Update chokidar to v4

[Fuzzyma]

I would like to propose to update chokidar to v4. It reduces its dependency count by 11 deps and also brings better type safety.
This is part of the ecosystem cleanup initiative (https://e18e.dev). We try to modernize and declutter the ecosystem as best as we can.
Tailwind is downloaded over 46 million times per month and reducing its dependencies will have a huge impact and will lead to faster install times in ci and on dev machines.

I would be happy to create a PR with the necessary changes if you are open for that :)

PS: Sorry for misusing the bug tracker. I didn't know where to put it

[philipp-spiess]

@Fuzzyma Hey! Really appreciate the help here, reducing the number of dependencies has been a huge focus for us too for v4.

I do think though that we'd rather focus on getting a release for Tailwind CSS v4 out instead. That should also take care of this issue since we're no longer relying on chokidar! I don't think it's going to be trivial either to upgrade to v4 since we're relying on the glob support heavily unfortunately (the Tailwind JS config API accepts a list of globs to the content array).

[Fuzzyma]

Oh I see! Yes, I should have figured that the content prop is powered by chokidars glob.
That's unfortunate :D.

But thanks for the prompt reply and explanation!

I am looking forward to the new tailwind version and I hope that most people will upgrade right away :).
My idea was that a lot of people will not bother upgrading and therefore v3 should be as lean as possible. But I agree that this is not an easy thing to do here.

Keep up the good work!!

[talentlessguy]

a suggestion on how it could work: use is-glob and tinyglobby and do something like this:

chokidar.watch(paths.map(p => isGlob(p) ? tinyglobby(p) : p))

[benmccann]

Good suggestion. One more on top of that: you don't need is-glob because you can use isDynamicPattern from tinyglobby, which does the same thing

[Fuzzyma]

@philipp-spiess I looked into the code of v3 and it seems like the glob functionality is not actually used. Instead the content paths are resolved via fastGlob. I can be totaly wrong here. But if thats the case, the update should be trivial (which would be great)