David Heinemeier Hansson

It's fine for it to return nil when the command is not found. It's falsy.

I don't understand the concern you're raising: ``` irb(main):001> system "command -v yarn > /dev/null" => true irb(main):002> system "command -v yorn > /dev/null" => false ```

Ah, I see. So you're hitting this problem running it in a Windows shell? Does the rest of everything work otherwise? We typically don't spend much/any time on compatibility with...

This explores similar territory to https://github.com/rails/importmap-rails/pull/57. I've been hesitant on #57, though. It feels like a lot to solve a little. Maybe we could simply allow `.js(m|x)?` as the extension...

Please do explore this!

I'd be happy to see this.

I know this is a million years old, but just stumbled over it. It strikes me as odd this is only happening in the memory store. That implies to me...

Ha. Makes sense. The initial report was ancient but hadn't seen any updates. Will close. Thanks @pixeltrix 🙏

Seems like its better to raise on a missing attribute than to silently return nil? Seems like a way you'd get screwed on spelling errors?

This seems like a sound strategy to me. I don't think there's anything hacky about a double-pass rendering process. The caveat re duplicate keys is similar to the caveats we...