custom mapping fix for issue deepset-ai#293

haystack/database/base.py

@@ -42,28 +42,24 @@ def to_dict(self):
         return self.__dict__
 
     @classmethod
-    def from_dict(cls, dict):
-        field_map = {}#{"context": "text"}
+    def from_dict(cls, dict, field_map={}):
         _doc = dict.copy()
-        #import pdb
         init_args = ["text", "id", "query_score", "question", "meta", "embedding"]
         if "meta" not in _doc.keys():
             _doc["meta"] = {}
         # copy additional fields into "meta"
         for k, v in _doc.items():
             if k not in init_args and k not in field_map:
                 _doc["meta"][k] = v
-        #pdb.set_trace()
         # remove additional fields from top level
         _new_doc = {}
-        for k,v in _doc.items():
+        for k, v in _doc.items():
             if k in init_args:
                 _new_doc[k] = v
             elif k in field_map:
-                temp = k
                 k = field_map[k]
                 _new_doc[k] = v
-        #pdb.set_trace()
+
         return cls(**_new_doc)
 
 

haystack/database/elasticsearch.py

@@ -128,6 +128,14 @@ def _create_label_index(self, index_name):
         }
         self.client.indices.create(index=index_name, ignore=400, body=mapping)
 
+    # TODO: Add flexibility to define other non-meta and meta fields expected by the Document class
+    def _create_document_field_map(self) -> Dict:
+        return {
+            self.text_field: "text",
+            self.embedding_field: "embedding",
+            self.faq_question_field if self.faq_question_field else "question": "question"
+        }
+
     def get_document_by_id(self, id: str, index=None) -> Optional[Document]:
         if index is None:
             index = self.index
@@ -161,7 +169,7 @@ def write_documents(self, documents: Union[List[dict], List[Document]], index: O
             index = self.index
 
         # Make sure we comply to Document class format
-        documents_objects = [Document.from_dict(d) if isinstance(d, dict) else d for d in documents]
+        documents_objects = [Document.from_dict(d, self._create_document_field_map()) if isinstance(d, dict) else d for d in documents]
 
         documents_to_index = []
         for doc in documents_objects: