Spit an additional .nrepl-port file?

[vemv]

For clojure-emacs/cider#3400, it would be nice if babashka-nrepl spat a file named e.g. .bb/nrepl.port in addition to its current, vanilla one, so that it's uniquely identifiable as coming from bb.

That way, M-x cider-connect would be able to display something like mydir:34567 (bb) (see screenshot in issue as well)

wdyt?

Cheers - V

[borkdude]

There is no current vanilla one AFAIK. There have been issues about this:

babashka/babashka#1032
babashka/babashka#157

One of the issues is: you can have multiple nREPL servers running simultaneously AND there could be Clojure JVM nREPL servers running.

[vemv]

Oh, got it

I would be OK with a some sort of bespoke, conflict-free format e.g. a single .bb-nrepl.edn file with info regarding all current bb nrepl ports open within the current directory.

IDK if these nrepl processes can have a name associated somehow, in which case that info could also be included.

Final, ideal result: mydir:34567 (bb, backend) | mydir:34568 (bb, frontend)

[vemv]

There is no current vanilla one AFAIK

btw, after clojure-emacs/cider#3399 we don't always need .nrepl-port files - they're used, but we also ps for nrepl-like processes.

Current command is ps u | grep java | grep -v leiningen | grep nrepl.cmdline, if we'd need to expand it to include bb processes, please let us know

[borkdude]

babashka nrepl is started using bb nrepl-server usually, so grepping like that could work. But that invocation only works on linux-like OS-ses I guess?

[borkdude]

Writing one .edn file for all processes is a bit complex, since one process can overwrite while one process is busy writing the file

Would it be a solution to write .bb-<pid>-nrepl-port, which would be cleaned up after babashka exited?
The pid would take care of conflicts between multiple processes. If there are multiple on disk, I don't know how CIDER would select the right one though.

[vemv]

(sorry for the delay)

Would it be a solution to write .bb--nrepl-port, which would be cleaned up after babashka exited?

It sounds good to me. We'd have to ls but it seems no big deal

If there are multiple on disk, I don't know how CIDER would select the right one though.

It's no issue, we let users choose (see clojure-emacs/cider#3400 screenshot)

For the users to make an optimal decision, it would be nice if .bb-<pid>-nrepl-port could have contain not only a port but an identifier ("backend", "frontend"), assuming that's a piece of info that can be available somehow.

[vemv]

Not going to be that needed after clojure-emacs/cider#3489