Re: Re: Re: [Chicken-users] performance issue in xml-rpc

[Daishi Kato]

On 12/15/06, felix winkelmann <address@hidden> wrote:
> On 12/15/06, Daishi Kato <address@hidden> wrote:
> >
> > I'm still waiting for a comment from felix about read-line not using
> > read-string.
>
> Sorry, I know I'm too slow.
>
> read-line can't directly use read-string unless I introduce some buffering,
> so this is not exactly trivial to do in the context of chicken's I/O system.
> Please give some time.

Sure.
Are you going to add unread-string in port class? Or exposing the buffer?

> >
> > In the meanwhile, I'm fixing the ssax egg to use ##sys#read-char-0
> > like the following.
> >
> > Index: ssax-core.scm
> > ===================================================================
> > --- ssax-core.scm       (revision 2646)
> > +++ ssax-core.scm       (working copy)
> > @@ -804,6 +804,10 @@
> >     (gambitize (eof-object? port))
> >     ;(gambitize (string-append a b))
> >     )
> > + (chicken
> > +   (define-macro (read-char port) `(##sys#read-char-0 ,port))
> > +   (define-macro (peek-char port) `(##sys#peek-char-0 ,port))
> > +   )
> >   (else #t))
> >
> >
> > How's this?
> >
>
> Does this improve performace? And: Can you assure thar the arguments
> are really ports? (otherwise you'll get crashes).

The performance is improved just a little bit, but better than nothing.
The arguments are always ports, and code for Gambit has the same syntax.
And,I got no error with my simple xml-rpc test.
I can commit the patch if you or anyone responsible wants.

Daishi