Re: SV: strange macro "or" behavior in CormanLisp2.01

At 12:39 PM 3/5/03 +0100, Pavel Grozman wrote:
>CormanLisp tries to evaluate constant subexpressions.
>Without constants it works as it should:
>
>(setq x 0)
>0
>(or 13 (/ 1 x))
>13
>

More precisely, Corman is not doing evaluation.  Corman always
compiles everything, even things you type into it interactively.

So the Corman compiler is walking over the form (/ 1 0), noticing that
it is composed completely of constant forms, and trying
to collapse the expression into a single value, much as it would
collapse (+ 1 1)  at compile time into '2'.

But it doesn't currently test whether its attempt to collapse the
expression succeeds or fails; if the expression it is collapsing
causes an error, that error is allowed to be signalled at compile
time.

As Roger and I have discussed elsewhere, this is not correct
ANSI Common Lisp semantics.  The compiler should produce
code which will cause the error condition to be raised when the
function is run, and it should probably generate a warning about
the offending expression.

> > I've noticed the following when I was doing a excercise in
> > Paul Graham's ANSI Common Lisp:
> >
> > [Corman Common Lisp 2.01]
> > (or 13 (/ 1 0))
> > ;;; An error occurred in function /:
> > ;;; Error: Divide by zero error: 1 / 0
> > ----snip----
> >
> > [Allegro Common Lisp 6.2]
> > CL-USER(1): (or 13 (/ 1 0))
> > 13
> > CL-USER(2):

Note that the compiler is not currently collapsing expressions
such as

(or 13 (/ 1 0))

to

'13'

but it legitimately could, just as it could (but does not now) collapse
something like

  (if t (foobar x) (foobaz x))

into a call to FOOBAR.

(Allegro does collapse this expression, to the point of not even
warning you that FOOBAZ is an undefined function, while warning
you that FOOBAR is undefined!)


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Get 128 Bit SSL Encryption!
http://us.click.yahoo.com/LIgTpC/vN2EAA/xGHJAA/SyjtlB/TM
---------------------------------------------------------------------~->

To unsubscribe from this group, send an email to:
cormanlisp-unsubscribe@xxxxxxxxxxxxxxx

 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/