Re: [DotGNU]RDF based discovery system

[James Michael DuPont]

--- Norbert Bollow <address@hidden> wrote:
> James Michael DuPont <address@hidden> wrote:
> 
> > Lets say that I have a md5 digest, and want to find all the
> functions
> > that operate on that. That would be a valid case. 
> > 
> > Or lets say, I am looking for a function to to a FFT on an array of
> > data. This would also be something to be discovered.
> 
> Such a "function discovery system" sounds very useful.  How will it
> work?

Well, the basic idea is that of relation.

For example, you have a Concept "FFT", this is described by a set of
webpages. They are related to the concept. You have datatypes that
represent instances of the FFT. Functions that operate on them. 
You have emails, cvs, and webpages that talk about the source code.

All of these would be indexed and given a search system that has the
following priority :
1. conceptual model.
2. source code identifiers
3. emails/cvs/
4. Documentation

These sources are pulled up as needed. Wordnet and other synonyms
systems can be used to create broader searches.

Then you need a way to search the conceptual web. 
I imagine it like a wiki system with lots of data and a better backend.

The next set is the mapping of internal declarations into this
conceptual web. This is done by first extracting all the things in the
program into a set of statements : 

<http://foo.bar/test.tgz/include/ffz.h#FFT> is-declared-in
<http://foo.bar/test.tgz>.
<http://foo.bar/test.tgz/include/ffz.h#FFT> is-a <gcc/datatype>.
<http://foo.bar/test.tgz/include/ffz.h#init_fft> is-a <gcc/function>.
<http://foo.bar/test.tgz/include/ffz.h#init_ffm> operates on >FFT>.

You have a mapping function : 
RealWorld(ProgramDecl) -> ConceptDecl 

this maps a program decl into a concept space,
maybe on the fsedu server :
<http://foo.bar/test.tgz/include/ffz.h#FFT> <realworld>
<fsedu.org:/math/algoritms#fft>  

that will map a single Program declaration into a concept. 

A program declaration has an identifier and it has a type, functions,
variables, constants, datatypes are all declarations.

The next set is to map emails and documentations onto the program. That
uses the identifiers of the program to index the mailing lists, the cvs
and the docs. 

Function : 

DocumentedBy(ProgramDecl)-> Document Usages.
<http://foo.bar/test.tgz/include/ffz.h#FFT> <DocumentedBy>
<http://foo.bar/docs/FFT.html>

The next step is to use the concepts to index the Documentation
RealWorld(DocumentIdea)-> Conceptual Model.
http://www.geocrawler.com/archives/3/7323/2000/12/0/4765458/
<RealWorld> <fsedu.org:/math/algoritms#fft>  

Of course this needs to have a Conceptual Model to begin with, a way to
index the various document files in the program, a way to extract all
the declarations out of the source code, and an indexing system.

Mike

=====
James Michael DuPont
http://introspector.sourceforge.net/

__________________________________________________
Do you Yahoo!?
Yahoo! Tax Center - File online, calculators, forms, and more
http://tax.yahoo.com