[Devel] default stem width and height for autohinting

[Werner LEMBERG]

The function `ah_hinter_compute_widths' relies on the existence of
glyph `o' in the font.  What about searching for alternative glyphs
depending on the font?  Especially in non-latin fonts, using `o' is a
bad idea, since there a lot of fonts (especially CJK ones) which have
extremely ugly shapes for ASCII since it is never used.  Even worse,
the typographic properties can differ enormously between such ASCII
glyphs and the real glyphs the font is used for.  The same problem
exists for computing blue zones in `ah_hinter_compute_blues'.

Perhaps it makes sense to derive different global autohinting
properties depending on the character range.  For example, a
hyphothetical Near East font contains glyphs for Hebrew, Arabic, and
the latin script.  If a Hebrew glyph has to be rendered, global
autohinting properties derived from Hebrew glyphs only should be used.
The same holds for Arabic and latin glyphs.

It shouldn't be too difficult to implement that.  The `blue_chars' has
to become two-dimensional in some way, and we have to define leading
characters for the various Unicode ranges to derive default stem width
and height values.  For CJK, to name an example, a typical glyph could
be the character for `ten': `十'.

Has anybody already done some tests whether this really improves
autohinting results?


    Werner