I’ve no idea how to get the “right thing”. So the easiest thing to do seems to be to evaluate the macroexpand form in the context and grab the result.

B/c we know that calling macroexpand in the REPL does the right thing

A source buffer with a cljs connection would yield the same eval results.

The only difference is that for REPLs the connection type is always static, but for other buffers it’s inferred dynamically.

I don’t understand. Clearly there is some context for that is not there for . Maybe some dynamic vars need to be bound? What I mean is that there’s no need to call cljs.analyzer/macroexpand-1. cljs.core/macroexpand-1 should work, if we provide the right context.

Probably you’re right, but keep in mind that in the middleware everything is Java code and that’s why everything we can do there is just use the cljs compiler state. That’s also how completion is implemented for Cljs. I’m not sure how we can evaluate real ClojureScript code there. Maybe looking to piggieback’s code would be informative?

Anything that can be done in cider-nrepl? Or does it have to be nREPL itself?