Robert Dodier writes:
> Hello,
>
> I've taken another look at src/cl-info.lisp and tried tracing through
> the cl-info functions but I've come up empty-handed. I'll keep trying.
You already discovered reason for describe misbehaviour under clisp
a while ago -- DOS end-of-lines in *.texi files (Input.es.texi and
Rules.es.texi). I committed the fix for 5.9.3 and HEAD
> Something that has occurred to me again is that cl-info is really
> doing far more work than it needs to.
> If you trace cl-info::show-info and/or cl-info::string-match I think
> you'll agree.
>
> It seems like it would simpler (and therefore easier to debug -- an important
> consideration at the moment) and faster to build a lookup table,
> such that the keys of that table comprise the set of topics
> that can be searched, with the value associated with each key
> being something like (info-filename byte-offset).
>
> I built an implementation of this idea some time ago but put it
> aside because it used Unix command line tools to build the lookup table.
> But I'm tempted to redo it now, as an in-memory hash table,
> in hopes of getting to the bottom of this "can't find node" business.
>
> Getting describe to work is the most important consideration,
> but the time spent waiting for describe to return something is
> nonnegligible; it takes several seconds on my development box.
> That's long enough to make people wonder what's going on.
> It should take no more than 1 second to get help text or to return
> to the input prompt if no text is found.
>
> Comments?
>
Very good idea! To have some pre-build index table instead of
searching whole *.info each time again and again is the only
right way to go. This becomes more and more important since
volume of Maxima documentation grows. All known to me help
systems has some kind of inex facilities -- Windows
help file, Mathematica on-line documentation system etc.
--
Vadim V. Zhytnikov
<vvzhy at mail.ru>
<vvzhy at netorn.ru>