Github user aweisberg commented on a diff in the pull request:
https://github.com/apache/cassandra/pull/224#discussion_r195811425
--- Diff: src/java/org/apache/cassandra/locator/Replicas.java ---
@@ -50,6 +50,30 @@
public abstract int size();
protected abstract Collection<Replica> getUnmodifiableCollection();
+
+ public boolean equals(Object o)
--- End diff --
So I want to try and be clearer again :-P
ReplicaList and ReplicaSet both work fine the way they are today. What
doesn't work is Replicas functions that return an ImmutableContainer because
those don't implement equals or hashCode for their contents. We should never
return a Replicas that isn't either an instance of either ReplicaSet or
ReplicaList with the implied equality and hashCode behavior.
Java has the Collections class with singleton (which is singleton set)
singletonList, singletonMap etc. Guava has ImmutableXYZ and the .of methods.
I think Guava got it right having those methods be in the type specific
class rather than just a generic Collections class.
Otherwise it's all good. ReplicaSet and ReplicaList are wrappers for actual
set and list implementations so it's not clear we would ever have more than
just that. I think optimized singleton implementations should be a later
optimization unless you feel driven.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
