Leo Butler <l_butler at users.sourceforge.net> writes:
> Maybe we should re-think the info system. IIRC, there have been at
> least 3 significant revisions in the last 3-4 years, none of which
> were incorporated into master. I think that the stumbling block in
> each case was testing on different combinations of os+lisps.
Indeed. I don't know what happened to the previous efforts, but I think
mine foundered on lack of Windows testing. At the time I was writing it,
it seems that pretty much *no-one* was actually able to build Maxima on
Windows and I agree that it would have been unwise to have messed up the
build further. I hope that I'll be able to address that lack of testing
with my code this weekend.
> Here is a thought: Emacs' info-mode has its own info parser that
> assembles the information needed. Since maxima-index.lisp is packaged
> with Maxima, some small amount of elisp hacking would let us use
> info-mode's tools to create maxima-index.lisp.
>
> The advantage of using Emacs' info-mode is that the Emacs' developers
> have a significant vested interest in making info work and fixing
> problems (encoding, changes in texinfo, cross-platform issues, etc.)
> promptly.
>
> My suggestion would be to replace the existing Perl build_index.pl
> with a bit of elisp code to run the relevant bits of info-mode and
> export the data structure in a format suitable for maxima-index.lisp.
One fundamental problem that my patch set tried to solve was that the
different lisps genuinely disagree on what it means to jump to "position
N" in a variable character width stream (byte N? character N? something
else??). The beauty of using these very lisps to generate the positions
is that we don't have to care about what "N" means, because it's right
by definition. I don't believe that getting elisp (or anything other
than CL) to generate offsets is going to be able to get around that
problem.
> Users of Emacs might want to chime in on how good info-mode is in
> dealing with Maxima's info files. In my experience, barring minor
> hiccups, under debian linux info-mode has weathered the changeover in
> texinfo very well.
That is true.
> Finally, I am not proposing to discard the work(s) done on CL info
> parsing. These could be incorporated as share packages that extend
> Maxima.
Gaargh! That last paragraph doesn't seem like a good idea at all - I'd
much prefer to ditch the code entirely than let it bit-rot, part of the
distribution but never used. Maybe I misunderstood you?
Rupert
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 315 bytes
Desc: not available
URL: <http://www.math.utexas.edu/pipermail/maxima/attachments/20131205/673217ca/attachment.pgp>;