Fix issues #289 and #280 to have year and cohorts applied to multivar…

…iate analysis endpoint (#293)

* fixed issue #289

* make multivariate analysis endpoint to honor cohorts

icees_api/features/sql.py

@@ -536,16 +536,7 @@ def count_unique(conn, table_name, year, *columns):
         ).fetchall()]
 
 
-def select_feature_matrix(
-        conn,
-        table_name,
-        year,
-        cohort_features,
-        cohort_year,
-        feature_a,
-        feature_b,
-):
-    """Select feature matrix."""
+def create_cohort_view(conn, table_name, cohort_features):
     if not isinstance(cohort_features, list):
         cohort_features = [
             {
@@ -576,7 +567,28 @@ def select_feature_matrix(
         )
         conn.execute("DROP VIEW if exists tmp")
         conn.execute(view_query)
-        table_name = "tmp"
+        # return the view table name 'tmp' for subsequent operations
+        return 'tmp'
+    else:
+        return table_name
+
+
+def drop_cohort_view(conn, cohort_features):
+    if cohort_features:
+        conn.execute("DROP VIEW tmp")
+
+
+def select_feature_matrix(
+        conn,
+        table_name,
+        year,
+        cohort_features,
+        cohort_year,
+        feature_a,
+        feature_b,
+):
+    """Select feature matrix."""
+    table_name = create_cohort_view(conn, table_name, cohort_features)
 
     # start_time = time.time()
     # cohort_features_norm = normalize_features(cohort_year, cohort_features)
@@ -735,8 +747,7 @@ def select_feature_matrix(
 
     # association_json = json.dumps(association, sort_keys=True)
 
-    if cohort_features:
-        conn.execute("DROP VIEW tmp")
+    drop_cohort_view(conn, cohort_features)
 
     # start_time = time.time()
     # conn.execute(cache.insert().values(digest=digest, association=association_json, table=table_name, cohort_features=cohort_features_json, feature_a=feature_a_json, feature_b=feature_b_json, access_time=timestamp))
@@ -791,9 +802,13 @@ def select_feature_count_all_values(
     return count
 
 
-def get_feature_levels(feature):
+def get_feature_levels(feature, year=None):
     """Get feature levels."""
-    return get_value_sets().get(feature, [])
+    feat_levs = get_value_sets().get(feature, [])
+    if year and feature == 'year' and int(year) in feat_levs:
+        # only include the pass-in year in the corresponding year feature level list
+        feat_levs = [int(year)]
+    return feat_levs
 
 
 def apply_correction(ret, correction=None):
@@ -990,7 +1005,7 @@ def compute_multivariate_table(conn, table_name, year, cohort_id, feature_variab
     associations = []
     # get feature_constraint list from the first feature variable
     feat_constraint_list = []
-    levels0 = get_feature_levels(feature_variables[0])
+    levels0 = get_feature_levels(feature_variables[0], year=year)
     for level in levels0:
         non_op_idx = 0
         for lev in level:
@@ -1015,7 +1030,7 @@ def compute_multivariate_table(conn, table_name, year, cohort_id, feature_variab
                 "feature_name": feature_variables[index],
                 "feature_qualifiers": list(map(
                     lambda level: {"operator": "=", "value": level},
-                    get_feature_levels(feature_variables[index]),
+                    get_feature_levels(feature_variables[index], year=year),
                 ))
             }
         ]
@@ -1024,7 +1039,7 @@ def compute_multivariate_table(conn, table_name, year, cohort_id, feature_variab
                 "feature_name": feature_variables[index + 1],
                 "feature_qualifiers": list(map(
                     lambda level: {"operator": "=", "value": level},
-                    get_feature_levels(feature_variables[index + 1]),
+                    get_feature_levels(feature_variables[index + 1], year=year),
                 ))
             }
         ]
@@ -1053,7 +1068,7 @@ def compute_multivariate_table(conn, table_name, year, cohort_id, feature_variab
     if index < feat_len:
         feature_qualifiers = list(map(
             lambda level: {"operator": "=", "value": level},
-            get_feature_levels(feature_variables[index])
+            get_feature_levels(feature_variables[index], year=year)
         ))
         more_constraint_list = []
         for feature_constraint in feat_constraint_list:
@@ -1066,6 +1081,10 @@ def compute_multivariate_table(conn, table_name, year, cohort_id, feature_variab
                 more_constraint_list.append(base_dict)
         feat_constraint_list = more_constraint_list
     # compute frequency for each feature constraint
+    if cohort_features:
+        # compute frequency on the cohort view
+        table_name = create_cohort_view(conn, table_name, cohort_features)
+
     if len(feat_constraint_list) > 0:
         columns = list(feat_constraint_list[0].keys())
         result = count_unique(conn, table_name, year, *columns)
@@ -1078,4 +1097,6 @@ def compute_multivariate_table(conn, table_name, year, cohort_id, feature_variab
         for fc in feat_constraint_list:
             fc['frequency'] = get_count(result, **fc)
 
+    if cohort_features:
+        drop_cohort_view(conn, cohort_features)
     return feat_constraint_list