Re: [Groff] Re: Adding styles to DESC

[Werner LEMBERG]

> This is exactly what I'm hoping can be accomplished somehow with
> groff, but without major changes to the way groff functions
> presently.  And without changing the way font/devps is currently set
> up.  I worry that there are groff users who may have already rolled
> their own solutions to the family/weight/shape problem by working
> within the constraints groff currently imposes, and that any change
> to groff may break their setup.

There are nice tricks available.  Consider this:

  .fp \n[.fp] TimesRoman TR
  .fp \n[.fp] TimesBold  TB
  .
  .fp 0 dummy TR
  .ft 0
  .
  .sty \n[.fp] Roman
  .
  .fam Times
  .ft Roman                   \" -> TimesRoman = TR

LaTeX is doing something similar in its font definition files: The
external names are completely separated from the internal ones.

> > As trivial as the font selection mechanism in groff seems to be,
> > it can be tricky to catch errors, providing meaningful defaults
> > and warning messages.
> 
> Indeed.  The major issue, as I see it is: what constitutes a
> "meaningful default"?  If a user calls .fam <m> when style <n> is
> active, but fam <m> has no style <n>, how should groff behave?  By
> selecting a (hypothetical) first "style" that does exist for family
> <m>, or by reverting to a something like the dummy font?

You have to decide on the priority of the font selection parameters.
If `family' is stronger than `style', then any style of family <m>
will be a proper replacement if style <n> isn't available.  You might
register a `fallback font' for a given family which is always used if
the user selects a nonexisting font shape.

In LaTeX you can provide substitutes for nonexisting font shapes in
the font definition files.  Example: If the user asks for
Helvetica/Medium/Italic, which doesn't exist, he gets
Helvetica/Medium/Slanted because of a corresponding substitution
entry.

> > What do you think about adding two conditional tests like this:
> > 
> >   .if F <font>   \" true if <font> can be found by troff --
> >                  \" the font isn't loaded
> >                  \" if <font> is a style, and the style is registered,
> >                  \" and <family> is the current family,
> >                  \" test whether concatenation of <family> and <font>
> >                  \" exists
> 
> I like this a lot.  But I'm thinking that what's also needed is a
> string-valued register containing the name of the current style,
> [...]

Good idea.  This should be rather easy to implement, I think.  Stay
tuned.

> >   .if S <style>  \" true if <style> has been registered, either via
> >                  \" the DESC file or with .sty
> 
> Not sure I understand the need for this, but then I'm a bit
> brain-fried today. :)

You need this to check whether e.g.

  .ft foo

selects a real font `foo' or a style `foo'.  In the former case, the
font family is ignored which must be handled as a special case.


    Werner