[Axiom-developer] Re: lisp lisp noweb speed issue

[C Y]

--- address@hidden wrote:

> > is this something you think can be used in Axiom?
> 
> Yes, I do. This debate has raged before on this list but I
> believe that we need a lisp version of noweb for several reasons.

What would you consider as the minimal features required to include
cl-notangle (or cl-noweb as the case may be) in Axiom?  I would like to
know what the key minimal requirements are so I know what to shoot for
first.

> 1) the )compile command needs to understand pamphlets directly

Agreed.

> 2) we need tighter integration of the interpreter and documentation

Not sure quite how that works but it sounds good.

> 3) the build process will eventually be lisp-only (controversy)

This is my main reason for wanting notangle in lisp.  Is the eventual
goal of building in lisp controversial?  My understanding is
all-in-lisp is regarded as the best (most portable) way to handle this
process, but the tools were/are lacking - perhaps with notangle in lisp
(and eventually noweb) we can achieve this.  The only catch is that we
can't be lisp only if we need to build GCL first, but at least once we
have a running lisp everything else can be lisp.  The news that we are
close to replacing old boot with new boot (which is simpler to get
working in ANSI lisp) and an sbcl port are two major pieces of the
puzzle, so I'm trying to add notangle and an asdf based loading system
as the other missing pieces.  

> 4) we need to be able to experiment with pamphlet formats for new
> ideas

Yes - I would like to see the default Axiom pamphlet output a little
different (fancier) than it is now (now there's controversy ;-)

> 5) noweb slows considerably as the files get larger, apparently due
> to
>    the shell pipe of data from process to process.
>    (my work pamphlet is now 6M lines, (4k pages of doc, 67k lines of
> lisp))

Tim, out of curosity how does this version (suboptimal I know) compare
with noweb on that really large file?  I'm sure the read commands will
be a killer but it shouldn't be piping data around, at least.  (Dunno
if it works in GCL, I've only tested it on sbcl).

Cheers,
CY


 
____________________________________________________________________________________
Sucker-punch spam with award-winning protection. 
Try the free Yahoo! Mail Beta.
http://advision.webevents.yahoo.com/mailbeta/features_spam.html