On Wed, 5 Apr 2023 09:19:57 GMT, Viktor Klang <d...@openjdk.org> wrote:
> Adds overrides for common Map operations to avoid having to allocate the
> entrySet for Collectors$Partition maps.
Another thing I noticed is that `keySet()` of this map is a constant. If we
want to preserve the order, we can use something like this:
@Override
public Set<Boolean> keySet() {
return KeySet.INSTANCE;
}
private static class KeySet extends AbstractSet<Boolean> {
private static final KeySet INSTANCE = new KeySet();
public Iterator<Boolean> iterator() {
return List.of(false, true).iterator();
}
public int size() {
return 2;
}
@Override
public int hashCode() {
// Boolean.hashCode(true) + Boolean.hashCode(false)
return 1237 + 1231;
}
@Override
public boolean equals(Object o) {
return o instanceof Set<?> s &&
s.size() == 2 &&
s.contains(false) && s.contains(true);
}
public boolean isEmpty() {
return false;
}
public boolean contains(Object k) {
return k instanceof Boolean;
}
@Override
public void forEach(Consumer<? super Boolean> action) {
Objects.requireNonNull(action);
action.accept(false);
action.accept(true);
}
@Override
public String toString() {
return "[false, true]";
}
}
Overengineering, probably? On the other hand, it might be useful to expose this
set somewhere, like `Collections.booleans()`? This would allow using
`for(boolean b: Collections.booleans()) {...}`. In our codebase, I see quite
many loops like `for (boolean b : new boolean[]{false, true})`. Declaring it as
a `SequencedSet` would be even more useful, as one may control the order via
`Collections.booleans().reversed()`.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/13347#issuecomment-1497432746
