1-d roots (error handling)

1-d roots (error handling)

[Peter S. Christopher]

Hi list,

	I'm using brents algorithm to find one dimensional roots, although
I don't think my question is restricted to brents algo. In the multi-root
package, the user defined function can return GSL_FAILURE to indicate to
the solver that a critical error has occurred. Then the solver returns an
appropriate error code. Can this be done with the one-d root
algorithms? The current interface has the 1-d user defined function
returning a double; if I return GSL_NAN from my user defined will the the
solver handle that like a critical error? Should it handle this like a
critical error?

Thanks in advance,
Pete Christopher  

Re: 1-d roots (error handling)

[Brian Gough]

Peter S. Christopher writes:
 > 	I'm using brents algorithm to find one dimensional roots,
 > although I don't think my question is restricted to brents algo. In
 > the multi-root package, the user defined function can return
 > GSL_FAILURE to indicate to the solver that a critical error has
 > occurred. Then the solver returns an appropriate error code. Can
 > this be done with the one-d root algorithms? The current interface
 > has the 1-d user defined function returning a double; if I return
 > GSL_NAN from my user defined will the the solver handle that like a
 > critical error? Should it handle this like a critical error?

If the function returns a NaN/Inf it should lead to an error code of
GSL_EBADFUNC. If it doesn't please send a bug report (either for the
code or documentation). Thanks.