Rubish: A Unix shell written in pure Ruby
github.comI'm simultaneously amazed and horrified (by the strange but amazing love child you've created between bash and ruby). I spent years (nearly a decade) trying to blend ruby and bash to make the perfect shell, and after never being quite satisfied, I eventually gave up and embraced bash. This does get closer/further than I ever could, and is a fascinating project. I'm going to give it a spin, though I can already imagine the biggest obstacle I'll hit: rubish not being available in the remote environments I need it to, meaning I'll either have to install it (and it's not lightweight currently), or I'll have to live in two different worlds, which isn't usually sustainable.
Kudos though, and great work! I can tell you put a lot of thought and effort into it.
> I can already imagine the biggest obstacle I'll hit: rubish not being available in the remote environments I need it to
The ubiquitousness of bash is among the few reasons why it continues to endure. It will be eternal if nobody tries to replace it.
Many, many people have tried...
Yeah. I will probably join their ranks at some point.
Bash maintainer actually implemented the library feature I suggested and it's already dramatically cut down the amount of unsightly bash code I need to keep around and maintain.
I'm getting pretty tired of coping with old stuff just because it's there though. Went through this phase with GNU make too.
I struggle with this too. On the plus side, the devil you know is often better than the devil you don't know, and anything new will require re-learning a lifetime's worth of muscle memory. It's also nice to know that your bash scripts are going to be hyper-portable and will still work even many years later. The muscle memory is also real. However it isn't great to be constrained with unsightly code for sake of extreme backwards compatibility. I've found a nice balance personally where I use ruby if I need anything that bash isn't good at, but it's never a perfectly clean split.
> It's also nice to know that your bash scripts are going to be hyper-portable
Doubt. I'm up to my neck in bashisms, and I require the very latest bash on top of that.
The -p flag for source landed in bash 5.3.import() { local f for f in "$@"; do [[ -v loaded[$f] ]] && continue loaded[$f]=1 source -p "${HOME}/.local/lib/bash" "${f}" done } import arguments terminalWell yes, if you're using newer features, it's not going to be available on older systems that lack a newer bash version with those features available. I think that's pretty reasonable, otherwise we'd have to freeze the language and never add anything. But your older scripts will be very portable between future systems, and across different distros once they update. If you need to target an older system, you can't use newer features, but that's true of everything so I wouldn't expect any different from bash.
I'm recently seeing more and more Ruby projects that are at least partly vibe-coded, and I'm kind of torn. On the one hand I appreciate that this allows people to create stuff that they maybe wouldn't have the time to do otherwise. On the other, the code itself makes it harder for people to contribute, especially those, like me, who don't use coding agents.
A random example:
https://github.com/amatsuda/rubish/blob/master/lib/rubish/pa...
Where are the interface boundaries? Why are there methods that are 200 lines long? This is not a dis at the author, and it's not really about "code quality" per se, whatever that means. It's just that if someone would like to study the code and be able to improve it or add features, how would one go about it? Does this mean you have to use a coding agent in order to contribute? I felt the same about the recent Ruby compiler from matz [1]. The code looks impenetrable. What does this bode for the future of OSS?
I think i can handle this code by hand in fact it’s better than code I have handled by hand. (at a cursory glance.)
In my day - I think it was around 2000 – I was handed a 5000 line perl script that both responded to CGI bin requests to run a store and kicked off fulfillment of the orders. Inside that script, it had two 1500 line long subroutines that sometimes navigated internally via goto.
We refactored, and added new features while a profitable business ran on top of the code. You don’t get quite the velocity you do on good code, but it’s manageable.
Honestly I don't know why would you choose ruby for vibecoding.
This is a language that explicitly sacrifices important stuff like the strength of automatic checks possible and performance in lieu of developer ergonomics. Even if you support that particular choice, chosing the language when you won't be writing or reading most of the code is a pretty poor tradeoff.
The interesting part is that agents are good at adding the safety layers (type safety) that exist for Ruby, but which add developer/cognitive overhead (such as Sorbet).
I actually find, for some reason, that LLMs seem to be able to be more "creative" when it comes to Ruby (having used LLMs across 4-5 languages). I don't mean hallucinating, but crafting solutions I would not have thought of, even if I've ensured that I've inserted my original thinking at the beginning.
I wonder if there is something about the combination of the expressiveness of Ruby and the way LLMs are closely tied to human language that brings that out. Of course, usual caveat: n of 1 on my own experience, and a dose of bias.
Well it may not life up to uncle Bob's clean code standards but it does fit the repo's name, doesn't it?
It's especially unfortunate because there are great tools (like rubocop) that coding agents can respond to, and actually generate very readable, maintainable, and contributable code.
I think this will improve, but I also think your comment is important for people using agents to read. Speaking for myself, I want people like you to be able to read/understand/contribute to my projects should you desire, so this is a great reminder for me.
Tangential:
I would love to see more interpreted languages offer shells with native constructs for operating as daily drivers shells (not just REPLs). When I first started learning Ruby I used `rush`[0] as my main shell. Being immersed in the language, even if there were a few helpers for shell operations, really helped me reason better about Ruby and think in the language. `scsh`[1] was enlightening as well. Ultimately the ergonomics of both pushed me back to more conventional variant but they were really helpful learning mechanisms.
0: https://github.com/adamwiggins/rush 1: https://github.com/scheme/scsh
Not sure if this is related, but i'd love to see more scripting languages (mostly Python) offer facilities which let them take over from shell script for more scripts and one-liners.
Think about what it would take to write this in Python right now:
With a few handy variables and functions predefined, this could be something like:for wmv_file in $(find $1 -name '*.wmv'); do echo -n "${wmv_file} " ffmpeg -i $wmv_file ${wmv_file%.wmv}.mpg 2>&1 | grep kb/s: || echo "ERROR $?" donefor wmv_file in find(argv[1], glob="\*.wmv"): print(wmv_file, end=" ") result = do("ffmpeg", "-i", wmv_file, basename(wmv_file, ".wmv") + ".mpg") if result: print(grep(str(result), "kb/s:")) else: print("ERROR", result.status)have you seen https://sh.readthedocs.io/en/latest/
Upvoted just for the name.
Akira showed me this work in progress in January and I was pretty amazed by it, but I have to be honest—as someone who prides himself on clever OSS repo names, he just absolutely put me to shame. English as a second language but he's a first class punner.
Oh my god this is the best name to the application. You win the weekend.
Exactly. Very clever name!
People think Ruby is a slow language, but little do they know Ruby is a slower language than Go. But ruby these days is faster than Python.
> little do they know Ruby is a slower language than Go
Isn't it generally expected for a feature-packed interpreted language to be slower than a minimal compiled language?
But who cares really? I am not using Ruby for HPC. I use it for prototyping, oneliners for ETL and to glue different moving parts in a system or network together. That's it. Its not doing the heavy lifting anyways.
when ruby was trendy the 1.9 branch was still cooking so in a lot of people's mind it is veery slow
Yea. Modern Ruby is "fast enough", but it's very real that when Ruby was hitting its peak it was dog slow. It's hard to shake those sorts of reputations (similar to the "can't scale" reputation that Rails got because of Twitter)
People think slugs are slow animals... But slugs these days are faster than snails.
The speed argument never convinced me in general, in that whether it is perl, ruby or python, they are all slower than C. So the comparisons really are odd to me.
The "scripting" languages should of course not try to be slow, but people rarely use them for speed-reasons; they use these languages for gains in productivity and ease of writing code, adding features and so forth. That should be the primary focus point.
In the future we may no longer have such a speed penalty anyway.
I would “up” this a little and say that scripting languages “should” be slow in comparison to low level compiled languages. We want eg. runtime evaluation of multis dependent on type. For (cod) example:
subset Even of Int where * %% 2; subset Odd of Int where * !%% 2; multi foo(Even $i) { ‘fizz’ } multi foo(Odd $i) { ‘buzz’ } say foo for ^9;slow is relative and frequently irrelevant. If you're just always waiting for network, or for results from postgres or redis or something, then a 100x speedup in your code won't change the user experience. And if you're doing computationally hard work in ruby or especially python, you're doing it wrong because either someone already wrote a native library to do it or you should.
There's also "slow compared to C" and "slow enough that you notice when using it as an interactive shell". Running something like `Dir.each_child('.') {|x| p x}` in the interpreter completes in 1.3 milliseconds, which includes all the separate print calls. It could be much faster if we compute the string to print first and then only issue a single print call, but this is deliberately inefficient to show it doesn't matter in this usecase.
I wouldn't use Ruby for high performance computing. But for scripting (where runtime is not critical), web services (where transport latency will usually far outstrip the few milliseconds your handler takes) or shell use (where humans aren't fast enough to issue a new command every millisecond anyway), Ruby is more than fast enough.
Very clever name!
As an avid Rails Console (basically an application-aware Ruby REPL) user, this seems familiar. Nice work.
I write a horrifying amount of PowersHell and I've always been craving something like that - rather than pwsh reinventing every wheel, just "bash but also with objects".
Usually when I see a project flaunting its language like this it elicits a sigh. (You probably know what I am talking about.) This is a happy exception since this project actually promotes a deep integration with its language of choice, so the title and name are fully warranted. Kudos for that.
I have to confess, seeing Claude as a contributor made me sigh, but I still skimmed through and it looked quite thorough and well thought out. So, I don't know if this sits on the thin line between a vibe coded project and an LLM assisted project.
Any software tool requires intelligence to be used well, including AI.
Is it just me of did others also read rubbish instead of rubi-sh...I think that might be the joke. hm...
For a while the preferred templating engine for .erb files was “erubis” which is the Japanese pronunciation for Elvis
The repository owner is a true ruby hero. I am not sure if the name is a joke, and he was just fooling around, but the code is real.
Cool project!
Just for fun, looking at code count as a rough measure of complexity.
rubish: 26,842
rc (plan9 shell): 5,888
To be fair, rubish does a lot more than rc. rc is pretty minimal.
rc source:
https://github.com/9front/9front/tree/front/sys/src/cmd/rc
Measures below:
$ wc -l `find . -name '*.rb'`
1124 ./rubish/execution_context.rb
43 ./rubish/frontend.rb
260 ./rubish/builtins/hash_directories.rb
510 ./rubish/builtins/echo_printf.rb
834 ./rubish/builtins/bind_readline.rb
182 ./rubish/builtins/directory_stack.rb
299 ./rubish/builtins/read.rb
324 ./rubish/builtins/trap.rb
129 ./rubish/builtins/arithmetic.rb
862 ./rubish/completion.rb
988 ./rubish/expansion.rb
431 ./rubish/completions/git.rb
114 ./rubish/completions/ssh.rb
530 ./rubish/completions/bash_helpers.rb
453 ./rubish/completions/help_parser.rb
167 ./rubish/ast.rb
46 ./rubish/frontend/tty.rb
1179 ./rubish/runtime.rb
127 ./rubish/lazy_loader.rb
63 ./rubish/data_define.rb
1163 ./rubish/runtime/command.rb
153 ./rubish/runtime/job.rb
7270 ./rubish/runtime/builtins.rb
306 ./rubish/config.rb
2442 ./rubish/repl.rb
1316 ./rubish/codegen.rb
1180 ./rubish/lexer.rb
742 ./rubish/history.rb
1169 ./rubish/parser.rb
67 ./rubish/startup_profiler.rb
848 ./rubish/prompt.rb
47 ./rubish/data/readline_config.rb
716 ./rubish/data/builtin_help.rb
251 ./rubish/data/shell_options.rb
53 ./rubish/data/completion_data.rb
5 ./rubish/version.rb
248 ./rubish/shell_state.rb
140 ./rubish/arithmetic.rb
61 ./rubish.rb
26842 total
$ wc -l *.c *.h *.y
547 code.c
1173 exec.c
234 getflags.c
259 glob.c
240 havefork.c
137 here.c
301 io.c
436 lex.c
169 pcmd.c
78 pfnc.c
494 plan9.c
539 simple.c
74 subr.c
37 trap.c
190 tree.c
420 unix.c
109 var.c
85 exec.h
72 fns.h
7 getflags.h
28 io.h
167 rc.h
92 syn.y
5888 totalI much prefer the pipe to method chaining.
Could this be elaborated?
Like the other commenter elaborated for me. My mental model of how programs are composed much prefers the pipe symbol rather than chaining. There's also less typing too. But each to their own.
Apparently a number of people disagree with me, or the way I initially expressed myself, judging from the amount of downvotes I've had. Weird how that happens; tabs and spaces.
ls | grep file.txt vs ls().grep("file.txt")
Your comparison is not quite optimised as you use () which is not necessary. But I understand the comparison you make.
But, you can write an optimised pipe in ruby too. I actually did that, because I could not want to be bothered to be restricted via ruby's syntax for pipe-like operations.
Even aside from that, the original claim was about pipes versus method chaining. To me these are not orthogonal to one another; they are very similar. Just with the pipe focusing on tying together different programs and focusing on input-output functionality. Method chaining in ruby is a bit more flexible, we have blocks, and usually the methods chained occur in one class/object or the toplevel namespace (less frequently though, usually). Even the pipe comparison is not ideal, because traditional UNIX pipes don't support e. g. data manipulation via an object-oriented focus. And I want that (see avisynth, but extend the idea there via a) nicer syntax and b) data manipulation for EVERYTHING).
I don't see pipe as being exclusive over method chaining or reverse.
One interesting idea was to add |> elixir's pipe-like operator to ruby. I like that, but indeed, the net-gain in ruby is quite minimal since method-chaining + blocks already offer a ton of flexibility, so I am not sure how |> would fit into ruby 1:1. Still I like the idea, but anyone proposing |> needs to come up with really convincing ideas to matz here. Because people WILL ask what the real difference is to method chaining. Even fail-safe method chaining in ruby though I absolutely hate the syntax via ? there ... it reads like garbage to me. Example:
https://github.com/ruby/ruby/blob/trunk/test/ruby/test_threa...
(It has moved since then, so the above link no longer works, been some years since I first saw it. Upon seeing it my brain instantly cancelled any use of "&.", even though I understand the rationale. It is just ugly to no ends. I still like the |> syntax in Elixir though, even though I can not really see what this should do in ruby.)t1&.kill&.joinI'm coming from the point of view of nushell, or powershell, both of which can be used for data manipulation (tables and objects respectively). I love programming in nushell compared to something like this.
The docs literally says that the () on the first ls is required.
Note that I just elaborated what I thought was being asked. Parantheses - see what switching professionally to Ruby wonderland to Python does to a person! Just in about half a year needed..
Hmm. The name is a bit awkward since people can call it "rubbish". The idea is also not quite new in that many years ago people worked on an ruby-like shell with OOP support from the get go and they used a ncurses drop down box too. I forgot the name, but it must have been before 2010 already, as I vaguely remember it from talking on IRC back in those days. I think the main developer was from South Africa, but I don't remember that much anymore.
A few years ago irb got a facelift, so rubish probably represents a more modern take on the shell concept. I tested it and it works too. I wonder how much the everything-is-an-object idea is extended here. Many years ago I learned avisynth + virtualdub and I always liked how they approached filtering. Ffmpeg is great, but I absolutely hate the filter system it uses and the ABSOLUTELY horrible syntax. The ffmpeg devs do not seem to know avisynth, or any alternatives here - so I want object manipulation with a convenient syntax at all times, not just for audio/video data but literally for any data. Naturally ruby would be a good fit by default, but I am unaware of many ruby developers even wanting to go that route. If there are still any ruby developers left that is - ruby has been tanking hard in the last few years, approaching extinction level, just like perl did before.
There has to be a better influx of new users; the old +50 years generation isn't going to keep languages alive really.
Edit: Also I forgot: the idea and implementation is fine, I just think we need much more of that in general. Ruby is kind of in a patchy patchwork situation. Where are the epic projects? Rails is also ancient already.
> Rails is also ancient already.
I think Rails both boosted Ruby and killed it. When I ask people about why they dislike Ruby it's usually due to something specific to Rails (plus some comments around syntax which are easily dismissed or accepted).
I used to be a pretty heavy Ruby user and I still love the language, though I have only used Rails sparsely and not by choice.
I had the opportunity to work on a Ruby project for a couple weeks a few years ago and it was such a pleasure to read through the code and interpret it! It was unfortunately another project that was being replaced with something else because Ruby skills were harder to find.
Indeed, I love Ruby, I find rails to be adequate and powerful, but it largely feels like a different language to me. Rails is so heavy on the "magic" while regular Ruby typically isn't. I use ruby a ton for scripts and small applications (especially micro-services in Sinatra) and it's so readable, expressive, and understandable, often even to people who don't know ruby all that well!
Good April 1 article.
But why would it be a first april article? Are there any arguments to be made for this statement? Because the shell works, I just tested it. It may not be everyone's cup of tea but that's always the case for any given software. The primary reason I use bash over, say, zsh, despite thinking zsh is more advanced, is that I use bash mostly because it is very simple. I like simplicity. (Bash could be even simpler, I would not mind. I don't use shell scripts for instance, ruby or python are much more convenient than shell scripts.)
It is a clever title. It would be funny to have an April fool sounding hook but backed by a legit project.