Adapters (check, quota, APAs) should communicate back to mixer on what input fields they used.

[guptasu]

For APAs and for other adapters too, we need a mechanism to figure out what input fields did the adapter used within a scope of a given request. This information along with operator configuration will help Mixer identify the smallest set of attributes used to serve a request, which indirectly will help reduce the attribute based cache key size and might result into higher cache hit rate.

For APAs, when we call the APA we need to tell it which of its outputs are actually going to be consumed and hence need to be populated. This will allow the adapter to avoid generating stuff that no one looks at, and will help the adapter reduce the set of input attributes it touches.

[guptasu]

@geeknoid do we already have an issue for this, I couldn't find it and therefore created it and referenced it from my APA to template model migration work (#1477)

[geeknoid]

I don't think we'll get much leverage for this idea for check & quota adapters. I'm much more interested in APAs at the moment.