Often it's useful to spawn tasks for each dependency result yielded. Got the idea from coala/coala-bears#1871:

we need a "link status bear" or "link bear" which extracts all links from a file and stores information about those links in a re-usable manner.

We could parallelize link-wise in this case with a bear dependency. But the general idea was born from that, that for each dependency we spawn a new task, instead of each file.

I'm not sure how the task generation shall exactly look like, there are multiple possibilities (because self.dependency_results is a dict):

• Pass every dependency-result to the analyze function: def analyze(self, bear, dependency_result)
• Provide every dependency result at once for each bear: def analyze(self, bear, dependency_results)
• Configurable behaviour allowing all of these (either using new bear classes or allowing to pass some settings).
• ...

Currently I'm favoring the 1st one.

The name of DependencyBear might change :)