Hi Florian,

This is an excellent suggestion - SVDPACKC has always been a bit
difficult, and in
fact just this morning I was running some tests and (once again)
getting segmentation
faults with las2 when using gcc version 4.1.1. However, it seems to
work fine with
gcc 3.4.0 and other earlier gcc's, so I was just thinking about how to
address that
either in the installation instructions and/or the install script.
But, I think looking at
SVDLIBC is very good idea, and I will do that. The whole motivation
behind moving
to CPAN is to allow for an easier and cleaner install process by
offering a Bundle
that can be used to install SenseClusters and all dependent modules, and that is
working great. I even have a little script to go get and install the
Cluto binaries that
seems to be working pretty well. But, good old SVDPACKC remains the weak
link. So, let me look at this.

I will probably go ahead and release another development version on CPAN in the
next day or two, this is mostly for me to check formatting and install
issues, and
that probably will still have SVDPACKC in it, but my goal is to have a
1.01 release
sometime soon that will be much easier to install, so maybe that can
use SVDLIBC.
You are correct btw, all we ever use from SVDPACKC is las2, so SVDLIBC
does meet our needs in that respect.

Thanks very much, an excellent and very well timed suggestion!

Cordially,
Ted

On Fri, Mar 28, 2008 at 10:25 AM, Florian Petran
<[EMAIL PROTECTED]> wrote:
> I think the CPAN stuff is a great idea. But have you considered using
>  svdlibc ( http://tedlab.mit.edu/~dr/SVDLIBC/ ) instead? It's basically
>  a reimplementation of svdpackc with a cleaner interface, and it's in
>  sort of active development. The drawback is that it only implements
>  the lanczos2 algorithm, but I think SC only uses that.
>
>
>  Greetings,
>  Florian
>
>  2008/3/23, Ted Pedersen <[EMAIL PROTECTED]>:
>
>
> > Greetings all,
>  >
>  >  I will be working on a clean up release of SenseClusters - mainly
>  >  taking care of small
>  >  glitches in documentation, organization, testing, and installation
>  >  procedures. I don't
>  >  plan on adding any new functionality for this version, more or less
>  >  just cleaning up a
>  >  few lose ends.
>  >
>  >  If there are issues of that sort that you've noticed, this would be a
>  >  great time to mention
>  >  them, as it will be very easy to take care of them now. If there is
>  >  larger functionality
>  >  changes that you are interested in, please don't hesitate to mention
>  >  those as I might get
>  >  ambitious, and there will be another release coming along after this.
>  >
>  >  I am considering using CPAN as a distribution site in addition to
>  >  sourceforge, and
>  >  actually have uploaded SenseClusters there now - it would be called
>  >  Text::SenseClusters
>  >  at CPAN, just because they frown on introducing new high level names
>  >  there, and it
>  >  helps describe it for new users somewhat. The reason I'm thinking of
>  >  using CPAN is
>  >  to further automate the installation procedure - some time agree I
>  >  created a Bundle for
>  >  SenseClusters (Bundle::SenseClusters, although I might rename as 
> Bundle::Text::
>  >  SenseClusters) that if you install using the CPAN.pm module will check
>  >  for an load all
>  >   the prerequisite CPAN packages, by simply doing this...
>  >
>  >  cpan
>  >  > install Bundle::SenseClusters
>  >
>  >  I think we can include SenseClusters in the Bundle,
>  >  or have it call the Bundle, and reduce the Perl installation part of
>  >  SenseClusters to
>  >  a single command. Right now it's actually surprisingly easy, but I'd
>  >  like it to be easier
>  >  still. I'm also contemplating including a pre-edited version of
>  >  SVDPACKC with the
>  >  package, since new releases of that are very very unlikely, and it
>  >  appears to be permissible
>  >  under the terms of the license that it is distributed. That would just
>  >  leave cluto as an
>  >  external install, which isn't too bad since that comes as a binary.
>  >
>  >  The other thing I like about CPAN is that it makes the code and
>  >  documentation much
>  >  more "visible". You can browse around the source code and
>  >  documentation and really see
>  >  it, and that seems like it will lead to easier maintenance. This might
>  >  also make it possible
>  >  to reduce certain directories that would become redundant, like our
>  >  HTML tree (which
>  >  would be created and available, in effect, via CPAN). It's important
>  >  to stress that we
>  >  would still use sourceforge for distribution and CVS, so a possible
>  >  CPAN release is
>  >  in addition to what we normally do, not instead of.
>  >
>  >  Finally, this is the route that we took with NSP, and over time we
>  >  ended up with a package
>  >  that is much more object oriented, and I think that evolution was in
>  >  some sense encouraged
>  >  by being on CPAN.
>  >
>  >  The good news is that none of this is that essential, it's just nice
>  >  to do. I downloaded
>  >  SenseClusters on a new machine today, and was actually quite impressed
>  >  at how easy
>  >  it was to install and get running, and the demo scripts and test
>  >  scripts that we make
>  >  available with it remain extensive and more complete/thorough than we
>  >  probably have
>  >  for any other package. So, it remains a great example of how to put
>  >  together a really
>  >  complicated system and make it easy to use, and the above are all just
>  >  tiny tweaks on
>  >  what is an extremely solid body of work.
>  >
>  >  So, suggestions would be particularly helpful now, and if you are
>  >  interested in seeing
>  >  SenseClusters on CPAN you can see an experimental release there now :
>  >
>  >  http://search.cpan.org/dist/Text-SenseClusters/
>  >
>  >  This is version 0.96, where the even number implies development
>  >  version, so 0.95 remains
>  >  our stable release, and what you should use if you are interested in
>  >  doing experiments,
>  >  etc. with the package. I'll continue to use the even numbers for
>  >  development releases like
>  >  this, hopefully culminating in a 1.01 stable release. SenseClusters
>  >  really deserves to be
>  >  > 1 as well.
>  >
>  >  Thanks!
>  >  Ted
>  >
>  >  --
>  >  Ted Pedersen
>  >  http://www.d.umn.edu/~tpederse
>  >
>  >  -------------------------------------------------------------------------
>  >  This SF.net email is sponsored by: Microsoft
>  >  Defy all challenges. Microsoft(R) Visual Studio 2008.
>  >  http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
>  >  _______________________________________________
>  >  senseclusters-users mailing list
>  >  [email protected]
>  >  https://lists.sourceforge.net/lists/listinfo/senseclusters-users
>  >
>



-- 
Ted Pedersen
http://www.d.umn.edu/~tpederse

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
_______________________________________________
senseclusters-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/senseclusters-users

Reply via email to