Re: character by name on the gnome canvas, custom fontencoding

[Christopher Blizzard]

Owen Taylor wrote:
> On Mon, 2004-06-14 at 16:24, Jan Nieuwenhuizen wrote:
>
> > Owen Taylor writes:
> >
> >
> > > Chris Blizzard is working on some API enhancements to Pango to
> > > allow the application to control how this mapping works
> >
> > I've just noticed that Chris has already checked in quite a hack and
> > I'm wondering if you could point me to an example or documentation
> > about how to use this.
>
>
> "Quite a hack". Why?
>
> API docs are with the check-in. Perhaps Chris has a pointer to 
> an example usage.
>
> Regards,
>                                                 Owen

I've attached the small test stub that I used to test the code before 
checking it in.  It should shed a little light on how the code paths are 
supposed to work.  At a high level, here's how you would add a custom 
decoder:

Create an object that derives from the PangoFcDecoder that implements 
the two class methods: get_charset and get_glyph.  More on these 
callbacks later.

When your application starts up, call 
pango_fc_font_map_add_decoder_find_func with a callback function you've 
created.  What pango creates new fonts, it will call back into your 
function and allow you to register one of your custom decoders for that 
particular font.

So this means the code looks like this:

somewhere during startup:

pango_fc_font_map_add_decoder_find_func (fontmap, mycallback, NULL, NULL);

and the callback:

PangoFcDecoder *
mycallback (FcPattern *pattern, gpointer user_data)
{
    /* check to see if the pattern matches a font that we want to decode
     * by hand */
    if (is_custom_font(pattern))
      {
        return custom_decoder_new();
      }
}

and your decoder:

struct _CustomDecoder
{
    PangoFcDecoder parent_instance;
};

get_glyph should return a single glyph for a specific code point.  Note 
that there is no support here for doing more than one glyph for a code 
point.  As Owen put it, I think correctly, "write a shaper" if you want 
to do that.

get_charset should return an FcCharSet * that includes complete coverage 
information for your font.  That might include coverage in the usual 
unicode ranges, or in the PUA.  It doesn't really matter all that much, 
as long as the code points in the text you want to render matches up to 
those characters in the coverage information and the font description 
font name will resolve to the font for which you have the decoder 
registered.

--Chris