[math-font-discuss] Ugly drawn rules in Adobe Reader

[Taco Hoekwater]

Hi Johannes.

In the end, we came to the conclusion that probably the best idea
is to generate an internal font in pdftex on-the-fly. That font would
contain just a single character, that is xscaled and yscaled as
needed to create a fake rule of the right dimensions. Because no new
primitives would be needed in such an approach (except perhaps the
possibility to turn it on/off) it would be completely transparant
to the user. If I had 48-hour days, I would have started on this
already ;-)

Taco

Johannes Kuester wrote:
> Dear all (well, Taco and Karl mainly),
> 
> one problem I do see with a \mathrulechar character is a heavier 
> fraction bar (as it is used mainly in schoolbooks, to make clear which 
> is the main fraction bar of stacked fractions etc.). How could this be 
> handled? Vertical scaling?
> 
> Another problem: Now the rule is drawn exactly to the desired width
> (of radicand, over-/underlined character, numerator / denomitar)
> How could this be handled? Maybe by overlapping the mathrulechar with 
> itself accordingly?
> 
> Making the mathrulechar very narrow might not help here, as it might 
> lead to other rendering problems: the glyph needs an overlap of about 10 
> units (in Type 1), otherwise a bar might appear dashed.
> 
> Taco Hoekwater wrote:
>  >
>  > Karl Berry wrote:
>  >
>  >> So in practice some \newfam would have to be defined, right?  Since
>  >> obviously no existing fonts have this character.
>  >
>  > Right, but for e.g. the newmath encoding and Latin-Modern math, such
>  > a character could be included right from the start.
> 
> Yes, of course we could include a codepoint in newmath, and I could 
> include it in LatinModern Math.
> 
> Another possibility to handle these bars (and some other extensibles as 
> well) would be "Type 1 on demand"
> (where TeX calculates the necessary glyph dimensions in its first pass, 
> then a special font is created, and the glyphs are inserted in TeX's 
> second pass). I think something like this has been done for Arab math 
> typesetting, and I expect that it wouldn't be very difficult to 
> implement with LatinModern Math and MetaType1,
> but quite certainly this will never be available for all math fonts.
> 
> 
> Johannes
> 
>