Package Private Record Fields

[Ty Young]

On 5/3/20 3:56 PM, Brian Goetz wrote:
> Making the raw version available by default within the package would 
> be a sharp edge on which users would cut themselves over and over.
>

How is this any different then default? One way or the other, unless you 
override the default accessor method, you're getting an object that, 
while the reference is final, the data it contains might not be. Lists, 
Map, Optional, etc all have the same issues. This is something that can 
and probably will be overlooked... although IDEs can probably fix this.


Is it expected that everytime such an object is a field of a record that 
you need to override the default accessor in order to defensive copy? 
Doesn't this violate the "transparent holders" part of records, and 
therefore mean that you are trying to use records for something it isn't 
designed for?


and for that very same lack of transparency, wouldn't the use of 
interfaces, a public one and a private(not exposed, extends the public 
one), be better here anyway? For example:


public interface FooPublicInterface

{

     public int[] clonedInts();

}


and


public interface FooPrivateInterface extends FooPublicInterface

{

     public int[] ints();

}