On 9/18/17 2:12 PM, Mike Miller wrote:
Recently tickets have been opened dealing with Guava in Accumulo (see ACCUMULO-4701 through 4704), in particular the use of Beta classes and methods. Use of Guava comes with a few warnings... From the Guava README: *1. APIs marked with the @Beta annotation at the class or method level are subject to change. They can be modified in any way, or even removed, at any time. If your code is a library itself (i.e. it is used on the CLASSPATH of users outside your own control), you should not use beta APIs, unless you repackage them (e.g. using ProGuard).2.Deprecated non-beta APIs will be removed two years after the release in which they are first deprecated. You must fix your references before this time. If you don't, any manner of breakage could result (you are not guaranteed a compilation error).* I think it is worth a discussion on how to handle Guava dependencies going forward across the different versions of Accumulo. The goal would be to allow use of a newer version version of Guava in client applications with the current supported versions of Accumulo. Ideally, we could just eliminate any use of Beta Guava code. But there are Beta classes that are very useful and some which we already have integrated into released Accumulo versions. There seem to be 3 ways to handle Guava dependencies: 1 - jar shading
+1 (favoring #3 too, when not intrusive). We stop "advertising" that we include Guava on the classpath and it's no longer our problem. As the other part of the thread alludes, if Hadoop brings in a version, fine. Accumulo specifically should stop trying to rely on something specific coming down from its dependencies and "control its own destiny".
FWIW, HBase has been actively moving to this model and, IMO, it's been working well.
2 - copy Guava code into Accumulo 3 - replace Guava code with standard Java We may have to handle it differently with each version of Accumulo. For example, 1.8 has more widespread use of Beta annotated code than 1.7.
