Re: What's New in Scala 2.8: Collections API
Posted: Aug 28, 2010 4:36 AM
> From a quick first look I had, it seems the new
> collections API offers a lot of collection types, I'd dare
> say almost all widely known collection types.
> Without being a Scala user and only knowing the very
> basics of the language, I'd like to ask:
> 1) why the emphasis on immutable data structures? what are
> the benefits? aren't those slower than the mutable types?
The library is symmetric wrt mutability. Every major type has mutable as well as immutable implementations. Immutable ones are available directly whereas mutable ones need an explicit qualification or import.
The reason for preferring immutable is that immutable is safer. So in the spirit of avoiding premature optimizations
we made immutable the default.
Immutable collections do have persistent implementations which keep the cost overhead fairly low, and which even give an advantage over mutable for small collections (an immutable set or map of up to 4 elements takes only a single object; compare to a mutable map backed by a hashtable).
> 2) when a collection's type is a primitive, does
> autoboxing happen? for example, an Array[int] is actually
> an Array[Integer]?
For general collections, yes. But for arrays, no. Scala arrays are represented exactly the same as Java arrays. There's a section on arrays in the collections document.
The next versions of Scala will specialize further collection types iso that primitive types need not be boxed. That's an ongoing process which was started with the availability of @specialized annotation in Scala 2.8.