Announce the new features: Detailed Action Indicator, Highlighting and Cycling of the Target at Point

[minad]

[oantolin]

Yes, an announcement is desperately needed. Definitely some changes to the README, maybe also a reddit post?

[oantolin]

Also a new tag release (although I want to wait a couple of days for that in case of undiscovered bugs).

[minad]

Okay, do you mind if I make a post with the recent news? I think it would be good to have a post now while the recent developments are still new. And then after you tagged a new release another announcement post listing more changes in detail. I am kind of excited by the recent changes and I am eager to hear what people think.

[minad]

If you want to post it only once that's okay - some projects post a lot of updates, some only releases. I am just unreasonably excited by the new developments :sweat_smile:

[oantolin]

Go ahead! To post on reddit I don't think we need to wait.

[oantolin]

These three things together really add a whole new level of polish to Embark. It feels more professional now, like Consult. 😉

[minad]

Posted here: https://www.reddit.com/r/emacs/comments/osx5t9/recent_embark_developments_detailed_action/

Well Consult is professional, I mean already the name. It is supposed for enterprise Emacs installs or professional academic installs :smile:

[minad]

Hopefully we get some interesting feedback :)

[oantolin]

What is going on at the top now? I see two copies of the gif.

[minad]

I abused github as a gif filehoster :-P (I really don't understand how this reddit works)

[jdtsmith]

These are great! So glad my suggestion of highlighting the match made it in. This really adds to the functionality/discoverability. Now if we can pre-highlight forward/backward showing matches (like isearch does) I'll never need M-s . C-s again. Perhaps n / p can enable temporary highlighting of the symbol at point?

[oantolin]

I'm not sure how much special treatment the next/previous symbol actions should get...

In the meantime, if you want to highlight occurrences of the symbol at point there is a toggle highlight action bound to capital H.

[jdtsmith]

I saw H which looks useful, but it leaves the highlighting active and then exits, whereas I'd usually be after the opposite: temporarily highlight and then immediate nearby navigation with n/p. (I love that you can then convert this into a consult-line!). Now that embark has non-exiting actions available from outside the minibuffer (what the * means?), perhaps it makes sense to allow non-exiting actions with prefix. E.g. C-u H to turn on hi-lighting but not exit. That would get me part the way there.

[oantolin]

For now, you can add embark-toggle-highlight to embark-repeat-commands if you never want it to exit embark-act. If you want to have the choice of exiting or not, you can define an alias and make that one repeatable

(defalias 'embark-toggle-highlight-noexit 'embark-toggle-highlight)
(add-to-list 'embark-repeat-commands 'embark-toggle-highlight-noexit)
(define-key embark-identifier-map (kbd "M-h") 'embark-toggle-highlight-noexit)

I realize this is not as comfortable as having the highlighting automatic, since you need to turn it on and off.

As for having the prefix indicate non-exiting, as in your C-u H example, I think that's probably not a good idea. Currently the prefix argument means, pretty logically, to run the action with a prefix argument, and some actions definitely need the option of using their prefix argument. Maybe the thing to do would be to have a separate key function as a kind of prefix argument for this, maybe M-c for "continue", so M-c before any action would mark it as temporarily repeatable.

[oantolin]

Another thing you could do, that is more automatic, is to have the symbol target finder highlight all occurrences of the the symbol at point and unhighlight them in a post-command hook:

(defun temporarily-highlight-symbol-at-point (target)
  (when target
    (let ((regexp (find-tag-default-as-symbol-regexp))
          (hook (make-symbol "hook")))
      (fset hook (lambda ()
                   (remove-hook 'post-command-hook hook)
                   (unhighlight-regexp regexp)))
      (add-hook 'post-command-hook hook)
      (highlight-symbol-at-point))
    target))

(advice-add 'embark-target-identifier-at-point
            :filter-return 'temporarily-highlight-symbol-at-point)

This is pretty close to what you want for just symbol targets. As far as I can tell the only problem is that the highlighting sticks around even if you cycle to other targets. Oh, and that if you select embark-toggle-highlight as the action it turns the highlighting off, of course!

[minad]

We should not add special treatment for symbols here. This would go very much against the Zen of Embark as I see it ;)

But we could consider adding general hooks target-activated-hook? It could even be an alist with the type as key. In principle this is very much the same as the indicator. What about making embark-indicator an alist with type as key and hooks as value? Then we can add a symbol indicator which highlights all the symbols after a delay and it could reuse the hi-lock functionality as in your temporarily-highlight-symbol-at-point function.

[minad]

I really like this idea btw, right now embark--highlight-target is ad-hoc functionality, it should instead also be implemented as a general action indicator, such that people can turn off this annoying highlighting if they want ;)

EDIT: Since I like the generalization idea so much, I took the liberty to create issue https://github.com/oantolin/embark/issues/330.

[jdtsmith]

Currently the prefix argument means, pretty logically, to run the action with a prefix argument, and some actions definitely need the option of using their prefix argument.

I wondered about that. In fact I had a related issue regarding embark and prefixes. It doesn't seem prefixes to export get passed on to the exported command. For example I regularly use (with isearch/occur) M-s ., M-5 M-s o to get 5 lines of context in an occur buffer. When exporting from consult-line etc. to an occur buffer, I had hoped a M-5 C-, E would accomplish the same thing, but the prefix info doesn't seem to make it through.

Thanks both for your work on this (which I mostly missed while messing around with consult-imenu!).

[jdtsmith]

I'll open a separate issue. Update: #337