#_ comments are not marked as comments (e.g. via `syntax-table` text property)

[dottedmag]

Expected behavior

#_ comments should mark the following expr as a comment (e.g. via syntax-table text property).

Actual behavior

It does not happen.

Steps to reproduce the problem

Enable rainbow-delimiters. See that the rainbow delimiters are not disabled in the form following #_:

See the corresponding (invalid) issue in rainbow-delimiters: https://github.com/Fanael/rainbow-delimiters/issues/29

Environment & Version information

clojure-mode version information

clojure-mode (version 5.7.0-snapshot)

Emacs version

GNU Emacs 25.1.90.1 (x86_64-apple-darwin15.6.0, NS appkit-1404.47 Version 10.11.6 (Build 15G1217))
 of 2017-02-06

Operating system

OS X 10.11

[arichiardi]

Maybe this regex needs tuning?

[Malabarba]

Actually, I think the regex is fine. My guess would be that the code that uses that regex is just not setting the syntax-table text property. Should be easy to fix.

[sogaiu]

Is this diagnosis still correct? That is, does it still appear to be an easy fix? Any further hints or pointers on how this might be fixed?

[dottedmag]

Yes, yes, and see the comment from @Malabarba.

[sogaiu]

I saw the comment but didn't succeed in getting enough from it to investigate in a meaningful way :)

It wasn't clear here what "the regex" refers to as, IIUC, the original link no longer points to a regex, and consequently I'm not succeeding in interpreting what "the code that uses that regex" points to.

Any hints / clarifications on these things?

[sogaiu]

I checked what might have been or was close to the master branch around June of 2017. Assuming what turned up was not off by much, the regex in question may be:

(defvar clojure--comment-macro-regexp
  (rx "#_" (* " ") (group-n 1 (not (any " "))))
  "Regexp matching the start of a comment sexp.
The beginning of match-group 1 should be before the sexp to be
marked as a comment.  The end of sexp is found with
`clojure-forward-logical-sexp'.

and the code that uses it (indirectly via clojure-comment-regexp) may be the function clojure--search-comment-macro-internal (and may be its friend clojure--search-comment-macro).

Perhaps an appropriate call or calls to put-text-property would be appropriate? Any help / hints on the precise argument values and where such a call or calls ought to go?

[sogaiu]

Does putting something like:

        (put-text-property start (1+ start)
                           'syntax-table (cons 11 nil))
        (put-text-property (point) (1+ (point))
                           'syntax-table (cons 12 nil))

right before the t at the end of clojure--search-comment-macro-internal function seem close?

Will (setq parse-sexp-lookup-properties t) be necessary too?

Also, do those text properties need to be removed if the leading #_ is removed / disabled?

~~On a side note, to use edebug to investigate, it seems doing (setq font-lock-support-mode nil) first seems to help a lot...~~

[Malabarba]

Sorry I can't be of more help here, but it's been a while since I've dealt with syntax tables and my memory is not helping.

What I can say is that it is possible to specify certain text properties that should be cleared when font-lock runs on a region. So, in theory, you don't have to manually clear the value if the #_ is removed. Look into font-lock-extra-managed-props.

On a side note, to use edebug to investigate, it seems doing (setq font-lock-support-mode nil) first seems to help a lot...

Oh my. I've had to edebug font-locking functions before and I can say I do not envy you. ;-)

[sogaiu]

Thanks for the hints, font-lock-extra-managed-props sounds promising.

As to debugging font-lock-related functionality, it would be much worse if I had not found font-lock-studio ;)

[sogaiu]

The following appears to help here:

diff --git a/clojure-mode.el b/clojure-mode.el
index 5d06272..150b6b0 100644
--- a/clojure-mode.el
+++ b/clojure-mode.el
@@ -532,6 +532,7 @@ replacement for `cljr-expand-let`."
   (setq-local lisp-doc-string-elt-property 'clojure-doc-string-elt)
   (setq-local clojure-expected-ns-function #'clojure-expected-ns)
   (setq-local parse-sexp-ignore-comments t)
+  (setq-local parse-sexp-lookup-properties t)
   (setq-local prettify-symbols-alist clojure--prettify-symbols-alist)
   (setq-local open-paren-in-column-0-is-defun-start nil)
   (setq-local beginning-of-defun-function #'clojure-beginning-of-defun-function))
@@ -688,6 +689,12 @@ what is considered a comment (affecting font locking).
         ;; Data for (match-end 1).
         (setf (elt md 3) (point))
         (set-match-data md)
+        (put-text-property start (1+ start)
+                           'syntax-table (cons 11 nil))
+        (put-text-property (1+ start) (point)
+                           'syntax-table (cons 14 nil))
+        (put-text-property (point) (1+ (point))
+                           'syntax-table (cons 12 nil))
         t))))
 
 (defun clojure--search-comment-macro (limit)
@@ -971,6 +978,8 @@ highlighted region)."
 (defun clojure-font-lock-setup ()
   "Configures font-lock for editing Clojure code."
   (setq-local font-lock-multiline t)
+  (setq-local font-lock-extra-managed-props
+              (cons 'syntax-table font-lock-extra-managed-props))
   (add-to-list 'font-lock-extend-region-functions
                #'clojure-font-lock-extend-region-def t)
   (setq font-lock-defaults

@dottedmag @arichiardi @Malabarba any time / interest in giving it a review / try?

[andreyorst]

Can confirm that the patch above works:

screenshots with rainbow delimiters

before patch:

with patch:

Although it introduces another bug, which results in wrong matching of ignored parentheses with show-smartparens-mode:

On the other hand, this doesn't happen with show-paren-mode, so maybe this is a bug in smartparens