Re: new subplot bug

[Ben Abbott]

On Wednesday, February 18, 2009, at 12:15PM, "Ben Abbott" <address@hidden> 
wrote:
>On Wednesday, February 18, 2009, at 11:59AM, "John W. Eaton" <address@hidden> 
>wrote:
>>On 18-Feb-2009, Ben Abbott wrote:
>>
>>| I found a few errors in my prior changeset.
>>| 
>>|     http://hg.savannah.gnu.org/hgweb/octave/rev/4142982c66c6
>>| 
>>| The most severe were not offsetting the outerposition from the postion  
>>| property, and calculating the lower/bottom inset correctly.
>>| 
>>| There are some surprises with the proprietary solution. First, the  
>>| outerposition properties for subplot with more than 4 rows overlap. In  
>>| Octave's implementation this results in new subplots deleting  
>>| neighboring ones.
>>| 
>>| I fixed that by trigger deletion when the position properties overlap.
>>| 
>>| A second surprise is the interaction of Matlab's parser and the  
>>| subplot command. See the example below.
>>| 
>>|  >> subplot(1,5,1);get(gca,'position')
>>| ans =     0.13         0.11      0.12374        0.815
>>| 
>>|  >> subplot(1,5,1);
>>|  >> get(gca,'position')
>>| ans =     0.13         0.11      0.12132        0.815
>>| 
>>| I characterized the proprietary behavior using scripts. Thus, it  
>>| presently respects the first example. I verified the propriety  
>>| behavior for Matlab 2007b as well as for 2008b. I'd prefer we not try  
>>| to duplicate this behavior.
>>| 
>>| As I already missed several errors, please check this over. If it is  
>>| ok, please push (I won't be able to until later).
>>
>>OK, this seems to improve things, so I checked it in.
>>
>>Now, here is another problem that I have no idea about.  Try running
>>
>>  x = -10:0.1:10;
>>  for i = 1:2; subplot (2, 1, i); pause; plot (x, sin(x)); endfor
>>
>>At the first pause, I see the empty top axes.
>>
>>At the second pause, both axes are visible and empty.  I would expect
>>the first one to have a sine wave plot.
>
>At the second pause, I see the sin in the first.
>
>>After the loop completes, There is a third axes object that overlaps
>>both subplots and has a single sine wave plot (see the attached image).
>
>I don't see the third axes. After the last pause, I get the expected result.
>
>>Any clues about this one?
>>
>>Thanks,
>>
>>jwe
>
>I haven't built since yesterday.
>
>I'll do so now and see if I can duplicate what you see.
>
>Ben
>

oh-oh ... now I see the same problem.

octave:1> x = 0:0.01:10;
octave:2> for i = 1:2; subplot (2, 1, i); pause; plot (x, sin(x)); endfor 
octave:3> findobj (gcf, "type", "axes")
ans =

  -11.9012
   -6.8679
   -1.0786

octave:4> h = findobj (gcf, "type", "axes");
octave:5> get (h, "tag")
ans =

{
  [1,1] = 
  [1,2] = gnuplot_drawnow
  [1,3] = 
}

octave:6> get (h, "visible")
ans =

{
  [1,1] = on
  [1,2] = on
  [1,3] = on
}

hmmm, This is quite strange.

clf; for i = 1:2; subplot (2, 1, i); disp(numel(findobj (gcf, "type", 
"axes"))); plot (x, sin(x)); endfor
1
2

clf; for i = 1:2; subplot (2, 1, i); pause; disp(numel(findobj (gcf, "type", 
"axes"))); plot (x, sin(x)); endfor
 2
 3

The "pause" appears to result in all "visible" properties being set to "on"

Ben