> On March 19, 2014, 12:52 p.m., Mike Drob wrote: > > core/src/main/java/org/apache/accumulo/core/client/impl/InstanceFactory.java, > > line 37 > > <https://reviews.apache.org/r/19409/diff/1/?file=528072#file528072line37> > > > > Is this exception class used? If not, it should be removed. > > Bill Havanki wrote: > It's not, but I wanted to leave a way for future factory implementations > to predictably throw exceptions, at the factory level of abstraction > (Effective Java 2nd ed. Item 61). > > John Vines wrote: > I like that. Could we add a basic coding to the exception type to make > various error cases easier to add. > > And is there a particular reason you're making that exception type extend > Runtime instead of just Exception? > > Sean Busbey wrote: > I presume it's because if it was not a RuntimeException it would have to > go in the method signature. Since the class isn't public, it would have to > throw Exception, which is not good. > > Bill Havanki wrote: > The InstanceFactoryException class is in fact public because all members > of interfaces are public (JLS 6.6.1, also I had to double-check with javap :) > ). I definitely intend for it to be public. > > I made it a RuntimeException because I judged that recovering from > failing to get an Instance would be unrecoverable. I could be wrong, and if > so, I'm happy to make it checked. > > John, is there any particular technique for adding a coding that you > like? Numeric code? Enum? I'll check around for any existing patterns for > this sort of thing. > > Mike Drob wrote: > Ah. Yes, you are correct, it is public. I'm just not used to seeing > members in interfaces. > > An Instance is just a container for holding a uuid (or name) and a list > of zookeepers, right? I'm not sure I see how that can fail. Unless there's > some sort of lookup that needs to happen that is susceptible to IOException? > Yea, that's probably unrecoverable then.
I figure having the exception available just covers our bases. We're also looking at a client API refactoring, so who knows what Instance will be after that? To satisfy John's wish for adding error codes, I was going to use the commons-lang ContextedRuntimeException, which lets you throw in key(String)/value(Object) pairs. Super keen. http://commons.apache.org/proper/commons-lang/javadocs/api-3.2.1/org/apache/commons/lang3/exception/ContextedRuntimeException.html It requires commons-lang 3, but that can co-exist with commons-lang 2.x, which we currently use, without conflict. - Bill ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/19409/#review37731 ----------------------------------------------------------- On March 19, 2014, 12:44 p.m., Bill Havanki wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/19409/ > ----------------------------------------------------------- > > (Updated March 19, 2014, 12:44 p.m.) > > > Review request for accumulo. > > > Bugs: ACCUMULO-2497 > https://issues.apache.org/jira/browse/ACCUMULO-2497 > > > Repository: accumulo > > > Description > ------- > > To help in testing, this commit adds an InstanceFactory interface with some > implementations. > > To avoid changing the public API, some classes are in o.a.a.core.client.impl. > > > Diffs > ----- > > > core/src/main/java/org/apache/accumulo/core/client/impl/InstanceFactory.java > PRE-CREATION > > core/src/main/java/org/apache/accumulo/core/client/impl/ZooKeeperInstanceFactory.java > PRE-CREATION > > core/src/test/java/org/apache/accumulo/core/client/impl/ZooKeeperInstanceFactoryTest.java > PRE-CREATION > > server/base/src/main/java/org/apache/accumulo/server/client/HdfsZooInstanceFactory.java > PRE-CREATION > > server/base/src/test/java/org/apache/accumulo/server/client/HdfsZooInstanceFactoryTest.java > PRE-CREATION > > Diff: https://reviews.apache.org/r/19409/diff/ > > > Testing > ------- > > Unit tests pass. > > > Thanks, > > Bill Havanki > >
