Re: [eclipse-clp-users] Question on EclipseCLP and viability for job-shop scheduling

On 07/11/2013 18:12, Chandrakanth R. Terupally wrote:
> I'm trying to investigate if Eclipse CLP software would be viable to solve many large CLP problems
> my client has. Currently an alternative approach is being used, but its ridden with issues on
> multiple fronts. My question specifically is (1) how capable is Eclipse CLP for crunching large
> datasets (with schedule tasks in the tune of 0.5M-5M)

As others have commented, there is no technology today to schedule this number of tasks to
optimality.  My estimate is that exact CLP techniques are good for 10s-100s of jobs, approximate
hybrids of CLP and local search techniques for 100s-10000s, and beyond that you probably can't do
much more than some sorting and local improvement.


> (2) how easy is programming it to read input
> data from text data files. The second question is as important as the first one as the constraints
> are made available by means of text files.

This should be easy, there are libraries for cvs and xml reading.  For more complex input formats
you can use the built-in DCG (definite clause grammar) mechanism.  It is also often possible to
preprocess data into a form that can be read directly with the ECLiPSe parser.


> Also, do you know if its possible or if anyone has tried
> accessing the kernel of the software for its algorithms from another software, say e.g. Matlab,
> java, etc?

We regularly embed ECLiPSe components into Java host applications.  A Matlab connection could be
built using ECLiPSe's "remote interface protocol" described in
http://eclipseclp.org/doc/embedding/embroot.html, but I don't know if anyone has done it.


-- Joachim