README

README (July 12, 1996)

This package contains the complete Scheme implementation of the
reflective language Black.  For more details on Black, see [1].
The file black.tar.gz (or .Z) contains the following files:

  README	: this file
  env.scm	: environment manipulation functions
  black.scm	: the main body of Black
  black-with-delta.scm
		: same as black.scm except that it supports the
		  delta abstraction introduced in Blond[2].
  int.scm	: user observable metalevel interpreter
  break.blk	: a sample program which install two new special
		  forms: break and inspect.
  compare	: a simple shell script which compares black.scm
		  with int.scm (except for the insertion of meta-apply)

Since the implementation uses no special constructs except for
cons-stream and pp, it should run on almost any Scheme by typing:

(define scheme-apply apply)
(load "env.scm")
(load "black.scm") or (load "black-with-delta.scm")
(black)

Followings are the Scheme system on which we have tested our
implementation together with some specific definitions required:

Chez Scheme Version 5.0a:
	(define head car)
	(define (tail x) (force (cdr x)))
	(define-syntax cons-stream
	  (syntax-rules ()
	    ((cons-stream a b) (cons a (delay b)))))

	(define scheme-apply apply)
	(load "env.scm")
	(load "black.scm") or (load "black-with-delta.scm")
	(black)

	To compile, you will have to include the definition of
	cons-stream, so that it is properly compiled as a macro.

MIT-Scheme Version 7.3.0:
	(define scheme-apply apply)
	(load "env.scm")
	(load "black.scm") or (load "black-with-delta.scm")
	(black)

Gambit Scheme Version 2.2:
	(define head car)
	(define (tail x) (force (cdr x)))
	(##define-macro (cons-stream a b)
	  `(cons ,a (delay ,b)))

	(define scheme-apply apply)
	(load "env.scm")
	(load "black.scm") or (load "black-with-delta.scm")
	(black)

	To compile, you will have to include the definition of
	cons-stream, so that it is properly compiled as a macro.
	You will also have to declare:

	(##declare (r4rs-scheme))

	so that delay is compiled as a special form.

SCM Version 4e1:
	(define head car)
	(define (tail x) (force (cdr x)))
	(define cons-stream
	  (procedure->macro
	     (lambda (x env)
		(let ((a (cadr x))
		      (b (caddr x)))
		  `(cons ,a (delay ,b))))))

	(define scheme-apply apply)
	(load "env.scm")
	(load "black.scm") or (load "black-with-delta.scm")
	(black)

elk:
	(define head car)
	(define (tail x) (force (cdr x)))
	(define-macro (cons-stream a b)
	  `(cons ,a (delay ,b)))
	(autoload 'pp 'pp)

	(define scheme-apply apply)
	(load "env.scm")
	(load "black.scm") or (load "black-with-delta.scm")
	(black)

	It seems that elk does not handle mutual tail recursion correctly.
	Thus, stack overflow will eventually occur.

To quit Black, type an interrupt key.  Black has no command to quit
the system.  (exit 0) will merely exit the current level.

The extensions that are not described in the paper are as follows:

1. Special forms not described in the paper (cond, let, etc.) are
   supported.

2. Special forms delay, force, cons-stream, and primitive-EM are
   supported, so that the implementation is meta-circular.  You can
   execute Black on top of Black by typing the followings in Black:

   (define head car)
   (define (tail x) (force (cdr x)))
   (load "env.scm")
   (load "black.scm") or (load "black-with-delta.scm")
   (black)

3. (if <bool> <then>) without <else> is supported.

4. The error "Wrong number of arguments" (for lambda closures) is
   supported.

When errors such as "Unbound variables" and "Not a Function" occur,
the current level is properly terminated, and the control moves to the
upper level.  However, some errors cause the Black system to stop.
Such errors include the ones occured in the application of primitives,
e.g., (car 0).  To handle these errors, we have to check if the
application of primitives leads to an error every time a primitive is
applied.  Because it would make the apply procedure much longer and
tedious, we do not remedy this but leave it as an alternative way to
quit the Black system.

For questions, comments, etc., contact asai@is.s.u-tokyo.ac.jp.

References

[1] Kenichi Asai, Satoshi Matsuoka, and Akinori Yonezawa
    "Duplication and Partial Evaluation
     --- For a Better Understanding of Reflective Languages ---"
    Lisp and Symbolic Computation, 9, pp.203-241, Kluwer Academic
    Publishers, Boston (1996).

[2] Olivier Danvy and Karoline Malmkjaer
    "Intensions and Extensions in a Reflective Tower"
    Lisp and Functional Programming, pp.327-341 (1988).