Yes but with eval-string ns is reset to the original binding, similar to load-string in Clojure

The sci.async ns has an eval-string+ function which returns the last ns binding as well. Sci/parse-next + eval-form allows you to save the last ns binding in case you are at the end of the stream

Eval-form doesn't reset the ns binding and requires you to bind ns yourself

Ohh, that makes sense. Thanks for the help! 🙏

How would I setup *1, *2, *3 and *e with eval-string+?

With sci/binding

Or in cljs maybe just use sci/alter-var-root

Since when using async those bindings go out of scope which sucks

I think parse-next + eval-form is what I'm looking for 👌

If you want sync eval, then use parse-next + eval-form. We could introduce a sci.core/eval-string+ as well

That would be nice, but I'm also cool with the more manual way

Everything is working perfectly!!! Thanks for the help ❤️

Cool! What are you doing, portal?

Yeah, working on exposing sci in portal via nrepl / socket repl

ah nice! you can check out nbb for nrepl / socket repl examples as well

also scittle (in browser nREPL)

So I'm adding middleware to an existing nrepl server, how does scittle do it?

scittle implements an nrepl server from scratch, so does nbb

but I'm curious to learn what you are doing, it sounds like it's something I haven't explored

Wait, how do clients connect to the browser?

Is there a proxy involved?

yes, there is a tiny proxy which is running in a bb process

https://github.com/babashka/scittle/tree/main/doc/nrepl

Interesting :thinking_face: I went the piggieback/shadow-cljs route and implemented nrepl middleware that intercepts evals 😆

I hadn't even considered that. I'm sure there are interesting trade-offs

Yeah, I'm not a big fan of middleware but I did get something working. Might try https://github.com/babashka/sci.nrepl instead.