Matcher dictionaries for lookup

[msschwartz21]

• Issue 1: We are constantly re-creating dictionaries from gt to pred and pred to gt nodes (and edges).
  • Proposal: Store those dictionaries in the matched object. Or… on the nodes and edges? What is the best way to represent a matching between two graphs? I know we were wary of storing redundant information, but I think the efficiency trade-off is probably worth it, since we often compute the dicts anyways.
• Issue 2: I need edge matching to implement my matcher. It is not fully implied by/equivalent to a node matching.
  • Proposal: Store an edge matching along with the node matching
• Dictionary from gt → pred and pred → gt nodes and edges in matched object

[DragaDoncila]

Would we consider just storing the matched node/edge on the graph itself after matching? I guess the issue there arises where you do multiple matching, since we don't want to copy the graph, just store a reference. Still the matcher could store an attribute which is the match_key that tells you the name of the node/edge attribute to query. I ask because otherwise to dump all the metric info to file, one has to dump the graphs and the dictionaries, whereas if stored on the graph, you could just dump the graph. And the match key I guess... Lol