[racket-dev] single-instantiation trick: kludgy workaround for planet packages?

I seem to have mangled my response right before sending it. Here it is 
again, hopefully straightened out.

On 10/05/2010 03:59 PM, John Clements wrote:
> A couple of weeks ago, you showed me the trick that rackunit uses to allow single-instantiation of a modules.  I'm now trying to do this
> for a planet package, and it looks like I have to change the required module from being a relative to being an absolute path.  That is:
>
> (require (prefix-in drlink: private/drracket-link))

I assume (hope) you meant
   (require (prefix-in drlink: "private/drracket-link.rkt"))

for the relative require.

> must become
>
> (require (prefix-in drlink: (planet "drracket-link.rkt" ("clements" "rsound.plt") "private")))
>
> This works, but seems extremely fragile.  Does it make sense to you that this would be required?  Without it, I get the error below:

 From what I recall, the call to namespace-module-attach must use the 
correct absolute module path, but the require can be relative. I 
confirmed that that's how the last PLaneT version of schemeunit that 
included the tool (version (2 12)) did it.

It looks like you're using a PLaneT development link. Perhaps that's the 
source of the problem?

How about dropping the following code into drracket-link.rkt so it 
prints out its (absolute) resolved module path when it gets required 
from the tool.

   (eprintf "drracket-link.rkt = ~s\n"
            (resolved-module-path-name
             (variable-reference->resolved-module-path
              (#%variable-reference))))

;; --

Another thing to consider, if you're going to be distributing this code 
via PLaneT, is that users will have to restart DrRacket after installing 
the package so the tool can take effect.

Ryan