On 26 November 2013 17:35, Martin Buchholz <marti...@google.com> wrote: > I haven't looked in depth, but I agree with Stephen's analysis. This API > and its javadoc needs work. > E.g. It's not clear that the purpose of Map.compute is to *update* the > mapping for key in the map.
I actually felt that the names of all four methods felt wrong. compute and merge seem like unfortunate choices. > Instead of "The default implementation makes no guarantees about > synchronization or atomicity properties of this method." we should boldly > say that the default implementation is *not* atomic, even if the underlying > map is. Saying that the default implementation is not atomic sounds uncontroversial. Stephen
