- Big refactor of graph stuff.

  - It should be more correct now
  - I think it's slower
  - The first two tests succeed now, but:
  - With lots of concepts (e.g., 20,000), it takes forever and the whole
    backend won't respond while it's processing
    - The problem isn't in connect_nodes anymore, it's in all_paths when
      it's looking for simple paths to root. Those paths can get long
      and there can be a lot of them. Maybe a strategy of looking for
      predecessors would work?
  - It needs a lot more inline documentation
- I may have fixed the duplicate test thing by putting the loop through
  tests inside the loop through environments rather than other way
  around.

.run/backend.run.xml

@@ -5,6 +5,9 @@
     <module name="TermHub" />
     <option name="INTERPRETER_OPTIONS" value="" />
     <option name="PARENT_ENVS" value="true" />
+    <envs>
+      <env name="PYDEVD_USE_CYTHON" value="NO" />
+    </envs>
     <option name="SDK_HOME" value="" />
     <option name="WORKING_DIRECTORY" value="$PROJECT_DIR$" />
     <option name="IS_MODULE_SDK" value="true" />

backend/api_logger.py

@@ -42,7 +42,7 @@ async def start_rpt(self, request: Request, params: Dict):
                 if len(v) > 20:
                     # change any params with len > 20 to just log the len
                     rpt_params[k + '_len'] = len(v)
-                elif k == 'codeset_ids' or k == 'id':
+                elif k in ['codeset_ids', 'id', ]:
                     # put codeset_ids in a separate column (is this going to be helpful?)
                     codeset_ids = v
 

backend/db/ddl-19-apirun_groups.jinja.sql

@@ -25,8 +25,8 @@ WITH RankedGroups AS (
         SELECT
             *,
             ROW_NUMBER() OVER (PARTITION BY api_call_group_id ORDER BY timestamp::timestamp DESC) AS rn
-        FROM
-            public.api_runs
+        FROM public.api_runs
+        WHERE api_call_group_id IS NOT NULL
     )
 SELECT
     host,

backend/db/ddl-20-concept_graph.jinja.sql

@@ -1,14 +1,25 @@
 DROP TABLE IF EXISTS concept_graph CASCADE;
 
 CREATE TABLE IF NOT EXISTS concept_graph AS (
-     SELECT ancestor_concept_id AS source_id, descendant_concept_id AS target_id
+     SELECT ancestor_concept_id AS source_id,
+            -- 'Child of' AS relationship_id,
+            descendant_concept_id AS target_id
      FROM concept_ancestor
      WHERE min_levels_of_separation = 1
+     /*
      UNION
-     SELECT concept_id_1, concept_id_2
+     SELECT concept_id_1,
+            relationship_id,
+            concept_id_2
      FROM concept_relationship
-     WHERE relationship_id = 'Is a'
+     WHERE relationship_id IN ('Is a', 'Replaces')
+      */
 );
+
+CREATE INDEX cg_idx1 ON concept_graph(source_id);
+
+CREATE INDEX cg_idx2 ON concept_graph(target_id);
+
 /*
         # load_csv(con, 'relationship', 'dataset', schema='n3c')
         # rels = sql_query(con, f"""
@@ -38,7 +49,4 @@ CREATE TABLE IF NOT EXISTS concept_graph AS (
             ) y
         """
  */
-
-CREATE INDEX cg_idx1 ON concept_graph(source_id);
-
-CREATE INDEX cg_idx2 ON concept_graph(target_id);
+ SELECT 1; -- in case ending on a comment breaks the ddl parser