matplotlib-devel Mailing List for matplotlib

Hi all,
I implemented the following method in ft2font.cpp, and although it
works I don't know if it does the *right thing*.
char FT2Font::draw_rect_filled__doc__[] =
"draw_rect_filled(x0, y0, x1, y1)\n"
"\n"
"Draw a filled rect to the image. It is your responsibility to set the\n"
"dimensions of the image, eg, with set_bitmap_size\n"
"\n"
;
Py::Object
FT2Font::draw_rect_filled(const Py::Tuple & args) {
 _VERBOSE("FT2Font::draw_rect_filled");
 args.verify_length(4);
 long x0 = Py::Int(args[0]);
 long y0 = Py::Int(args[1]);
 long x1 = Py::Int(args[2]);
 long y1 = Py::Int(args[3]);
 FT_Int iwidth = (FT_Int)image.width;
 FT_Int iheight = (FT_Int)image.height;
 if ( x0<0 || y0<0 || x1<0 || y1<0 ||
 x0>iwidth || x1>iwidth ||
 y0>iheight || y1>iheight )
 throw Py::ValueError("Rect coords outside image bounds");
 for (long j=y0; j<y1; ++j) {
 for (long i=x0; i<x1+1; ++i) {
 image.buffer[i + j*iwidth] = 255;
 }
 }
 return Py::Object();
}
Basically, I copied the existing draw_rect method and changed the code
a bit. The above code draws a filled rectangle in the image buffer. I
use it mathtext for drawing a line (for a fraction etc.). What I'm
interested is the for loop:
for (long j=y0; j<y1; ++j) {
 for (long i=x0; i<x1+1; ++i) {
 image.buffer[i + j*iwidth] = 255;
 }
 }
I'm not sure that this is the right code - probably some pixel isn't
getting drawn. Anyone has some ideas?
Thanks,
Edin

Hi all,
I just wanted to get others some insight about what I did in
mathtext2. Who knows - it might turn out to be useful to someone ;)
Also, any comments/thoughts are appreciated.
A TeX math expression:
$\sum_{i = 0}^{\infty}$
gets translated to a pure Python list, consisting of Python builtins:
[u'\\sum', u'_', [u'i', u' ', u'=', u' ', u'0'], u'^', [u'\\infty']]
This gets fed to a token parser that produces the TeX equivalent of
hboxes. Every unicode character (including "\sum" - u'\u2211') gets
translated to a TexCharClass instance. TexCharClass (I'm not very good
at naming things :) is a class that completely handles rendering of a
single character. It does this based on the information available from
the font file (TrueType). More about fonts can be found in the
excellent FreeType lib docs.
http://www.freetype.org/freetype2/documentation.html
A combination like a_b^c (or a^c_b) gets translated to a Scripted
instance, which, again, handles the rendering of the sub/supercript
and the nucleus. Similar for a fraction: a Fraction instance has a
numerator and denumerator, which are used for rendering.
The above python list
[u'\\sum', u'_', [u'i', u' ', u'=', u' ', u'0'], u'^', [u'\\infty']]
and any other sublist, like
[u'i', u' ', u'=', u' ', u'0']
get translated to a Hbox instance.
TexCharClass, Scripted, Fraction, Hbox (but also Line, Kern) are all
sublclasses of Renderer, a class that defines an __init__ method and a
render method. The data attributes of a Renderer instance (initialized
to 0 by the Renderer's __init__ method) are xmin, xmax, ymin, ymax,
bearingx, bearingy, width, height - for now. Basically, these are
taken from the FreeType implementation of glyphs. In other words,
every instance of Renderer, that is: an instance of a subclass of
Renderer (Renderer is a virtual class) behaves like a glyph.
http://www.freetype.org/freetype2/docs/glyphs/glyphs-3.html
Unfortunately, although FreeType allows drawing of vertical text, I
haven't implemented it yet. This would be also a base for Vbox, so,
for example, \frac 1 2 could be translated to Vbox([u'1', Line(width,
height), u'2']). Supporting vertical text involves adding additional
attributes to the classes like, vertical bearingx, vertical bearingy
(see the FreeType docs).
Cheers,
Edin