You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mahout.apache.org by Jake Mannix <ja...@gmail.com> on 2010/03/05 20:20:19 UTC

0.3 Patches

Hey all,

  Our "flash-freeze" has unthawed considerably, but I've been trying to be
good and not check in stuff with functionality improvements I've been
wanting to check in.

What do you folks say about me checking in fixes for

MAHOUT-312 </jira/browse/MAHOUT-312> DistributedRowMatrix iterateAll() and
iterate() don't work on multi-part SequenceFiles </jira/browse/MAHOUT-312>

MAHOUT-313 </jira/browse/MAHOUT-313> DistributedRowMatrix needs
times(Vector) implementation as M/R job </jira/browse/MAHOUT-313>

which together basically solve

MAHOUT-310 </jira/browse/MAHOUT-310> LanczosSolver and
DistributedLanczosSolver always assume rectangular input, but should also
handle symmetric eigensystems. </jira/browse/MAHOUT-310>

and a nice freebie I've got a patch for is

 MAHOUT-314 </jira/browse/MAHOUT-314>DistributedRowMatrix needs a sparse
DistributedRowMatrix times(DistributedRowMatrix other)
implementation</jira/browse/MAHOUT-314>

Unit tests are included, and all current tests continue to pass.  These
changes make DistributedRowMatrix a much-more fully-featured Matrix
implementation, and will allow for our users to come up with their own ideas
for uses of a nice big HDFS-backed sparse matrix, instead of simply using it
for SVD (which is basically all it's useful for without this patch).

Can I get these in for 0.3?  I could commit today if it's ok with the team.

  -jake

Re: 0.3 Patches

Posted by Jake Mannix <ja...@gmail.com>.
Ha!  Perpetual code freeze to get new features, now there's a concept!

Three +1's, ok if I don't get any negative feedback before I get back to a
computer, I'll check in.

I've just added more wiki pages for me to write, too I guess...

  -jake

On Mar 5, 2010 2:53 PM, "Jeff Eastman" <jd...@windwardsolutions.com> wrote:

Robin Anil wrote: > > Seems we are most productive when its a code freeze :)
+1 from me as well. > Y...
Maybe we should just declare a perpetual code freeze <grin>. But you guys
are really on a roll so I'm +1 too
Jeff

Re: 0.3 Patches

Posted by Jeff Eastman <jd...@windwardsolutions.com>.
Robin Anil wrote:
> Seems we are most productive when its a code freeze :) +1 from me as well.
> You have time till Hadoop resolves 6617, assuming nothing gets broken
>   
Maybe we should just declare a perpetual code freeze <grin>. But you 
guys are really on a roll so I'm +1 too
Jeff

Re: 0.3 Patches

Posted by Robin Anil <ro...@gmail.com>.
Seems we are most productive when its a code freeze :) +1 from me as well.
You have time till Hadoop resolves 6617, assuming nothing gets broken

Robin

On Sat, Mar 6, 2010 at 12:50 AM, Jake Mannix <ja...@gmail.com> wrote:

> Hey all,
>
>  Our "flash-freeze" has unthawed considerably, but I've been trying to be
> good and not check in stuff with functionality improvements I've been
> wanting to check in.
>
> What do you folks say about me checking in fixes for
>
> MAHOUT-312 </jira/browse/MAHOUT-312> DistributedRowMatrix iterateAll() and
> iterate() don't work on multi-part SequenceFiles </jira/browse/MAHOUT-312>
>
> MAHOUT-313 </jira/browse/MAHOUT-313> DistributedRowMatrix needs
> times(Vector) implementation as M/R job </jira/browse/MAHOUT-313>
>
> which together basically solve
>
> MAHOUT-310 </jira/browse/MAHOUT-310> LanczosSolver and
> DistributedLanczosSolver always assume rectangular input, but should also
> handle symmetric eigensystems. </jira/browse/MAHOUT-310>
>
> and a nice freebie I've got a patch for is
>
>  MAHOUT-314 </jira/browse/MAHOUT-314>DistributedRowMatrix needs a sparse
> DistributedRowMatrix times(DistributedRowMatrix other)
> implementation</jira/browse/MAHOUT-314>
>
> Unit tests are included, and all current tests continue to pass.  These
> changes make DistributedRowMatrix a much-more fully-featured Matrix
> implementation, and will allow for our users to come up with their own
> ideas
> for uses of a nice big HDFS-backed sparse matrix, instead of simply using
> it
> for SVD (which is basically all it's useful for without this patch).
>
> Can I get these in for 0.3?  I could commit today if it's ok with the team.
>
>  -jake
>

Re: 0.3 Patches

Posted by Ted Dunning <te...@gmail.com>.
Tentative +1 from me.

On Fri, Mar 5, 2010 at 11:20 AM, Jake Mannix <ja...@gmail.com> wrote:

> Can I get these in for 0.3?  I could commit today if it's ok with the team.




-- 
Ted Dunning, CTO
DeepDyve