[changeset] bug concatenating vectors in 3.1.51+

[Ben Abbott]

On Tuesday, October 14, 2008, at 02:51PM, "John W. Eaton" <jwe at bevo.che.wisc.edu> wrote:
>On 13-Oct-2008, Ben Abbott wrote:
>
>| 
>| On Oct 12, 2008, at 3:51 PM, Jaroslav Hajek wrote:
>| 
>| > On Sun, Oct 12, 2008 at 8:59 PM, Ben Abbott <bpabbott at mac.com> wrote:
>| >> I attempted a fresh build last night. My tip is
>| >>
>| >> changeset:   8210:a10397d26114
>| >> tag:         tip
>| >> user:        John W. Eaton <jwe at octave.org>
>| >> date:        Fri Oct 10 15:10:58 2008 -0400
>| >> summary:     graphics.cc (base_graphics_properties::adopt): place new
>| >> child at front of list, not end; make children a column vector  
>| >> instead
>| >> of a row vector
>| >>
>| >> The bug is fairly easy to see
>| >>
>| >>       [ones(1,0); zeros(5,1)]
>| >>       error: number of columns must match (1 != 0)
>| >>
>| >> In Matlab
>| >>
>| >>>> [ones(1,0);zeros(5,1)]
>| >>
>| >> ans =
>| >>
>| >>     0
>| >>     0
>| >>     0
>| >>     0
>| >>     0
>| >>
>| >> Presently Octave will not build for me.
>| >>
>| >> if [ -f gplot.eps ] ; then gs -dBATCH -dEPSCrop -dNOPAUSE -q -
>| >> sDEVICE=pdfwrite -sOutputFile=gplot.pdf gplot.eps ; fi
>| >> ../../run-octave -f -q -H -p . --eval "sparseimages ('grid', 'eps');"
>| >> error: number of rows must match (79 != 1) near line 937, column 23
>| >> error: called from:
>| >> error:   /Users/bpabbott/Development/mercurial/octave-3.1.51/scripts/
>| >> plot/__go_draw_axes__.m at line 937, column 6
>| >> error:   /Users/bpabbott/Development/mercurial/octave-3.1.51/scripts/
>| >> plot/__go_draw_figure__.m at line 59, column 8
>| >> error:   /Users/bpabbott/Development/mercurial/octave-3.1.51/scripts/
>| >> plot/gnuplot_drawnow.m at line 55, column 7
>| >> error:   /Users/bpabbott/Development/mercurial/octave-3.1.51/scripts/
>| >> plot/print.m at line 447, column 7
>| >> error:   /Users/bpabbott/Development/mercurial/octave-3.1.51/doc/
>| >> interpreter/sparseimages.m at line 231, column 5
>| >> error:   /Users/bpabbott/Development/mercurial/octave-3.1.51/doc/
>| >> interpreter/sparseimages.m at line 33, column 2
>| >>
>| >> make[3]: *** [grid.eps] Error 1
>| >> make[2]: *** [interpreter] Error 2
>| >> make[1]: *** [doc] Error 2
>| >> make: *** [all] Error 2
>| >>
>| >> The offending line in __go_draw_axes__.m is
>| >>
>| >>       kids = [obj.children, kids];
>| >>
>| >> Which with John's change from row to column vector for handles needs
>| >> to be
>| >>
>| >>       kids = [obj.children; kids];
>| >>
>| >> However, in this instance an empty vector with size [1, 0] fails.
>| >>
>| >> I replaced the line with
>| >>
>| >>         if (isempty (kids))
>| >>           kids = obj.children;
>| >>         elseif (! isempty (obj.children))
>| >>           kids = [obj.children; kids];
>| >>         endif
>| >>
>| >> and "make" and "make check" completed without further problems.
>| >>
>| >> I could offer a changeset to resolve the problem with
>| >> __go_draw_axes__, but expect the the proper solution would to fix the
>| >> concatenation of column vectors.
>| >>
>| >> Might someone with c++ skill take care of this, or if there are
>| >> substantial problem with that, say the word and I'll submit a
>| >> changeset for __go_draw_axes__.
>| >>
>| >> Ben
>| >>
>| >
>| > Ben,
>| >
>| > I've just tried in Matlab 2007a:
>| >
>| > Matlab gives zeros(5,1), yet along with this warning:
>| >
>| > Warning: Concatenation involves an empty array with an incorrect
>| > number of columns. This may not be allowed in a future release.
>| >
>| > Why ones(1,0) does not gripe, I don't know. I guess it's a bug in  
>| > Matlab.
>| > Anyway, I think that in light of this fact it doesn't seem reasonable
>| > to bother with getting Matlab compatible here. And even if we do, I
>| > think that code relying on such a behaviour is a bad idea.
>| >
>| > regards
>| 
>| I'd thought I responded to this already, but I don't see it on Nabble,  
>| so ....
>| 
>| In Matlab 2008a
>| 
>|  >> [ones(2,0);zeros(5,1)]
>| Warning: Concatenation involves an empty array with an incorrect  
>| number of columns.
>| This may not be allowed in a future release.
>| 
>| ans =
>| 
>|       0
>|       0
>|       0
>|       0
>|       0
>| 
>| So Jaroslav you are a practical prophet ;-)
>| 
>| I've attached a changeset for __go_draw_axes__.m
>| 
>| Ben
>
>
>Thanks, I applied the changeset.
>
>Also, sorry for breaking things.  I noticed that the orientation of
>the children property was different in Matlab and Octave, but I didn't
>think about the consequences of changing it.  Oops.
>
>jwe
>

No problem here. It was the right thing to do. With all the handle related work coming up it would have been much more painful later.

In fact, I'm looking at the graphic m-files for instances where how "handles" are handled might be improved (hence the other thread).

Ben