I wonder if it's worth showing brew --cellar to accommodate alternative brew prefixes?

RUBIES+=("$(brew --cellar ruby)"/*)

On the other hand, the hardcoded /usr/local version is prettier.

I didn't think about that option.

Not sure if it could be a custom path for it. However, as it's documentation, not a script, it could have both options documented.

In my case, I rather set the path, as this would be one command less to execute every time I start a terminal session.

Those are good points. I think it's fine to hard code to the brew prefix.

It'd be nice to add engine-version to RUBIES rather than just version. With Ruby 2.6.3 for example, it'd currently be 2.6.3 instead of ruby-2.6.3.

I also wonder about other Ruby engines that brew has packages for. Is it worth adding those too, or maybe just separate examples for Ruby and JRuby?

for ruby_engine in jruby mruby ruby; do
	for ruby in /usr/local/Cellar/$ruby_engine/*; do
		RUBIES+=("${ruby%/*}/$ruby_engine-${ruby##/*/}")
	done
done

I can't tell about the jruby ones, as I just use homebrew for it. Sounds good to me, but as I can't test it, I wouldn't dear :D

I was going to propose you this:

RUBIES+=(
  /usr/local/Cellar/jruby/*
  /usr/local/Cellar/mruby/*
  /usr/local/Cellar/ruby/*
)

But here is my experience...

First of all, I installed (throw homebrew) jruby and mruby to check it out.

The binary directories would be then:

• /usr/local/Cellar/ruby/2.6.5/bin/
• /usr/local/Cellar/jruby/9.2.9.0/bin/
• /usr/local/Cellar/mruby/2.0.1/bin/

If I run your loop, it would check for this:

/usr/local/Cellar/jruby/jruby-9.2.9.0
/usr/local/Cellar/mruby/mruby-2.0.1
/usr/local/Cellar/ruby/ruby-2.6.5

Which is not right, as you can see, so you can just run it like:

for ruby_engine in jruby mruby ruby; do
	for ruby in /usr/local/Cellar/$ruby_engine/*; do
		echo "${ruby%/*}/${ruby##/*/}"
	done
done

But what would be the point? It would do the same as what I propose:

RUBIES+=(
  /usr/local/Cellar/jruby/*
  /usr/local/Cellar/mruby/*
  /usr/local/Cellar/ruby/*
)

But there is a problem here:

> chruby
   2.6.5
   9.2.9.0
   2.0.1
> chruby 2.6.5
> ruby --version
ruby 2.6.5p114 (2019-10-01 revision 67812) [x86_64-darwin19]
> chruby 2.0.1
chruby: /usr/local/Cellar/mruby/2.0.1/bin/ruby not executable
> chruby 9.2.9.0
chruby: /usr/local/Cellar/jruby/9.2.9.0/bin/ruby not executable

You see what it's doing? and why it fails? is because it is trying to get the {path}/bin/ruby binary, but in the case of jruby and mruby, the binaries are called jruby and mruby, so it fails.

In conclusion:

1. I don't see the point of adding a for loop, I think that it would be better to use just the * wildcard.
2. It will not work anyways, unless there is a way to tell chruby the name of the binary (I don't now if there is a way or not), in which case, this is not the purpose of this PR, I think.

So I suggest to get stick with the current proposal.