pcolor and "interp" [changeset-2]

[Kai Habel]

Ben Abbott schrieb:
> On Thursday, October 02, 2008, at 08:55AM, "Kai Habel" <kai.habel at gmx.de> wrote:
>   
>> Ben Abbott schrieb:
>>     
>>> On Oct 1, 2008, at 8:42 PM, Ben Abbott wrote:
>>>
>>>       
>>>> On Sep 30, 2008, at 4:53 PM, Ben Abbott wrote:
>>>>
>>>>         
>>>>> On Tuesday, September 30, 2008, at 04:40PM, "Ben Abbott"
>>>>> <bpabbott at mac.com> wrote:
>>>>>           
>>>>>> On Tuesday, September 30, 2008, at 04:32PM, "Kai Habel"
>>>>>> <kai.habel at gmx.de> wrote:
>>>>>>             
>>>>>>> John W. Eaton schrieb:
>>>>>>>               
>>>>>>>> On 30-Sep-2008, Kai Habel wrote:
>>>>>>>>
>>>>>>>> | I would support interpolated surfaces only for gnuplot
>>>>>>>> 4.2.4+(5). In
>>>>>>>> | that case "illusions" esp. for pcolor are not required. Users
>>>>>>>> who really
>>>>>>>> | need interpolated surfaces can use the cvs version(s) of gnuplot.
>>>>>>>>
>>>>>>>> In that case, do we even need to check the gnuplot version here?
>>>>>>>>                 
>>>>>>> I don't think so.
>>>>>>>               
>>>>>>>> What
>>>>>>>> is the result if we do not?  Does gnuplot crash, or simply fail to
>>>>>>>> interpolate the colors?
>>>>>>>>                 
>>>>>>> In my understanding it should just fail to interpolate, but I have
>>>>>>> not
>>>>>>> tried this for myself. Maybe, Ben has a complete overview?
>>>>>>>               
>>>>>> I'll need to rebuild my gnuplot without the patch applied and find
>>>>>> out.
>>>>>>
>>>>>> I'll get back to the list in a few hours.
>>>>>>
>>>>>> Ben
>>>>>>             
>>>>> Ok ... it only took a few minutes. My package manager still had the
>>>>> deb files or unpatched version :-)
>>>>>
>>>>> I think Kai is correct. If I disable the check the old version of
>>>>> gnuplot appear to produce the same results they always did. That
>>>>> seems more proper to me. I'll modify the changeset and run a "make
>>>>> check" and run some more demos. If all looks good I'll post the
>>>>> corrected one later tonight.
>>>>>
>>>>> Ben
>>>>>           
>>>> My prior changes have been checked in. The changeset needed to
>>>> implement the version suggested by Kai is attached.
>>>>
>>>> Ben
>>>>         
>>> Trying again. This time with the facecolor bug reported by Michael
>>> included.
>>>
>>>       
>> Hello Ben,
>>
>> the interpolation for surfaces works here pretty well. But I noticed one
>> gnuplot issue, related to the interpolation.
>>
>> Try this:
>> h=pcolor(peaks(12));
>> set(h,'edgecolor','w');
>>
>> If you now toggle between:
>> set(h,'facecolor','interp');
>> set(h,'facecolor','flat');
>> the white grid changes its  spacing, where it should not.
>>
>> I don't think we can change this behavior easily. So, I would consider
>> this as minor problem and leave it as it is.
>>
>> Kai
>>     
>
> hmmm ...
>
> I don't see this problem in 3.0.2 (can someone verify).
>
> Kai, do you think this is a result of my changes or due to some thing else?
>
> Ben
>   
Interesting,

I have this behavior with a very recent checkout of octave (development
version) plus your last changeset and gnuplot (4.3 cvs). I think it is a
'feature' of gnuplot's interpolate.

Kai