Re: [changeset] colorbar bug

[dbateman]

On Sun, Mar 01, 2009 at 10:04:42PM -0500, Ben Abbott wrote:
> 
> On Mar 1, 2009, at 7:57 PM, Ben Abbott wrote:
> 
> >
> >On Mar 1, 2009, at 1:52 PM, dbateman wrote:
> >
> >>
> >>Ben Abbott wrote:
> >>>
> >>>This changeset addresses two bugs I encountered.
> >>>
> >>>(1) Two colorbar commands produced two colorbars.
> >>>
> >>>(2) colorbar ("off") did not turn off the colorbar.
> >>>
> >>>Demos are included which verify the intended behavior.
> >>>
> >>>This fix was trivial, and has been pushed to savannah.
> >>>
> >>>Ben
> >>>
> >>
> >>I'm not sure (1) is a bug. Yes I noticed this behavior when I  
> >>reimplement
> >>colorbar in 3.2, but kept it as that is similar behavior to matlab  
> >>when I
> >>tested it. However, I think its probably better to only have a single
> >>colorbar so the fix is fine.
> >>
> >>D.
> >
> >I should have checked what Matlab does.
> >
> >>> clf
> >>> set(0,'handleVisibility','on')
> >>> plot(1:10)
> >>> hc1 = colorbar
> >
> >hc1 =          173
> >
> >>> hc2 = colorbar
> >
> >hc2 =          173
> >
> >Thus Octave's present implementation shoud not delete colorbar  
> >replace it with a new one, but should update it.
> >
> >When given a new location, a second colorbar is created.
> >
> >>> hc3 = colorbar('westoutside')
> >
> >hc3 =          203
> >
> >I did not expect this behavior. The fix should be trivial. I'll look  
> >at it.
> >
> >Ben
> 
> David, the job is a bit more complicated than I'd expected. It appears  
> to me that a rewrite of update_colorbar is needed.
> 
>       addlistener (ax, "dataaspectratio", address@hidden, cax})
>       addlistener (ax, "position", address@hidden, cax})
> 
> Updating a colorbar axis when other colorbars are present will require  
> a significant change to how this presently works. I'd (1) add the  
> listener to the data axis, (2) pass the original data axis position  
> through the listener, (3) have the data axis updater find all  
> associated colorbars and set the position and size of data axis and  
> all associated colorbar axes, (4) preserved the original data axis  
> position value, by taking care that the listener is not modified as  
> subsequent colorbars are added.
> 
> I'm still studying how  your present implementation is working. Does  
> my approach sound proper, or can you recommend a something better?
> 
> If not, I'm not sure that respecting compatible behavior in this  
> instance is important enough at the moment (there are some more  
> important backend issues that need attention for the 3.2 release), so  
> I'm likely to delay working further on this.
> 
> In any event, a partial changeset is attached below. It preserves the  
> handle when colorbar is called multiple times for the same location,  
> but only allow one colorbar per axes (a demo has been added to  
> illustrate this behavior). I've also attempted to ensure that non- 
> gnuplot backends will work.
> 
> Can you take a quick look?


Frankly I can't imagine a good reason to have two colorbars and so if 
its too hard just leave it as is after your change

D.