Allow single semicolon comments on a line

[jeaye]

Expected behavior

Single semicolon comments can be on their own line without messing up the indentation.

This was originally reported to SO here in 2010. I recognize that the recommended Lisp style is to have double semicolons for dedicated comment lines and single semicolons only for lines with code. However, even the core Clojure code and some of the most popular Clojure libraries have uses of single semicolons on dedicated lines (clojure.core, cljs.core, instaparse). So, I implore you to support this as a non-default option.

Furthermore, this is a pain point on my team at work, since some folks are in Emacs (using clojure-mode) and others are in Vim or IntelliJ where those semicolon indentation rules aren't a problem. The Clojure codebase was written before we had Emacs users on the team, so we have 15K+ lines of code using single semicolons and now the Emacs folks are having issues. We could change all of the code, but then everyone not using Emacs would need to change the way they write comments.

Concretely, applying indentation to a file containing this should not change anything (assuming this option is enabled):

; See below.
(defn foo []
  ; Wow, cool!
  (println "cool"))

Actual behavior

There is no configuration for this, so it's not currently possible to support. Lines containing single semicolon comments get indented by huge amounts when applying indentation to the line, range, form, or file.

Following the concrete example, the file becomes:

                                        ; See below.
(defn foo []
                                        ; Wow, cool!
  (println "cool"))

Steps to reproduce the problem

Create a new file in Clojure mode, add a single semicolon comment, like:

; uh oh!

and then apply indentation to the line or file (by pressing Tab or Enter, for example). The line will be indented a huge amount, rather than left exactly as it was. This also applies to evil mode (hitting ==, for example).

Environment & Version information

clojure-mode version information

This returned nil for me. I'm using Spacemacs, which may be interfering here. I'm a Vim user, primarily, but trying to do my best to repro.

clojure-mode (version nil)

Emacs version

26.1

Operating system

This has been reproduced on:

• Arch Linux
• Ubuntu Linux
• macOS

[bbatsov]

I'm aware of the problem, but unfortunately this requires significant changes that I don't have time to do myself right now.

The core of the problem is that clojure-mode uses a lot of code straight from lisp-mode and that's code is kind of hard to change, without us just duplicating it. If it wasn't for this I would have added this config option long ago.

I'd gladly merge a PR if someone decides to tackle this.

[bbatsov]

Just noticed I created a task for me regarding this over 4 years ago. (#270) Long standing problem indeed.

[yuhan0]

Doesn't seem to be that big of a change - here's the relevant code adapted from lisp-indent-line minus special handling for comments and spliced in place into clojure-indent-line

(defun clojure-indent-line ()
  "Indent current line as Clojure code."
  (interactive)
  (if (clojure-in-docstring-p)
      (save-excursion
        (beginning-of-line)
        (when (and (looking-at "^\\s-*")
                   (<= (string-width (match-string-no-properties 0))
                       (string-width (clojure-docstring-fill-prefix))))
          (replace-match (clojure-docstring-fill-prefix))))

    ;; `lisp-indent-line', but without special handling of comments
    (let ((pos (- (point-max) (point)))
          (indent (progn (beginning-of-line)
                         (calculate-lisp-indent (lisp-ppss)))))
      (when indent (indent-line-to indent))
      ;; If initial point was within line's indentation,
      ;; position after the indentation.  Else stay at same point in text.
      (if (> (- (point-max) pos) (point))
	        (goto-char (- (point-max) pos))))))

You might also want to change the value of comment-indent-function and comment-add in clojure buffers:

(add-hook 'clojure-mode-hook
          (lambda ()
            (setq comment-indent-function 'comment-indent-default)
            (setq comment-add 0)
            (comment-normalize-vars)))

I could open a PR for the above but it's not clear what the defaults should be, or if there should be some sort of configuration toggle?
It does seem like the 1-semicolon style is pretty common in the wild, and in clojure core libs.

[bbatsov]

Well, the default should be what it has always been, as it's generally a bad idea to change something that has been the default since day 1.

Generally we need some toggle - "clojure-semantic-comments" that's true by default.

Doesn't seem to be that big of a change - here's the relevant code adapted from lisp-indent-line minus special handling for comments and spliced in place into clojure-indent-line

But I think your change will break other commands relying on the current behavior - e.g. the various insert comment at point commands that Emacs has. Anyways, I'm glad you've decided to tackle this, but you should be careful as the behavior from lisp-mode trickles down to many places and we don't want to break something big.

                     (calculate-lisp-indent (lisp-ppss)))))

I'd prefer it if we copied those functions as well and tweaked them for Clojure if needed, so we can achieve true separation. I'm not sure how much code it's reasonable to share with lisp-mode, my preference would be as little as possible, but I don't have time to dwell on this much.

[yuhan0]

Thanks for the reply - what I'll do (and welcome others to) is use this change in my local config for the next couple of days or weeks and see if it breaks anything else.
Do you have any hunch on what sort of places this change might trickle down to? It seemed pretty self-contained but I'm not at all familiar with all the intricacies of lisp-mode.

                 (calculate-lisp-indent (lisp-ppss)))))

I had a peek at that function but it's a huge 150-line body with other nested calls and what looks like plenty of mutable state to untangle - probably the sort of thing that's best left untouched :p

[bbatsov]

Do you have any hunch on what sort of places this change might trickle down to? It seemed pretty self-contained but I'm not at all familiar with all the intricacies of lisp-mode.

You and me both. :D Nothing comes to mind expect commands like M-;, but I guess those should be easy to get right.

Let's go with your idea and see how well it will work.

[bbatsov]

@yuhan0 Any progress with this?

[yuhan0]

Hi, thanks for the follow up - I've been using the above change for more than a week now and noticed nothing out of the ordinary :)

(It actually helped fix a bug in my personal config, where I had bound indent-sexp to a custom key, not knowing that it was a lisp-mode specific command - the standard binding C-M-q is bound to prog-indent-sexp in clojure-mode which behaves as expected.)

[Frozenlock]

I tried clojure-indent-line as provided by @yuhan0 but it breaks in my Emacs once I press tab to indent:

calculate-lisp-indent: Symbol’s function definition is void: lisp-ppss

[yuhan0]

Hmm, turns out that was something added relatively recently in Emacs 26:
https://git.savannah.gnu.org/cgit/emacs.git/commit/?id=4713dd425beac5cb459704e67dcb8f6faf714375

I assume you're using an older version of Emacs, It should work if you simply remove the (lisp-ppss) form and use the 0-argument call to calculate-lisp-indent.

[bbatsov]

Small update - I've decided to tackle this myself, so expect some fix in the next couple of days. I'm mostly wondering how deep down the rabbit hole do I want to go (how much code to inline in clojure-mode).

[Frozenlock]

Any news on this? If the problem is harder than expected, I could move to Emacs 26 and try the patch mentioned above.

[bbatsov]

I've been super busy lately and I dropped the ball on this. It's not super complex, but I need to find some time for it.