Hi folks! By a happy coincidence I spotted "extended include" rolling
by in a log of mail which gets archived without me ever getting to it
these days, which includes TeX mail. I've been submerged in other
nonsense for ages, that is why Robin hasn't heard from me and no more
work has been done on the newclude package that he mentioned.
Includex was a hack and I began to write newclude as a more
respectable user- and developer-friendly piece of code once I thought
I had solved the stickier problems. It turned out I couldn't quite
make the grade, and I never ended up with a solid solution. I was
working out two entirely different implementations, one that was
relatively complicated and used multiple aux files (static stream
assignment was simple, it was the dynamic assignment that got hairy),
and one that was simpler and used but a single aux file. There were
certain functional limitations on each solution, so I worked on both,
though the user interface was the same. As I remember, I ran up
against two difficulties, though I don't remember how they were
distributed between the two methods. One was the page break problem
that's been mentioned in this thread. It sounds like Frank has a far
stronger grasp on what to do about that than I ever had. The second
problem occurred when a user aborted the LaTeX run under certain
conditions, leaving an unclosed group in the aux file, which would
break the next LaTeX run.
I agree with Frank that it's an important question whether doing this
work in LaTeX is worth it. With the universality of Perl and other
text processors these days, the portability argument is weakened.
I suggest I try to publish to this group the code, documentation, and
notes I have lying around on this subject. I'll bring includex up to
date with Robin's fixes. He's right that despite it being such a hack
it still does things newclude doesn't do. And I'll release newclude;
it's functional, clean, and documented, I just never got it robust
enough to want to release it.
Best wishes to all