Re: [Groff] recursive subdirectory search in GROFF_TMAC_PATH

[Steffen Nurpmeso]

Werner LEMBERG <address@hidden> wrote:
 |> And then think about the special path "//" which' behaviour is
 |> implementation defined as well according to POSIX.
 |
 |Do you actually know the kpathsea library?  Without this library, TeX
 |file management would be a nightmare.  And kpathsea works on all major
 |platforms and is very mature.

Not really, no.  I didn't use TeX since.. 2002?  I've searched and
read a bit once you brought it up, however.  And / but i also had,
until that day (2002-08-17+x), my own macro package and only needed

  address@hidden ]$ git grep -i 'inputifexists{' master|sed -e '/doc/d'
  ..               \qInputIfExists{#1.ddf}% ddf = dimension-definition-file
  ..       \qInputIfExists{#1.fdf}% fdf = font-definition-file
  ..               \qInputIfExists{#1.lrf}% lrf = language-ressource-file
  ..               \qInputIfExists{\jobname .aux}% labels, write .toc,..
  ..       #1\makeatletter\qInputIfExists{\jobname .toc}\makeatother#2%
  ..       \qInputIfExists{\jobname .ind}%
  ..       \qInputIfExists{\jobname .bib}%
  ..\qInputIfExists{epsf.tex}%
  ..\qInputIfExists{colordvi.tex}%

epsf.tex and colordvi.tex (plus, of course, Knuth's base file, but
i have forgotten the name of it -- tex.tex?), that is really quick.
And then the REAL boooooster:

 % 00-05-31: new scheme to avoid waste. now a font is init only if it's used.
 % 2Compare (Cyrix 166+, 49MB, Linux 2.2.13-12, X 3.3.5):
 % | OLD                               | NEW                                   |
 % |-----------------------------------|---------------------------------------|
 % | 1178 strings out of 13013         | 1413 strings out of 13013             |
 % | 13106 string characters of 122154 | 17026 string characters of 122154     |
 % | 35574 words of memory of 263001   | 54220 words of memory of 263001       |
 % | 2086 multiletter of 10000+0       | 2321 multiletter of 10000+0           |
 % | 80647 font info for 276 fonts     | 20674 words of font info for 70 fonts |
 %   (out of 400000 for 1000)
 % and 38s of 4000, instead of 37s. Note however that the additional use of the
 % (test) font-package, 'ec.fdf', let to "200000 font info"! Today it is 28689
 ..
                       (TeX stucked for one to two seconds before, but now it
 % runs straight and __the complete test is faster than the loading of the fonts
 % was before__).

I.e., lazy creating fonts à la

  address@hidden@hugestrm{%
  address@hidden@address@hidden@address@hidden address@hidden@rm}%

That actually makes me wonder when exactly i became a Eunuch,
and even more, why.  I think a deep inspection of that topic will
turn out something positive.

 |>|kpathsea library for groff's file management.  In particular, it
 |>|would allow for multiple font configuration files.
 |>
 |> This is all too hardwired and restricted, and then scattered at the
 |> same time.  Nothing a normal person can deal with easily.
 |
 |Well, if I want to add a font to groff, I would prefer to *not* modify
 |the system-wide font config file but a local one...

I think a normal user would prefer something like

  .font tahoma.ttf

and get-set-go.  For other use cases i would have no problem with
a single global .ini-style configuration file and a per-user one
with the same syntax.  And then, e.g., a "[eqn]" section instead
of an eqnrc somewhere deep in an intransparent machinery.  Ok,
that is possibly a no-go given the content of eqnrc, but why not,
likewise hypothetic

  [startup-files]
  eqn = INSTALLPATH/eqnrc

So then the user looks /etc/XY.rc and then ~/.XY.rc and knows,
especially with some hinting comments.
And yes i have read the second paragraph of eqn(1).  But with that
no path search at all.  And that is the best of all path searches,
in my humble opinion.
Regarding my own efforts this is a long road to go, however.
Ciao, and a nice rest weekend to you and all lurkers,

--steffen