Originally, dataflow processor was written as a two-way fold, but this produced problems when trying to resolve function calls
which require information regarding the calling and definition context. While this only is a problem for late bindings as they happen
with functions (and probably quote'd R-expressions), it is still a problem that must be dealt with.
Therefore, the dataflow processor has no complete control over the traversal and merge strategy of the graph, with each processor being in
the position to call the other processors as needed for its children.
Now this method can be called recursively within the other processors to parse the dataflow for nodes that you cannot narrow down
in type or context.
Originally, dataflow processor was written as a two-way fold, but this produced problems when trying to resolve function calls which require information regarding the calling and definition context. While this only is a problem for late bindings as they happen with functions (and probably quote'd R-expressions), it is still a problem that must be dealt with. Therefore, the dataflow processor has no complete control over the traversal and merge strategy of the graph, with each processor being in the position to call the other processors as needed for its children.
Now this method can be called recursively within the other processors to parse the dataflow for nodes that you cannot narrow down in type or context.