syntax-local-lift-require is permitted in a 'module-begin context in Racket 7, but does not actually work

[lexi-lambda]

The following program fails with an error:

#lang racket/base

(require (for-syntax racket/base
                     syntax/strip-context)
         syntax/parse/define)

(provide #%module-begin)

(define-syntax-parser #%module-begin
  [(_ . body)
   (syntax-local-lift-require
    'racket
    (datum->syntax #f (cons '#%module-begin (strip-context #'body))))])

(module* test (submod "..")
  (match-define _ #f))

expand: namespace mismatch; cannot locate module instance
  module: #<module-path-index:racket/match/match>
  use phase: 0
  definition phase: 0
  for identifier: #<syntax:bad.rkt:16:3 match-define>

It fails on Racket 6.12 as well, but much earlier and with a different (less confusing) error:

syntax-local-lift-requires: could not find target context
  context...:
   standard-module-name-resolver

I would actually very much appreciate it if syntax-local-lift-require was made to work properly in a 'module-begin context, but if that doesn’t make any sense, it should signal an error like it does in Racket 6.12.

[mflatt]

I've pushed a repair to restore v6.12 behavior.

So far, trying to make syntax-local-lift-require work in a 'module-begin context doesn't sound like a good idea. For the purposes I can imagine, a trampoline to a macro within #%plain-module-begin seems better. However, you may have an example that I haven't imagined, so I'm open to the possibility that syntax-local-lift-require should work in a 'module-begin context.

[lexi-lambda]

I think you’re right that I can make a trampoline work in my case with a little extra work, so I’ll give it a try, and if it works, I’ll close this issue as resolved.

For context, what I’m doing is a bit unorthodox. I’m interested in essentially expanding the body of one module in a module with a different module language, and I want to be able to introduce all the bindings of the original module language before expanding its #%module-begin. However, what I can probably do is expand to my own #%plain-module-begin to get a 'module expansion context, expand the other module’s contents in a 'module-begin context, and manually extract the subforms from the resulting #%plain-module-begin form to splice them into the form I’ve already expanded to.

[mflatt]

In case you haven't looked at it already, I think the "hyper-literate" package does something like that, but with local-expand on a synthesized submodule (if I remember correctly).

[lexi-lambda]

Yes, it looks like I got a trampoline that works for my needs. I think I can also apply this technique to fix something in Hackett, but that’s for another day. Thanks for your help. :)