knownSize should return Option[Int]

[phrmoy]

The following code prints -1

@main def runPlay = {
  println(List(1,2,3).knownSize)
}

I was reasonably surprised by this behavior and though this may not (or maybe) qualify as a type safety issue, I walked away with the sense of lack of type safety.

Rationale/Ask: If the length cannot be produced in constant time, then do not return an incorrect value, instead, return None since there is nothing to be returned (the value does not exist in constant time effort). If there is some value to be returned, then return a value of type Some[Int].

I searched for knownSize and did not find anything related to it.

Thank you.

Pedro

[nicolasstucki]

It makes sense to define knownSize as an Int for performance reasons. This avoids the creation of new objects for some core operations and may be helpful for the JIT compiler optimizations.

Just like we have head and headOpt, we could consider having a konwnSizeOpt in the library. Scala 3 uses the Scala 2 library for compatibility. This would need to be added to the 2.13.12+ Scala library in https://github.com/scala/scala.

[phrmoy]

Thank you for your reply. I understand the performance concern. Now with that said, here are some thoughts:

1] I do like the knownSizeOpt idea as hopefully for a newcomer like me to Scala, I will see both options with the assistance of an IDE and might guess that knownSize may return something unexpected since there is a knownSizeOpt.

2] The comparison between head/headOpt and knownSize/knownSizeOpt is not quite the same because head/headOpt both have "correct" behavior. That is, head throws an exception on an empty sequence.

3] Having knownSizeOpt is better than not at all even if knownSize may return a -1. I would prefer if knownSize also threw an exception like head does for consistency and accuracy in behavior (but then this may implicate your performance issue again?). If I had the power, I would then create a _knownSize or unsafeKnownSize (or anything else to alert the user, specially noobies like me) and allow the inconsistent behavior in it.

Just my thoughts. Thank you.

Pedro

[som-snytt]

knownSize is not a noob-friendly feature. It's a subtle and suitably obscure mechanism to avoid forcing lazy collections. It's performance-sensitive because used for other ops.

The doc could suggest the quaint sizeIs DSL for beginners, or perhaps sizeCompare, for times when size or length is inadvisable.

Compare the huge comment on (deprecated) hasDefiniteSize.