Edge/Node constraints

[rjawesome]

biothings/biothings_explorer#795

Note: for subclass edges, the constraint is applied to the underlying edge. Ie. if a subclass edge represents A--treats-->B--subclass-->C then the constraints are applied to the A->B edge (and A/B nodes).

Example Query:

{
  "message": {
    "query_graph": {
      "edges": {
        "e0": {
          "predicates": [
            "biolink:treats"
          ],
          "subject": "n0",
          "object": "n1",
          "attribute_constraints": [
            {
              "name": "evidence_count_constraint",
              "id": "biolink:evidence_count",
              "operator": ">",
              "value": 30
            }
          ]
        }
      },
      "nodes": {
        "n0": {
          "categories": [
            "biolink:ChemicalEntity"
          ],
          "constraints": [
            {
                "name": "syn",
                "id": "biolink:synonym",
                "operator": "matches",
                "value": ".*MG.*"
            },
            {
                "name": "xr",
                "id": "biolink:xref",
                "operator": "matches",
                "value": "DrugCentral:1.*"
            }
          ]
        },
        "n1": {
          "categories": [
            "biolink:DiseaseOrPhenotypicFeature"
          ],
          "ids": ["MONDO:0002050"]
        }
      }
    }
  }
}

[tokebe]

@rjawesome Reviewed&tested and everything looks good. Hoping to get to dev today/monday.

Regarding your comment about node constraints possibly behaving weirdly with subclassing, could you explain what ambiguities arise?

[rjawesome]

@tokebe
Since the filtering for constraints occurs before subclass edges are created, constraints aren't applied to the subclass edge. Rather, the constraints are applied to the edge from the record (ie. the edge pointed to in the subclass edge's support graph).

ie. if a result contains the edge A-C-via_subclass which in the support graph points to the edge A-B and the edge B-subclass_of-C. the constraint is applied to A-B instead of A-C-via_subclass

[tokebe]

Thanks! I think that'll probably be fine for now, since the majority of things being constrained on would be represented in A-B rather than A-C-via_subclass.

[rjawesome]

@tokebe This PR should be good to go, with merge conflicts fixed, as well as working properly with subclassing.