re: Issue SYNTACTIC-ENVIRONMENT-ACCESS

[Kim A. Barrett <IIM@ECLA.USC.EDU>]

I think you may be missing the point of AUGMENT-ENVIRONMENT.  First, it can't
possibly introduce retroactive local bindings.  It does not side-effect the
environment argument, it returns a new environment which contains everything
in the environment argument, plus the stuff specified by the keyword arguments.
Perhaps an example (as you requested) might help.

I can't think of any really simple examples of using augment-environment off
hand.  This is a handler for the LET special form from a hypothetical code
walker.  I don't guarantee that the semantics implemented by this function
match any particular varient of Common Lisp, though I was thinking in terms of
X3J13 CL.  Also, this example is far from self-contained, containing numerous
references to helper functions which are not described here, though hopefully
they shouldn't be too confusing.

(defun walk-let (form env walk-function)
  (multiple-value-bind (body decls)
      (extract-declarations (cddr form))
    (multiple-value-bind (special-decls type ftype)
	(parse-declarations decls)
      (let ((vars ())
	    (inits ()))
	(dolist (binding (second form))
	  (cond ((symbolp binding)
		 (push binding vars)
		 (push nil inits))
		(t
		 (push (car binding) vars)
		 (push (walk-form (cadr binding) env walk-function) inits))))
	(let ((lexicals ()))
	  (dolist (var vars)
	    (or (member var special-decls)
		(variable-globally-special-p var env)
		(push var lexicals)))
	  `(let ,(mapcar #'list (nreverse vars) (nreverse inits))
	     ,@decls
	     ,@(let ((new-env (augment-environment env
						   :lexical lexicals
						   :special special-decls
						   :type type
						   :ftype ftype)))
		 (mapcar #'(lambda (form)
			     (walk-form form new-env walk-function))
			 body))))))))

Note that this example points out a possible misfeature of the current
description of AUGMENT-ENVIRONMENT.  Namely, it might be better to simply
specify all the variables being bound, along with a list of local special
declarations, and let augment-environment figure out the split between lexicals
and specials. 

I discussed the question of the extent of these things in a recent message
regarding the WITH-COMPILATION-UNIT issue.

kab
-------