The current package requires a try-catch block in every mapper, combiner and reducer to handle those exceptions. I've seen them get thrown in the mapper by the occasional bad input data records - since they are coming from outside our world - and just ignore them so as not to abort the whole job.
As I noted in MAHOUT-6 earlier, I made the exceptions checked so that users would need to think about this use case and decide what to do. I agree with Ted that there is little in terms of corrective action that is possible, except to ignore the records. His patch only changes the superclass of the exceptions and leaves them in the interfaces and implementations. This retains their documentation value but removes the requirement to explicitly handle them. I think there is some value in requiring users to handle them explicitly, but can see both arguments. I do not feel strongly about this one. +-0 Jeff Jeff Eastman, Ph.D. Windward Solutions Inc. +1.415.298.0023 http://windwardsolutions.com http://jeffeastman.blogspot.com > -----Original Message----- > From: Karl Wettin [mailto:[EMAIL PROTECTED] > Sent: Monday, March 31, 2008 6:51 AM > To: mahout-dev@lucene.apache.org > Subject: Re: [jira] Updated: (MAHOUT-22) Several matrix exceptions are > checked exceptions, but should be unchecked > > Ted Dunning (JIRA) skrev: > > [ https://issues.apache.org/jira/browse/MAHOUT- > 22?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] > > > > Ted Dunning updated MAHOUT-22: > > ------------------------------ > > > > Attachment: MAHOUT-22.patch > > I don't get contact with the JIRA via SSL right now. > > If nobody objects or beats me (calling Jeff, hello?) to it I'll review > and probably commit this in a few days. > > > karl
