[Kst] kdeextragear-2/kst/kst

[George Staikos]

I just wanted to point out the floating point issues with the current code.  
I'll let you guys decide what you want as a solution to the problem as I 
really have to work on a few other major items.  Just keep in mind the 
limitation of floating point arithmetic (including the fact that precision 
internally in the CPU is different than what is stored).

On Wednesday 04 August 2004 17:48, Barth Netterfield wrote:
> It is a very tricky problem!
>
> I do not want to limit the dynamic range one can plot over unnecesarily,
> but, the label problem has to be fixed...
>
> A good usage case to think about (one we face often) is - we are plotting
> time_t on the x axis, and are zoomed into a x range of 200ms.  So
> basically, the X axis as it is now is completely useless, though at least
> we can see the data.
>
> One thing I saw somewhere for this situation was:
> the value for Xmin is displayed to full precision at the left side of the
> plot, below the axis label line, then the relative offset from XMin is
> shown at each major tick mark.
>
> +0		+0.05		+0.10		+0.15		+0.20
> 1064757578.752				(Time)
>
> We could move to this notation automatically if we hit the dynamic range
> limit.
>
> cbn
>
> On August 4, 2004 05:29 pm, Andrew Walker wrote:
> > The ideal solution would be to limit the zoom range to what can be
> > encompassed by the labels. Most architectures have 14 signifincant
> > figures for doubles, while the labels give (at best) 9 significant
> > figures.  Thus, at first glance, limiting to 9 sig. figs. would seem
> > appropriate.
> >
> > However, this zoom range would change depending on the value that we are
> > zoomed around. Thus, if the user is zoomed in to the minimum range
> > permitted near 0.0 this would then be less than the minimum zoom range
> > near 10.0. So scrolling around would present problems, unless we
> > dynamically change the zoom range as the user scrolls (probably not
> > desirable).
> >
> > The current code is better than it was, but far from ideal. Suggestions
> > are welcome.
> >
> > ---------- Original Message ----------------------------------
> > From: George Staikos <staikos at kde.org>
> > Date:  Wed, 4 Aug 2004 13:02:23 -0400
> >
> > >On Tuesday 03 August 2004 22:36, Andrew Walker wrote:
> > >> CVS commit by arwalker:
> > >> Fix for 86524.
> > >>
> > >> -#define X_MIN_DIFFERENCE 0.000001
> > >> -#define Y_MIN_DIFFERENCE 0.000001
> > >> +#define X_MIN_DIFFERENCE (MINDOUBLE*10.0)
> > >> +#define Y_MIN_DIFFERENCE (MINDOUBLE*10.0)
> > >
> > >  I'm not sure that this can be correct anywhere except around 0.0.
> > > MINDOUBLE is defined to be the smallest value represented by double.
> > >
> > >  The zoom logic looks like this:
> > >
> > > if (ymax_in > ymin_in + Y_MIN_DIFFERENCE) { // only accept new range if
> > > valid YMax = ymax_in;
> > >    YMin = ymin_in;
> > >    return true;
> > >  }
> > >
> > >   However floats are not evenly distributed.  This means that while the
> > >difference between 0 and [first_positive_neighbor_of_0] is MINDOUBLE,
> > > further out it could be very many times MINDOUBLE.
> > >
> > >--
> > >George Staikos
> > >KDE Developer			http://www.kde.org/
> > >Staikos Computing Services Inc.	http://www.staikos.net/
> >
> > ________________________________________________________________
> > $0 Web Hosting with up to 120MB web space, 1000 MB Data Transfer
> > 10 Personalized POP and Web E-mail Accounts, and much more.
> > Get It Now At www.doteasy.com
> >
> >
> >
> > _______________________________________________
> > Kst mailing list
> > Kst at kde.org
> > https://mail.kde.org/mailman/listinfo/kst
>
> _______________________________________________
> Kst mailing list
> Kst at kde.org
> https://mail.kde.org/mailman/listinfo/kst

-- 
George Staikos
KDE Developer				http://www.kde.org/
Staikos Computing Services Inc.		http://www.staikos.net/