Re: [PATCH] Implement local-eval, local-compile, and the-environment

[Andy Wingo]

On Tue 03 Jan 2012 11:52, Mark H Weaver <address@hidden> writes:

> +(define-record-type lexical-environment-type
> +  (make-lexical-environment wrapper names boxes others module-name)
> +  lexical-environment?
> +  (wrapper     lexenv-wrapper)
> +  (names       lexenv-names)
> +  (boxes       lexenv-boxes)
> +  (others      lexenv-others)
> +  (module-name lexenv-module-name))

Why a module-name and not a module?

> +(define-syntax-rule (box v)
> +  (case-lambda
> +   (() v)
> +   ((x) (set! v x))))

This is nice.

If you want to make a hack, you can get at the variable object here,
using program-free-variables.  It is fairly well guaranteed to work.
Dunno if it is needed, though.

> +(define-syntax-rule (box-lambda* (v ...) (other ...) e)
> +  (lambda (v ...)
> +    (let-syntax
> +     ((v (identifier-syntax-from-box v))
> +      ...
> +      (other (unsupported-binding 'other))
> +      ...)
> +      (if #t e))))

I would fix the indentation here.  What's the purpose of the (if #t e) ?

> +(define-syntax-rule (capture-environment
> +                  module-name (v ...) (b ...) (other ...))
> +  (make-lexical-environment
> +   (lambda (expression)
> +     #`(box-lambda*
> +     #,'(v ...)

Why not just (v ...) ?

> +     #,'(other ...)

Likewise.  And again, the indentation :)

> +(define-syntax-rule (identifier-syntax-from-box b)
> +  (let ((trans (identifier-syntax
> +             (id          (b))
> +             ((set! id x) (b x)))))
> +    (set-procedure-property! trans
> +                          'identifier-syntax-box
> +                          (syntax-object-of b))
> +    trans))

Ew, identifier-syntax as a value :)  But OK.

> +;; XXX The returned syntax object includes an anti-mark
> +;;     Is there a good way to avoid this?
> +(define-syntax syntax-object-of
> +  (lambda (form)
> +    (syntax-case form ()
> +      ((_ x) #`(quote #,(datum->syntax #'x #'x))))))

This has a bad smell.  Anything that is outside psyntax should not have
to think about marks.  Perhaps syntax-local-value could serve your
purpose?

> +    (global-extend 'core 'the-environment

This one is really nasty, and I'd like to avoid it if possible.  Are
there some short primitives that psyntax could export that would make it
possible to implement `the-environment' in a module?

It seems to me that some primitive to return a list of all syntax
objects visible at a given point of a program, together with a primitive
to retrieve the module that corresponds to a syntax object, could be
sufficient.

WDYT?

>  static void
> +test_scm_local_eval ()

Thanks for going all the way with documentation and test cases.  This is
great.

Cheers,

Andy
-- 
http://wingolog.org/