Hacker News Reader

Comment by akkartik | Hacker News Reader

I eagerly tried this out, and unfortunately it doesn't fix the biggest stumbling block for me: Emacs's tab behavior. Coming from Vim, I never. ever. want to hit Tab and see nothing happen. Regardless of how strongly my editor thinks it knows best, hitting Tab should insert some spaces. I'll even accept indenting the current line the first time I hit Tab. But if I hit Tab a second time, follow my other preferences and insert some character(s).

And in 5 years I can't for the life of me configure this! No solution transparently sets this default for all time. Every new fucking file type I try to edit needs me to go around in circles fixing this stupid behavior.

Sigh. Back to Vim, I guess. I don't understand why this doesn't come up more often in all the blog posts about transitioning to/from Emacs.

View on HN

Fighting against idiomatic Emacs keybindings is an uphill battle, but the hill is not very steep. The recipe that works for me:

1. Bind the behavior you want in the global map. (global-set-key (kbd "<key-here>") #'the-action-you-want)

2. Any time you encounter a mode which overrides this key, press C-h k <key-here>. It will tell you the name of the map in which the key is bound (usually foo-mode-map). Then, add: (define-key foo-mode-map (kbd "<key-here>") nil). This will unmap the key from the mode's map, revealing the binding in the global map.

Thank you! This is very actionable.

>I don't understand why this doesn't come up more often in all the blog posts about transitioning to/from Emacs.

Because most people who use Emacs like this behaviour.

Also considering that this post promises a relatively vanilla editing experience, you shouldn't expect a solution to your particular problem.

> And in 5 years I can't for the life of me configure this!

It is not that hard to solve this. It is extremely easy to map any key in Emacs. You can just map Tab key to insert a tab if that is what you want. It will be like two lines of elisp.

Edit: The reason most Emacs users don't find the tab behaviour a problem is: a, Emacs indents correctly more often than not and b, you can insert a literal tab character when you want using 'C-q Tab'.

You can't just blanket remap tab as it's overloaded for org-cycle, helm-select-action, expand in ido and so on.

You can remap it in global map, or prog-mode map, so it would keep the customized behaviour in other modes.

Thank you! global-map did the trick for me, though I seem to have tried it a few years ago without success: https://news.ycombinator.com/item?id=29726398#29731027

You want to create a minor mode (minor mode overrides major mode key bindings) and have the minor mode enabled globally. use-package has a bind-key* function that does this I think. Something like (check the syntax, I haven't tested it)

      (require 'bind-key)
      (bind-key* (kbd "tab") (lambda () (insert "    ")))

And you should probably construct the number of spaces from the tab-width variable.

Thanks! I'll look into this in combination with https://nullprogram.com/blog/2013/02/06 the next time I try out Emacs.

No, bind-key* should do this for you automatically. If you wanted to do it yourself you would make a minor-mode manually.

There's also general-override-mode-map from the "general" package.

edit: Looking at the code I don't think bind-key* creates a minor-mode but it still does what you want (and is probably the recommended way).

Oh! Let me try that.

Here's what I have so far:

* I installed Emacs 26.3 using `sudo apt install emacs`

* I installed Melpa by following https://github.com/melpa/melpa/blob/master/README.md#usage

* I installed use-package using `M-x package-install use-package`

Now I'm trying to make sense of your instructions while reading https://github.com/jwiegley/use-package#readme. I don't see a bind-key*, but I do see:

- bind-key

- bind-keymap

- bind-keymap*

Should I try one of those instead? They all seem to be in the context of a `(use-package ...)` declaration.

Edit: Never mind, I do see it now at https://github.com/jwiegley/use-package/blob/master/bind-key...

Edit 2: I just tried it and it didn't work. Turns out I've tried this approach in the past: https://lobste.rs/s/whaez0/emacs_everywhere#c_l6ez9w (edit 2 there, heh)

Wait, the suggestion at the top of that comment now works for me on a few different file extensions!

    (setq-default tab-width 2)
    (setq-default indent-tabs-mode nil)  ; to insert a tab anyway: C-q TAB

    (define-key global-map [tab]
      (lambda () (interactive) (insert-tab)))

Thank you!

I think the problem in your lobste.rs post is that you have (bind-key* "Tab" ...) instead of (bind-key* (kbd "tab") ...). I don't think your define-key solution works because I think it can be rebound by major or minor modes. I think bind-key.el is required by use-package so if you have use-package you should have it. If you don't have use-package you should install it.

I'm not at a computer with an emacs installation so I can't try it out myself. I would try binding to a regular key (e.g. "q") while debugging to try to reduce the possible problems.

I think you'll find that remapping tab everywhere like that is going to break a lot of things though (navigating fields/links, expanding sections, etc). You might want to just remap the key by explicitly listing the modes you care about.

The #emacs channel on libera (previously freenode) used to be amazing but since IRC waned it's become less good but if you're lucky and during busy times you might still find someone there that can give you a better solution.

Thanks a lot! Yeah, I'm already realizing I need to be careful what I wish for, because I can't tab complete anymore after C-x C-f. I think I'm going to follow the process outlined by CyberShadow: https://news.ycombinator.com/item?id=29726398#29728696. I'm not hung up on getting this perfectly automated, just want to not have to think about this the next time I run into it.

I think I can now give Emacs another try!

Someone above pointed out that you can remap it for prog-mode which (I think) programming modes derive from so (again, I think) if you remapped it there you would cover the programming modes. Then you could add the few other modes that aren't covered. You would define a list of the modes and loop over them in your init.el to remap them.

The other thing you might want to try is to remap the function indent-for-tab-command to insert-tab.

No! I've somehow never encountered `tab-always-indent`, according to my detailed past records. It actually looks like `(setq tab-always-indent nil)` should get me pretty close, according to the documentation.

"If the value is t (the default), the command normally just indents the current line. If the value is nil, the command indents the current line only if point is at the left margin or in the line’s indentation; otherwise, it inserts a tab character." (https://www.gnu.org/software/emacs/manual/html_node/elisp/Mo...)

Unfortunately this doesn't immediately work as described (Mac OS, GNU Emacs 27.2). When the buffer x.c looks like this (^ showing point):

    int main(void) {
      aaa
         ^

Pressing Tab now does nothing. I wonder if I need to do something to add this to c-major-mode. I also don't yet understand `setq-default`.

In the docs it says that (for some reason) C has its own separate variable c-tab-always-indent.

Also, does M-i (tab-to-tab-stop) do what you want?

Ah, yes, this ([c-]tab-always-indent) seems to work. On both my machines. I'm slightly leery of declaring victory, but this might do the trick for me. I think the major reason I care about this so much (I don't have any unique-snowflake indentation approach) is being able to indent prose inside comments. This approach would buy me that. Let me run with it and see how it goes. Thanks again.

We've hit comment depth limit, but check out M-i (tab-to-tab-stop), other people seem to declare that that is the solution for them (and they rebind it to tab per mode).

Thanks! I'll do that. I plan to carefully try each of the solutions mentioned in this thread.

(FYI, you can always get around the comment reply timer (I don't think it's a depth limit) by clicking on a comment's permalink (timestamp).)

Update: now nothing is working. I think maybe it's because I switched machines from Linux to Mac. Nothing is working on Mac OS running GNU Emacs 27.2 installed using `brew install`. I'll circle back in a few hours and post a more careful update describing what I see on each machine (if that matters) with each of the solutions.

> it's because I switched machines from Linux to Mac. Nothing is working on Mac OS

There is one obvious conclusion to draw from that.

Have you tried hooking self-insert-command to the modes you'd use it on? Like prog mode and fundamental. The only corner case I can think of is if you'd like to tab-insert in org-mode and also expand headings using tab.

Edit: Try this:

  (defun indent-for-tab-command ()
    (interactive)
    (insert-char #x20 4))

I too am interested in the Tab question

Reply on news.ycombinator.com