[Fontinst] reglyphing and kerns

[Lars Hellström]

Dr. Clea F. Rees skrev:
> I am working with an enormous font containing several latin and greek
> alphabets. To install a particular alphabet, I'm using reglyphing. For
> example:
>     \renameglyph{A}{some-other-A}
> This is fine but I have an issue with kerning. I seem unable to get rid
> of all the kerns I need to get rid of while retaining the ones I need.
> That is, I end up with kerning for the original 'A' applied to
> 'some-other-A'.

Then you should probably also

   \killglyph{A}

since you want to be completely rid of the original A. Without that, 
you will probably end up with two \set(scaled)rawglyph{A}... commands, 
and as you've noticed all the kerns of the old A applied to the new one.

> Using the trick from csckrn.mtx:
>     \resetcommand\iftokeep#1\then{\ifnum 0<#1}
> has helped considerably but this still retains kerns for any character
> which has been renamed. In particular, if the original font includes
> a kern from 'A' to 'some-other-A' then I end up with:
>     \setkern{A}{A}{25}
> and if the glyph 'minute' has been renamed to 'prime' and there is a
> kern in the original font from 'A' to 'minute', then I end up with:
>     \setkern{A}{prime}{-147}
> This is obviously not what I want because there is no kern from
> 'some-other-A' to either 'some-other-A' or 'minute' in the original
> font. On the other hand, I cannot turn off kerning altogether because
> there are kerns for 'some-other-A' which I need to retain
> (specifically, kerns to uni200B which need to be remapped to kerns to
> 'prime' as the skew character for OMS).
> 
> I'm totally fed up and frustrated. (This is not the fault of anybody so
> please don't think I am suggesting for one second otherwise. I'm simply
> explaining why I'm asking for advice.) I'd appreciate any advice anyone
> can give me for dealing with an enormous font which contains almost (but
> not quite) all the symbols and letters needed for a maths setup as well
> as many text characters.
> 
> In particular, I would like to know if it would be better to create
> either subsetted AFMs or subsetted fonts and, 

Could be, if you're running into resource limits (number of slots in a 
base font, TeX memory when running fontinst), but chopping things up 
also introduces a certain amount of administration.

> if so, what the best way
> to do that might be. But any other suggestions would be equally welcome.
> 
> afm2pl doesn't seem to be an option because I do need characters from
> more than one font (e.g. I need to take dotlessi and dotlessj from the
> separate text fonts for maths italic) and it doesn't seem to have a
> mechanism for inserting maths-specific slot commands such as those
> needed for OMX.

Lars Hellström