Support for multiple arguments in --before-build
#70
Comments
|
subshell |
|
direct |
|
|
|
@witoldsz Thanks for sharing your thoughts! It’s true, but I would argue you can put shell-specific syntax in a shell-specific executable. Accepting shell-specific syntax would introduce the ambiguity I mentioned: which shell do we use to run that? Also, if people want to share elm-live snippets, I’d rather they always work – not only in specific shells. |
|
"scripts" section of package.js (NPM) works fine, cross-platform, so this could work just fine. Also, most common things like pipes and redirects works everywhere (I guess)… |
|
Yeah, the scripts section works just fine – but many scripts don’t work cross-platform. I’ve worked in a cross-platform team before and npm scripts were one of the biggest pain points to keep portable. Despite following https://github.com/mattdesl/module-best-practices#task-running (I even helped write most of that), we constantly had to jump through hoops to keep it all humming along. Even in a pure Unix world, you still have different shells people use, and some of them (like fish) are not POSIX-compliant. As a closing point, npm scripts are designed to run once, so the performance penalty for spawning a subshell is only paid once. With elm-live, that penalty would be paid every time you save a file. But, let’s wait and see what others say. 🙂 I’ll leave the poll open for now, perhaps your comments would change the game. Thanks again for the valuable feedback! |
|
Poll results:
Seems like the direct option wins! I’d love to pick this up as soon as possible, this is a really nice feature! But it’s quite a crazy moment in my life – lots of things going on right now. So if someone needs this feature quickly and wants to jump in on this, I’ll be happy to support. 🙂 |
It could be worth mentioning that as a workaround for those who need some kind of shell, the "direct" can be used this way: |
|
Great suggestion @witoldsz! |
|
I have decided to remove both before and after built commands in the new alpha. If this is critical to your development workflow please open an issue with your use case |
We made a conscious decision to roll out an MVP
--before-buildand wait for feedback from our users. There are two options to take this further:Run the command inside a subshell, like this:
elm-live --before-build="my-command arg1 'arg 2'"This would probably feel pretty familiar for most of us. But this comes at a cost of poorer performance and implicitness (Which shell should we use to run the command? The current shell? A POSIX shell? cmd.exe?).
Spawn the command directly, like this:
elm-live --before-build=[ my-command arg1 'arg 2' ]This is more explicit and much faster, but may be a surprising syntax for you guys.
Where should we take this? You can vote below, we’ll be super grateful for your feedback!
The text was updated successfully, but these errors were encountered: