(def s "a string") font locks "a string" as a docstring

[oskarkv]

Expected behavior

Font lock the string "a string" as a string in (def s "a string").

Actual behavior

The string "a string" is font locked as a docstring in (def s "a string").

clojure-mode version information

clojure-mode (version 5.7.0-snapshot)

Proposed solution

I played around with this, and it can be solved by adding a check in this and https://github.com/clojure-emacs/clojure-mode/blob/master/clojure-mode.el#L898 in the function clojure-font-lock-syntactic-face-function. For example

(and ...
  (or (save-excursion
        (goto-char startpos)
        (not (= ?\) (char-after (+ startpos (length (sexp-at-point)) 2)))))
      (string= "defprotocol" firstsym)))

This, of course, checks for a closing paren after the string, and if so, it is not considered a docstring, except for defprotocol, which is the only kind of form that can have a docstring in the last position. Instead of checking for a paren, one could of course check for another element, but it is more involved.

Opinions, suggestions, etc?

[Macroz]

Your proposed fix works for me.

[vspinu]

Relatedly in

(s/def :some/keyword (some-spec))

:some/keyword is highlighted with font-lock-function-name-face instead of the keyword-face.

[bbatsov]

@vspinu That one should be easier to fix. Just a minor regexp tweak.

[bbatsov]

@oskarkv File a PR and we can discuss the potential fix there. I'm a bit concerned about potential performance implications on larger buffers, but overall the approach seems reasonable, although it seems to be it won't work for ns for instance, so maybe it should be scope to just def or something like this.

[bbatsov]

e.g. (ns some.ns "some docstring").

[Macroz]

Hmm this works for me in ns case also but it might be because I'm running a slightly customized version of clojure-mode with more specific faces for ns and defn as you can see from the screenshot.

[Macroz]

ns works if there are requires etc, not without them.

[bbatsov]

That's what I meant. :-)

[bbatsov]

Perhaps this should just check for a list of specific forms (with semantics similar to def) and that would be both safe and solves the problem.