Replace watcher.js with broccoli-sane-watcher.

[jschilli]

Replaces watcher.js with @krisselden's broccoli-sane-watcher

Should resolve #72 and replace #150

On our largish ember.js project we went from consistent 25% cpu with current watcher to roughly 0% using the sane-watcher

cc: @stefanpenner @rondale-sc @krisselden

[stefanpenner]

nice!

[stefanpenner]

I think... there might be a bug in the watcher.
broccolijs/broccoli-sane-watcher#3

We should confirm, if it is the case I imagine it wont be hard to fix.

[stefanpenner]

broccoli-sane-watcher has been updated.

[amasad]

As another data point of this vs Chokidar:

1. Chokidar depends on native modules (annoying at the least and makes it unusable in some settings were you can't install native modules)
2. I tried to use Chokidar at Facebook and was impossibly slow to startup and had a delay on change events when watching a lot of files.
3. Eats up a lot more resources than sane because creates a lot more watcher instances
4. Not sure if you care about that here but chokidar has no ready event so if you have a big project and it's crawling, creating instances, it's impossible to know when it's done

[krisselden]

That's a TODO, I'm not using the ready event yet, either outputing to the user that it is now in fact watching, or double checking that the tree is still the same as when it was added.

[krisselden]

I had planned broccoli-sane-watcher to become a PR for broccoli, not for broccoli to depend on it. Right now that is circular, since broccoli-sane-watcher depends on broccoli for printSlowTrees. If we go this route, printSlowTrees should be a module, and watching should be more pluggable with broccoli. /cc @joliss @stefanpenner

[stefanpenner]

@krisselden i'll extract the logging for you -> broccolijs/broccoli-sane-watcher#5 (assigned myself)

[krisselden]

@joliss should this be closed? You have a different plan for integration of broccoli-sane-watcher, confirm?

[stefanpenner]

@krisselden ya i believe she does.