Fwd: Re: Patch 1804820 : FIT_MAXITER<0

[<pl...@pi...>]

On Sat, 24 May 2008 01:03:49 +0200, Philipp K. Janert <ja...@ie...>
wrote:

>
> Ethan has asked me to take a look at this
> patch.
>
> I guess the purpose of this patch is to allow
> the user to enter a negative value for
> MAX_ITER, whereupon the fit command will
> calculate the residual using the starting
> values of all fit parameters and exits. (That is
> really zero iteration steps, but MAX_ITER==0
> has already been used to indicate unlimited
> iteration).
>
> I am not very happy with this patch:
> - I don't understand the purpose. Why do
> 	you care about the unfitted residual?
> - It leads to partially garbage output (the
> 	errors in the fit parameters are not
> 	defined for no iteration, etc)
> - We are (ab-)using the MAX_ITER parameter
> 	by introducing a magic value with
> 	non-trivial semantics (which is never
> 	a good idea in my mind).
>
> How does everybody else feel about this?
> Who would find this patch useful, and why?
>
> (If it is useful, can we find a better way of
> achieving the same goal, rather than shoe-horning
> semantics into a negative iteration count?)
>
> Best,
>
> 		Ph.
>

Hi,

I agree with your general comment , I find there is too much of this
"magic number" and syntax tweeking approach in gnuplot. It makes it very
difficult to remember and even harder to discover some very useful
functions.

It can get very aesoteric at times.

best regards, Peter.