Re: [Kawa-commonlisp-dev] [GSoC] Destructuring bind

[Per Bothner]

On 07/31/2012 03:32 PM, Charles Turner wrote:
> First buglet I'm not sure how to fix!
>
> DEFVAR appears to be broken. The spec  says it should establish it's
> name argument as a special variable. But it doesn't seem to:
>
> (defvar *avar* 10)
> (defun readit () (display *avar*))
> (let ((*avar* 15))
>    (readit))
> ;-> 10
>
> Whereas I think it should output 15. I don't see anything special
> being done in defvar.java. My question then is, is there a flag or
> something similar I can set in Declaration to get this behavior?

defvar is fine, I think.  The problem is let.
It translates to a LetExp.  However, in this case we want it
to translate to a FluidLetExp.

The complication is that in CommonLisp a single let may bind
some variables lexically  (normal case) and some fluidly (if
declared special).  We could merge the functionality of
FluidLetExp into LetExp, handling different variables
differently, but that's a fairly major change.  One can
simulate the effect this way - assume:

(declare (special v2 v4))
(let ((l1 e1)
      (v2 e2)
      (l3 e3)
      (v4 e4))
  body)

as:
(let ((l1 e1)
      (tmp2 e2)
      (l3 e3)
      (tmp4 e4))
   (fluid-let ((v2 tmp2)
               (v4 tmp4))
      body))

(Note you shouldn't need to actually generate symbols for tmp2 or tmp4;
just generate Declarations with a null name.)
--
        --Per Bothner
address@hidden   http://per.bothner.com/