Regardless of what the app does and whether the thing that does is particularly useful, powerful or important for what you need to do (or even well implemented), what is a command-line interface that you had a particularly good experience both learning and working with?
In other words, I’m thinking about command line interface design patterns that tend to correlate with good user experience.
“Good user experience” being vague, what I mean is, including (but not limited to)
- discoverability–learning what features are available),
- usability–those features actually being useful,
- and expressiveness–being able to do more with less words without losing clarity,
but if there’s a CLI that has none of those but you still like it, I’d be happy to hear about it.
Edit: Trying to stress more that this post is not about the functionality behind the tool. Looks like most of first responders missed the nuance: whether app x is better than app y because it does x1 ad x2 differently or better does not matter; I’m purely interested in how the command line interface is designed (short/long flags, sub-commands, verbs, nouns, output behaviors)…
A well-designed CLI? Maybe
sed. A badly designed CLI? Probably alsosed.sedis great. I use it all the time, and I love it, but sometimes, I hate it.sed -i 's/༼ つ ◕_◕ ༽つ/¯\\\_(ツ)\_/¯/g'Hahahaha. Best comment.
┬──┬◡ノ(° -°ノ)
I always thought openSUSE’s package manager
zypperhas quite a few neat ideas:- It offers two-letter shorthands for subcommands, so
zypper install→zypper in,update→up,remove→rm. - When it lists what packages it will install or remove, it will list them with the first letter highlighted in a different color, kind of like so:
fishgittexlive
This makes it really easy to visually scan the package list, and since it’s sorted alphabetically, it also makes it easier to find a particular package you might be looking for.
And while there’s separate lists for packages to be added vs. updated vs. removed, they also color those letters in green vs. yellow vs. red, so you can immediately see what’s what. - When it lists items (other than packages), it prints an ID number, too.
So,zypper reposgives you a list of your repositories, numberered 1, 2, 3 etc., and then if you want to remove a repo, you can runzypper removerepo 3. - When you run a
zypper search, it prints the results in a nicely formatted table.
Documentation: https://doc.opensuse.org/documentation/tumbleweed/zypper/
- It offers two-letter shorthands for subcommands, so
I like the trend of refining existing tools. You take tried-and-true commands and shave off the rough edges and quirks. I use
ripgrepinstead ofgrep,fdinstead offind,scm_breezeon top ofgit,dustinstead ofdu,dufinstead ofdf,zovercd, andxhinstead ofcurlman
man man

Me reading the man
Honestly, incus.
I know it’s not strictly a utility, but holy cow, Stephane Graber and his team have put the work into that product, such that anything you can do in the ui can be done in the CLI, and more.
Tab completion entries for all the resource types (storage, instances, image repos, etc), help entries for everything, it brings a tear to the eye.
I once thought it was cool to have standardised man entries, but even better is context-sensitive
--helpentries that work well. Almost all the discovery I’ve made using incus, I’ve made using the commands themselves.It’s a real testament to how putting in the documentation work might be tedious, but it is a boon to both users and devs.
I think it depends a lot on what kind of application you’re thinking of. Here are some examples of tools that I particularly remember for having nice CLI interfaces, but they are all slightly different in terms of the number of operations/entities they support.
Single operation, a lot of different types of data: httpie
httpie is a CLI HTTP client. It allows to generate fairly complex requests, choose which diagnostic information you want to see, etc.
Examples:
$ https httpie.io/hello $ http PUT pie.dev/put X-API-Token:123 name=John $ http -v pie.dev/getSeveral operations, a lot of parameters: Bazel
Bazel is a build system. There are only a handful of operations you typically perform (build, run, test, query), but they all follow the same pattern.
Examples:
$ bazel build //app $ bazel --quiet build //app -c opt $ bazel --quiet run //app -c opt -- --port=1234Multiple entites, multiple operations: LVM
I really liked the interface of Linux Volume Manager tools. They consist of multiple commands that follow the same pattern, and use the same flags for options.
Examples:
# vgcreate -n vg_name # vgs # pvcreate /dev/sda # vgextend vg_name /dev/sda # lvcreate vg_name -n lv_name -L5G # lvresize vg_name/lv_name -L+1G # lvsLVM is in your favorites? It’s essential command(s) for sure but every time I use it it’s trial and error because I didn’t use the correct flag case and/or argument order
Favourite CLI I’d say is Junos. Structured, hierarchical, flexible.
nmonThat, along with
tmuxandhtop, are installed on everything I have.nmonthenld-give me a system health page that shows me where the bottleneck is.It’s interesting to see how a system behaves when you’re doing something like a backup… it’s not always what you think.
Ncmpcpp. I’ve been using it for so long that using other cli music players is almost a no go. Learning a new muscle memory wouldn’t be worth it. Album art would be nice but I’m listening to music and staring at the album art for hours. The metadata editor is really nice. It’s old reliable.
Ncmpcpp
Gesundheit.
findandrename/perl-rename/prename(depending on your distro), are two of my favorite cli tools. I generally find both well designed and easy to use. For me, they are indispensable.I think
findUI is so bad every time I use it I think about hacking a script just to make it simpler for my use case. At the same time I am very reluctant to use one of this new versions of standard commands trying to reinvent the wheel.Some things I don’t link about
find:-
How the directory needs to be the first argument. I get the reasoning but it is such a pain, specially if you are using it with the same query repeatedly in different paths.
-
The parenthesis to set order of matches, you are doing it in the shell so you have to escape them which is never fun.
-
The fact that
-namedoes not match partial names and there is not a version that do so you have to keep doing stuff like-name "*foo*"and of course you have to escape that shit or risk you shell expanding it. Having the GLOB version is nice but there could have a more ergonomic way to do this type, which I assume is a very common use case. -
Actually, doing more complex logical matches is always a pain and it would be nice to have a easier way to do some common operations.
-
The fact that when you do some complex match then the
-printis not automatic anymore or the the behaviour is kinda weird. And is a pain to add it in all logical branches or do it in a way that you do not repeat a lot.
Anyway, sorry for the rant.
Hahaha. No apology needed. And honestly, all fair points.
-
Pure cli or also TUI? When it comes to TUI probably yazi is my most used tool right now, use it pretty much every day. For pure cli i would probably give my vote to sed. I use the crap out of it in a bunch of scripts. For example i switch my themes with it by replacing whatever import i had in the config to the desired theme, then reload the programs.
Btop, beautiful and functional.
Just recently switched to Linux and realized I had no htop so I went for btop instead (because I’m trying out loads of new stuff). I’m very happy with it and I love the doom aesthetic of the esc-menu.
No-one mentioned ‘jq’ yet.
Maybe there’s a reason for that!
does ncurses stuff count or is TUI cheating?
My RSS shell script that simply dumps the XML’s content nicely formatted on stdout. Seriously, why do all of them have to impose some opinionated TUI on me? Terminals can do bold, dim and underline, that’s enough.
Sorry no, not public. It’s faulty and ugly and is missing features (multiple sources for one) and i want to rewrite it in python since at least 2 years.






