“invalid context within quasiquote” when using unquote-splicing in macro body

[aartaka]

So I’m writing an STklos port of SRFI 279. I mostly copied the existing (Chibi) impelementation where it uses SRFI 253 define-checked to define most procedures. And returns a quasiquote-constructed lists. However, such style results in a file that’s basically unparseable for STklos, emitting a bunch of errors, one per such define-checked:

compiler-error: invalid context within quasiquote. Call was (define-checked (vector-properties (object vector?)) `((vector-length ,(vector-length object)) ,@(map (lambda (Si Se) (list Si Se)) (iota (vector-length object)) (vector->list object))))

I was able to isolate it to the use of unquote-splicing / ,@ and hyg:untag-quasiquote in mbe.stk where the error text appears. However, I cannot understand the purpose of this procedure—why strip off quasiquotes if these are valid procedure bodies?

Example code to test on:

;; Depends on SRFI 253, but can be replaced with any macro?
(define-checked (vector-properties (object vector?))
  `((vector-length ,(vector-length object))
    ,@(map (lambda (i e) (list i e))
           ;; Depends on SRFI 1
           (iota (vector-length object))
           (vector->list object))))

@egallesio?

[jpellegrini]

@aartaka I'm not sure I understand what happened. The file mbe.stk has the old "Macros By Example" implementation technique for syntax-rules. The tag/untag procedures are part of it (you can see why they exist by reading the comments in the file).

But what do you mean by strip all quasiquotes? Are we doing that in that file?

[aartaka]

@jpellegrini

But what do you mean by strip all quasiquotes? Are we doing that in that file?

I literally have no idea what happens in that file 😵 But it seems to parse the macro body to strip off some data. And, apparently, it doesn’t like unquote-splicing, erroring on encountering it. Almost unconditionally. Which feels really wrong. But I’m not sure who to attribute the wrong to—implementation or myself.

[jpellegrini]

The file tags the S-expressions... look at the comments:

;; hyg:tag uses gensym to rename variables (actually, tag the ones which will be
;; renamed).
;;
;;    e is the form
;;    kk are the keywords
;;    al is an assoc list
;;
;; returns a list with
;; CAR: the original form, rewritten so only the keywords stay with the same names,
;;      the rest is substituted.
;; CDR: is an assoc list that maps the generated symbols (G...) with their original
;;      names.
;;
;; the list al is appended to the output CDR, so the assoc list may be build
;; gradually.
;;
;; (hyg:tag '(f a b c) '(a c)          '() '...) => ((G984 a G985 c) (G985 . b)
;;                                                   (G984 . f))
;; (hyg:tag 1          '(fun x fun oh) '() '...) => (1)
;; (hyg:tag '(f a b c) '(a c)          '((G-something . y)) '...)
;;                                               => ((G1333 a G1334 c)
;;                                                   (G1334 . b)
;;                                                   (G1333 . f)
;;                                                   (G-something . y))

;;untagging
;;
;; e is the TAGGED expression, as produced by hyg:tag
;; al is the assoc list of tags -- se hyg:tag
;; tmps is a list of gensymed symbols NOT to be substituted back.
;;
;; (hyg:untag '(G1333 a G1334 c)
;;            '((G1334 . b) (G1333 . f) (G-something . y))
;;            '())                                             => (f a b c)
;; (hyg:untag '(G1333 a G1334 c)
;;            '((G1334 . b) (G1333 . f) (G-something . y))
;;            '(G1334))                                        => (f a G1334 c)

This is related to hygiene.

But untagging should be able to handle anything that tag produces, so we obviously have a bug there.