RFR: 8221473: Configuration::reads can use Set.copyOf

Peter Levart peter.levart at gmail.com
Wed Mar 27 11:25:04 UTC 2019


Hi,

On 3/26/19 7:44 PM, Daniel Fuchs wrote:
> Hi Peter,
>
> On 26/03/2019 18:01, Peter Levart wrote:
>> Would such method addition be worth it?
>
> Here is the impl of Set.copyOf:
>
>     static <E> Set<E> copyOf(Collection<? extends E> coll) {
>         if (coll instanceof ImmutableCollections.AbstractImmutableSet) {
>             return (Set<E>)coll;
>         } else {
>             return (Set<E>)Set.of(new HashSet<>(coll).toArray());
>         }
>     }
>

Yes, and that method is suitable for situations where there is a chance 
of one or the other case and the user is aware of performance 
implications of either case. There are other situations where the code 
guarantees that the Set at hand is already immutable and programmer 
might just want to verify that as a post-condition before returning it. 
In that case the non-immutability would be considered an error. I'm 
thinking of the following implementation:

     static <E> Set<E> requireImmutable(Set<E> set) {
         if (set instanceof ImmutableCollections.AbstractImmutableSet) {
             return set;
         } else {
             throw new IllegalStateException("Not an immutable Set");
         }
     }

Currently the immutable implementation classes are package-private so 
there's no way for 3rd party code (short of introspecting the class 
hierarchy of the implementation class against the FQN of the 
AbstractImmutableSet class, which might be costly and fragile) to 
implement a helper method like this.

Regards, Peter

> Best regards,
>
> -- daniel



More information about the core-libs-dev mailing list