From: M. Edward (Ed) Borasky <znmeb@ce...> - 2006-09-23 04:53:55
Joseph Wang wrote:
> Yes. I'm very interested in hearing anything about the R backend.
> Right now as far as I can tell, the test-suite compiles and executes
> for the R backend. The gotcha is that right now the test-suite
> doesn't verify that the answers are correct, merely that things compile
> and execute.
> Unless the R API changes, there shouldn't be any problems with 2.4 beta.
> I'm actively using the R-SWIG interface to compile a Quantlib frontend
> to do research on Shanghai warrants. This involves lots of algorthmic
> computation but the data sets are small, and so I don't know (and would
> be interested in finding out) how R-SWIG handles larger data sets.
I accidentally sent the whole log (all languages) to the list. It is
four megabytes and luckily the list filtered it out. In any event,
zipped it is only about 200 kbytes if anyone wants the whole run. It
includes everything SWIG found on my machine, which is several versions
of Scheme, Perl, Python, Ruby and Tcl, of course, Pike and Lua. GCC is
I'm going to re-run just the R tests and look at that, since I have a
rough idea what R should do on those cases.
My end goal in all of this is to build "scripting-language independent"
wrappers for some major math libraries, such as LinBox, Givaro, NTL, and
GiNaC/CLN. There is a project to do Python wrappers for GiNaC in SWIG,
but they are somewhat Python-specific and I'm trying to be generic. But
the two languages I'm most interested in are Ruby and R; I don't know
Python, Tcl, Pike or Lua and I'm tired of Perl. :)
How is your R interface to Quantlib different from the CRAN package that
interfaces to Quantlib?
From: Joseph Wang <joe@gn...> - 2006-09-24 04:33:37
R-QuantLib uses hand written wrappers whereas my implementation of
R-QuantLib-SWIG uses the general SWIG wrappers which get shared with all of
the SWIG frontends. This is a bit more scalable, and implements more of the