Re: [Chicken-users] Runtime arity?

[Kon Lovett]

On Jan 31, 2008, at 5:48 AM, felix winkelmann wrote:

> On Jan 29, 2008 6:25 PM, Kon Lovett <address@hidden> wrote:
> > Yeah, it seems we need a database of the built-in's "procedure-
> > information". Oh well, one more todo.
>
> Is it really necessary to allow accessing the lambda-lists of  
> primitives?
> The lambda info object is a string - parsing it all the time to figure
> out calling conventions is ugly at best and incomplete as well, since
> optional and keyword arguments are not visible (only that optional
> or rest arguments are at all available). Lambda-info is just a  
> debugging
> aid, not a reflection mechanism. In the interest of keeping code  
> and heap
> sizes small I recommend not extending this stuff.

Forgot that lambda-info is a string.

Having accurate arity would be nice for more sophisticated parameter  
checking. When taking procedure params verifying the correct param  
count is an extra level of warm-fuzzy. It keeps the error detection  
local, rather than somewhen & somewhere else in the computation, far  
removed from the error site.

I would leave accurate "lambda info" as a TRAC ticket. As you state,  
even for user lambdas, the DSSSL info isn't saved.

Leaves open the possibility of a "lambda creation hook" that can be  
used by some extension to do all sorts of wonderful things.

> cheers,
> felix

Best Wishes,
Kon