Dawid, To rephrase what I said in my other email,
Yes. These are going away. BUT, we plan to port/test the important functionality into the newer style. The rationale behind this is a) we need a marker for what has been made usable/safe/tested b) the Colt API was too tangled to allow new implementations of things like matrices c) adding tests has shown the Colt numerical API has proven to be relatively low quality in a few areas. I had to completely rewrite some of the statistical stuff to get more than a few significant figures. d) my own personal efforts to deal with the LUD code foundered on difficulties caused by the Colt code assuming internal representation details unnecessarily. The upshot is that our strategy is that all of the Colt code will eventually suffer one of the following fates: - delete - add tests, restructure to use tested API, rename to new style - completely re-implement in a maintainable style My initial preference was for the second fate as much as possible. That preference has changed a bit to prefer the first option with the third as a backup. Partly this is because we are scraping down to the less-used, lower quality parts of Colt. But having somebody who wants a capability is an excellent motivator! On Thu, Feb 3, 2011 at 9:04 AM, Dawid Weiss <[email protected]>wrote: > Just a clarification -- I'm working with the trunk (collections and > math); these bits (in spite of being marked as @Deprecated) are not > going to go away, are they? My quick scan through the issues after a > local upgrade show that the major problem for us is final/ hidden > impl. of DenseDoubleMatrix2D. I will just copy this class entirely to > our own code -- it didn't change that much. > > Will let you know if I find out anything else. > > Dawid > > On Thu, Feb 3, 2011 at 2:33 PM, Sean Owen <[email protected]> wrote: > > That sounds good. The bits that were removed were deprecated, without > tests, > > and without (known) users. If they're resurrected, we'll un-deprecate, > but > > therefore also need tests (and perhaps a short think about whether the > API > > is right). There's always the chance we all decide that a given bit of > > functionality is far enough afield from Mahout that it should just live > in > > Carrot2. > > > > On Thu, Feb 3, 2011 at 1:27 PM, Grant Ingersoll <[email protected]> > wrote: > >> > >> It seems in our clean up of Mahout Math, that we may have cut Carrot2 > off > >> a bit which is a downstream consumer of math. It appears that some of > the > >> methods C2 uses have been removed in 0.4 of Mahout. I'm CC'ing Dawid > and > >> Stanislaw here so perhaps they can chime in with what is missing. > >> > >> -Grant > > >
