[R-SIG-Finance] Quantlib in R (was RE:Financial Basket Options)

Dirk Eddelbuettel edd at debian.org
Thu Jan 24 16:00:23 CET 2008


Hi Brian,

On 24 January 2008 at 07:39, Brian G. Peterson wrote:
| Dirk Eddelbuettel wrote:
| > Sorry, I meant to chime in earlier on this. Did any of you look at QuantLib?
| > AFAICT it has a basketoption class allowing for multiple assets, Monte Carlo
| > pricers for american and european exercise as well as Stulz (1992) method.  
| > 
| > I often look at the available code via what's in the regression tests, so
| > here is the header test-suite/basketoption.hpp:
| > 
| > class BasketOptionTest {
| >   public:
| >     static void testEuroTwoValues();
| >     static void testBarraquandThreeValues();
| >     static void testTavellaValues();
| >     static void testOneDAmericanValues();
| >     static void testOddSamples();
| >     static boost::unit_test_framework::test_suite* suite();
| > };
| > 
| > My RQuantLib package currently does not wrap basket options, but has a few
| > other exotics one could take as a stanza.  Building the package is possible
| > on both Windows and Linux, but somewhat more tedious on the former (as you
| > need to build QL and Boost first).  My main contributor Dominick has a
| > tarball with that prebuilt if it is of interest.  If there is interest, we
| > could add this to RQuantLib.  
| 
| Well, I don't really use Windows as an operating system for running R, 
| but it seems that adding the required libraries to the Windows version 
| of RQuantlib would increase the likelihood of extensions.

I think we are talking about the same issue here, but maybe we are also
talking past each other, so just do be clear:

- Windows users R can just 'click on' RQuantLib and they get a binary ready
  to run; this includes a static library of QuantLib itself.  [ And Debian
  users  get RQuantLib via apt-get as usual, same for Ubuntu etc pp so those
  derivate distros may be a version behind

- Because R on Windows uses MinGW, one needs a MinGW build of QuantLib to
  build R against.  Such builds have in the past been provided by Dominick.
  However, these builds are typically not publically announced but instead
  have been given to Uwe to build CRAN's Windows version of RQuantLib.
  This allows _use_ of RQuantLib but not really development.
 
- It would be great of someone volunteered to provided this library on
  go-forward basis as Dominick indicated that he will no longer do this.
  
| On a related note:
| 
| I'm really curious whether there's any interest in extending and 
| formalizing the SWIG work that was done with Quantlib and R.  It seems 
| that the Quantlib team has some interest, as they have added some 
| documentation to their site, which is new since the last time I looked:
| 
| http://wiki.quantlib.org/twiki/bin/view/Quantlib/RSwigDocumentation
| 
| and the SWIG team's "quite incomplete" documentation, which overlaps to 
| a degree with the docs on the Quantlib page:
| 
| http://www.swig.org/Doc1.3/R.html
| 
| Joseph Chen-Yu Wang of the QuantLib team's paper on this may be found here:
| http://papers.ssrn.com/sol3/papers.cfm?abstract_id=965317
| 
| It seems that it would be good to spend a little time trying to package 
| at least some of the SWIG work (possibly into RQuantlib?) so that there 
| is a template for adding other functions to R as they are added to Quantlib.

I am obviously fully in favour, but I do not have the bandwidth to drive
this along with my current list of things to juggle.

The last time I tried to build this, shortly after Joe Wang finished a major
piece of his work, I found that

- his documentation was inadequate (at least for my Python and Swig skills)
  and wrong (and then offered to proof-read and sadly never got back to him)
  though I ended up getting it build with some help by Joe

- created one monstrously large dynamic library file you need to load;
  basically all of QL becomes one blob which is a tad unwielding if you just
  want one or two pricers

- however, this is really useful piece of code, and QuantLib is about to go
  to the long-awaited 1.0 release so the time may be right.

| Quantlib is such an extensive and useful collection, it seems that we 
| could all benefit from a little more work in that direction.

So despite of what I said above, maybe you and I need to clear a weekend and
make it a Chicago hackathon?  

Anybody else?  Chicago can be reached from just about anywhere, and I sure we
can find a place to stay for any and all volunteers :)  Barring that, we
could always try virtual meetings but real hackathons may be more effective.

Dirk

-- 
Three out of two people have difficulties with fractions.



More information about the R-SIG-Finance mailing list