Re: [Chicken-users] types.db incorrect

[Jörg F . Wittenberger]

On Sep 6 2011, Felix wrote:

> > What I would really have a use for was,
> > if
> > the assertions from types.db or whatever I pass with additional -types
> > parameters would be turned into runtime assertions (before call and
> > before
> > return) if, and only if, I pass a certain switch to the compiler.
>
> Can you give an example of this? (a piece of code and how you would
> like it to be transformed in the mode you are describing)

Easy enough:

(define (my-work-load a b) (values a 1 b))

And in types.db (that would be one private to the corresponding project):

(my-work-load (#(procedure #:clean #:enforce) my-work-load (fixnum fixnum) 
fixnum fixnum fixnum))

Now with the magic turned on the could expand to

(define (my-work-load a b)
; asserts come in by magic
(assert (fixnum? a))
(assert (fixnum b))
(receive (a b c)

  (values a 1 b)  ;  This is the only code I originally wrote!

 (assert (fixnum? a))
 (assert (fixnum? b))
 (assert (fixnum? c))
 (values a b c)))

So, the code would be wrapped by "contract verification" wrt. the types.db

The same could apply to the calling site.  With "types-enforces" instead
of simply compiling a call to my-work-load something like this could be
expanded:

(receive (a b c)
  (my-work-load
   (begin (assert (fixnum? p1)) p1)
   (begin (assert (fixnum? p2)) p2))
  (assert (fixnum? a))
  (assert (fixnum? b))
  (assert (fixnum? c))
  (values a b c))

Thus I could switch either site into debug mode.  When in this mode
I could be sure that successfully invoking a procedure means all parameters
have been checked.

/Jerry
........