Rework the REPL to use comint

[shosti]

This is a really small thing, but the current behavior for the CIDER repl is to always scroll to show maximum output. I don't really like this behavior, and it would be nice if it were configurable (I'm currently hacking around it by redefining cider-repl--show-maximum-output—let me know if there's already another way to turn this behavior off). A cider-repl-show-maximum-output variable would probably be enough, although it might be nice to have some more configuration options à la comint—what are your thoughts? (Also, are there any plans to eventually base cider-repl-mode off of comint-mode?) I'd be happy to put together a PR for this.

[vspinu]

👍 I would love if cider repl could behave similarly to comint buffers and would use comint customization variables.

Why exactly cider decided not to base its repl on comint?

[bbatsov]

The REPL code was originally copied from SLIME. It'd be a huge improvement if we switched to comint, but I don't have time to work on that. Help welcome!

[shosti]

I'd be happy to take a look at moving the repl code to comint, although it might take a little while (I'm a little busy right now and I'm not an expert in comint).

[bbatsov]

geiser's and ielm's source code can used as reference I guess, as they both currently employ comint.
Using comint would also solve issues like #326.

[bbatsov]

@Vitoshka I think you can find some useful ideas about the comint implemenation here. Sly is a new fork of SLIME using comint based REPL buffers.

[vspinu]

Thanks for the pointer. I will have a look at it when I move to comint next week. There are still a couple things I would like to push to CIDER before getting to comint.

[sooheon]

Another quality of life improvement if this were to happen would be interaction with evil-mode. Currently, actions such as escaping out of insert mode into normal mode place the cursor at the start of the REPL prompt. This is a problem when you then send snippets of code from the clojure buffer, it gets inserted before the prompt. Commands like evil-delete-line do not work, while comint based packages like monroe or inf-clojure handle this just fine.

[antonyshchenko]

Yes, integration with evil-mode in repl would be awesome.

[bbatsov]

Rebasing cider-repl-mode on top of comint-mode should solve your problems. I don't think we'll need any more evil-specific support.

[bbatsov]

This important ticket is still looking for a volunteer to tackle it.

[GordonGustafson]

@vspinu did you have any unfinished work on this I should look at if I decide to take a crack at this? Or should I start by looking at ielm source and the comint documentation?

[johnbendi]

What about @shosti the OP who promised a PR?

[bbatsov]

I think that nobody actually got anywhere with this task, so @GordonGustafson can definitely start from scratch.

[vspinu]

@GordonGustafson, I haven't done anything concrete.

[shosti]

@GordonGustafson yeah, never got around to looking at this (as you can probably surmise).

[johnv02139]

I think basing it off of comint is a great idea, but for the record, for the main complaint, I added a defcustom in January, cider-repl-scroll-on-output, which can be set to nil to avoid the irritating scrolling. I didn't see this issue at the time.

Is any work ongoing here to use comint?

[bbatsov]

To my knowledge no one is working on this, so without a volunteer to push it forward it's not going to happen any time soon.

[vspinu]

I am not even convinced that it's worth anymore Comint is old and rigid which makes it hard to build a richer interfaces on top of it. So, IMO, if there is ever an effort in this direction, it should be better spent on building a new comint system from scratch.

[johnv02139]

I would agree it doesn't seem worth reimplementing. I think your (@vspinu) original message made more sense, i.e., "use comint customization variables." I would think we could use the variables and emulate the behavior without sharing implementation.

[bbatsov]

OK, I'll close this ticket then. We've got more important tasks to do anyways.