[PATCH] Payoffs fully implemented in a foreign language via SWIG
Locked
[PATCH] Payoffs fully implemented in a foreign language via SWIG
 
|
I would like to post 2 patches (one for QuantLib, the other for QuantLib-SWIG) to allow users to
implement a payoff entirely in a foreign language (e.g. Java). These patches apply on top of the other patch I have just sent. By enabling directors the user can implement the "value" function in Java like it was a standard subclass of a C++ class. The goal here is not speed :-) but ease of use, by writing a payoff in a language like Java or C#. There is also a simple example of an option that pays the running average on every fixing date. Everything works fine with the exception of the following issue. But I thought to post the code since I could not find a better solution that what I have currently implemented. Basically I want to expose to Java a hierarchy like that class PathPayoff { virtual ....... = 0; // this is the payoff interface }; class PathOption { virtual vector<Date> fixingDates() = 0; virtual boost::shared_ptr<PathPayoff> payoff() = 0; <<<<<<<<< PROBLEM }; No problem for PathPayoff, but huge issues on how to implement PathOption::payoff() in Java. Basically the problem is how to create in Java a shared_ptr containing an object inherited from PathPayoff (which is implemented in Java as well). I know how to create a raw pointer (which is already managed by SWIG/Java), but I don't think this pointer can be safely stored into a shared_ptr (i.e. there would be double management). So I've changed and exposed only one class class ExternalOption { virtual ....... = 0; // this is the payoff interface virtual vector<Date> fixingDates() = 0; boost::shared_ptr<PathOption> convert(); // helper to convert an ExternalOption into a PathOption }; this class contains all functions required to describe a contract. Then there is a function that converts the ExternalOption into a PathOption using a custom wrapper. The problem now is that the wrapper contains a reference (C++ reference) to the ExternalOption, so that there could be problems if someone tried to use the Wrapped PathOption after the ExternalOption has gone out of scope of garbage collected. I wish I had a shared_ptr<ExternalOption> but I only have a raw pointer or a reference. It is a problem similar to the original, but I found it is easier to manage. It is all down to the use of shared_ptr in SWIG. If anyone has a comment or better solution I would be delighted to hear about it. Andrea ------------------------------------------------------------------------------ Throughout its 18-year history, RSA Conference consistently attracts the world's best and brightest in the field, creating opportunities for Conference attendees to learn about information security's most important issues through interactions with peers, luminaries and emerging and established companies. http://p.sf.net/sfu/rsaconf-dev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev  external.diff (7K) Download Attachment swig.diff (6K) Download Attachment JavaOption.java (5K) Download Attachment |
Re: [PATCH] Payoffs fully implemented in a foreign language via SWIG
|
|
Re: [PATCH] Payoffs fully implemented in a foreign language via SWIG
|
|
Re: [PATCH] Payoffs fully implemented in a foreign language via SWIG
|
|
On 20/01/10 11:15, Luigi Ballabio wrote:
> On Mon, 2010-01-18 at 22:14 +0000, Andrea wrote: >> On 17/01/10 21:31, Andrea wrote: >>> I would like to post 2 patches (one for QuantLib, the other for >> QuantLib-SWIG) to allow users to >>> implement a payoff entirely in a foreign language (e.g. Java). >>> These patches apply on top of the other patch I have just sent. >> >> I've actually found a better solution, but it requires a global change >> in QuantLib-SWIG. >> >> In SWIG there is already support for boost::shared_ptr<> that removes >> distinction between raw pointer, value, reference and smart pointer >> making it all easier. >> >> Does anybody know why it is not used? > > SWIG didn't provide it when we first wrote the interfaces, and I'm > afraid we haven't keep abreast of the latest developments. I'll have to > check it out. Would it result in an interface change as seen from the > host language? Basically one has to export the raw C++ inheritance. Then swig will remove all differences between pointers, references, shared_ptr, wrapping them when needed. I think it would simplify what Quantlib-SWIG does now, where it needs to expose classes inheriting from shared_ptr<I> rather than I itself. I think the idea is to expose things as they are, then SWIG will take care of all the shared_ptr details. If you compile the file I have attached, you can see what it looks like. It is very likely that the user will not notice any difference. Coming back to my original problem, unfortunately it seems that using shared_ptr and directors does not work (yet?) (SWIG generates invalid Java code). So for the time being the code I posted at the top of the thread is still my best solution. Andrea ------------------------------------------------------------------------------ Throughout its 18-year history, RSA Conference consistently attracts the world's best and brightest in the field, creating opportunities for Conference attendees to learn about information security's most important issues through interactions with peers, luminaries and emerging and established companies. http://p.sf.net/sfu/rsaconf-dev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: [PATCH] Payoffs fully implemented in a foreign language via SWIG
|
|
Re: [PATCH] Payoffs fully implemented in a foreign language via SWIG
|
|
Re: [PATCH] Payoffs fully implemented in a foreign language via SWIG
|
|
| Free forum by Nabble | Edit this page |