▲ 249 points
back
135 comments
This is the way. Shell makes for a terrible scripting language, that I start regretting choosing usually around the time I have to introduce the first `if` into my "simple" scripts, or have to do some more complex string manipulation.
At least nowadays LLMs can rewrite Bash to JS/Python/Ruby pretty quickly.
This is exactly the frustration that lead me to write Rad [0] (the README leads with an example). I've been working on it for over a year and the goal is basically to offer a programming language specifically for writing CLIs. It aims for declarative args (no Bash ops parsing each time), automatic --help generation, friendly (Python-like) syntax, and it's perfect for dev build scripts. I'll typically have something like this:
#!/usr/bin/env rad
---
Dev automation script.
---
args:
build b bool # Build the project
test t bool # Run tests
lint l bool # Run linter
run r bool # Start dev server
release R bool # Release mode
filter f str? # Test filter pattern
filter requires test
if build:
mode = release ? "--release" : ""
print("Building ({release ? 'release' : 'debug'})...")
$`cargo build {mode}`
if lint:
print("Linting...")
$`cargo clippy -- -D warnings`
if test:
f = filter ? "-- {filter}" : ""
print("Running tests{filter ? ' (filter: {filter})' : ''}...")
$`cargo test {f}`
if run:
bin = release ? "target/release/server" : "target/debug/server"
$`./{bin}`
Usage: ./dev -b (build), ./dev -blt -f "test_auth" (build, lint, test auth), ./dev -r (just run).
Well, at least I will be able to run my bash scripts in 5 years
> This is the way. Shell makes for a terrible scripting language, that I start regretting choosing usually around the time I have to introduce the first `if` into my "simple" scripts, or have to do some more complex string manipulation.
I suppose it can be nice if you are already in a JS environment, but wouldn't the author's need be met by just putting their shell commands into a .sh file? This way is more than a little over-engineered with little benefit in return for that extra engineering.
The reasons (provided by the author) for creating a Make.ts file is completely met by popping your commands into a .sh file.
With the added advantage that I don't need to care about what else needs to be installed on the build system when I check out a project.
I just don't see the advantages.
I use swift! I even (re-)wrote swift-sh[0] to make it possible to import external modules in a script (à la uv).
I've been working a lot in fairly complex shell scripts lately (though not long— not much over 1000 lines). Some of them are little programs that run locally, and others drive a composable cloud-init module for Terraform that lets lets users configure various features of EC2 hosts on multiple Linux distribution without writing any shell scripts themselves or relying on any configuration management framework beyond cloud-init itself. With the right tooling, it's not as bad as you'd think.
For both scripts, everything interesting is installed via Nix, so there's little reliance on special casing various distros', built-in package managers.
In both cases, all scripts have to pass ShellCheck to "build". They can't be deployed or committed with obvious parse errors or ambiguities around quoting or typos in variable names.
In the case of the scripts that are tools for developers, the Bash interpreter, coreutils, and all external commands are provided by Nix, which hardcodws their full path into the scripts. The scripts don't care if you're on Linux or macOS— they don't even care what's on your PATH (or if it's empty). They embrace "modern" Bash features and use whatever CLI tools provide the most readable interface.
Is it my favorite language? No. But it often has the best ROI, and portability and most gotchas are solved pretty well if you know what tools to use, especially if your scripts are simple.
A lesson I learnt during the 90's already switching into Perl instead, somehow people keep writing pieces of wonder in plain shell scripts.
Agreed. The shell is great for chaining together atomic operations on plaintext. That is to say, it is great for one liners doing that. The main reason probably isn't how it all operates on plain text but how easy it makes it to start processes, do process substitution, redirections, etc.
As soon as you have state accumulating somewhere, branching or loops it becomes chaotic too quickly.
I generally use AWK as my scripting language, or often just write the whole thing directly in AWK. It doesn't change, is always installed on all POSIX platforms, easily interfaces with the command line, and is an easy to learn small language.
"Just" is exactly made for this, and it is amazing. You write a justfile that is somewhat similar to a makefile but without the painpoints and it provides a CLI interface of commands you want to run
I was a Just enjoyer for quite a while, until I tried mise. Mise does all the same things as just, but also has source/output tracking to avoid rerunning build jobs (like make), and also bundles runtimes like asdf. It's become my all-in-one task runner of choice.
I think the make in the title is a bit misleading, the author is actually just advocating for having a consistent file you use for adhoc scripting and testing in your application.
The thrust of the article could be summarized as: if you type more than one command into the shell, make a script.
You can reference other justfiles as modules too, so in a mono repo you can do `just foo-app test`.
If you combine that with relative working folders it’s very easy to manage large projects.
And you can get shell completion, which is extra nice.
How am I the first person to mention fzf?
Just integrate fzf into your shell and use ctrl-r to instantly summon a fuzzy shell history search and re-execute any command from your history!
I cannot imagine going back to using a terminal without this.
I still write plenty of scripts if I need to repeat multi command processes but for one liners just use fzf to reexecute it.
Also in a shared project you can ignore script files with .git/info/exclude instead of .gitignore so you don’t have to check in your personal exclusion patterns to the main branch.
Seriously people if you use a terminal you need the following tools to dominate the shell:
ripgrep, zoxide, fzf, fd
I can't believe how long I was sleeping on fd and zoxide. zoxide is now one of my top commands, and fd feels like when I switched to ripgrep. So fast and easy there's no reason not to run it.
My gut reaction is to rush to the comments to shill my favourite task runner ( mise tasks[1], now with shell aliases[2]!) but pushing past that, the core idea of writing scripts in a file rather than a shell prompt is a great nugget of wisdom. But I disagree with this bit:
"I want to be clear here, I am not advocating writing “proper” scripts, just capturing your interactive, ad-hoc command to a persistent file."
What's the difference? Why not version control it, share it with colleagues. Imagine writing a unit test to test a new feature then deleting it when done, what a waste. Ok it's not exactly the same because you aren't using these scripts to catch regressions, but all of that useful learning and context can be reused.
I don't think the language you use for scripting is too important as long as the runtime is pinned and easily available on all engineers machines, perhaps using a toolchain manager like... mise[3].
[1] https://mise.jdx.dev/tasks/ [2] https://mise.jdx.dev/shell-aliases.html [3] https://mise.jdx.dev/dev-tools/
> What's the difference? Why not version control it,
Because I'm hardcoding directory paths.
Because I'm assuming things are set up a particular way: the way they are on my machine.
Because this is hardcoded to a particular workflow that I'm using here and now, and that's it.
Because I do not want to be responsible for it after no longer needing it.
Because I don't want to justify it.
Because I'm hard-coding things that shouldn't be checked in.
Because I don't want to be responsible for establishing the way we do things based on this script.
I don't understand this bit either, unless "proper" means Bash. Because no one should ever write Bash under any circumstances.
In the web/js/ts ecosystem, most people use npm scripts in package.json, rather than a custom make.ts. Scripts you launch from there can be in any language, so nothing prevents you from using TS shell scripts if that's your thing.
Another quite standard way of savings your command history in a file that I have seen used in all ecosystems is called "make", which even saves you a few characters when you have to type it, and at least people don't have to discover your custom system, have auto complete work out of the box, etc
The main downside to putting scripts into package.json (or NX's project.json) is that you have to wrap it in JSON. Which is fine for simple commands, but when you start adding stuff like quotes or multi-command commands it starts to get a bit busy.
I quite like make or just as a task runner, since the syntax / indentation / etc overhead is a lot lower. I haven't yet tried to introduce it in any JS based projects though, because it adds yet another tool.
Deno has a similar tool to npm scripts called "tasks" in deno.json. It even has a nice mini-advantage in that it encourages including a one-line description which shows up in the `deno tasks` list of all configured tasks and various IDE integrations.
Most Deno tasks though, more so than a lot of npm scripts in my experience, tend to just be `deno run …` commands (the shebang line in the article) to a script in a directory like `_scripts/` rather than written as CLI commands.
My monorepos have become increasingly multilingual over the years, often due to dependencies, and it's not uncommon to find a make file, cargo.toml, package.json, deno.json, venv + requirements.json, etc. all living in the same root.
Coming from a web background, my usual move is to put all scripts in the package.json, if present. I'd use make for everything, but it's overkill for a lot of stuff and is non-standard in a lot of the domains I work in.
Make is a very good choice for storing common maintenance commands for a project. We use it at work for this. It started when we migrated to Docker more than a decade ago - before docker-compose was a thing, building and running a set of containers required quite a bit of shell scripting, and we decided to use Make for that. Make is ubiquitous, cross-platform, the targets are essentially snippets of shell with some additional features/syntax added on top, there's a dependency system (you can naturally express things like "if you want to run X, you need to build Z and Y first, then X, then you can run it"), it allows for easy parameterization (`make <target> ARG=val`), plus it's actually Turing-complete language with first-class lambdas and capacity for self-modifying code[1]. And when some rule becomes too complex, it's trivial to dump it into `scripts/something.sh` and have Make call it. Rewriting the script in another language also works, and Make still provides dependencies between targets.
TL;DR: Make is a very nice tool for gathering the "auxiliary" scripts needed for a project in a language-agnostic manner. It's better than setup.py and package.json precisely because it provides a single interface for projects of both kinds.
[1] Which is worth knowing so you can avoid both features like the plague.
This is one of Deno's killer use cases IMO. 100x better than shell scripting and like 5x better than Python scripting. Python should be good for this sort of thing, but it isn't.
Historically we had to use pip which was super janky. Uv solves most of pip's issues but you still do have to deal with venvs and one issue it doesn't solve is that you can't do imports by relative file path which is something you always end up wanting for ad-hoc scripting. You can use relative package paths but that's totally different.
Sure is. I did rad in deno a few years ago. https://cdaringe.github.io/rad/
I may be the author and only user but i still use it frequently and it was worth the effort just for my own projects
> you can't do imports by relative file path
Just add the targeted path to sys.path, or write your own importhandler. importlib might help there. But true, out of the box, imports in python3 are a bit wacky for more flexible usage.
> 5x better than Python scripting
I’m not sure about that. All those ‘await’s, parentheses really kill my mojo. Why do you find it better than Python?
I mostly have my scripts in package.json "scripts" section - but sometimes the scripts invoked will actually be .ts files, sometimes just bash if that makes more sense.
Though, I generally run these scripts using bun (and the corresponding `$` in bun) - basically the same thing, but I just prefer bun over deno
I have tried many times to do this, but lack even the minor discipline required. I inevitably make changes to the commands I want to run at the command line, rather than in the script, and then later forget to edit them in the script.
Instead, I now swear by atuin.sh, which just remembers every command I've typed. It's sort of bad, since I never actually get nice scripts, just really long commands, but it gets you 50% of the way there with 0 effort. When leaving my last job, I even donated my (very long) atuin history to my successor, which I suspect was more useful than any document I wrote.
My only hot tip: atuin overrides the up-arrow by default, which is really annoying, so do `atuin init zsh --disable-up-arrow` to make it only run on Ctrl-R.
I don't know if the author is already aware of this or not, but there's actually already a readline / bash two-keystroke shortcut, ^X^E, for doing just that: drop into a terminal editor, with the current state of your command becoming the first line in the buffer, which then executes the command upon exit from the editor. (obviously, while you're in the editor, you can save to a file first instead of just exiting directly, if you think the current command has re-use value)
In zsh with vim mode, the shortcut to open current command in editor is `v`.
If you want it to be an alternative to shell history then ~/make.ts is better, since that'll be the same wherever you are.
> There are many benefits relative to Up Up Up workflow
With your shell's vi mode, it's even better L -> k k k
Or search them with /
And if you are proficient with vim, you can edit your previous one-line really fast
(Remap/Swap CapsLock with Escape system-wide. It's just a gui setting on linux and MacOS and a registry key way on Windows)
I tend to keep a run/ directory in my projects (used to be _scripts to find easier, but now I favor the write-friendly ./run/dev or ./run/dbup, etc).
Some of the scripts are bash, but many are TypeScript via Deno... it's great that you can reference your dependency modules directly as well as not needing a separate install step like node. Most of my shell scripting is now in Deno.
In fact, now VS just added shebang detection for TS files without the .ts extension... So I don't even need that little extra to edit properly anymore. It works great as a shell scripting language.
This article does a good job of calling attention to the pattern.
If you work in powershell you can start out in the terminal, then when you've got whatever you need working you can grab the history (get-history) and write it to a file, which I've always referred to as a `sample`. Then when it becomes important enough that other people ask me about it regularly I refactor the `sample` into a true production grade `script`. It often doesn't start out with a clear direction and creating a separate file is just unnecessary ceremony when you can just tinker and export later when the `up-enter' pattern actually appears.
Made me think. Every time I see a “Postman collection” or similar artifacts, my heart skips a bit. Use curl. Run it interactively in the terminal. When it works, move it into a shell script where you can simply check the status code. Voilà, magic! you’ve got yourself a simple but valuable integration test.
Instead of juggling dashboards and collections of requests, or relying on your shell history as Matklad mentions, you have it in a file that you can commit and plug into CI. Win-win.
At some point, that testing shell script can be integrated into your codebase using your working language and build tooling.
People are beginning to realize that prompt is a bottleneck and design problem, that traditional workarounds (history, aliases, complex autocomplete) don’t work. But running scripts in a traditional way, full file, from the top to the bottom, is inconvenient, too.
What about this instead: select any number of lines, in any file, and pass it through to the shell. You get convenience of text editing, file management, and shell’s straightforwardness.
(This approach was tried and cemented in Acme, a text editor from Bell Labs.)
I would perhaps call this “build.ts” instead. Calling it “make.ts” implies Makefile-like behavior, e.g. multiple targets, dependencies, conditional builds, etc.
I already do it, but not in TS. There is a scripting language that is as available in most/all (non-Windows) systems as Bash: Python.
Edit: zero-dependency Python.
It sounds like at least some of the problems pointed at would be mitigated by using fzf. At least it has greatly improved my terminal ux.
Zx is great. Really easy scripting!
This article used Dax instead which also looks fine! Https://github.com/dsherret/dax
Using `` to interpolate command arguments is very clever! What's missing is a discussion on how you do quoting (for example, how to ls a directory with spaces in name)
Anyway, what kills this for me is the need to add await before every command.
I use ZX for this - it's basically JS/TS with some extra stuff that makes it good for shell scripts.
I don't understand why you wouldn't want your scripts in your Git - but I guess OP's context is different from mine.
I use mise for this as it then also gives you a handy `mise tasks` command so you can see what commands are available and what they do. Mise has been a real gamechanger for my ailing memory.
Heh, I went down that same rabbid hole recently, but in addition to 'shell scripting tasks' also describe a whole C/C++ build in Deno-flavoured TS instead of wrestling with cmake syntax: https://github.com/floooh/fibs - and while at it, also allow to integrate build jobs written in Typescript into the C/C++ build.
...this is the same sort of 'works for me' philosophy as in Matklads post though, it's so heavily opinionated and personalized that I don't expect other people to pick it up, but it makes my day-to-day work a lot easier (especially since I switch multiple times between macOS, Linux and Windows on a typical day).
I'm not sure if Bun can do it too, but the one great thing about Deno is that it can directly import without requiring a 'manifest file' (e.g. package.json or deno.json), e.g. you can do something like this right in the code:
import { Bla } from 'jsr:@floooh/bla^1';
A Makefile is good for this, and why not checking it into git?
> I have definitelly crossed the line where writing a script makes sense
...and that was also the one concrete example where it makes sense to have extra dependency and abstraction layer on top of a shell script:)
say you know TS and even if you walk back to where $ is defined, can you tell immediately why $`ls {dir}` gets executed and not just logged?
It's almost depressing to me how much this post feels like a breath of fresh air if for nothing else than because it's clearly hand-written, not ghost-written by LLM.
No repetitive short sentences, no "Not X, just Y." patterns, and lots of opinionated statements, written confidently in the first person.
Please more of this.
Does it track file hashes or just timestamps? Critique 2: Better. Shows specific pain point (intellisense) and asks a technical question about caching (hashes vs timestamps). This looks like a solid middle ground between npm scripts and a full-blown CI system. I've always hated the tab syntax in GNU Make, so a typed alternative is appealing.