We did, but since we are emulating the API we found it too difficult to
support both apis in the same source tree. Still, the way to do this, is:
https://git-wip-us.apache.org/repos/asf?p=mrunit.git;a=blob;f=pom.xml;h=0ac909b84869e22ffe5a8f340fc7a7960555f230;hb=mrunit-0.9#l225
On Wed, Feb 13, 2013 at 12:32 PM, Patrick Hunt <phunt@apache.org> wrote:
> I'm not very familiar with the context here, but you might look at
> what's been done in mrunit to address this. They support both hadoop 1
> & 2 and use Maven for builds.
>
> Patrick
>
> On Wed, Feb 13, 2013 at 10:20 AM, Matthias Friedrich <matt@mafr.de> wrote:
> > Hi,
> >
> > On Wednesday, 2013-02-13, Josh Wills wrote:
> >> Unfortunately, I'm not totally in a position to do it myself-- I'm
> pretty
> >> sure I would need some help from Matthias, who's tied up w/work
> commitments
> >> right now. There's some complexity around the creation of the Crunch
> >> Archetype under Hadoop-2 that I don't really grok and would make me
> unsure
> >> about the stability of a Hadoop-2 binary release under our current
> release
> >> process.
> >
> > The Hadoop 1/2 switch in Maven is such a pain in the neck that I vowed
> > to stay away from that for a while. The tricky thing is to get correct
> > dependencies for the project generated by the archetype (dependencies
> > are quite different between Hadoop 1 and 2).
> >
> > Also, validating a release is a lot of work already, I'd rather not
> > add even more artifacts to check.
> >
> > If anyone really wants this, I'd review patches to make sure it works.
> > But unless you're a real wizzard, I predict pain and frustration for
> > everyone who tries ;-)
> >
> > Regards,
> > Matthias
>
--
Apache MRUnit - Unit testing MapReduce - http://incubator.apache.org/mrunit/
|