Re: tests in log-scale plots

[Ben Abbott]

On Sep 7, 2011, at 8:20 AM, bpabbott wrote:

> On Sep 07, 2011, at 07:48 AM, "c." <address@hidden> wrote:
> 
>> 
>> On 7 Sep 2011, at 13:44, bpabbott wrote:
>> 
>> > Carlo, what is the value for the 'ydir' property after the "loglog(a,b)" 
>> > plot?
>> > 
>> > Is is actually reversed in 2007b? Since -10 < -1, I expect the setting of 
>> > ydir is still normal
>> > 
>> > Ben
>> 
>> Ben,
>> 
>> Now I see what you mean:
>> 
>> >> loglog (a, b) 
>> >> get (gca, 'ydir')
>> 
>> ans =
>> normal
>> 
>> >> 
>> 
>> On the other hand, to reproduce the same plot using abs (b) I have to do
>> 
>> >> loglog (a, abs(b))
>> >> set (gca, 'ydir', 'reverse')
>> >> get (gca, 'ydir')
>> 
>> ans =
>> reverse
>> 
>> c.
>  
> I think the proper way to handle negative numbers, in log plots, is to have 
> the backend do the job.
> 
> For gnuplot this needs to be done in __go_draw_axes__.m
> 
> Ben

I've had a long discussion with Michael Godfrey on this problem, which led to 
the attached patch for graphics.cc and graphics.h.in.

The attached patch is intended to implement the fix the front-end. Changes are 
still required for __go_draw_axes__.m and gl-renderer.cc.

With this patch applied to the default branch ...

> >> x = logspace (-5, 5, 11);
> >> loglog (-x, -x);
>            line 0: x range must be greater than 0 for log scale
> 
> >> get (gca, "xtick")
> ans =
> 
>   -1.0000e+06  -1.0000e+04  -1.0000e+02  -1.0000e+00  -1.0000e-02  
> -1.0000e-04  -1.0000e-06
> 
> >> get (gca, "ytick")
> ans =
> 
>   -1.0000e+06  -1.0000e+04  -1.0000e+02  -1.0000e+00  -1.0000e-02  
> -1.0000e-04  -1.0000e-06

My c/c++ skills deserve skepticism. I'd appreciate it if someone would look 
over the patch.

Ben

log_and_negative_values.diff
Description: Binary data