Re: Feature request/RFC: proper highlighting of code embedded in comments

[Clément Pit--Claudel]

On 2016-10-17 09:12, Stefan Monnier wrote:
>>     (font-lock-add-keywords nil '(("^ *>>> \\(.*\\)" (0
>> (indirect-font-lock-highlighter 1 'python-mode)))))
> 
> IIUC you additionally want to handle the "... " lines.

Yup; this was just a POC :)

>> Stefan (and emacs-devel!), do you think I should add this to ELPA?
> 
> Feel free.

Will do it soon!

>> Are there downsides I should be aware of?
> 
> The usual:
> - running the major mode might run "side-effecting code" from the user's
>   hook.  I consider it a bug in the user's setup, but it does happen.
> - it fails to take into account local modifications of the font-lock rules.

This second point is interesting.  I ran into this issue with company-coq.  The 
problem was that entering the coq major mode took a long time (maybe .2 
seconds?), so fontifying small things on the fly was slow.  I ended up copying 
font-lock related variables.  Something like

    (defconst company-coq--font-lock-vars '(font-lock-keywords
                                 font-lock-keywords-only
                                 font-lock-keywords-case-fold-search
                                 font-lock-syntax-table
                                 font-lock-syntactic-face-function)
      "Font-lock variables that influence fontification.")

    (defun company-coq--fontify-buffer-with (&optional ref-buffer)
      "Fontify current buffer according to settings in REF-BUFFER."
      (cl-loop for var in company-coq--font-lock-vars
               do (set (make-local-variable var)
                       (buffer-local-value var ref-buffer)))
      (ignore-errors
        ;; Some modes, such as whitespace-mode, rely on buffer-local state to do
        ;; their fontification.  Thus copying only font-lock variables is not
        ;; enough; one would need to copy these modes private variables as well.
        ;; See GH-124.
        (font-lock-default-fontify-region (point-min) (point-max) nil)))

But this is a bit broken, because fontification rules can run essentially 
anything.  It would be nice to have a feature like "fontify as if this code was 
at position 0 in this buffer".  I don't know whether it would make sense.

Cheers,
Clément.

signature.asc
Description: OpenPGP digital signature