ScalaTest Scalactic-自定义带公差的双等号，包括Double.NaN大小写

我正在尝试创建一个自定义匹配器，该匹配器将考虑Double.NaN并对非NaN值使用容差

import org.scalactic.{Equality, TolerantNumerics}
import org.scalatest.Matchers

trait MoreMatchers extends Matchers {

  implicit def doubleEqWithNaNAndTol: Equality[Double] = new Equality[Double] {

    implicit val tolerance: Equality[Double] = TolerantNumerics.tolerantDoubleEquality(1e-6)

    def areEqual(expected: Double, actual: Any): Boolean = actual match {
      case number: Double if number.isNaN => expected.isNaN
      case _: Double => actual === expected
      case _ => false
    }

  }

不幸的是，它不起作用

assert(0.00226685508536916 === 0.0022668550853691587)  // failure - doesn't use tolerance
assert (Double.NaN === Double.NaN )  // success

如果在断言中输入容差，则如果存在NaN，则将失败

assert(0.00226685508536916 === 0.0022668550853691587 +- 1e-6)  // success
assert (Double.NaN === Double.NaN  +- 1e-6)  // failure - can't use tolerance when NaN

如果我只是像下面这样称呼它，那么它是有效的

implicit val tolerance: Equality[Double] = TolerantNumerics.tolerantDoubleEquality(1e-6)

def eq(expected: Double, actual: Double): Boolean = expected match {
    case x if x.isNaN => actual.isNaN
    case _ => actual === expected
}

然后称之为：

assert(eq(...,...))

我想知道是否有可能用第一种方法让它工作。你以前见过这种情况吗？你能提出一些解决办法吗？任何帮助都将不胜感激：）

谢谢，

---

ele

非常感谢eirikr@d6:），请参见中的解决方案

基本上，在上面的代码中，在比较非nan双精度时，您需要使用

公差.areEqual（预期值，数字）

，以便能够使用该比较中隐含的公差

import org.scalactic.{Equality, TolerantNumerics}
import org.scalatest.Matchers

trait MoreMatchers extends Matchers {

  implicit def doubleEqWithNaNAndTol: Equality[Double] = new Equality[Double] {

    implicit val tolerance: Equality[Double] = TolerantNumerics.tolerantDoubleEquality(1e-6)

    def areEqual(expected: Double, actual: Any): Boolean = actual match {
      case x: Double if x.isNaN => expected.isNaN
      case x: Double => tolerance.areEqual(expected, x)
      case _ => false
    }

  }

问候,， ele