Anyone done anything with QuantLib?
- Thread starter sle
- Start date
I'm a fan of it. What do you want to do - vol stuff?
Be careful of your ./configure settings flags so that you don't accidentally mix debug/release versions, O levels etc (./configure --help) and also using the right boost version, so checking the ./bjam flags when building. Hope you find it useful. There's plenty of help on QuantLib-users.
Be careful of your ./configure settings flags so that you don't accidentally mix debug/release versions, O levels etc (./configure --help) and also using the right boost version, so checking the ./bjam flags when building. Hope you find it useful. There's plenty of help on QuantLib-users.
Mostly light exotic pricing/delta and similar stuff. How easy is it create new payoff models? Is it easy to integrate with python if I write custom pricing models in it?
Dunno about Py, but RQuantLib is authored/maintained by Dirk Eddelbuettel, who's brilliant.
It's a wrapper around the C++ library, right? What I don't get is if I build a custom payoff, how easy is it to expose it through any of these wrappers?
QuantLib-Python is a SWIG based wrapper around the C++ Quantlib, the python bindings are generated by SWIG.
You could develop your custom payoffs in C++ using some of QuantLibs functionality (e.g day count conventions, yield curve generators etc) and then expose this functionality (your new functions/class member functions) using swig. The swig definitions file my only require one line of code to do this for each function or member function.
Or you could write your custom payoffs in python and use the QuantLib-Python for any standard analytics to reduce reinventing the wheel.
In either case, you would have to install/maintain: boost, QuantLib and QuantLib-Python.
You could develop your custom payoffs in C++ using some of QuantLibs functionality (e.g day count conventions, yield curve generators etc) and then expose this functionality (your new functions/class member functions) using swig. The swig definitions file my only require one line of code to do this for each function or member function.
Or you could write your custom payoffs in python and use the QuantLib-Python for any standard analytics to reduce reinventing the wheel.
In either case, you would have to install/maintain: boost, QuantLib and QuantLib-Python.
Yup. We got Fincad but it does not expose the guts enough for a couple products (like knock-in forward variance) that I want to price.
State of the art is a vague term
I got Fincad but some of the products we trade need a more involved model then Fincad allowes