Re: Improvements in the calculation of text extents

[Ben Abbott]

On Dec 22, 2010, at 6:43 AM, logari81 wrote:

> Currently, the calculation of the extent of text objects in Octave is
> very limited. For example it doesn't take into account the rotation
> angle:
> 
> Revision 11407 of Octave:
> 
> octave:1> backend fltk
> octave:2> hx = xlabel('x label');
> octave:3> get (hx, 'extent')
> ans =
>   1.6832e-17  -7.7629e-17   6.2500e-02   3.6252e-02
> octave:4> set(hx, 'rotation', 90)
> octave:5> get (hx, 'extent')
> ans =
>   1.6832e-17  -7.7629e-17   2.6210e-02   8.6447e-02
> 
> For this reason I have worked in cleaning up and improving the
> calculation of text extents and positioning of text. My modifications
> can be found in the attached changeset. The result of the previous
> example after these changes becomes rotation aware:
> 
> With the attached changeset:
> 
> octave:1> backend fltk
> octave:2> hx = xlabel('x label');
> octave:3> get (hx, 'extent')
> ans =
>  -0.033266  -0.039041   0.062500   0.036252
> octave:4> set(hx, 'rotation', 90)
> octave:5> get (hx, 'extent')
> ans =
>  -0.0020161  -0.0460123   0.0262097   0.0864473
> 
> For comparison see the corresponding result from ML:
> 
>>> hx = xlabel('x label');
>>> get (hx,'extent')
> ans =
>    0.4458   -0.1115    0.0982    0.0541
>>> set (hx, 'rotation', 90)
>>> get (hx,'extent')
> ans =
>    0.4761   -0.1521    0.0605    0.1521
> 
> The attached patch does not only fix this small problem but also
> prepares all necessary infrastructure for handling the tightinset
> property later.
> 
> The first two components of the extent property implemented in the
> attached changeset are not compatible with ML. The convention that ML
> uses for these two components makes sense only in 2D plots and I don't
> think that it is good idea to follow the ML convention in this case.
> 
> Please test the attached patch and give me some feedback in case it
> breaks anything. Otherwise I think it is a good idea to commit it since
> it cleans up a few things.
> 
> BR
> 
> Kostas 
> <text-extents.changeset>

This changeset doesn't apply for me. In particular, some lines from graphics.cc 
seem to be different than what is indicated by your changeset.

The current sources indicate ...

5079   elt = text_parser_none ().parse (get_string ());
5080 #ifdef HAVE_FONTCONFIG
5081   text_renderer.set_font (*this);
5082 #endif
5083   box = text_renderer.get_extent (elt, get_rotation ());

... but your changeset expects ...

-  elt = text_parser_none ().parse (get_string ());
-#ifdef HAVE_FONTCONFIG
-  text_renderer.set_font (get ("fontname").string_value (),
-                          get ("fontweight").string_value (),
-                          get ("fontangle").string_value (),
-                          get ("fontsize").double_value ());
-#endif
-  box = text_renderer.get_extent (elt, get_rotation ());

Ben