Support for multiple connections

[technomancy]

It seems like right now monroe only supports connecting to a single nrepl instance at a time.

Are you interested in adding support for more than one active *monroe* buffer, or is it better to use separate Emacs instances for that?

[sanel]

I was planning to add it (made a few attempts), just didn't have enough time to complete it. However, if you are able to do it, keeping the code simple, I'll be happy to accept the patch :)

[technomancy]

Good to know; thanks. No promises, but I might take a shot at this at some point.

[bbuccianti]

I'm thinking of making an attempt too. Maybe we can discuss a little about how to do it?

[sanel]

Feel free @bbuccianti :) Maybe check inferior-lisp mode (inf-lisp.el in Emacs) for multiple processes support.

[sogaiu]

@bbuccianti

For reference, if you want to see an existing attempt for a socket repl-based thing with multiple connections, I've had some luck w/:

• https://github.com/sogaiu/acrepl/blob/arcadia-and-reconnect/acrepl-connect.el
• https://github.com/sogaiu/acrepl/blob/arcadia-and-reconnect/acrepl.el

It's been working pretty well for me so far.

I started with sesman, but decided it was much more than I needed.

[bbuccianti]

@sanel @technomancy @sogaiu

Can you help me test this?

https://github.com/bbuccianti/monroe/

I'm sure there is a lot of ways to improve it, please help me!

[sogaiu]

Do you mean the multiple-connections branch?

https://github.com/bbuccianti/monroe/tree/multiple-connections

[bbuccianti]

This is the correct commit:

https://github.com/bbuccianti/monroe/commit/38a1488f54596ccdb35769c83ce0def2380b233c

[sogaiu]

Had success interacting with 2 separate processes from within a single emacs instance.

Opened two files -- each one from a different project -- ran M-x monroe in each and tried M-x monroe-eval-region. Seems to work :+1:

IIUC, there is at most one connection per project and at most one process per project. Is that correct?

[bbuccianti]

Yes, it's using the project folder as a pointer concatenating it with monroe and monroe-conn buffers.

I've created a pull request #33

[sanel]

Thanks @bbuccianti for your work and @sogaiu for fast response!

Is #33 will add support for multiple connections? Title of PR confuses me :)

Also, does it handle reading .nrepl-port from project tree?

[bbuccianti]

It still uses monroe-locate-port-file.

The only change is that instead of naming the buffers *monroe* and *monroe-conn* I conj the project folder to it. Then, when you try to send code to the nrepl it searches for the correct buffers using the project folder.

[bbuccianti]

@sanel I just improved the title of the PR

[sanel]

Thanks!

Here are few issues I found (my Emacs is 26.3).

After first expression is entered (in single or multiple REPL connections), prompt is not shown any more.

Next issue is connecting to REPLs outside of working directories. For example, you run REPL at localhost:7888 in project foo, then in external terminal run REPL in project baz at localhost:7889, but keep Emacs in project foo, it will connect to foo for both REPL and inputs and outputs will be shared.

The same applies if you connect Emacs to any multiple projects without entering to project folders.

Since you are calculating process name from working directory, I'd suggest to use combination of host+port instead, because they have to be unique on given system.

Also, I'd suggest to pop out exising monroe connection REPL if user attempt to connect to that session again. This will prevent multiple REPLs that will echo inputs and outputs between each other.

[bbuccianti]

I can't reproduce the issue with prompt. Maybe I'm evaluating trivial things, but seems to work fine for me.

https://github.com/bbuccianti/monroe/commit/05b37d975da409210762a343b131e9ccb41b9ab6

Now it's using the host:port combination to name buffers and also call (monroe-disconnect) before trying to attempt another connection.

Let me know!

[sanel]

Thanks @bbuccianti for your work so far! However, I still have issue with prompt. What Emacs are you using?

Here is how to reproduce it:

1. Run lein repl somewhere.
2. In new Emacs session, run monroe, it will ask for connection.
3. After you connect, you will see prompt. Enter something like (+ 1 2 3) in REPL. It will print result, but prompt is no longer visible.

Here is how it looks for me:

user=> (+ 1 2 3)                                                                                                                                                                                                    
6
;; no prompt here                                                                                                                                                                                                                   
(println "abc")                                                                                                                                                                                                     
abc
;; no prompt here

[bbuccianti]

Are you using this? https://github.com/sanel/monroe/pull/33

I'm currently using GNU Emacs 26.3 (build 1)

[sanel]

Yes, using PR #33 and same Emacs version. Tried even with emacs -Q.

Can you try loading PR with emacs -Q please?

[bbuccianti]

Well, that works for reproducing the issue. Any suggestion on how to fix it?

[sanel]

I'm curious how it worked for you - maybe some package "fixed" that :) This happened before for me and I believe it is up to comint-mode now as well (comint can break in weird ways if you messed up options or processes). I'll check this one.

[bbuccianti]

I've tested with commit https://github.com/sanel/monroe/commit/2f472fdc09c1b36c291ddb5ed9aecc331fd7e082 and it's not working in emacs -q !

[sanel]

Hm... does it work for you if you revert to this: https://github.com/sanel/monroe/commit/a7cf37e6e29a1e31d9a6e9ce19d5324eca09465a ?

[bbuccianti]

Yeah! Can confirm that commit a7cf37e it's working as expected.

Also I want to tell that seems like a callback done with monroe-make-response-handler isn't running with the "done" status. Then monroe-requests isn't cleaned so the prompt is never added again.

But is really strange, because status is clearly truthy but can't go inside the (when status ;cleaning) code. I'm really new to those kind of callbacks and don't know really how monroe-dbind-response is doing is magic, but I'm learning a lot. Would be great if you @sanel can give some advice in how to debug that part.

Thanks!

[bbuccianti]

@sanel no advances on this?

[sanel]

Hey Benjamín,

sadly I'm a bit busy currently :S If you have time to take a look at it, I'll be happy to review it.

[bbuccianti]

I force-pushed to #33

Thanks @sanel

[sanel]

Thanks @bbuccianti ! Will take a look :)

[sanel]

Implemented with https://github.com/sanel/monroe/pull/33

[sanel]

Sorry @bbuccianti but this still doesn't work. To demonstrate:

• Emacs 26.3
• In separate run new nrepl server with lein repl
• From emacs, run monroe and connect to nrepl address
• In monroe buffer, enter + and press enter
• It will print #object[clojure.core$_PLUS_ 0x4e0e22f4 "clojure.core$_PLUS_@4e0e22f4"] but prompt is not shown.

Worse, with multiple connections, it mess up where evaluation should go. To reproduce:

• Run 2 different nrepl servers it separate terminals.
• Connect from Emacs to both of them (running monroe and entering their addresses).
• Enter + in one monroe buffer, it will evaluate in second.

How you manage to get this working?

[sanel]

@technomancy after https://github.com/sanel/monroe/commit/1c2559b6862e1d4f8780d370973ee63b383b6122 prompt is no longer shown when you enter something like + in fresh monroe session. What is reasoning behind this change?

Does it work for you?