#285 do not recreate all components every time + optimize document co…

…llection filter

src/narraint/frontend/ui/views.py

@@ -42,7 +42,7 @@
 from narraint.queryengine.result import QueryDocumentResult, QueryDocumentResultList
 from narraint.ranking.corpus import DocumentCorpus
 from narraint.ranking.indexed_document import IndexedDocument
-from narraint.recommender.recommendation import apply_recommendation
+from narraint.recommender.recommendation import RecommendationSystem
 from narrant.entity.entityresolver import EntityResolver
 
 logging.basicConfig(format='%(asctime)s,%(msecs)d %(levelname)-8s [%(filename)s:%(lineno)d] %(message)s',
@@ -76,6 +76,7 @@ def __new__(cls):
             cls.explainer = EntityExplainer()
             cls.keyword2graph = Keyword2GraphTranslation()
             cls.corpus = DocumentCorpus()
+            cls.recommender = RecommendationSystem()
         return cls._instance
 
 
@@ -654,7 +655,8 @@ def get_query(request):
             results, cache_hit, time_needed = do_query_processing_with_caching(graph_query, document_collection)
             result_ids = {r.document_id for r in results}
             opt_query = QueryOptimizer.optimize_query(graph_query)
-            View().query_logger.write_query_log(time_needed, "-".join(sorted(document_collection)), cache_hit, len(result_ids),
+            View().query_logger.write_query_log(time_needed, "-".join(sorted(document_collection)), cache_hit,
+                                                len(result_ids),
                                                 query, opt_query)
 
             results = TitleFilter.filter_documents(results, title_filter)
@@ -1473,13 +1475,15 @@ def get_data_sources(request):
     except Exception:
         return HttpResponse(status=500)
 
+
 def get_classifications(request):
     try:
         available_classifications = ClassificationFilter.get_available_classifications()
         return JsonResponse(status=200, data=dict(classifications=available_classifications))
     except Exception:
         return HttpResponse(status=500)
 
+
 def get_recommend(request):
     results_converted = []
     is_aggregate = False
@@ -1498,7 +1502,6 @@ def get_recommend(request):
         if not request.GET.keys():
             return HttpResponse(status=500)
 
-
         document_id = int(request.GET.get("query", ""))
         query_collection = request.GET.get("query_col", "")
         query_trans_string = document_id
@@ -1553,7 +1556,7 @@ def get_recommend(request):
             valid_query = True
             logging.info(f'Requested recommendation for document id: {document_id}')
 
-            json_data = apply_recommendation(document_id, query_collection, document_collections, View().corpus)
+            json_data = View().recommender.apply_recommendation(document_id, query_collection, document_collections)
 
             results = []
             graph_data = dict()

src/narraint/recommender/first_stage.py

@@ -10,12 +10,13 @@ class FirstStage:
     Previously called FSConceptFlex, now default first stage implementation.
     """
 
-    def __init__(self, extractor: NarrativeCoreExtractor, document_collections: set):
+    def __init__(self, extractor: NarrativeCoreExtractor):
         self.extractor = extractor
-        self.document_collections = document_collections
+        self.document_collections = None
         self.session = SessionExtended.get()
 
-    def retrieve_documents_for(self, document: RecommenderDocument):
+    def retrieve_documents_for(self, document: RecommenderDocument, document_collections: [str]):
+        self.document_collections = list(document_collections)
         # Compute the cores
         core = self.extractor.extract_concept_core(document)
 
@@ -38,7 +39,10 @@ def retrieve_documents(self, concept: str, concept_type: str):
         # Search for matching nodes but not for predicates (ignore direction)
         q = q.filter(TagInvertedIndex.entity_id == concept)
         q = q.filter(TagInvertedIndex.entity_type == concept_type)
-        q = q.filter(TagInvertedIndex.document_collection.in_(self.document_collections))
+        if len(self.document_collections) == 1:
+            q = q.filter(TagInvertedIndex.document_collection == self.document_collections[0])
+        else:
+            q = q.filter(TagInvertedIndex.document_collection.in_(self.document_collections))
         document_ids = set()
         for row in q:
             document_ids.update(TagInvertedIndex.prepare_document_ids(row.document_ids))