RE: [Axiom-mail] Re: noweb

[Page, Bill]

Ralf, 

On Wednesday, August 09, 2006 9:17 AM you wrote:
> 
> > | ... If we distribute mnt/linux/src/algebra at all, then
> > | it should probably contain the original *.spad.pamphlet
> > | files, not the notangle'd files.
> 
> I was first thinking that we need that .spad files. But 
> hyperdoc does in most cases not find the corresponding
> .spad file anyway.

That is true. The same (incorrect) result is returned by
the ')sh' commmand and by 'asq'.

> So I think we should distribute .spad.pamphlet and try to
> fix the links. The fix might, however, take some time. Is
> there somebody knowledgable enough to look into such a fix?
> 

Good question. I do know that the solution must involve the
Axiom database (.daase files) in some way since that is where
the structural information about domains etc. is stored.
But we need also to extend the database with a mapping of
module names into *.pamphlet and <<chunk names>>. This is
problematic since by the time the database files are built
the information about the pamphlet names is gone.

So I supose the first step would have to be to extend the
compile command to include <<chunk names>> as I suggested
before

  )compile xxx.spad.pamphlet )chunk <<yyy>>

and also to capture both of these names in the NRLIB files
for the module. Then we can extend the database to load this
additional information. And the build would have to be changed
to use the above form of the compile command.

As Clifford Yap and Tim Daly have said a few times on this
list, that would mean providing a replacement for notangle,
built-in to the Axiom compiler (and also the interface to
Aldor). To do this we would need to write some Boot code (or
lisp as a last resort ;). Or I suppose that it would also be
possible just to call the noweb notangle as a system command
from within the Axiom code. The last option might be less
portable but easier to implement.

Regards,
Bill Page.