Bulk create Transcription

arkindex/images/importer.py

@@ -104,35 +104,56 @@ def bulk_transcriptions(image, page, items):
 
     # Calc needed TrBox to build
     needed = required.difference(existing)
+    if not needed:
+        return []
 
-    zones = []
-    transcriptions = []
     with transaction.atomic():
 
-        # Create transcriptions and linked zones
-        for n in needed:
-            tr = Transcription.objects.create(
-                line=n.line,
-                text=n.text,
-                score=n.score,
+        # Raw elements
+        elements = Element.objects.bulk_create(
+            Element(type=ElementType.Transcription)
+            for _ in needed
+        )
+
+        # Build transcriptions & zones instances at the same time
+        transcriptions, zones = zip(*[
+            (
+                Transcription(
+                    element_ptr_id=elt.id,
+                    line=n.line,
+                    text=n.text,
+                    score=n.score,
+                ),
+                Zone(
+                    element_id=elt.id,
+                    image=image,
+                    polygon=n.box.to_polygon(),
+                )
             )
-            transcriptions.append(tr)
+            for elt, n in zip(elements, needed)
+        ])
+
+        # Create transcriptions using a low-level bulk_create
+        # as multi table is not supported yet by Django
+        Transcription.objects.none()._batched_insert(
+            transcriptions,
 
-            zones.append(Zone(
-                element_id=tr.id,
-                image=image,
-                polygon=n.box.to_polygon(),
-            ))
+            # Here is the magic: we need only to insert the fields from documents_transcription
+            fields=set(Transcription._meta.concrete_fields).difference(Element._meta.concrete_fields),
+
+            # Default
+            batch_size=None,
+        )
 
-        # Build zones in bulk
+        # Create zones in bulk
         Zone.objects.bulk_create(zones)
 
     # Create all links between transcription and page
     max_order_dl = ElementLink.objects.filter(parent=page).order_by('-order').first()
     max_order = 0 if max_order_dl is None else max_order_dl.order + 1
     ElementLink.objects.bulk_create(
-        ElementLink(parent=page, child=tr, order=i)
-        for i, tr in enumerate(transcriptions, max_order)
+        ElementLink(parent=page, child=elt, order=i)
+        for i, elt in enumerate(elements, max_order)
     )
 
     return transcriptions