] = Arbitrary.arbString.arbitrary
3
votes

Ok, I figured it out. This is what works for me:

def unicodeChar = Gen((p: Gen.Params) => {
    var c = 0
    do {
      c = util.Random.nextInt(0xFFFF)
    } while (!Character.isDefined(c))
    Some(c.toChar)
  })

Pretty simple, actually. What I didn't get is that you can create an arbitrary generator of type T by passing a function Gen.Params => T to Gen.apply().

0
votes

Have you tried suchThat instead of filter?