[cairo] Font interface problem

[Carl Worth]

On Mar 9, Maarten Breddels wrote:
 > I think there are some problems with the current font interface.

There certainly are. This code hasn't seen a lot of heavy use, so it's
good to be getting some good feedback like this.

 > The font backends only knows about the font matrix (which is multiplied
 > with the CTM before any call to the font backend). I think it's better
 > to pass the CTM as argument to the font backend because it is really
 > needed.

Please feel free to re-work backend interfaces as necessary to get
things working.

A larger concern of mine is fixing bugs in the user-visible portion of
the API. There are problems that I've already found,
(eg. cairo_scale_font modifies fonts that the user previously grabbed
via cairo_current_font). Also, there are some ambiguities in the
current interface that need to be nailed down. Some of these will
require experience from developers trying to use cairo in non-trivial
applications.

 > Currently _cairo_gstate_text_extents does a bit of a fix to
 > change the device coordinates back to the user coordinates(allthough i'm
 > not sure if it's the right way, since it doesn't do rotation, which
 > 'bearing' and 'advance' need). I also noticed
 > _cairo_gstate_current_font_extents doesn't try to transform the device
 > coordinates to user coordinates (is that a bug?).

Your descriptions sound like bugs. If you can generate minimal test
cases to demonstrate the problems, that would help,
(cairo-demo/png/text.c might be a good starting point). Of course,
even better would be patches to fix the bugs.

 > And then there is a problem with glyphs, i think the cairo_glyph_t x and
 > y members are supposed to be in user coordinates,

Yes.

 > Another thing is that (0,0) is the upper left corner, with y going down
 > the screen, but fonts are mirrored in x, so they appear upright with an
 > unmodified CTM and font matrix.

Yes, the default graphics convention is for Y to increase from
top-down, while the default font convention is for Y to increase from
bottom-up. This leads to plenty of confusion at the boundary of these
two systems, (such as cairo_text_extents). Concrete proposals to
address the problem would be quite helpful.

 > Ok, it's not a big issue, but to keep things simple, i think it's
 > better to just draw them the 'normal' way.

Could you describe more specifically what you would like to change
here?

Thanks again for your feedback,

-Carl