pythonnut/quark-emacs

:rocket: An incredible wonderland of code

Quark Emacs

My personal Emacs config!

Featuring:

• A lot of weird stuff you might not see in other configs.
• Text editing built around vim emulation (using evil) including many custom textobjects, motions, and operators.
• Comprehensive version tracking backup system, even for buffers that do not visit files.
• Persistence of:
  • File major modes
  • File locations (thanks to saveplace)
  • Registers, minibuffer histories, kill ring, jump list (thanks to savehist)
  • Undo history (thanks to undo-tree)
• Widespread flx integration:
  • Helm (fuzzy file finder, goto definition, etc.)
  • Company (code autocomplete)
  • Ivy (minibuffer completion)
  • Isearch
• Extremely aggressive file lazy loading, and fast startup times (~0.4s).
  • Made much easier by straight.el!
  • Packages are incrementally loaded during idle time immediately after init. By the time you open a file, most major packages will already have been loaded, but you won't notice the delay!
• An emphasis on correct code and robustness without the need for explicit configuration.
  • Config should automatically determine system capabilities and adjust.

Portability

This config should be pretty portable, as long as your OS is UNIX-y. I currently run it on various Arch Linux, Fedora, macOS, as well as Android phones and Windows Subsystem for Linux.

One thing that it does not support is older versions of Emacs. Right now, the minimum version is 27.1.

Supported Languages

I actively work in the following languages:

• Elisp (of course)
• LaTeX
• Python
• Julia

quark-emacs's Issues

Mode line doesn't switch color when connecting via emacsclient

In Emacs 25.x, the initial frame is automatically resized incorrectly

The height of the frame appears unchanged, but the final width of the frame is only a few lines, which is obviously nonoptimal.

Fuzzy completion in C/C++

C/C++ names can get quite long, so this could be a big win if this could be worked into company-irony.

Get better at "fill"ing

Try to setup a reasonable system with auto-fill-mode or something similar.

Intelligently detect language for *.h files

The language could be c or c++. Since c++ is something like a superset of c, I've been defaulting to c++, but now it's time to make the process of choosing between the two automatic, since we don't want to allow c++ when it's invalid.

Flyspell wants to spellcheck C headers names?

This might be related to the fact that even when C headers are enclosed in angle brackets, they are still highlighted as strings.

helm-omni freezes periodically

Session can be recovered with C-g, however the helm session is lost in the process.

C++ keywords are completed even when semantically invalid

This mean that company-keywords should probably be removed for C/C++ at the very least, and all currently supported languages probably (to be re-added on a per-language basis).

package-upgrade-all no longer prints messages while upgrading

Probably due to a change in emacs-async?

master is broken

And that seems to break all of the stability contracts I had in mind.

Polymode deferred installation doesn't work

See #21.

Polymodes require two passes to get right. However, find-alternate-file is sufficient, while it is not for TeX.

Pasting can break when DISPLAY becomes invalid.

If DISPLAY was previously invalid, it's possible that xclip-mode will break, and block pasting, which can be very inconvenient.

exchange-point-and-mark doesn't work in visual state

Report package upgrade progress

In package-upgrade-all, no messages are printed now that the byte compilation is silent, even though Emacs is blocked for the duration of the upgrade. We should print messages indicating what is happening.

In company-irony, functions are completed as call templates when passed as pointers

I'd rather this not happen.

"<t> is undefined"

This error sometimes appears on init after attempting a keybinding, but it's non-deterministic and non-fatal.

Migrate to use-package

use-package is almost certainly better than my hand-rolled macros. Migrate to it.

This will happen at the same time as #49.

Migrate to straight.el

I've been around for much discussion during the development of straight.el, and I can say that it presents a much more compelling workflow to me than package.el does.

Complete class/struct members on -> in C/C++

Would be nice to not have to type a character to see what members a given struct or class has in C/C++. s. and s-> should be sufficient.

File completion hangs on Raspberry Pi

Remove old history relating to ELPA

Window splitting logic seems wrong

This is mostly a discussion of the function split-window-sensibly. Right now, it's logic can be roughtly summarized as follows.

1. Can we split vertically, if so split vertically.
2. Otherwise, can we split horizontally, if so do so.
3. If the window is the only window, try splitting vertically.

I think this logic is backwards. As far as I'm concerned, splitting vertically is always permissible, as the worst case of less vertical space can be overcome easily by vertical scrolling. However horizontal scrolling is not, as we want to keep a minimum horizontal size for windows so things don't get squished.

Therefore, I think cases 1. and 2. should be swapped.

Yasnippet snippets are sometimes expanded in the middle of a word

For example test-if<tab> will expand the if snippet, which is clearly nonsense.

Don't use flx to sort in helm-fasd

fasd entries are already in an optimal order.

Support remacs

See remacs/remacs#137.

C-x, C-c, and C-v should be supported for clipboard actions

• https://stackoverflow.com/questions/3323630/should-i-start-out-learning-emacs-with-ergoemacs
• https://ergoemacs.github.io/cua-conflict.html

Do something with the window title

emacs@<hostname> on <hostname> is starting to get old.

Find a way to intelligently reorder the kill-ring on pop

We want to fix the issue described here: https://stackoverflow.com/questions/5823495/emacs-how-to-yank-the-last-yanked-text-regardless-of-subsequent-kills

Need a new branch workflow

The fact that master is still stuck in the stone age (and almost certainly broke when 25.1 was released, if not earlier, as evidenced by #2) brings into question the entire structure of my branching workflow.

Current workflow:

• master should never break.
• Tags should never be broken
• dev should be where the development is done.
• Features are developed on topic branches and rebased onto dev.
• Merges to master should be explicit.

Since my config is a collection of horrifying hacks, I've never been comfortable with the dev -> master merge. And since master and tags play the same role here, they essentially have redundant meaning.

New workflow:

• master may break. (But try to keep breakage to a minimum)
• Tags should never be broken
• dev should be where the development is done.
• Features are developed on topic branches and rebased onto dev.
• Work on dev is rebased onto master.

Analysis

Admittedly, this change is subtle, but it has the following consequences:

• master should never be far behind dev, while also having little breakage. For non-critical systems, this lets me sit on master and not worry too much about breakage from whatever I'm currently experimenting with, while also not being stuck in the stone age.
• Tags exists as a fallback. If things go wrong, checkout the latest tag, which should be guaranteed to work.

Config autorecompiler doesn't seem to be functioning

As far as I can tell, the config autorecompiler never recompiles the files, even when they're obviously out of date. This should be fixed, because M-x byte-recompile-config RET is a pain.

TRAMP auto-root can't create parent directories automatically

Face related woes with emacs --daemon

• Mode line is often not the correct color upon connecting to the daemon for the first time.
• Highlighting in ivy is wrong when using the daemon from the command line.

Add "go to closest flycheck error" function and bind

(defun goto-closest-flycheck-error ()
  (interactive)
  (let ((next (flycheck-next-error-pos 1 nil))
        (prev (flycheck-next-error-pos -1 nil)))
    (cond
     ((and next prev)
      (goto-char (if  (< (abs (- (point) prev))
                         (abs (- (point) next)))
                     prev
                   next)))
     (next (goto-char next))
     (prev (goto-char prev))
     (t (user-error "No Flycheck errors")))))

YASnippet <tab> takes precedence over company

This means that it's impossible to complete while inside a snippet, which can be very inconvenient.

Echoe message interrupts message screen

I copied this to my personal init.file from your config

(require 'cl-lib)
(setq gc-cons-threshold 200000000)      ;was 100000000
(cl-letf* (;; In fact, never GC during initialization to save time.
           (gc-cons-threshold most-positive-fixnum)

           ;; Also override load to hide  superfluous loading messages
          ...


  (message "[              ]")
...
(message "[=             ]")
(load-library "thor_themes")
(message "[==            ]")
...

it works fine, the message progress bar is shown at startup but interrupet by this message in the minibuffer area:
Loading bookmarks from ~/.emacs.d/bookmarks...done [yas] Prepared just-in-time loading of snippets successfully.

the message pops up and than loading screen goes on.
How can I fix this?

P.S. thanks for this great config!!

Regards
Thorsten

Startup performance

Thank you for sharing your emacs-conifg, I am learning many neat things from it.

In regards to startup performance, I like the idea of not adding a dependancy to optimise loading performance, and I was wondering if you had cosidered use-package or other similar tools?

Thanks!

Scrolling quickly changes buffers instead of scrolling

When scrolling quickly in the GUI, Emacs cycles through buffers instead of scrolling.

Why this happens is a mystery to me.

AucTeX deferred installation doesn't work

Something special about AucTeX means that it currently doesn't work with the deferred installation system. Two passes are required to fully install it.

Move yasnippets to a fork

Versioning them here is a bad idea. I could easily keep up with upstream if I simply forked the yasnippet snippets repo and included it with quelpa.

Add a textobject for LaTeX objects

Expressions are poorly defined, but some delimiters:

• =
• \\
• \item
• \begin{...}
• \end{...}
• &
• %
• \label

vlf deferred installation is broken

For some reason, the system appears to have trouble extracting the downloaded .tar file when installing vlf on-demand.

Breaks big-time on 25.0

recentf seems to be broken

No new files have been added for several weeks now. Something is preventing the recentf system from working.

Integrate avy with dired

We should be able to jump to a file with avy.

Highlighting for CUA rectangles is broken

Editing large files is very slow

Editing large (e.g. > 100MB) files is now very slow, even with vlf. It's likely some new minor mode is messing me up here.

Silence mc-lists loading

I don't want to see

Loading /home/pythonnut/.emacs.d/data/.mc-lists.el (source)...done

Every single time I load my Emacs.

<tab> moves to next yasnippet field when company is active

Region may deactivate when moving cursor accross blank lines

Steps to reproduce:

1. Open a decently-sized C file.
2. Vjjjjjjjjjjjjjjjjjjjjjjj
3. Watch as the mark becomes inactive

Quickly switch from C/C++ header to source and back

I'm pretty sure there's some package or something that will let me do this quickly.