Re: [Axiom-developer] Axiom and streams

[Stephen Wilson]

Camm Maguire <address@hidden> writes:

> Greetings!  This stuff sounds great, and likely can be included in the
> gcl distribution proper even though it is an ansi extension.  I think
> slime depends on gray streams, right?  That would be reason enough.
> But perhaps support should be in the form of an autoloadable module.

OK.  Thats great you support this!  

Yes, Slime requires Gray streams to the best of my knowledge.

The main implementation specific bit is to connect
'fundamental-stream, the root of the Gray stream hierarchy, into the
CL type system.  We need to have both 'cl:stream and
'cl:standard-object as superclasses, so this might need to be done at
the PCL level.

An autoloadable module would be ideal, for at that time further
extensions to the CL name space could be made (we would need to
redefine cl:close, cl:streamp, and a few others, as generic
functions).

I will see what I can come up with and hopefully provide an initial
proposal for GCL.

Thanks,
Steve


> 
> Take care,
> 
> Stephen Wilson <address@hidden> writes:
> 
> > C Y <address@hidden> writes:
> > > Are they mutually exclusive?  
> > 
> > In a sense, yes.  Take a look at the Simple streams doc and see how
> > the traditional common lisp functions are redefined to support both
> > approaches. The starting point would probably be here:
> > 
> >  http://www.franz.com/support/documentation/6.2/doc/streams.htm#cl-funs-2
> > 
> > > Perhaps it would be useful to have both available, as I can see both
> > > byte-based operations and abstract objects being of interest.
> > 
> > One of the main advantages of Simple streams is that they do not
> > require a call through a generic function to do things like read-char,
> > write-byte, etc.  Its not that Gray streams cant do these things, its
> > just that the interface is somewhat heavy.
> > 
> > > I'm guessing at the current stage the Simple Streams approach would be
> > > more useful, and later on the more abstract objects will become
> > > interesting.
> > 
> > I tend to agree.  I ported the SBCL Gray streams code over to GCL this
> > morning in the form of a user library (that is, its not redefining CL
> > functions or sub-classing CL's stream type -- its portable code similar
> > to Corman Lisps implementation).  Next will be to attempt the same
> > with SBCL's Simple stream implementation.  Once both portable versions
> > are running I can look at how they can be embedded in GCL so as to
> > take advantage of implementation specifics.  Hopefully a lot of the
> > details, pros/cons, etc, will become clear by taking that approach.  
> > 
> > 
> > Take care,
> > Steve
> > 
> > 
> > 
> > 
> 
> -- 
> Camm Maguire                                          address@hidden
> ==========================================================================
> "The earth is but one country, and mankind its citizens."  --  Baha'u'llah