'let' to intialize variable

[xeoneox]

I'm pretty happy with your simplified shell syntax, except for having to use blanks during variable init, like:

_, tags = `git tag | head -1`

That makes sense to Go and Lua devs, but not for shells. In GDscript, which is syntactic sugar to make Lua look more like Python, it would be:

let tags = `git tag | head -1`

I think that makes more sense with your intended audience. What do you think? Have you looked at GDscript before?

[amterp]

except for having to use blanks during variable init, like

To clarify, you don't usually have to in rad. Variable initialization is simply

name = "bob"

The _, syntax you're seeing is because shell commands return 3 variables:

exit_code, stdout, stderr = $!`git tag | head -1`

You don't need to declare all 3, so if you're only interested in the exit code and stdout, you can just do

exit_code, stdout = $!`git tag | head -1`

But, if you're not interested in the exit code and just want stdout, you can do

_, stdout = $!`git tag | head -1`

where _ is communicating "I don't care about this variable; I won't use it" and exists purely to absorb the first returned value so you can get to the later ones you care about.

I don't think a let keyword will change that, unless I'm missing what you mean? I've not looked at GDScript, but I can see it shares similarities with Python & friends, so open to there being some good ideas in there!

[amterp]

Tangent, but I'm also considering changing shell commands a bit. I implemented a catch keyword semi-recently in f01081b (I've not documented it on the docs website yet), and I think we can perhaps use it for shell commands as well so we don't have that unique recover/fail syntax (dunno if you've come across that in docs). Anyway, dw, just flagging as potentially relevant, if we're discussing shell command syntax :)

[xeoneox]

I mean it makes sense, that's why we have blank assignments, I'm just saying they should be hidden in something like Rad. Like if not equal 0, then return stderr instead, with a text header saying 'err' or something. It's boilerplate I think can be simplified 90% of the time.

Try / Catch sounds good too

[amterp]

Gotcha, yeah I can get behind changing the output of shell commands, at least the ordering, since I've also found the most common usage pattern to be treating them as critical and then only really using stdout. When I had initially implemented them, I leant quite hard into safety e.g. giving error code first, encouraging recover/fail pattern, but I don't think that makes sense in hindsight.

Would changing the ordering to make error code the last output work for you? I'm not certain returning a map/struct is overall gonna be better, outside of the particular case where you just want the last returned value. But am open to it!

[xeoneox]

Yes, I think so. I am not the language designer here, my only thing is on-boarding users with something easy and gradual, and it's jarring to have have placeholders and what not before you need them. So people won't have to look up the standard error until they need it, etc.

Feel free to handle it however

[amterp]

I've made some changes here such that blank assignments no longer are necessary. You can instead assign with specifically stdout or stderr, and that's what will get captured:

stdout = $`make build`

Can read more here in the guide: https://amterp.github.io/rad/guide/shell-commands/#named-assignment

Let me know if you have thoughts!