PayoffInterpreter: specify new pay-offs and price them without code recompilation
Locked
PayoffInterpreter: specify new pay-offs and price them without code recompilation
 
|
Hi
the idea to specify new pay-offs and price them without code recompilation by using a script language is floating around for some time. To achieve this goal I used an interpreter for a C / Fortron90 like language two colleagues and I have written some time ago. As an example I've connected this interpreter with the MonteCarlo pricing engine. The interpreter itself is running in a "Sandbox" with a C++ interface. Therefore one has full control over the interpreter during the runtime (e.g. reading/modifiying the variable stack, run self defined functions etc.). If you are interest in this little project please feel free to download it from www.spanderen.de/PayoffInterpreter.zip or www.spanderen.de/PayoffInterpreter.tar.gz Makefiles for Linux/Unix and VC++ are included. My colleagues have given permission to put the interpreter under the Quantlib licence. Feedback whether this project is off any use and should be developed further on are welcome. cheers content of the tar ball: langauge/ : the interpreter incl. Bison/Flex files math/ : multi dimensional array and index array util/ : utils docu/ : html documentation of the language testsuite/ : small C++ testsuite using the Sandbox examples/ : QuantLib MonteCarlo pricer using the PayoffInterpreter pilib/ : PayoffInterpreter example programs and tests console/ : simple console for the PayoffInterpreter -- Klaus Spanderen Ludwig Erhard Str. 12 48734 Reken (Germany) Email: [hidden email] (remove NOSPAM from the address) ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: PayoffInterpreter: specify new pay-offs and price them without code recompilation
|
|
Hi Klaus
> If you are interest in this little project please feel free to download it I don't have time enough to take a look at it, but it would surely an interesting addition. I know others have been working on similar projects and I wonder if anybody knowledgeable about payoff scripting could summarize the current consensus on this issue I would be interested in knowing what are the approaches used on the street by commercial or proprietary parsers, any ad-hoc open source library which could help, design decisions, etc ciao -- nando ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: PayoffInterpreter: specify new pay-offs and price them without code recompilation
|
|
Hi Nando,
the language itself is quite comprehensive C derivative with some MatLab/ Fortran90 extensions, e.g. vector/matrix as build-in types and the interpreter has a C++ interface. The "problem" is that the project consists of +20K lines of code and I don't want to blow up QL until I know that a PayoffInterpreter is of any use for other users;-). The parser itself is written using GNU Flex/Bison. On Tuesday 18 September 2007 10:45 am, Ferdinando Ametrano wrote: > I would be interested in knowing what are the approaches used on the > street by commercial or proprietary parsers, any ad-hoc open source > library which could help, design decisions, etc At least on the equity side stucturer are using script/interpreter languages together with MonteCarlo scenorio engines.(e.g. Reech). cheers -- Klaus Spanderen Ludwig Erhard Str. 12 48734 Reken (Germany) E-Mail: [hidden email] (remove NOSPAM from the address) ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: PayoffInterpreter: specify new pay-offs and price them without code recompilation
|
|
On Sep 19, 2007, at 9:39 PM, Klaus Spanderen wrote: > the language itself is quite comprehensive C derivative with some > MatLab/ > Fortran90 extensions, e.g. vector/matrix as build-in types and the > interpreter has a C++ interface. The "problem" is that the project > consists > of +20K lines of code and I don't want to blow up QL until I know that > a > PayoffInterpreter is of any use for other users;-). Well, I wouldn't put in into the core library anyway; it would be great to have it available as an additional module. The point is that I wouldn't commit to that one as "the" payoff interpreter. People using, say, the Python or Ruby bindings might want to define such an interpreter in their native language---it could be done easily enough using SWIG. Luigi ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
Apologies if this seems to be off topic, though this is somewhat related to a pipe dream I had regarding QuantLib. If it is possible to allow some external scripting language to define a new payoff structure, why not go ahead and allow the interpreter to extend QuantLib in general?
Perhaps an embedded version of Python could work or some other method of allowing "add-in" style hooks/call backs. :-) ------------------------------------------- Well, I wouldn't put in into the core library anyway; it would be great to have it available as an additional module. The point is that I wouldn't commit to that one as "the" payoff interpreter. People using, say, the Python or Ruby bindings might want to define such an interpreter in their native language---it could be done easily enough using SWIG. Luigi ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list QuantLib-dev@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
In reply to this post by Klaus Spanderen-2
Hi,
i tried to you your tool and running the example EquityOption.cpp. It compiles but crashes at run time. Did you have such a feedback before ? (0xC0000005: Access violation reading location 0x2183120a.) thanks
|
Re: PayoffInterpreter: specify new pay-offs and price them without code recompilation
|
|
In reply to this post by Luigi Ballabio
Hi Luigi,
> Well, I wouldn't put in into the core library anyway; it would be great > to have it available as an additional module. The point is that I > wouldn't commit to that one as "the" payoff interpreter. The intention is/was to have "a" payoff interpreter at hand to be able to do some prototyping when playing around with new payoffs. > People using, > say, the Python or Ruby bindings might want to define such an > interpreter in their native language---it could be done easily enough > using SWIG. BTW: I know that one can call QL functions from Python etc. via. SWIG. Is it also possible to call a Python-Script (which might define a payoff) from C++? regards -- Klaus Spanderen Ludwig Erhard Str. 12 48734 Reken (Germany) E-Mail: [hidden email] (remove NOSPAM from the address) ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: PayoffInterpreter: specify new pay-offs and price them without code recompilation
|
|
In reply to this post by HFQuant
On Thursday 20 September 2007 5:21 pm, HFQuant wrote:
> Hi, > > i tried to you your tool and running the example EquityOption.cpp. It > compiles but crashes at run time. > Did you have such a feedback before ? (0xC0000005: Access violation reading > location 0x2183120a.) > > thanks > arrg. No hadn't had this feedback before;-(. What operating system are you using? On windows you might want to include the payoff interpreter project into your quantlib project to avoid any parameter clashes. I checked the source code on Linux using g++-4.1 and on Windows with Visual Studio Express. On Linux the memory checker valgrind didn't report a memery leak or any access violation. I'm compiling against a recent QL version from the SVN head. I cheers -- Klaus Spanderen Ludwig Erhard Str. 12 48734 Reken (Germany) E-Mail: [hidden email] (remove NOSPAM from the address) ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: PayoffInterpreter: specify new pay-offs and price them without code recompilation
|
|
In reply to this post by Klaus Spanderen-2
Hi Klaus, On Sep 21, 2007, at 7:51 PM, Klaus Spanderen wrote: >> Well, I wouldn't put in into the core library anyway; it would be >> great >> to have it available as an additional module. The point is that I >> wouldn't commit to that one as "the" payoff interpreter. > > The intention is/was to have "a" payoff interpreter at hand to be able > to do > some prototyping when playing around with new payoffs. Don't worry, I'm aware of the good intention. > BTW: I know that one can call QL functions from Python etc. via. SWIG. > Is it > also possible to call a Python-Script (which might define a payoff) > from C++? Yes, with some code in the SWIG interface files. I'll try and send an example when I get some time. Luigi ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
In reply to this post by Klaus Spanderen-2
Hi,
I am using VS 2005 professional edition and the last Quantlib version 0.8.1. So i changed your include statements in the EquityOption.cpp file, basically including QuantLib.hpp and deleting the others. thks
|
Re: PayoffInterpreter: specify new pay-offs and price them without code recompilation
|
|
Hi
strange. Do you compile it (the whole project) as part of the quantlib? If you want I'll send you me VS project. On Monday 24 September 2007 11:34 am, HFQuant wrote: > Hi, > > I am using VS 2005 professional edition and the last Quantlib version > 0.8.1. > > So i changed your include statements in the EquityOption.cpp file, > basically including QuantLib.hpp and deleting the others. > > > thks > > Klaus Spanderen-2 wrote: > > On Thursday 20 September 2007 5:21 pm, HFQuant wrote: > >> Hi, > >> > >> i tried to you your tool and running the example EquityOption.cpp. It > >> compiles but crashes at run time. > >> Did you have such a feedback before ? (0xC0000005: Access violation > >> reading > >> location 0x2183120a.) > >> > >> thanks > > > > arrg. No hadn't had this feedback before;-(. What operating system are > > you using? On windows you might want to include the payoff interpreter > > project > > into your quantlib project to avoid any parameter clashes. > > > > I checked the source code on Linux using g++-4.1 and on Windows with > > Visual > > Studio Express. On Linux the memory checker valgrind didn't report a > > memery > > leak or any access violation. I'm compiling against a recent QL version > > from > > the SVN head. I > > > > cheers > > > > -- > > Klaus Spanderen > > Ludwig Erhard Str. 12 > > 48734 Reken (Germany) > > E-Mail: [hidden email] (remove NOSPAM from the address) > > > > > > ------------------------------------------------------------------------- > > This SF.net email is sponsored by: Microsoft > > Defy all challenges. Microsoft(R) Visual Studio 2005. > > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > > _______________________________________________ > > QuantLib-dev mailing list > > [hidden email] > > https://lists.sourceforge.net/lists/listinfo/quantlib-dev -- Klaus Spanderen Ludwig Erhard Str. 12 48734 Reken (Germany) E-Mail: [hidden email] (remove NOSPAM from the address) ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: PayoffInterpreter: specify new pay-offs and price them without code recompilation
|
|
In reply to this post by Klaus Spanderen-2
What would be nice is if there were some standard C++ interfaces which people
could using to "hook in" function evaluation systems from any language. Something very simple would be a "function adapter" which defined a function that took a double and returned a double, The trend I've seen in production code is to move away from special built payoff engines and toward general purpose scripting languages, because special built payoff engines tend in the long run to be unmaintainable. The other trend I've seen is to try to be interpreter agnostic since a typical situation is one in which you have two or three custom payoff interpreters which are used in various places, and there is an effort to try to start merging the code base. Also as far as payoffs, one thing that needs to be defined is the ability to specify a payoff schedule. One final thing, what would be useful is to create an object in python, matlab, R, java or what have you, and this would generate a C++ object which would look and act like any other C++ object out there. This would be useful not only in payoffs, but also generally useful both inside and outside of quantlib. ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: PayoffInterpreter: specify new pay-offs and price them without code recompilation
|
|
On Tue, 2007-09-25 at 00:00 -0400, Joseph Wang wrote:
> What would be nice is if there were some standard C++ interfaces which people > could using to "hook in" function evaluation systems from any language. This can be done already with a combination of polymorphism and the scripting language API. In Python, for instance, it would be enough to define a C++ class like the following one and export it to Python via SWIG: class PyPayoff : public Payoff { PyObject payoff_; // The payoff defined in Python. It can be // a function or function object. public: PyPayoff(PyObject* payoff) : payoff_(payoff) { // Store it... Py_XINCREF(function_); // ...and make sure Python doesn't } // garbage-collect it. ~PyPayoff() { Py_XDECREF(function_); // Release it when we're done. } Real operator()(Real price) const { // Call the Python function using the API... PyObject* pyResult = PyObject_CallFunction(payoff_,"d",price); // ...convert the result from Python... Real result = PyFloat_AsDouble(pyResult); // ...clean up... Py_XDECREF(pyResult); // ...and return. return result; } ... }; Once the above is exported (which can be done in a few lines of SWIG interface) the payoff can be defined from Python and used, as in: def my_payoff1(S): # a regular function return sqrt(S) class my_payoff2: # a function object def __init__(self,K): self.K = K def __call__(self,S): return (S-K)**2 option1 = SomeOption(process, PyPayoff(my_payoff1), exercise, engine) option2 = SomeOption(process, PyPayoff(my_payoff2(42.0)), exercise, engine) I'm sure the same can be done in other languages. Later, Luigi -- Poets have been mysteriously silent on the subject of cheese. -- Gilbert K. Chesterton ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
«
Return to quantlib-dev
|
1 view|%1 views
| Free forum by Nabble | Edit this page |