Re: [Quantlib-dev] [QuantLib-svn] SF.net SVN: quantlib: [11606] trunk/QuantLib

[Luigi B. <lui...@gm...>]

On Tue, 2007-06-19 at 08:53 -0700, fd...@us... wrote:
> Log Message:
> -----------
> QL :
> minValueAfter of Discount curve traits modified to allow the use of
> CubicSpline. This method returns half of the previous DF value.

No, this is not correct. If the nodes are far apart, there's nothing
that forbids the discount at a node to be less than half of the previous
value; yet you're preventing such a curve from bootstrapping.

If the minimum at 0 was wrecking havoc on the spline, we could use the
other solve() method in Solver1D, namely, the one that takes a guess and
a step and tries to bracket the solution on its own. It's going to be
slower, but probably more robust.

Luigi


---------------------------------------- 

The purpose of abstraction is not to be vague, but to create a new 
semantic level in which one can be absolutely precise. 
-- W.E. Dijkstra