11

In Scala (2.9.2) I am trying to create a function that provides some other value when accessed

() => Any

I thought I was successfully doing this via a by name / currying solution:

def byName(theValue : => Any)() : Any = theValue
val myHolder = byName(myValue)_

So my holder is of the correct type. However I find that in the course of creating this curried function the by name parameter is evaluated. If I do the following instead it works as intended:

def byName(theValue : => Any) : Any = () => theValue
val myHolder = byName(myValue)

I conclude some part of the currying process is referencing the first parameter list and triggering its evaluation. Can anybody confirm that and/or explain why or if there are any guidelines around using by name parameters with multiple parameter lists?

Improve this question

2 Answers 2

Reset to default
3

This seems like a bug to me. I could trigger this strange behaviour in scala 2.9.1, but not in scala 2.10 RC1 so I guess this was fixed at some point.

In scala 2.9.1:

Welcome to Scala version 2.9.1.final (Java HotSpot(TM) Client VM, Java 1.6.0_27).
Type in expressions to have them evaluated.
Type :help for more information.
scala> def byName(theValue : => Any)() : Any = theValue
byName: (theValue: => Any)()Any
scala> def myValue: String = { println("Computing myValue"); "hello" }
myValue: String
scala> val myHolder = byName(myValue)_
Computing myValue
myHolder: () => Any = <function0>
scala> myHolder()
res0: Any = hello

In scala 2.10-RC1:

Welcome to Scala version 2.10.0-RC1 (Java HotSpot(TM) Client VM, Java 1.6.0_27).
Type in expressions to have them evaluated.
Type :help for more information.

scala> def byName(theValue : => Any)() : Any = theValue
byName: (theValue: => Any)()Any
scala> def myValue: String = { println("Computing myValue"); "hello" }
myValue: String
scala> val myHolder = byName(myValue)_
myHolder: () => Any = <function0>
scala> myHolder()
Computing myValue
res0: Any = hello
Improve this answer
Sign up to request clarification or add additional context in comments.

1 Comment

Well spotted. From the comments it seems that it was fixed in scala 2.10-M4.
2

The issues are https://issues.scala-lang.org/browse/SI-302 and https://issues.scala-lang.org/browse/SI-5610

Personally, I find what is now the "old" behavior more intuitive: Partial application means something is applied.

Or as Odersky had it:

No, that's how eta expansion is defined. You don't just wrap a lambda around the expression, you evaluate what you can first.

But conversely, it's easier to force evaluation than to devise a syntax for suspending it.

Improve this answer

3 Comments

Yes, but what does "apply" mean in the context of a by name parameter? What if you were to pass a Function0 as a parameter, would "apply" mean that the function has to be eagerly evaluated? The natural answer is no, and I would think that it is sensible to have the same conclusion for a by name parameter. In short, I personnally tend to agree with the counter arguments of Matt Hellige: issues.scala-lang.org/browse/…
The spec on by-name calls it a "use." To me, f(myValue) looks like a use of myValue, whatever the result of the expr. If f is overloaded as f(v:V) and f(v:=>V), you can't tell by looking that a by-name myValue is evaluated and passed to the first f. (But I'm not virulently opposed to the change in semantics; but to me, the old spec was not as crazy and counter-intuitive as other puzzlers.)
Hmm - both sides make some sense. Thanks for your comments.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.