Github user mattyb149 commented on the issue:
https://github.com/apache/nifi/pull/2351
Thanks for the commit, great stuff! Do you think the processing of the
stack for each record will be ok in terms of performance impact? I wonder if
we'd be better off building a tree (basically a flow graph model) when building
a component map, along with another "index" map from component id -> node in
the tree. It might replace the need for other "inheritance" maps or property
maps (as the node could hold the properties). Then we can get the component and
traverse to the root during the filter?
---
