gnuplot-beta

--enable-thin-spline

[Petr M. <mi...@ph...>]

Looking into gnuplot's rpm file on OpenSUSE Linux, its ChangeLog says that 
they compile gnuplot with --enable-thin-spline. What about having this 
option default?

---
PM

Re: --enable-thin-spline

[Ethan M. <merritt@u.washington.edu>]

On Tuesday 05 December 2006 11:14 am, Petr Mikulik wrote:
> Looking into gnuplot's rpm file on OpenSUSE Linux, its ChangeLog says
> that they compile gnuplot with --enable-thin-spline. What about
> having this option default?

I have no personal experience with it, so I do not know how much
difference it makes, or whether there are any problems with it.

I am reluctant to change the defaults for the 4.2 release at this
late date, just because I assume 99% of the testing so far has used
the existing set of defaults.

On the other hand, if SuSE has been shipping it that way for some
years and we have not received any bug reports on it, that may be
evidence that it is in good shape.

Could somebody provide an example plot where this option makes
a difference?

-- 
Ethan A Merritt
Biomolecular Structure Center
University of Washington, Seattle WA

Re: --enable-thin-spline

[<HBB...@t-...>]

Ethan Merritt wrote:

> On the other hand, if SuSE has been shipping it that way for some
> years and we have not received any bug reports on it, that may be
> evidence that it is in good shape.

--enable-thin-spline differs from most other options in gnuplot in that 
it doesn't just enable or disable a feature, but rather chooses a 
different method to implement the same feature: set dgrid3d.

Thin-plate splines provide a radically different way of producing a 
continuous surface from non-gridded data that can then be sampled to 
yield a grid.

I.e. it's not really a question of receiving bug reports, it's one of 
getting different output from the same gnuplot script, caused by a build 
option reported only where most people never look: in 'show version 
long', in a form they won't understand.

> Could somebody provide an example plot where this option makes
> a difference?

Any plot using dgrid3d is expected to come out different with 
--enable-thin-splines.  Some will differ more than others.

In the longer term, --enable-thin-splines really should be replaced by 
an 'algorithm' option in 'set dgrid3d'.

And for what it's worth: I'm with Ethan on this: now is certainly not 
the time to change the default state of this setting.

Re: --enable-thin-spline

[Petr M. <mi...@ph...>]

> Any plot using dgrid3d is expected to come out different with 
> --enable-thin-splines.  Some will differ more than others.
> 
> In the longer term, --enable-thin-splines really should be replaced by 
> an 'algorithm' option in 'set dgrid3d'.
> 
> And for what it's worth: I'm with Ethan on this: now is certainly not 
> the time to change the default state of this setting.

Thanks for explanation. Please consider adding such an option into 4.3.

---
PM