racket-base-requires does not preserve requires for submodules

[florence]

#lang racket
(module+ test list->set)

gets rewritten to

#lang racket/base


(module+ test list->set)

which will then error, because list->set comes from racket/set not racket/base

[greghendershott]

Yep. This is a known limitation.

But I notice that this caveat in the doc string for racket-trim-requires:

Note: This only works for requires at the top level of a source
file using `#lang`. It does *not* work for `require`s inside
`module` forms.

Did not get copied into the doc string for racket-base-requires, too.

At the very least, I should update the doc string. Last I had looked at it, though, it was not clear to me how to remove this limitation. I can take another look, in case it's more clear to me, now.

[florence]

Ahhh. I see.

Could you use rackets module->imports to find the needed requires? Or does the problem lie elsewhere?

[greghendershott]

The limitation is with show-requires, it doesn't seem to consider sub-modules.

Maybe I could traverse the file to discover sub-modules (and sub-sub-modules and so on), run show-requires on each, and combine the results. Maybe that would be reliable, I don't know.

Another tack would be to submit a PR against show-requires. :)

Honestly so far I've found it only a mild nuisance, sometimes needing to manually add back a missing require. I feel like the feature is usually saving me a dozen manual changes, and 1 step backward isn't a big deal. So I haven't felt super-motivated to fix it, especially if it's a lot of work. However I agree it's technically not correct. And I might feel differently working with very submodule-heavy source files.

[florence]

I've just been playing with show-requires called directly on submodules and it... explodes. Looks like an internal error with how show-requires calls get-module-code.

Gah, I see why this is difficult. Oh well. :) Maybe a warning message when either command is run (and submodules are present?) saying "hey you might have broken submodules" is in order, to stop anyone else from being surprised?