|
|
>I will make sure the stuff compiles on GNU/Linux.
this would be great!
>We need a curve builder :-)
The yield curve base class is TermStructure (/Include/termstructure.h)
This is an almost pure interface class, stating the methods every
TermStructure has to implement (zeroYield, discount, forward, etc.)
A curve builder has to produce an instance of a class derived from
TermStructure.
Three abstract adapter classes are derived from TermStructure:
ZeroYieldStructure, DiscountStructure, and ForwardRateStructure.
These classes might help, since by using them the curve builder only has to
produce a zero yield curve OR a discount curve OR a forward curve.
This means that toy term structures are already available right now.
A curve builder with bootstrap on deposits is already there for a piecewise
constant forward curve (Include\TermStructures\piecewiseconstantforwards.h
and Sources\TermStructures\piecewiseconstantforwards.cpp).
We have also provided a simple example of flat constant forward
(QuantLib/Include/TermStructures/piecewiseconstantforwards.h in the current
cvs) for didactic purposes (to replicate Hull numerical examples, etc.)
Everyone can implement his favorite bootstrapping method and contribute it
to QuantLib. I welcome any contribution.
Luigi and I are working on adding bootstrapping on swap rates. The design
is quite clear already (piecewise constant forward, maybe later piecewise
linear forward), but we need to implement swaps first - which will take a
little time, especially since the whole date/payment schedule must be designed.
A strong point of the builder we're working on (and a strong point of
QuantLib as a whole) will be the fact that it uses the actual QuantLib
financial instruments (deposit, futures, FRA, swap) in the bootstrapping
procedure.
This should guarantee that the curve building will always reprice inputs
exactly, whatever the evolution of QuantLib financial instruments will be.
Our work shouldn't stop anybody from providing different curve builders,
since the date/payment schedule is a crucial point that will take a while.
You know, the devil is in (boring) details.
> Then the fun can start!
Yeah, and you can't imagine how excited I am!
Bernd, I would ask you to post your messages/comments on the quantlib-dev
and/or quantlib-users mailing lists.
This will help to have a discussion going on, so that everybody may
contribute. That's why I cc my reply to the lists
thank you
ciao -- Nando
|
Locked
