参见[官方文档](https://github.com/syl20bnr/spacemacs)，就不再赘述

装好之后，打开 Emacs，

find-file ~/.spacemacs

要想正常使用软件，第一件事就是配好代理，Emacs 里面的包管理，shell，都是需要代理的，所以第一步就是要配好代理

在 dotspacemacs/user-init() 中

(setq url-proxy-services
      '(("no_proxy" . "^\\(localhost\\|10.*\\)")
        ("http" . "proxy_host:port")
        ("https" . "proxy_host:port")))

C-x，C-e and you are good to go. 这里需要一个 http 的代理，如果用 shadowsocks 转一下就好。

字体不对简直不想写代码，所以先配好字体

英文字体

(setq-default dotspacemacs-default-font '("Essential PragmataPro"
                                          :size 14
                                          :weight normal
                                          :width normal
                                          :powerline-scale 1.1))

中文字体

(dolist (charset '(kana han symbol cjk-misc bopomofo))
  (set-fontset-font (frame-parameter nil 'font)
                    charset (font-spec :family "Source Han Sans SC"
                                       :size 13)))

接下来是比较麻烦的部分，至于为什么要在 Emacs 中用 Fish，没办法，Emacs 党就是喜欢 one to rule them all。

首先安装 spacemacs 提供的 shell layer 并配置一下默认 shell

(shell :variables
       shell-default-term-shell "/usr/local/bin/fish"
       shell-default-height 30
       shell-default-position 'bottom)

然后打开 ansi-term，你会发现几个问题：

1. prompt 下面会多出一个回车符号，在 dotspacemacs/user-init() 里加入

(add-hook 'term-mode-hook
          (lambda ()
            (toggle-truncate-lines)
            (make-local-variable 'transient-mark-mode)
            (setq transient-mark-mode nil)))

1. prompt title 显示问题

   find-file ~/.config/fish/config.fish

加上

function fish_title
    true
end

1. fish 的自动提示出现时，会显示一个 4m 并换行，同样在 dotspacemacs/user-init() 中

(setq system-uses-terminfo nil)

暂且到另一个 terminal 中，如果 emacs 是用 Homebrew 安装的，执行以下命令

tic -o ~/.terminfo /usr/local/share/emacs/24.5/etc/e/eterm-color.ti

其他安装方法，请自行找到类似目录

目前 kotlin 官方并没有支持 Emacs，只在 github 上搜到了两个 repo，[这个]()以及[这个]()，目前没有发现有什么区别。

参照 spacemacs 自定义 layer 的方法并使用以上两个中任意一个即可获得 kotlin 代码的语法高亮显示。

magit 的安装非常简单，安装 spacemacs 自带的 git layer 即可，唯一的问题是文档中提到的默认将 magit-status 的 buffer 打开全屏的设置方法失效了，经过一番查询后，现在需要一个第三方的 extesion 才能实现。

这个在.spacemacs 中搜索一下加在对应位置即可

(setq dotspacemacs-additional-packages'(fullframe))

dotspacemacs/user-config() 中

(fullframe magit-status magit-mode-quit-window)

重新同步 packages 即可。

1. 补充链接

2015.12.12 initial version 2015.12.13 kotlin-mode 和 magit

Emacs 作为一个 Lisp Machine，启动时间虽说比其他操作系统快得多，但是由于经常会无意间被关机，导致大家对其启动速度印象深刻。虽说 emacs --daemon 可以部分解决这个问题，但还是无法拯救喜欢乱按 space-q-q 的我（spacemacs 用户），在这样的情况下呢，启动慢已经不是最让我烦心的事了，由于 emacs --daemon, 所以大部分的东西我都会用 emacscliet -c 来当作编辑器，如果 daemon 不小心被我自己干掉了，那么这些操作全部会中断。又得手动 emacs --daemon ，总之一个字，烦！

于是就有了这么一个脚本

#!/bin/sh
# making fcitx working in emacs
client="emacsclient"
# I dont want to start a new frame if there is one
frame_count=`emacsclient -e '(true-frame-count)' 2>/dev/null || echo -1`
# start server if not started
if [ $frame_count -lt 0 ]; then
    LC_CTYPE=zh_CN.UTF-8 emacs --daemon
    client="$client -c"
    # make frame if no param coming
elif [ $frame_count -lt 1 ]; then
    client="$client -c"
fi
$client $@ 2>/dev/null
```
and in emacs init script:

```
(defun true-frame-count()
  (length
   (remove-if
    (lambda(f) (string-equal (terminal-name f) "initial_terminal"))
    (frame-list))))

因为如果以 –daemon 模式启动，就会有这样一个叫做 initial_terminal 的 frame 强行说自己 visible。

and in i3 config:

# emacs
focus_on_window_activation focus

大概会看看用 Automator 怎么在 OSX 哦不，是 macOS 上实现类似的功能。（改名字真好玩，反正改名字的人不用搜索）。

kwmc bindsym ctrl-1 space -fExperimental 1
kwmc bindsym ctrl-2 space -fExperimental 2
kwmc bindsym ctrl-3 space -fExperimental 3
kwmc bindsym ctrl-4 space -fExperimental 4

需要注意的是这个功能会把 Mission Control 搞坏

kwmc rule properties={float="true"} owner="App Store"
kwmc rule properties={float="true"} owner="System Preferences"

可根据需求酌情添加

kwmc bindsym cmd+alt+ctrl-q quit
kwmc bindsym cmd+alt+ctrl-r config reload

这里上面一个命令虽然说是退出，但是因为我们是以 daemon 模式启动的，所以它会自动重启，我一般用这种方式来更新配置，因为下面一个命令按了之后 UI 没有反应，不知道有没有生效…

根据我这一两个星期的使用，大量使用的快捷键大概就这些了。

虽说现在已经到可用的程度，不过问题还是有一些的，比如有时候窗口跳来跳去，有些藏在后台的窗口会影响布局等…但是一般这种情况按下重启的快捷键也就好了…

现在拖后腿的反倒是用了 cVim 的 Chrome 呢，感觉不久后我又会启动一波 Firefox 体验活动了…

到 Openwrt 找到路由器对应的 SDK 下载条目，我路由器是 WNDR4300，版本是 lede 17.01.4，SDK 名字是 lede-sdk-17.01.4-ar71xx-nand_gcc-5.4.0_musl-1.1.16.Linux-x86_64.tar.xz 之所以先看名字是想确定 gcc 的版本，据经验来看，使用其他版本的 gcc 编译会很惨（此处省略 200 字抱怨）。

由于这时候还没有翻墙，所以编译最好在一个墙外的 VPS 上搞定。（安装 docker 这种步骤就先跳过了）。

#+NAME prepare gcc

docker run -it --name lede gcc:5.4 bash

cd /usr/src
curl $SDK_URL -O
tar xvf $SDK_FILENAME
mv $SDK_DIR lede-sdk
cd lede-sdk

这两位 openwrt-shadowsocks & luci-app-shadowsocks

git clone https://github.com/shadowsocks/openwrt-feeds.git package/feeds
git clone https://github.com/shadowsocks/openwrt-shadowsocks.git package/shadowsocks-libev


git clone https://github.com/shadowsocks/luci-app-shadowsocks.git package/luci-app-shadowsocks

pushd package/luci-app-shadowsocks/tools/po2lmo
make && make install
popd

make menuconfig

make package/shadowsocks-libev/compile V=99
make package/luci-app-shadowsocks/compile V=99

这里用的是 dnsmasq-china-list

git clone https://github.com/felixonmars/dnsmasq-china-list.git
cd dnsmasq-china-list
scp *.conf root@lede:~/

ssh root@lede
mkdir /etc/dnsmasq.conf.d
mv *.conf /etc/dnsmasq.conf.d/

# opkg install vim
vim /etc/dnsmasq.conf
# insert conf-dir=/etc/dnsmasq.conf.d/

opkg remove dnsmasq && opkg install dnsmasq-full
/etc/init.d/dnsmasq restart

然后再配置一个没有污染的 DNS，最好在 luci 里面，方便修改。

最后是 chnroutes

vim /usr/bin/update-ignore-list

# insert
#!/bin/sh

set -e -o pipefail

wget -O- 'http://ftp.apnic.net/apnic/stats/apnic/delegated-apnic-latest' | \
    awk -F\| '/CN\|ipv4/ { printf("%s/%d\n", $4, 32-log($5)/log(2)) }' > \
    /tmp/ignore.list

mv /tmp/ignore.list /etc/

if pidof ss-redir>/dev/null; then
    /etc/init.d/shadowsocks rules
fi

# end

chmod +x /usr/bin/update-ignore-list

update-ignore-list

crontab -e
30 4 * * 0 /usr/bin/update-ignore-list >> /var/log/update_ignore_list.log

Note to self： 下次最好升级之前先弄好。

最常见的情况是当前编辑的文件中有需要修改的变量名，比如说一个 typo，由于自动补全工具的强大（比如说我在用的company-mode），如果你写错一次，那么多半这个 typo 基本就会到处都是了。

最容易想到的办法是使用 (replace-string) 查找替换，不过这样的局限性也很强。一是不够直观，二是在手动输入查找替换字符串中难免再次出错导致需要再来一瓶。

而使用 iedit ，只需要选中需要修改的字符串并执行 (iedit-mode): C-; ，当前 Buffer 所有的相同字符串都会高亮：

这时候修改高亮的字符串，其他的字符串也会同步修改。

这个过程中随时都能看到全局的变化，也不容易再次犯错。

如果你觉得这样还不够清晰，也可以使用下文提到的 ivy-occur + wgrep 的方式修改，将搜索的范围限定到当前文件即可。

有些时候需要替换的是一个函数内部局部变量的名字，但是其他函数里面也有同名的变量我不想修改，这时候只用 iedit 的话就有点麻烦了，不过 Emacs 内置的 narrow 功能就派上了用场，需要注意的是这个功能默认是被禁用了，因为新手如果不小心按到快捷键会不知道发生了什么。第一次使用时 Emacs 弹出警告按下 y 即可。

简单来说， narrow 就是指限定当前编辑的区域，如果你按下 M-x 并输入 narrow<Tab> 会看到自动补全的三个方法分别是：

1. narrow-to-defun (C-x n d)

这个方法一般只在 emacs-lisp-mode 有用，将当前编辑区域限定为当前的 (defun ()) 区域。

1. narrow-to-page (C-x n p)

将当前编辑区域限定为当前的 page(?)，关于 Page 相关的定义我还没有完全理解，根据我的 use case 基本跟 widen 的功能一致。

1. narrow-to-region (C-x n n)

这个是我最常用到的，先选定一个区域然后将编辑区域限定为选定的区域。

当然还有恢复到完整显示的方法 widen (C-x n w) 。

处于 narrow 状态时，不管你怎么修改当前区域的内容都不会影响到文件的其余部分，所以就愉快的用 iedit 改掉局部变量的名字吧。

最惨烈的情况当然是一个项目里多个文件里面都需要修改，（包括但不限于 leader 对你起的变量名不满意（x 等情况）。这里就需要一些额外的工具，包括和 ivy 。如果你还没用过 projectile 的话，建议花一点时间熟悉。另外 ivy 并不是必须的，不过非常推荐。

首先查找用 ivy 提供的 counsel-rg 配合 projectile

(defun projectile-smart-search()
  (interactive)
  (let ((text (substring-no-properties (or (thing-at-point 'symbol) ""))))
    (if (projectile-project-p)
        (counsel-rg text (projectile-project-root))
      (counsel-rg text))))

Tips: text 部分的作用是将光标位置所在的单词直接当成搜索字符串，不过这不重要。

当搜索结果出现在 minibuffer 之后，执行 (ivy-occur): C-x C-o 将所有查找结果在另一个 buffer 显示。

如图的情况，假定我现在需要将 struct EmacsRime 改名为 ERime , 而另一个类型 EmacsRimeCandidates 不需要修改。我需要做的是将光标移动到 EmacsRimeCandidates 行执行 (ivy-occur-delete-candidate): C-d ，保证剩下的都是需要修改的：

使用 (ivy-wgrep-change-to-wgrep-mode): w 切换到 wgrep-mode 使用 iedit 完成修改。按下 C-c C-c 保存即可。

Tips: 使用 C-x s 可以一次性保存所有修改过的 buffer。

虽然这边举例都是用修改代码来说明，不过这并不意味着这些工具只能在编写代码时有用。比如我需要将本文中的 Emacs 全部替换为 Vim，我需要的操作仅仅是：

1. 移动光标到其中一个 Emacs
2. C-;
3. M-d
4. Vim
5. C-;

org-roam supports org-protocol, so you need to setup org-protocol and roam beforehand, luckily, doom-emacs setup this nicely for me, all I have to do is change the $doomdir/init.el like this:

(org                ; organize your plain life in plain text
  +dragndrop         ; drag & drop files/images into org buffers
  +hugo              ; use Emacs for hugo blogging
  +roam
  +journal
  ;;+jupyter         ; ipython/jupyter support for babel
  ;;+pandoc          ; export-with-pandoc support
  ;;+pomodoro        ; be fruitful with the tomato technique
  ;;+present)          ; using org-mode for presentations

Of course, only the roam part is necessary.

And a desktop file $HOME/.local/share/applications/org-capture.desktop to handle org-protocol:// links.

[Desktop Entry]
Name=Org Capture
GenericName=Text Editor
Comment=Edit text
MimeType=x-scheme-handler/org-protocol;
Exec=emacsclient -a emacs -c -F '((name . "doom-capture") (transient . t))' %u
Icon=emacs
Type=Application
Terminal=false
Categories=Development;TextEditor;
StartupWMClass=Emacs
Keywords=Text;Editor;

The name for the new frame is for identify in window manager, and the transient is for close the new frame after capture is done. Since I’m using awesome wm, so I didn’t setup the frame size.

Then follow the guide in org-roam documents, add a bookmarklet with following content:

javascript:location.href =
'org-protocol://roam-ref?template=r&ref='
+ encodeURIComponent(location.href)
+ '&title='
+ encodeURIComponent(document.title)

or if with Firefox and tridactyl, add a shortcut in $HOME/.tridactylrc:

bind ;r js javascript:location.href = 'org-protocol:/roam-ref?template=r&ref=' + encodeURIComponent(location.href) + '&title=' + encodeURIComponent(document.title)

Now when I press ;r in any site, an emacs frame opens up, with title and #ROAM_KEY ready, I can just put my note and thoughts about it and hit C-c C-c.

roam-capture buffer opens in a popup buffer in doom emacs, by default it’s too small for taking notes, so I added a popup rule to increase it.

in $doomdir/config.el

(set-popup-rule! "^CAPTURE" :side 'bottom :size 0.90 :select t :ttl nil)

As I said above, I’m using awesome wm, like many tiling wm user, I place my Emacs window in a specific tag, the first one, for me, and Firefox in a specific tag, and it’s the second for me. But this is not very convenient for capture workflow, every time I press ;r in Firefox, the emacs frame is opened in the first tag, so I need to switch to it and take notes, since the name of the capture frame is fixed doom-capture, so I just exclude it from the rule.

ruled.client.append_rule {
    id         = "emacs",
    rule   = {
        class = "Emacs",
    },
    except = { name = "doom-capture"},
    properties = {
        tag = '1'
    }
}

It’s working for me pretty well, I take more notes in two weeks than last 4 years, and my tabs in Firefox is much less since I actually read them. Although it’s not the only thing org-roam offers, but it’s a good start.

Another thing pop up in the discussion is zstd, pacman switches to zstd from xz on Friday, Dec 27, 2019, and claims it had a 1300% speedup. Since we can also use it in makepkg now, so it’s a great time to switch. Just change one line in /etc/makepkg.conf to make it use zstd.

- PKGEXT='.pkg.tar.xz'
+ PKGEXT='.pkg.tar.zst'

The main complain is xz is only using one thread when there are 8/16 available. It’s because the commands is not setup to use multi-thread by default. we can change it as documented in the Makepkg wiki page. This change can change zstd to use multi-thread (similar to xz).

- COMPRESSZST=(zstd -c -z -q -)
+ COMPRESSZST=(zstd -c -z -q - --threads=0)

If you have spare memory, you can also use it for speed up build. This is also documented in the Makepkg wiki page. Just uncomment one line in /etc/makepkg.conf

- #BUILDDIR=/tmp/makepkg
+ BUILDDIR=/tmp/makepkg

These steps should give you a 1000%+ speedup with aur package compressing(according to Robin Broda). I have so many times starring at aur package install stuck at Compressing Package step, but never thought of how to optimize it. #TodayILearned

• 2020-04-30 add title explanation.

This is pretty straightforward, just run gpg --full-gen-key and follow the instructions, I’m using RSA and RSA with 4096 bits keysize and no expire options.

Do this twice to generate two keys, one for personal use, and one for sign git commits for work.

Find out the <new-key-id> with gpg --list-secret-keys.

First check for .gpg-id file in the password store directory, if it exists, it should contain the old gpg key-id, just remove it. The just run a pass init with newly generated key to re-encrypt all password files in it.

cd .password-store
rm .gpg-id
pass init <new-key-id>

references: pgp - How to change the gpg key of the pass password store - Ask Ubuntu

I have multiple git config with gitdir: config, just update all .gitconfig files, replacing the old key id with the new one. If only one git config is used, it can also be changed in terminal.

git config user.signingkey <new-key-id>

Get a public key export by running

gpg --export --armor <new-key-id> | pbcopy

pbcopy is a wrapper I use for save stdout to clipboard.

#!/bin/bash
if [[ $(uname) == "Darwin" ]]; then
    /usr/bin/pbcopy
else
    xclip -i -sel c -f |xclip -i -sel p
fi

Then go to GitHub/GitLab or any other service needed, add the new pubkey and remove the old key from GPG Key Settings.

Just do a decrypt and encrypt.

gpg --decrypt -o authinfo.txt .authinfo.gpg
gpg --encrypt -o .authinfo.gpg --default-key <new-key-id> authinfo.txt
rm authinfo.txt

gpg --decrypt -o mailpass.txt .mailpass.gpg
gpg --encrypt -o .mailpass.gpg --default-key <new-key-id> mailpass.txt
rm mailpass.txt

Backup keys in case of something missing from the above steps.

gpg --export --armor <old-key-id> > old_pubkey.asc
gpg --export-secret-keys --armor <old-key-id> > old_privkey.asc
gpg --delete-secret-and-public-key <old-key-id>

There will be a lot of popups, just hit delete 🙂️.

It’s quite easy I’d say, definitely easier than replacing an Email address or phone number. Be safe.

Kindle给这本推理小说的副标题是"全球销量第一的侦探小说，古典推理的不朽名作！", 虽然不知道准不准确，不过就我的水平来看还算是精彩的，阿加莎克里斯蒂娜的推理小说特点就是“暴风雪山庄”，所以凶手必定是在出现的人物之中，所以说猜凶手和手法就是阅读这类小说的最大乐趣了。

黄金时代是王小波先生的几篇小说合集，其中的《革命时期的爱情》、《似水流年》我大概已经看过好几遍了，不过这次来看合集也还是看得津津有味。

我摘选了一些自觉比较有意思的段落放在这里。跟上面不同，这里完全没有剧透也不会影响阅读。

“我自己也没兴趣”。

“配种站吗？我找郭主任。不！我什么都不送……我自己也没兴趣……我们公的母的都有。郭二，我们要去了。现在不是节气，只能看看样子了。刚才接电话的是谁？”

伟大的诗人。

现在我是诗人。虽然没发表过一行诗，但是正因为如此，我更伟大。

经历过的一切。

似水流年是一个人所有的一切，只有这个东西，才真正归你所有。其余的一切，都是片刻的欢娱和不幸，转眼间就已跑到那似水流年里去了。我所认识的人，都不珍视自己的似水流年。他们甚至不知道，自己还有这么一件东西，所以一个个像丢了魂一样。

生孩子没屁眼。

出这主意的那位首长，后来生了个孩子没屁眼，是我妈动手术给孩子做了个人工肛门。这个故事告诉我们，随着医学的发展，干点缺德事不要紧，生孩子没屁眼可以做人工肛门，怕什么？

亩产十万斤。

假设有一头猪，在一般饲养条件下每天只能长八两的话，本发明能让它长到一斤半，其法是用一斤花生油，加鸡蛋黄两个对它做肌肉注射。据说这样喂出的猪不光肥胖，肉质还十分细嫩。当时我就想到了这个发明虽好，但还不是尽善尽美。应该再打点酱油和料酒进去，使它不等挨刀子就变成一根巨大的广东香肠。

KPI。

指标这种东西，是一切浪漫情调的死敌。假如有上级下达指标令我每周和老婆做爱三次的话，我就会把自己阉掉。

我看到的不是我想看到的。

虽然有确凿的证据说明在这个炉筒里什么都没有，但是我仍然相信假如不是我哥哥踢开了挡门的树粧，而是我自己爬了进去，情况就会有所不同。

原始的时代。

等到六七年的武斗发展到了动枪时，我离开了“拿起笔做刀枪”回家去了。有人可能会说我胆小，但我决不承认。因为用大刀长矛投石机战斗，显然需要更多的勇气。就以我们院为例，自从动了枪，就没有打死过一个人。这一点丝毫不足为怪，因为在历史上也是刀矛杀掉的比枪炮多得多。原子弹造出来已经有四十多年了，除了在日本发了两回利市,还没有炸死过一个人。

别觉得自己老。

我只有四十岁,人生的道路还相当漫长。我不能总是心怀厌倦吧。

倘若王小波先生能活到现在，我大概也能当一回脑残粉吧。

About 2 years ago, I found this awesome Firefox theme, I’m astonished, because who doesn’t like a cat wagging his tail on their screen? But the extension is somewhat outdated, cat looks blurry on hidpi screens, and I also want it on my emacs! So why not?

Then I extracted the images from the extension, converted to png and svg (didn’t remember how, should have taken notes). And dragged them into Xcode since I was using a macbook back then.

Then time passed, I mainly use my desktop running Linux (I use Arch Gentoo BTW), I totally forgot this cat (sorry) for almost 2 years, until I opened it again on my macbook last Monday, I was thinking, damn, I needed it on my desktop too.

It’s a rather simple application, which, just rendering images, change it every 1/10 seconds, in another word, a frame animation. But in order to display nicely on the desktop, it needs some unordinary features.

• window needs to be transparent, so it looks nice.
• better if we can click through it so it won’t block the mouse events for the application underneath.

These two requirements sounds easy enough, and can simply implemented with the cocoa version, but I find them very challenging for its cross platform implementation.

First come to mind is to use electron, I tried that 2 years ago but had some issue about transparency and I don’t really like electron that much. So I decide to try other frameworks.

Creating an animation is really easy with the PixbufSimpleAnim API, and it handles svg out of box.

let anime = PixbufSimpleAnim::new(WIDTH, HEIGHT, 10.0);
for i in 1..22 {
    let data = Cats::get(&format!("svg_cat{}.svg", i)).unwrap();
    let buffer = ByteBuffer::from_bytes(&data);
    let pixbuf = Pixbuf::from_read(buffer).unwrap();
    anime.add_frame(&pixbuf);
}
anime.set_loop(true);

I wanted the images bundled in the binary, so rust-embed comes to help.

#[derive(RustEmbed)]
#[folder = "cats"]
struct Cats;

And we need gtk to use the compositor so it can show transparent pixels.

let screen = win.screen().unwrap();
win.set_visual(screen.rgba_visual().as_ref());

The rest of the code is just passing some params to create and display the window, that’s it.

The repo is at Gitlab, please try it out!

All credits goes to the author of the Firefox theme!

虽然自己的英文写作水平比较低，不过还是希望坚持一下多写一点。由于不再使用 Hugo 的翻译功能，所以给中英文的博客分别打上了标签。

我在 Emacs 里一直在使用 pangu-spacing， 这次给博客也加上了 pangu.js。

不过感觉可能作用有限了，由于在 Org-mode 里不中英文不加上空格，spell-check 会变得相当混乱，所以我现在都把 pangu-spacing-real-insert-separtor 又打开了，以后的博客内容应该都会自带空格。

之前给博客主题添加了一个简单的 Light/Dark Mode 切换按钮，用到了几行 JavaScript，这次在偶然关掉浏览器 JavaScript 的情况下打开之后发现就定在 Dark Mode 了，狠狠的伤害了我自己的眼睛。并且右上角的切换按钮在没有 JavaScript 的情况下也毫无作用。

更新后在没有 JavaScript 的情况下，会使用浏览器的 prefers-color-scheme 值设定 Light/Dark Mode，并不再显示切换按钮。

下一步准备给博客增加一个 "Live Section"，大概目的是用来放新弄的 Literate Doom Emacs 配置，随时更新，并且发布出来也会让我给配置补充文档增加点动力，参考 tecosaur 的配置。

或许还有一些其他用处，比如用来更新自己的读书笔记？（说得好像会读书）。

If you are using Emacs 29, you can use the built-in Tree-sitter support for TSX/TS. See Typescript/TSX in (Doom) Emacs 29+.

If you’re using emacs 29, feature/tree-sitter just got merged, and you can use the built in

View The Final Result or use emacs-mac (if you are using macOS).

I have a desktop with Ryzen 3900X and a MacBook Pro 2019(I know, I’m sorry). So this problem didn’t occur to me until I opened the project on the Macbook Pro. Emacs lags so much when navigating in the .tsx files. And I remembered that a former colleague was complaining about this a couple of days ago, and I was like: “It works for me.”. It’s not now, so I started digging.

Doom Emacs uses a major mode derived from web-mode to handle .tsx files, because typescript-mode does not support TSX. I agree with their decision after I read part of the web-mode code, like this function handles the indentation. A 700+ lines code function is called just for I press enter in a web-mode buffer. web-mode is a great package and I understand that it has to supports soooo many template languages, but it’s just too complicated.

In the typescript-mode thread, someone mentioned using typescript-mode + Tree-sitter to handle .tsx files caught my attention. Since I was running Tree-sitter in my configuration for weeks, and it works pretty well. So I tried it.

I already configured Tree-sitter in my configuration, copied from a pending PR from Doom Emacs.

• Update: there are another PR working for better Tree-sitter integration to Doom.

packages.el

(package! tree-sitter
  :ignore (null (bound-and-true-p module-file-suffix)))

(package! tree-sitter-langs
  :ignore (null (bound-and-true-p module-file-suffix)))

config.el

(use-package! tree-sitter
  :when (bound-and-true-p module-file-suffix)
  :hook (prog-mode . tree-sitter-mode)
  :hook (tree-sitter-after-on . tree-sitter-hl-mode)
  :config
  (require 'tree-sitter-langs)
  (global-tree-sitter-mode)
  (add-hook 'tree-sitter-after-on-hook #'tree-sitter-hl-mode)
  (defadvice! doom-tree-sitter-fail-gracefully-a (orig-fn &rest args)
    "Don't break with errors when current major mode lacks tree-sitter support."
    :around #'tree-sitter-mode
    (condition-case e
        (apply orig-fn args)
      (error
       (unless (string-match-p (concat "^Cannot find shared library\\|"
                                       "^No language registered\\|"
                                       "cannot open shared object file")
                               (error-message-string e))
         (signal (car e) (cadr e)))))))

So I added the suggestion from the issue thread in typescript-mode.

(use-package! typescript-mode
  :mode ("\\.tsx\\'" . typescript-tsx-tree-sitter-mode)
  :config
  (setq typescript-indent-level 2)
  (define-derived-mode typescript-tsx-tree-sitter-mode typescript-mode "TypeScript TSX"))

(after! tree-sitter
  (add-to-list 'tree-sitter-major-mode-language-alist '(typescript-tsx-tree-sitter-mode . tsx)))

And it works! JSX syntax highlighted, lsp-mode works, but(this is the last “but”; I promise.), since typescript-mode does not support JSX, it doesn’t know how to indent the code, of course, you can still manually press tab, but that’s not what we want from a “smart” editor, right?

I started to figure out the indentation. At first, I found this package tree-sitter-indent.el, it seems in a fairly early stage, only supports rust right now, it does have some instruction to adding new language support. But it requires some deep knowledge in Tree-sitter and the language parser that I don’t have, and I also only want to handle the JSX part, leave the TypeScript part to typescript-mode, so I spent 4h to come up with this…

(defun typescript-tsx-tree-sitter-indent-line ()
  (if-let* ((col (typescript-tsx-tree-sitter--proper-indentation))
            (offset (- (current-column) (current-indentation))))
      (progn
        (indent-line-to col)
        (move-to-column (+ offset col)))
    (typescript-indent-line)))

(defun typescript-tsx-tree-sitter--in-tsx (node-type parent-type)
  (--some? (s-starts-with? "jsx" it)
           (--map (format "%s" it) (list node-type parent-type))))

(defun typescript-tsx-tree-sitter--proper-indentation ()
  (save-excursion
    (back-to-indentation)
    (-let* ((node (tree-sitter-node-at-point))
            (parent (tsc-get-parent node))
            (node-type (tsc-node-type node))
            (parent-type (tsc-node-type parent)))
      ;; only handles jsx related indention
      (when (typescript-tsx-tree-sitter--in-tsx node-type parent-type)
        ;; (message "current: %s" node-type)
        ;; (message "parent: %s" parent-type)
        (cond
         ((and (member node-type '("{" "}"))
               (eq 'jsx_expression parent-type))
          (+ (cdr (tsc-node-start-point (tsc-get-parent parent))) typescript-indent-level))
         ((eq 'jsx_expression parent-type)
          nil)
         ((eq 'jsx_attribute parent-type)
          (+ (cdr (tsc-node-start-point (tsc-get-parent parent))) typescript-indent-level))
         ((eq 'jsx_expression node-type)
          (+ typescript-indent-level (cdr (tsc-node-start-point node))))
         ((member node-type '(">" "/"))
          (cdr (tsc-node-start-point parent)))
         ((eq 'jsx_closing_element parent-type)
          (cdr (tsc-node-start-point (tsc-get-parent parent))))
         ((eq 'jsx_self_closing_element parent-type)
          nil)
         ((eq 'jsx_self_closing_element node-type)
          (+ (cdr (tsc-node-start-point node)) typescript-indent-level))
         (t
          (+ (cdr (tsc-node-start-point parent)) typescript-indent-level)))))))

It’s kind of working, but so much worse than the web-mode version. And I believe I missed lots of use cases.

Before I got frustrated, I started wondering: “Why .jsx files’ indentation works so well?”.

The answer is rjsx-mode, it’s a lot simpler than web-mode, and of course, does not support TypeScript, but I only want it to handle the indentation. It should be easy, right?

The answer is YES! rjsx-mode provides a rjsx-minor-mode which parses JSX syntax into AST, it can then be used in other rjsx-mode function, in this case, rjsx-indent-line. It will probably comes with a performance cost, but I feel it’s worth.

Here is the working configuration I ended up with, even some bonus “electric” action from rjsx-mode.

(use-package! typescript-mode
  :mode ("\\.tsx\\'" . typescript-tsx-tree-sitter-mode)
  :config
  (setq typescript-indent-level 2)

  (define-derived-mode typescript-tsx-tree-sitter-mode typescript-mode "TypeScript TSX"
    (setq-local indent-line-function 'rjsx-indent-line))

  (add-hook! 'typescript-tsx-tree-sitter-mode-local-vars-hook
             #'+javascript-init-lsp-or-tide-maybe-h
             #'rjsx-minor-mode)
  (map! :map typescript-tsx-tree-sitter-mode-map
        "<" 'rjsx-electric-lt
        ">" 'rjsx-electric-gt))

(after! tree-sitter
  (add-to-list 'tree-sitter-major-mode-language-alist '(typescript-tsx-tree-sitter-mode . tsx)))

All the above is I trying to improve TSX support in Doom Emacs with emacs 28 w/ native-compile on a macbook.

A day after that, I’m trying out the emacs-mac port, turns out it can massively improve the Emacs performance on macOS, way much than emacs 28 w/ native-compile. I guess the biggest problem is not web-mode or lisp after all, but I feel it still helps. The rjsx electric magic is fantastic.

There is another take on integrating tree-sitter into emacs core, I hope this can make tsx or in general web development better in Emacs.

在使用 docker 作为 docker-compose 的后端时，默认创建的 network 中，services 之间可以通过名字相互访问。但是在使用 podman 作为后端时，默认无法实现这个功能，而是需要用到 dnsname 这个 cni plugin。

由于 gentoo 没有提供 podman 相关插件的 package, 所以需要手动安装一下。 Update 2022-04-01: added to my overlay.

git clone https://github.com/containers/dnsname
cd dnsname
make
make install PREFIX=/usr/local

NOTE: 如果安装前尝试过 docker-compose up，需要清理旧的 container 和 network, 否则

1. container 无法获取 dnsname plugin 给到的 resolv.conf.
2. 创建的 network 没有正确配置 dnsname plugin.

# after docker-compose down
sudo podman network rm xx_default # it should be removed, but just in case
sudo podman container prune

NOTE 2: 默认的 dnsname 配置，在同一 network 中访问其他 service 的域名为 [service name].dns.podman, 作为比较，使用 docker 时只需要用 service name 就行。

附上使用的 docker-compose.yml 以供参考。

version: "3"
services:
  nginx:
    image: nginx:stable-alpine
    ports:
      - "80:80"
    volumes:
      - ./nginx/nginx.conf:/etc/nginx/nginx.conf:ro
      - ./nginx/sites-available:/etc/nginx/sites-available:ro
      - ./nginx/sites-enabled:/etc/nginx/sites-enabled:ro
      - ./nginx/nginxconfig.io:/etc/nginx/nginxconfig.io:ro
    depends_on:
      - tube
  tube:
    build: ./cloudtube
    ports:
      - "10412:10412"
    volumes:
      - ./config.js:/workdir/config/config.js:ro
  newleaf:
    build: ./NewLeaf
    volumes:
      - ./configuration.py:/workdir/configuration.py:ro

如果有某些 傻逼 奇怪的 Dockerifle 用到一些 docker 特定的语法，需要指定 --format=docker.

WARN[0017] SHELL is not supported for OCI image format, [/bin/bash -c] will be ignored. Must use `docker` format

但是这个没有办法在 docker-compose.yml 里面写，因为是 buildlah 的参数，还好也可以通过环境变量覆盖。

export BUILDAH_FORMAT=docker

第一次用 Go 可能要回到 5 年前，刚从一个 Android/iOS Developer 转型，给公司写一个新的后端项目。在用习惯了充满语法糖的 kotlin 和 swift 之后再来使用这样一个“大道至简”的语言确实不太习惯。其中的一点当然就是泛型。关于泛型的讨论在 Go 社区应该从未停止过，这当中还有不少支持者是让我比较难以相信的，这也算是我第一次接触到“语言宗教”这样一个概念吧。不过这就扯远了，我简单说一下我不喜欢的理由吧。

没有泛型必然导致很多“泛用”方法需要针对不同类型重复定义，在没有泛型的时候，大概有这样三种 workaround(AFAIK)。而这三种方法和我不喜欢的主要理由如下：

• 标准库作弊，比如说 append, len 等。相当于是对于特定数据结构进行了“特殊处理”，如果你有类似的自定义数据结构，却没办法保证一致的处理模式。
• 使用 interface{} 然后通过反射处理。既享受不了动态语言的好处，却要承担动态语言的坏处。
• 代码生成。虽然重复，但又没那么重复。生成的代码是代码还是数据？应不应该放在仓库里？

他们还有一个共同的问题就是会导致代码的可读性或可浏览性变差，虽然没有盲目使用其他语言的“设计模式”来得伤害那么大，但也算是相当难受了。咳咳，阴阳怪气就到这里吧。毕竟后来官方就宣布“已经在做了”。

快进到前段时间突然在 HN 看到了 go-generics-example，发现本来预订要在 Go 2 出现的泛型，竟然在 Go 1.17 已经可以实验性开启，也就是说不需要等到很大概率不会保持向后兼容的 Go 2 了。

泛型成为 a thing 之后，似乎 Golang 团队对如何修改标准库开始犯难了，作为一个长期向后兼容的语言，如何合理的进化标准库确实是一个很难的话题。详情可见这些讨论：

• how to update APIs for generics
• go: don't change the libraries in 1.18.

抱着知道写这些代码多半不会有太大用处的心情，我们还是从 clojure.core 以及 clojure.core.async 中精心(随机)挑选了一些 API 之后，就开始模仿(抄袭)了。

Go 1.17 中的实验性泛型功能无法定义 public function, 所以需要用到 gotip 来获取 master 分支的 Go. 比较惊喜的是 gopls 也同步更新着，所以工具支持上没有任何问题。

go get -v golang.org/dl/gotip
gotip download
gotip get golang.org/x/tools/gopls@master golang.org/x/tools@master

在 发稿日 之前，2022/3/15 Go 1.18 正式 release 了，所以 gotip 不再必要。

实现这些方法倒是也不难（比较难的是我不确定只写了针对 int 类型的测试算不算完全覆盖），不过在写了一点注释后尝试通过 go doc 生成文档时倒是发现了一个奇妙的 bug，那就是 所有返回 []T 的方法都会被 go doc 忽略 ，对于针对 slice 写的 dash 包几乎可以说是全灭了。

勤奋的 ahxxm 同学即日就阅读了 go doc 源码并提交了 issue 和 pr，不过最后因为某团队 Code Review 效率问题，最后还是被内部解决了。

Go 其实不算是一个能让人感觉 exciting 的语言，增加泛型这次算是个例外吧。写这个没用的库倒是也能学到一些东西，比如在 pkg.go.dev 可以随意申请添加 package。看到 go doc 生成的 html 之后我对从注释生成文档也没那么讨厌了（治好了我的 javadoc 后遗症）。

不过比起这些我更希望标准库能尽快更新泛型版本，以后堆代码也能开心一点。就目前来看 1.18 里包含的 golang.org/x/exp 只能算聊胜于无吧，并且也不太可能在生产代码中使用。

I don’t have an email workflow, I just use it for tracking news or receiving notifications. I’ve been using mbsync & mu4e for almost 2 years, never got to learn any of the “advanced” features, only read & archive.

I also subscribed to the emacs-devel mail list, which basically made the Zoho mail web UI (or any other web mail UI) unusable. I don’t chase all the threads too often, but I just feel happy to see emacs development is so active, xD.

After messing around my own emacs config for quite some time, I’m now settled with doom emacs, it provides a nice-looking mu4e integration, that’s where I start to use it. I’m not entirely happy with it, I had to write long queries to filter out the mailing list emails in the bookmarks setting, but as I said I’m not a heavy email user, so I just live with it.

I know about notmuch a long time ago, even before I tried to use mu4e, I remember this cuz its logo in the notmuch-hello buffer is so memorably, but obviously, I didn’t stick to it for long. I’m just walking through some old emacs news as usual yesterday, and opened this link and then this video by Protesilaos Stavrou.

After I finished the video (well, while I’m finishing it), I started trying it out, the ability to search quickly with tags and add tags on the fly while reading emails attracted me. Luckily I already have all my emails on my maildir with mbsync.

So I enabled the :email notmuch and disable :email mu4e in doom. I can feel immediately that this module is not well maintained like the mu4e one, e.g., it opens up a *notmuch* workspace, but the *mu4e-hello* buffer is just a popup to the left, and the workspace remains when I closed the buffer.

The dashboard is just as I remembered with the balance logo, so far the experience is just like the last time I tried, until I learned the real magic, initial tagging.

The one on notmuch’s website is helpful but not as clear as in Prot’s video, anyway I basically copied his hook script from his video and did some manual tagging like tag all email in the archive folder with archive, and now I’m getting a much cleaner email view, and way easier to configure as well.

#!/usr/bin/env bash
notmuch tag -unread -new -- tag:replied
notmuch tag +inbox +unread -new -- tag:new

notmuch tag -new -unread +sent -- '(from:"*@luois.me* not to:"*@luois.me*)'

notmuch tag -inbox +list +emacs-devel -- from:emacs-devel@gnu.org or to:emacs-devel@gnu.org or subject:[emacs-devel]

(after! notmuch
  (set-popup-rule! "^\\*notmuch" :ignore t)

  (setq notmuch-saved-searches
        '((:name "inbox" :query "tag:inbox not tag:trash" :key "i")
          (:name "todo" :query "tag:todo" :key "t")
          (:name "flagged" :query "tag:flagged" :key "f")
          (:name "sent" :query "tag:sent" :key "s")
          (:name "drafts" :query "tag:draft" :key "d")))

  (setq +notmuch-mail-folder "~/Maildir"
        +notmuch-sync-backend 'mbsync))

As I mentioned, the notmuch module in doom emacs is not quite as good as the mu4e one, but I’m planning to improve it myself a little bit, port some of the features from the mu4e module, like the dedicated workspace thing:

(defvar +notmuch--old-wconf nil)

(add-hook 'notmuch-hello-mode-hook #'+notmuch-init-h)

(defun +notmuch-init-h ()
  (add-hook 'kill-buffer-hook #'+notmuch-kill-notmuch-h nil t))

(defun +notmuch-kill-notmuch-h ()
  ;; (prolusion-mail-hide)
  (cond
   ((and (modulep! :ui workspaces) (+workspace-exists-p +notmuch-workspace-name))
    (+workspace/delete +notmuch-workspace-name))

   (+notmuch--old-wconf
    (set-window-configuration +notmuch--old-wconf)
    (setq +notmuch--old-wconf nil))))

I’m happy I took 30min to watch the video, email never felt so easy to me, maybe I can have an email workflow now, let’s see.

ref: https://github.com/Yubico/pam-u2f

There are official guides on yubico website for how to use yubikey to setup login on windows and macOS, but not for linux.

Got those steps after some search and reading the project readme.

1. install pam-u2f with package manager
2. generate key handles and public keys

pamu2fcfg -uusername -P -N | sudo tee -a /etc/u2f_mappings

1. edit pam config to use it, adjust the options for your need, my key plugged in the back of my desktop, so I don’t want to touch it every time I tries to sudo.

:/etc/pam.d/system-auth

auth    sufficient      pam_u2f.so  authfile=/etc/u2f_mappings pinverification=1 userpresence=0

On my gentoo system, both /etc/pam.d/sudo and /etc/pam.d/login includes /etc/pam.d/system-auth, so I just need to change this.

1. test with sudo and switch to another tty for login, success! I can finally change my login password to a auto generated string.

Org Capture is a powerful tool for quickly creating notes in Emacs. To use it for web clipping, we’ll need to install the org-web-tools package. This package allows us to easily insert web pages into our note-taking system.

Org Roam has builtin org protocol support, which includes two types of protocols. For web clipping, we’ll be using the roam-ref protocol to preserve the link as a “ref”, you can find more information about this in the org roam manual.

Define a capture template for web clips:

(setq org-roam-capture-ref-templates
        '(("w" "ref" plain "%(org-web-tools--url-as-readable-org \"${ref}\")"
           :target (file+head "clips/${slug}.org" "#+title: ${title}\n")
           :unnarrowed t)))

However, since the function org-web-tools--url-as-readable-org is not autoloaded, make sure to load the package before capture:

(use-package! org-web-tools
  :commands org-web-tools--url-as-readable-org)

As mentioned in my previous post, I use tridactyl in Firefox, with the following configuration:

bind ;r js javascript:location.href = 'org-protocol://roam-ref?template=w&ref=' + encodeURIComponent(location.href) + '&title=' + encodeURIComponent(document.title)

Now I can pressing ;r in Firefox would bring up a capture frame. The web page is then converted to a plain readable org format with org-web-tools, although I do need to do some minor cleanup after the capture, such as removing base64 images from the result.

Overall, this is a highly efficient and streamlined way of clipping web pages and integrating them into my note-taking workflow.

This post is written with the help of chatgpt. What can I say, it writes English waaayy better than me :). I do think it makes the article looks too “formal” to me, but still a quite good experience.

Tree-sitter is a big deal in text editors lately. And now even Emacs is getting in on the action.

In the Emacs 29 release branch, we not only have built-in Tree-sitter support, but also many Tree-sitter powered major modes, including TypeScript/TSX.

These new major modes are developed by the maintainers of the typescript-mode package, who have stopped maintaining the old typescript-mode package. As a result, we can expect the new Tree-sitter based major mode to be more stable and feature-rich. IMO, it’s already better than the old typescript-mode package, and it’s the best new Tree-sitter based major mode in Emacs 29.

The new major modes works great, but there is this one thing I miss from the rjsx-mode: the “electric pair” feature, which automatically insert the closing tag when you type the opening tag.

I tried to do the same hack as I did in the previous post, but it doesn’t work, since rjsx-minor-mode relies on js2-mode to build the syntax tree, and I don’t think it works well with the Typescript syntax.

I really hope I can get this feature in the new typescript-tsx-mode, maybe even trying to implement it myself, let’s see.

This post discusses how to use it in Doom Emacs, but it doesn’t use any Doom-specific features or packages other than the configuration macros, so it should be easy to port to vanilla Emacs.

As you can see, this is way easier to work with typescript/tsx in emacs 29 now, and all the hack config I mentioned in my previous post is no longer needed, and I’m happy to delete them.

My view of what makes a beautiful blog has changed over time, but always circles back to simplicity. I used to build my blog with Next.js with server site rendering — already fairly minimal with limited JavaScript—but now I question whether I even need a framework. While I still enjoy working with React, it seems an overkill for a blog site. And sourcehut and bearblog shows how elegant true minimalism can be.

I’m already using org-publish to generate the posts last time, and I know that org-publish is capable of generating fully featured websites. It’s an obvious choice for me.

However, since my last blog post, I made a lot of changes with my note workflow, notably, I have switched my note taking system from org-roam to denote, and all my blog posts are still in the roam collection, so I need to take care of the migration too.

So the overall plan looks like this:

• make a org-publish project that can generate
  • pages: index page, about page etc
  • posts: all past and future blog posts
  • rss: I hecking love rss
  • styles: custom css styles
• migrate all past notes from roam to denote
• tweak the styles until I feel good about it

Time to dive in!

This new blogging setup feels just right for me. Did you notice all the code blocks in this post have syntax highlighting? And the best part - no custom CSS or JavaScript required! It even uses your Emacs theme when exporting.

Here it is: a clean, fast blog with minimal CSS, completely generated from Org files. And I’m happy with it.

• <2025-04-27 Sun> add Denote to title

The approach involves using a key file to unlock the disk, while we still need to input the passphrase for Grub, This key file handles the second decryption during boot. While many guides exist for other distros (easily found with a quick search), there’s a lack of documentation for Guix—which motivated me to write this post (Even before I complete my “migrate to Guix” post).

Okay, enough rants, here are the steps.

I have checked out some dotfile repos out there. I think everyone has their own opinions on this, for example, one would consider the whole Guix System config as dotfiles, but here in my title, I’d only count the configuration files for my commonly used tools, such as Emacs, fish (Bash), terminal emulators like alacritty/ghostty, tmux, window managers sway/niri, and so on. And nowadays, agent tools (I’ve been using pi-coding-agent recently).

Guix Home also provides a wide variety of options for managing dotfiles, for example:

• home-files-service-type
• home-dotfiles-service-type
• home-xdg-configuration-files-service-type

There are also application-specific services, like

• home-openssh-service-type
• home-bash-service-type
• home-fish-service-type
• home-mpv-service-type

… and a lot more

They’re all great tools, especially home-dotfiles-service-type; I’m able to dump my existing dotfiles right into Guix Home, but they all share the same problem that hurts the whole flow: it symlinks the config file to the immutable Guix store.

That’s not wrong; that’s the whole point of Guix/Nix: it manages everything in the store, and you can use it to control the “generations”, roll-back and switch at any point. That’s an absolutely killer feature, and it works so well with

• system config - reboot into disaster? Just select a previous working generation!
• package management - never forget about which package to install.
• background daemons - which systemd unit file did I just change?

But not for the dotfiles that you want to change often, I tweak my Emacs config every now and then, sometimes I want to try different keybindings in niri, and many times I want to write a new extension for pi, and I still want them to be git-tracked. With the Guix Home managed services, I need to

1. go to my Guix configuration repo
2. make the change
3. guix home reconfigure
4. try it out
   1. it works, good, commit the change.
   2. it doesn’t work, well, tweak it again and go back to step 2

Oh did I mention reconfigure takes forever? Guix often likes to contact substitute servers when I make no change to my packages (this might be a solvable problem but I didn’t bother). Imagine wanting to add/change a niri keybinding (which has config hot reload by default): instead of verifying it in 1s, you now need a 1min loop. And good luck if you happen to have just done a guix pull but forgotten to reconfigure, only to find out you have to wait for LibreWolf to compile because it’s not available in the substitute server yet…

And yes, all of that has happened to me, and I fail to think of a good reason to keep doing this. The benefit of being able to roll back can come from a git-tracked .dotfiles repo too, symlinked into place with stow. So the other day I just spun up a coding agent, asked it to migrate those back to a standalone dotfiles repo, and baam, just 5 mins later, I can edit my dotfiles freely again, and every change is still tracked.

Guix (and Nix) are great pieces of software that changed the way I manage my servers and desktop, but the “all in one” mindset actually constrained me for a long time, just like someone trying to apply a “design pattern” or “tech stack” to every project – a wise man once said: it’s always about trade-offs.

I still use a lot of Guix Home services, to manage those configurations I rarely change, like gpg-agent, GTK themes etc, but I finally achieved freedom of dotfiles after breaking out of the “all in one” mindset (pardon the exaggeration).