Hello everyone!

So, i’ve been using Nushell as my daily driver for three months now. my previous shell was Fish, which i liked because it works well out of the box and is fast. Completions worked great and there’s almost no setup. Nushell’s different. I started using it just for programming/dev stuff, but then switched completely. its not without quirks, the initial setup takes more effort than Fish, some defaults feel like they should be built-in, Fish completions are more snappier.

So, why switch? Because of how it handles data. If your work touches structured stuff (json, csv, command output that you usually massage with awk/grep), Nushell is in a different league. it treats data like a proper language, not a blob of text. Also, its supported on Linux, Mac and Windows as well.

it’s a trade-off. u sacrifice a little bit of that raw, instant speed for a big win in clarity and power for data-y tasks. not for everyone, but absolutely worth a spin to see if it fits your flow.

text -> structure

traditional shells spit text. you end up parsing columns that shift if the moon is in retrograde. nushell returns typed data (tables, lists, records).

bash/fish:

ls -l | sort -k5 -nr | head -n 5

i can't seem to remember these commands

nu:

ls | where type == file | sort-by size | reverse | first 5

its more verbose but readable and stable. columns have names (name, type, size), sizes are actual sizes, dates are dates. you stop doing text surgery and start querying data.

the pipeline

in nushell, the pipe doesn't just send text forward, it’s send a table/list/record forward.

top 5 largest files (current dir):

ls | where type == file | sort-by size | reverse | first 5

processes (>10% cpu):

ps | where cpu > 10 | sort-by cpu | reverse | select pid name cpu mem

usage examples

curl,

http get "https://api.github.com/repos/nushell/nushell" | get stargazers_count
http get "https://api.github.com/repos/nushell/nushell" | get owner.login

yes, nushell has built-in http client

or, maybe let

let nu_git_data = http get "https://api.github.com/repos/nushell/nushell"
$nu_git_data | get stargazers_count

csv/calc in the shell

open data.csv | where status == "active" | group-by team | wrap rows | each { |it|
  { team: $it.column0, count: ($it.rows | length) }
} | sort-by count | reverse

sqlite

open foo.db | query db "SELECT * FROM Bar"

the trio: where, get, select

• where filters rows: history | where command =~ "git"
• get grabs fields: ls | get name
• select keeps columns: ls | select name size modified

discoverability

help commands | where name =~ "split"
help str
help date now

log slicing

open app.log
| lines
| where $it =~ "ERROR"
| first 50

errors

the config (config.nu)

Here, i am using sqlite for shell history, carapace as external completions provider, and starship for prompt. this was my starter config.

$env.config = {
    show_banner: false

    history :{
        file_format: "sqlite"
        max_size: 1_000_000
        sync_on_enter: true
        isolation: false
    }

    completions: {
        case_sensitive: false
        quick: true
        partial: true
        algorithm: "fuzzy"
        sort: "smart"
        use_ls_colors: true
    }

    cursor_shape: {
        emacs: "line"
        vi_insert: "line"
        vi_normal: "block"
    }
}

# completions ----------------
$env.CARAPACE_BRIDGES = 'inshellisense,fish,zsh,bash'
mkdir ~/.cache/carapace
carapace _carapace nushell | save --force ~/.cache/carapace/init.nu
source ~/.config/carapace/init.nu

# alias ----------------------
alias ll = ls -l
alias la = ls -a
alias lla = ls -la
alias sl = ^ls
alias mk = mkdir
alias cat = open --raw
alias nv = nvim

# for python venv ------------
# here some functions are defined that can be called from the shell
# venv_activate .venv - for activate python virtual environment (.venv by default)
# venv_deactivate - deactivating py venv

export def --env venv_activate [a_venv_path: path = "./.venv"] {
    let venv_path = $a_venv_path | path expand
    let bin_path = $venv_path | path join "bin"
    let python_path = $bin_path | path join "python"
    let python3_path = $bin_path | path join "python3"

    if not (($python_path | path exists) or ($python3_path | path exists)) {
        print $"Error: is not python venv path! ($a_venv_path)"
        return false
    }
    if ("_OLD_VIRTUAL_PATH" in $env) {
        $env.PATH = $env._OLD_VIRTUAL_PATH
        hide-env _OLD_VIRTUAL_PATH
    }

    $env.VIRTUAL_ENV = $venv_path
    $env.PATH = $env.PATH | prepend $bin_path
    $env._OLD_PROMPT_COMMAND_RIGHT = $env.PROMPT_COMMAND_RIGHT

    return true
}

export def --env venv_deactivate [] {
    if "VIRTUAL_ENV" in $env {
        let venv_bin = $env.VIRTUAL_ENV | path join "bin"
        hide-env VIRTUAL_ENV
        $env.PATH = ($env.PATH | where {|x| $x != $venv_bin})
        return true
    }

    return false
}

# prompt ---------------------
$env.STARSHIP_LOG = "error"
mkdir ($nu.data-dir | path join "vendor/autoload")
starship init nu | save -f ($nu.data-dir | path join "vendor/autoload/starship.nu")

trade-offs

• compat: you can’t copy-paste bash scripts and expect them to run in nu. use the escape hatch: %%CODE0%%
• completions: good for nu-native. carapace helps.

trading POSIX universality for structured power and readability. if your work is data-heavy or you script a lot, it’s worth it.

nu vs fish

I'd rather put them in separate buckets. Fish is great for quick, interactive use with minimal setup, and its POSIX-like syntax feels familiar. Nushell, written in Rust, trades some of that instant familiarity for structured data pipelines that shine in data-heavy tasks. I still keep Fish installed and use it sometimes. Nushell has active development and a growing communit...

[Post too long, continue reading on site]

Procrastinated for 7 (almost 8) months

But, am back writing today, also did a little update to the site, grabbed a domain through github students, played around with cloudflare a bit, used github actions, some tweaks here n there. good stuff

Thought I’d finally talk a bit about my setup~never done that before.

The setup

Arch Linux with Hyprland (yeah wayland life). I don’t rice for aesthetics but rather for functionality~with a side of "this doesn’t look like garbage."

Interestingly, I tried i3 for a hot minute and thought, "Huh, this actually looks... better, than what i have got?" So, naturally, i did the logical thing~stole its visual soul and put it into Hyprland. Most of my configs are just pulled from public anyway. that's just being efficient.

Tools:-

Hyprland: but not the animation heavy kind, no wobbly windows. though i keep a keybind to activate all animations and stuff, just in case :)

Kitty: its got everything i ever needed in a terminal. ghostty is bulky meanwhile.

Nushell: i deal with data a lot, so i can do something like: open data.csv | where year > 2020 | sort-by name

Zathura: PDF/E-book viewer, its the best, and got vim keybinds

Pandoc: document conversions are ofc a requirement

mpv/imv: mpv has yt-dlp support, so i watch videos from internet directly on it, cuz browsers are ugly and bulky

ffmpeg/imagemagick/sox: these are so good tools, ffmpeg especially is what i use most, images are edited usually in krita meanwhile. sox, well its there is enough.

Neovim: I dont really recommend it to everyone. its real power plays when you extend its functionality, creating your own environment n stuff. I actually wrote some plugins myself (for tasks that can be performed from the editor itself) i am indeed writing this blog post in neovim, which i created using a plugin i created for notes, journals, and recently added blogs thingy.

Aria2: yeah, i download a lot, like a lot really.

Newsboat: for RSS. using this since like 2 days now (on friend's recommendation) havent customized yet, so no idea, its TUI though, thats nice.

But why

well, just cuz i need a fully controlled environment, which is easy to make if i use these tools. These tools are my puppets, I pull the strings.

That’s it for now. Just wanted to break the silence. Might write again soon, or in another 7 months~we’ll see.

Anupam here!

So, is this a blog, nah.. I will just post something every now and then (sounds like a blog huh). Who's gonna read it all anyway, well actually I only know just 4 people who will every now and then come looking for new posts. Prolly I will make new friends.

I created the site only cuz I wanted to, though I also want to have a domain name and host it on my own server. There time guys; maybe someday when I earn a living.

I will post stuff that I do, and my interests. Honestly, I have no idea what this will turn into.

For today, I am gonna leave a tip here:

Procrastination is opportunity's assassin.

So get up anon, what are you waiting for.

You can reach me on X.