> A small drawback is that just isn’t readily available on developer’s machines (unlike bash and make)
Looking back, this was usually a substantial stumbling block for adoption of new tooling. Things have changed. All one has to do is look at what JS programmers will go through to scratch build entire toolchains on the back of npm. IMO, the bar has been moved up to allow _some_ installation friction rather than none.
IMO, if the tool is a solid improvement and installation is a one-liner, use it and promote it loudly so it continues to get support.
Even for projects where Make isn't part of the build pipeline, like Rails or C++, I have a standardized makefile I put in the project root, using includes for "secrets.mk", so I can do things like "make deploy" and it Just Works(tm).
Another neat and simple trick it to have the first target in the Makefile print out a helpful message that tells the reader what the project is and how to use it. One of those things that is useful both when you are working on the project and when you later return to it, years later.
I use this pattern for my makefiles. First item is for "help", and it dynamically scrapes all the other commands (assuming format <command>: ## <description>). So if I do 'make help' it'll list all the commands in the makefile without much extra work from me.
I wish mise has better secrets management or allows for split config e.g. a repo wide config that's to be shared and a personal config that can override whatever's necessary. It's difficult to commit mise.toml because the [env] section almost always contains sensitive variables.
Historically, I did it the other way around: have some autogenerated shell commands (dev, test, pull...etc) on my local machine that detect which language the project in the current folder is using, and calls commands accordingly.
I did it this way because, in my case, it was challenging to consolidate multiple projects across multiple organizations under a single tool.
I like to use a handful of stable shell scripts as the entry points, this way you can change the tooling under the hood without having to update related callers. E.g. you can break up with make or mise or whatever, but you don't need to update your CI. Run.sh runs, build.sh builds, etc.
Also had this issue. I have made a meta-task runner to make discovery and running of already defined make/just/task/npm/etc task targets easier called dela[0].
The ergonomics of running tasks are very similar to the article as you can run a task simply by typing the name of the task in your shell without saying "make" or "npm run" or whatever else that is the actual task runner. The difference is that you just have to install dela without having to change anything in the underlying project repository.
I write a Python script for this stuff. argparse with subparsers makes (fairly) light work of creating a command line tool with git-/svn-/etc.-style subcommands, and you get very useful autogenerated --help output. (The fact you have to statically specify your program's command line interface up front is maybe a bit un-pythonic, but it's all the better for it.)
Python is always measurably more lines than the Makefile or shell script equivalent, but just think of the effect on your productivity metrics!
The standard shutil and subprocess modules are pretty convenient to use as-is. They're a bit syntactically tedious for anything that's a very basic sequence of shell operations, but in exchange you'll never forget to quote the arguments correctly. And I've always found that these processes to tend to grow over time anyway - something that's much easier to handle in Python! There's also far more scope for additional error checking and nicer UX.
I was on the "just use makefiles" boat way too long, except it was Just at the time that was calling for my attention. At first glance, they look similar. Make's already there, so what's the big deal?
And then I use Just for a week and become a diehard fan. Yeah, it's similar to make, except optimized for developer ergonomics instead of feature completeness. What if... the tool actually supported and encouraged you instead of making you fight it?
Later I came to prefer mise over just. It does more common developer stuff like managing tools and env vars, and I prefer the way it handles embedded scripts. In any case, I vastly prefer either of them to make for the kinds of languages I mostly use.
Even if I did prefer make, which one? The GNU make that ships with macOS is ancient, so I'd probably have to install a newer one, and if I'm going to have to install a tool anyone, might as well make the effort to make that tool be mise (or just or just about anything else).
I have been using just for some projects and really like it. I looked at mise to replace direnv in the few places I use it, but between mise trying to do so many things and the in terminal advertisement for a problematic company I want nothing to do with, I have halted all plans I had for learning more about mise.
What in terminal ads do you mean? I don’t recall ever seeing an ad of any kind in mise. Not saying you’re wrong, just that our experiences are different here.
I'm not going to argue since I haven't used mise, but how beginner friendly is it? The nice thing about "just" is literally the "can learn in an afternoon". When I was in a team of, well, less motivated developers, that mattered. It meant they could modify and maintain the justfile without everything depending on me.
Of course, if those aren't dynamics you have to deal with, definitely go with what works.
There's also invoke (https://www.pyinvoke.org/). I suspect it too is more powerful than just, and I know teams that use them.
One issue with Just is that by default each line in a recipe is executed in a sub shell. Set a variable in line 1 and you can’t use it in line 2. Mise `run` blocks are embedded shell scripts. Alternatively, you can pull those scripts out into individual files so that you can edit them with syntax highlighting etc with any editor that understands shell scripts. Perhaps you can with just now, too, but mise makes it trivially easy. You can get away without a mise.toml at all if you just put your scripts in the right directory.
> One issue with Just is that by default each line in a recipe is executed in a sub shell. Set a variable in line 1 and you can’t use it in line 2. Mise `run` blocks are embedded shell scripts.
just supports the same. If you begin a recipe with a shebang (#!/bin/bash), then it dumps the whole recipe to a file and executes it as a shell script.
Not limited to bash - you can do the same with Python, etc.
As for scripts in their own files vs embedded in a justfile: I've noticed people's opinions vary. Even in my team we'd debate the issue. I prefer them in the justfile so I can read them in one place - as long as it's not huge. Some team members preferred them in their own files.
There was something about the shebang method that made someone else that I wanted to use not work but I forget the details. In any case, just is very nice — I just like mise more at the moment.
The own-file bit comes in nice when you want to run them standalone (CI/CD; maybe someone doesn’t want to use the runner) or you’re fighting against escaping rules. I like having them in the mise file when they’re just a couple of lines. If it turns into a full-on thing, then I prefer to extract it.
Looking back, this was usually a substantial stumbling block for adoption of new tooling. Things have changed. All one has to do is look at what JS programmers will go through to scratch build entire toolchains on the back of npm. IMO, the bar has been moved up to allow _some_ installation friction rather than none.
IMO, if the tool is a solid improvement and installation is a one-liner, use it and promote it loudly so it continues to get support.