Separate Flow DB

[bkuczenski]

FlowDB should stand alone. Workflow should be something like this:

• FlowDB is really a quantity DB. Has categories / compartments, quantities, and flow synonyms, as currently implemented. Maybe leave out the intermediate flow categories, allowing them to be project-specific (or better curate them).
• Flow entities store their own characterizations. The main operation of the FlowDB is to provide characterization values to flows when they request them.
• FlowDB stores "master lists" of characterizations, meaning fully location-specified.
• Queries to FlowDB take the form of current queries: given a flow and a quantity, and optional location, return the characterization value, which the flow can cache locally.

Meanwhile, processes and LCIA do not involve the flow DB. They strictly use the characterizations stored in the flows, which are pointed to by the exchanges, which are owned by the processes. LCIA computation is self-contained within the process.

• strike all references to flowdb in ForegroundManager and LcProcess
• these LCIA methods should uniformly use the process LCIA method
• bg_lcia requires some elaborate computation that has not been written yet.

[bkuczenski]

This is hard because the flowdb is how the foreground knows which flows are elementary. Why is that? because the FlowDB keeps track of the full compartment hierarchy.

implication: compartment manager should be separated from FlowDB. Foreground manager needs to know compartments but not DB. Compartment tree should be an input argument to flowdb, not defined within flowdb.