mflatt: looks like the kernel error is that resolve-module-path-index returns '#%kernel rather than ''#%kernel

but that violates the return contract

IRC bridge is down again

@jeapostrophe '#%kernel seems like a fit for the contract of resolve-module-path-index, which promises to return a path or a symbol or an encoding of a submodule reference; possibly, you want module-path-index-resolve, or maybe you want to use collapse-module-path-index

mflatt: collapse-module-path-index solves the #%kernel problem, but not the way to figure out what self is

Normally, you’d do that with variable-reference->module-path-index, but I don’t see a way to use that for a require transfomer

Or does the single-argument form of collapse-module-path-index work in this case?

Or maybe you’re looking for (module-path-index-join #f #f) as the “self” index

I just updated the example and I get collapse-m-p-i to violate its own contract — https://gist.github.com/jeapostrophe/ab3e1837bf6c7627fe3b07555ba49bea

i’ll try the self idea

ya, i don’t know how to make variable-reference->module-path-index useful

I think the contract of collapse-module-path-index with a single argument will have to be changed to allow a #f result if the given module path index is a “self” index. You can work around the bug by checking specifically for that before calling collapse-module-path-index

it turns out that I can also pass in the thunk to lookup self and throw an error there

no idea on how to get self though? (thinking ahead: suppose I could, I assume that make-import will fail if it tries to import from itself?)

I assume it’s clear what this transformer does? I’m pretty happy with it generally

Right - it makes no sense for a module to import from itself, and it’s not expressible as an import struct

No, I haven’t yet tried to figure out what the transformer does

Ah, if you look down at the example, (default-in m [id def]) imports id from m if m provides it OR it binds id to def

right now it works assuming that def is actually imported from somewhere

but it would be nice if it didn’t have to be and could be locally defined

(remix makes its so that when you require a module m, you automatically put (m:#%required) in your module, so it can mess with your environment)

You can use syntax-local-lift-module-end-declaration during a require transformer, if that helps

I just tried that :) for some reason I thought that wasn’t allowed

Thanks!

I’ll just change collapse-...’s contract to allow #f

I’ll push a repair soon

Ok, I’ll wait

(will need to rebuild, since I’m days behind)

np