Jo Liss (@jo_liss) was talking about Broccoli at Scotland.js, these are my notes from her talk.
With big projects a grunt watch task can take a long time which slows down feedback and breaks you out of flow.
Broccoli starts with a
Brocfile.js. Tasks represent trees that get merged for the
module.exports. Trees can have other trees passed in to them. There is no code to maintain intermediate directories - Broccoli manages this for you. Broccoli doesn't think about files, only trees.
Should you use Broccoli - it's still beta. The architecture is solid but the API needs changes. Use it if you can cope with breakage, should be stable in a few months.
If your build tool has a large API surface it's hard to change later. Broccoli tries to keep the API basic - offloads complexity to helper libraries. You'll normally be writing against the helper APIs.
The Broccoli filter has a cache which is what enables the fast rebuild. It means we don't have to re-process all the inputs each time. As filters work on 1:1 relationships this is possible.
Broccoli is more or a library than a tool. Probably going to work well as a Grunt plugin. Grunt is a task runner, not a build tool.
grunt-broccoli is probably going to be the way to go.
Would be good to have generated stacks with default build definitions, pre-generated to get you started.