[Fontinst] LIG plus kerning

[Ulrich Dirr]

> Ulrich Dirr wrote:
> 
> > Lars Hellström wrote:
> > 
> >>>But it looks like that the kerning information for, e.g.,
> >>>'zero'+'ellipsis' is discarded. It will only be kerned when the
> >>>'ellipsis' char is entered directly but not with the ligature
mechanism.
> >>>Is there a way to get this nevertheless?
> 
> Use a kern for zero + period?

Definitely not! Just look at '7' + 'period' vs '7' + 'ellipsis'. The first
needs almost always negative kerning while the second needs positive
kerning.

 
> >>Not really, since this is a fundamental limitation of the TeX
ligaturing
> >>mechanism: kerns and ligatures are formed in a single pass over the
text,
> >>so the right hand side of a kerning or ligature pair will always be a
raw
> >>character. (Maybe you can get it right if you do something really
clever
> >>with the more exotic ligaturing instructions and additional dummy
> >>characters, but that it most likely not worth the trouble.)
> >>
> >>
> >>>Can I combine this 'two-step'-ligatures with individual kerning pairs?
> >>
> >>Kerns to its right will work as usual. Kerns on its left have no effect
> >>since the ligature has not yet been formed when that kern position is
> >>considered. 
> >>
> >>Similar problems occur in the T1 encoding for the >> and << ligatures.
> > 
> > 
> > That's bad news. So I've to find a 'nice' looking char for direct input
> > (the ANSI dec133 is already used by another glyph). And a macro
solution
> > like \dots would make the input file unreadable because 'ellipsis' is
used
> > heavily.
> 
> Well, experience shows that for typical texts with ...
> this: '...' --> '..' +  '.' --> 'ellipsis' or '. . .' (kerned dots)
> works reasonably well and provides an easy automatic + transparent 
> improvement compared to the text where authors simply didn't 
> care about...

I don't understand this. I've made the lig mechanism '.' + '.' --> '..' +
'.' --> 'ellipsis'. But as Lars has pointed out I will lose the kerning.

 
> Kerned dots do not require a "dummy" character '..' but you will not 
> have the 'ellipsis' in the PDF for copy/paste (search?/Unicode).
> 
> You could add for very special cases a '0period'-slot and decide then 
> what happens if there follows another period: '0period' + 'period' /-/ 
> '0' 'period' 'period' &c. But that won't work for lots of kerns to be 
> considered...

The best way I've found is to use a translation table of the editor which
replaces the 'ellipsis'-char with '...' when read in, and the other way
round when writing the doc. Probably \textellipsis has to be redefined
because of different font designs.

Best regards,
Ulrich Dirr