Message-Id: <9303011608.AA18573@sc.zib-berlin.dbp.de>
Reply-To: Mailing list for the LaTeX3 project <LATEX-L@vm.urz.Uni-Heidelberg.de>
In-Reply-To:  <01GV5N6TML82GW2E6N@MATH.AMS.ORG>
Date:         Mon, 1 Mar 93 17:08:25 CET
From: Michael Downes <MJD@MATH.AMS.ORG>
Sender: Mailing list for the LaTeX3 project <LATEX-L@vm.urz.Uni-Heidelberg.de>
To: Multiple Recipients of <LATEX-L@vm.urz.Uni-Heidelberg.de>
Subject:      Re: subdocuments
Status: R


> Incidentally, Nelson's example can be done with some clever
> coding. This would be useful for all sorts of applications and in
> fact I have a 2.09 solution half-finished (it should be done in a
> couple of weeks). It can be done almost transparently for any
> article-class style and a few unintrusive additions to a document
> style fix many of the deficiencies.

I would be interested to see your solution. How do you handle the
full-volume subject index? In a similar project in the past I settled
on a processing method of separate LaTeX runs for each article in the
proceedings, with \includeonly{articlename} in the driver file for one
article at a time; it then seemed natural to generate the index in two
steps through the .aux file, in the same way as the table of contents
is currently done (if you don't do something, standard LaTeX has each
article overwriting the .idx data from the previous article, leaving
you with an .idx containing only the index terms from the last one
processed).  Other approaches such as appending the .idx file for each
run to a master .idx file could be done but seemed less attractive.
Either you have to call a (system-dependent) operating system command
externally after each LaTeX run (which is maybe OK, because it's
usually a good idea anyway to have a system-dependent .bat or .com or
other batch file to make running the entire volume easy; but then for
consistency the .toc should probably be done the same way), or you
have to make TeX read and write each line of the .idx files to append
them.

If articles are not processed in separate LaTeX runs, but instead
processed together with each article enclosed in a group, then any
\global assignments in an article are time bombs that can explode
unexpectedly in some later article in the same run.  There might also
arise `TeX capacity exceeded' errors for save stack that would not
occur when running the articles separately.  On the other hand,
running each article separately requires careful handling of
\newcounter declarations in individual articles (typically from
\newtheorem), and of the indexing.

Michael Downes                              mjd@math.ams.org (Internet)