fix: support long texts for labels in ElasticsearchDocumentStore (#3346)

Author: anakin87

haystack/document_stores/elasticsearch.py

@@ -521,8 +521,8 @@ def _create_label_index(self, index_name: str, headers: Optional[Dict[str, str]]
             "mappings": {
                 "properties": {
                     "query": {"type": "text"},
-                    "answer": {"type": "flattened"},  # light-weight but less search options than full object
-                    "document": {"type": "flattened"},
+                    "answer": {"type": "nested"},
+                    "document": {"type": "nested"},
                     "is_correct_answer": {"type": "boolean"},
                     "is_correct_document": {"type": "boolean"},
                     "origin": {"type": "keyword"},  # e.g. user-feedback or gold-label

test/document_stores/test_document_store.py

@@ -799,6 +799,30 @@ def test_labels(document_store: BaseDocumentStore):
     assert len(labels) == 0
 
 
+@pytest.mark.parametrize("document_store", ["elasticsearch", "opensearch"], indirect=True)
+def test_labels_with_long_texts(document_store: BaseDocumentStore):
+    document_store.delete_index("label")
+    label = Label(
+        query="question1",
+        answer=Answer(
+            answer="answer",
+            type="extractive",
+            score=0.0,
+            context="something " * 10_000,
+            offsets_in_document=[Span(start=12, end=14)],
+            offsets_in_context=[Span(start=12, end=14)],
+        ),
+        is_correct_answer=True,
+        is_correct_document=True,
+        document=Document(content="something " * 10_000, id="123"),
+        origin="gold-label",
+    )
+    document_store.write_labels(labels=[label], index="label")
+    labels = document_store.get_all_labels(index="label")
+    assert len(labels) == 1
+    assert label == labels[0]
+
+
 # exclude weaviate because it does not support storing labels
 @pytest.mark.parametrize("document_store", ["elasticsearch", "faiss", "memory", "milvus1", "pinecone"], indirect=True)
 def test_multilabel(document_store: BaseDocumentStore):