M


On Tue, Apr 1, 2014 at 4:13 PM, Ted Dunning <[email protected]> wrote:

> On Tue, Apr 1, 2014 at 11:32 AM, Dmitriy Lyubimov <[email protected]>
> wrote:
>
> > *Corollary to that are 2 arguments :*
> > *(1) doing what you say on AbstractMatrix or AbstractVector hierarchy
> > is not possible without a "nuclear option" on the api, which will send
> > a ripple effect inside and outside Mahout (my outside code in
> > particular too);*
> >
>
> I don't think that you have produced any evidence of this yet.
>

Or you chose to ignore it? Parallelize this:

Ted, what about parallelizing most often used pattern --

for (el:v.iterateNonZero()) {

   ... do something with element


}
I don't see how you can do it without functional programming -- maybe
there's a trick i don't know.

or another most used style:

for ( i = 0; i < v.size; i++) v.setQuick(i, ...)

What is  a reason DistributedRowMatrix is not AbstractMatrix? I bet it is
the same reason. Please directly address this issue (how this code can
execute in parallel) without starting deprecating apis etc.

(I give you the hint: parallel version of this code is using FP closures,
i.e.

v.foreach( (i,v) => ... )

but of course this requires API change, i.e. what i call "nuclear option".

Please address this issue. Otherwise your arguments are just ad nauseam,
i.e. just simply repeating "I don't believe it because i don't".

Be concrete and specif within given examples.

Reply via email to