CreateTranscriptions endpoint

arkindex/documents/api/ml.py

@@ -3,8 +3,8 @@ from django.db import transaction
 from django.db.models import Q, Count
 from rest_framework import status
 from rest_framework.generics import (
-    GenericAPIView, ListAPIView, ListCreateAPIView,
-    CreateAPIView, UpdateAPIView, RetrieveDestroyAPIView, RetrieveUpdateDestroyAPIView
+    GenericAPIView, ListAPIView, ListCreateAPIView, CreateAPIView,
+    RetrieveDestroyAPIView, RetrieveUpdateDestroyAPIView
 )
 from rest_framework.exceptions import PermissionDenied, ValidationError
 from rest_framework.response import Response
@@ -13,9 +13,10 @@ from arkindex.documents.models import \
 from arkindex_common.ml_tool import MLToolType
 from arkindex.documents.serializers.ml import (
     ClassificationsSerializer, ClassificationCreateSerializer, ClassificationSerializer,
-    TranscriptionSerializer, TranscriptionCreateSerializer, ElementTranscriptionsBulkSerializer,
-    DataSourceStatsSerializer, ClassificationsSelectionSerializer, ClassificationMode,
-    CountMLClassSerializer, AnnotatedElementSerializer
+    ClassificationsSelectionSerializer, ClassificationMode,
+    TranscriptionSerializer, TranscriptionCreateSerializer, TranscriptionBulkSerializer,
+    ElementTranscriptionsBulkSerializer, AnnotatedElementSerializer,
+    DataSourceStatsSerializer, CountMLClassSerializer
 )
 from arkindex.images.models import Zone
 from arkindex.project.filters import SafeSearchFilter
@@ -293,21 +294,16 @@ class ElementTranscriptionsBulk(CreateAPIView):
         return annotations
 
 
-class TranscriptionBulk(DeprecatedMixin, CreateAPIView, UpdateAPIView):
+class TranscriptionBulk(CreateAPIView):
     '''
-    Create multiple transcriptions at once, all linked to the same page
-    and to the same recognizer.
+    Create multiple transcriptions at once on existing elements
     '''
-    # Force DRF to ignore PATCH
-    http_method_names = ['post', 'put', 'head', 'options', 'trace']
     openapi_overrides = {
+        'operationId': 'CreateTranscriptions',
         'tags': ['transcriptions'],
     }
-    deprecation_message = (
-        'Creating or updating transcriptions with their own zones is now deprecated. '
-        'Please use CreateElementTranscriptions to push transcriptions in bulk '
-        'attached to sub-elements.'
-    )
+    permission_classes = (IsVerified, )
+    serializer_class = TranscriptionBulkSerializer
 
 
 class CorpusMLClassList(CorpusACLMixin, ListCreateAPIView):

arkindex/documents/serializers/ml.py

@@ -370,6 +370,55 @@ class AnnotatedElementSerializer(serializers.Serializer):
     created = serializers.BooleanField(default=False)
 
 
+class TranscriptionBulkItemSerializer(serializers.Serializer):
+    # Element retrieval and checks is done in the BulkSerializer to avoid duplicate queries
+    element_id = serializers.UUIDField(
+        help_text='ID of an existing element to add the transcription to'
+    )
+    type = EnumField(TranscriptionType, help_text='Type of the transcription')
+    text = serializers.CharField()
+    score = serializers.FloatField(min_value=0, max_value=1)
+
+
+class TranscriptionBulkSerializer(serializers.Serializer):
+    worker_version = serializers.PrimaryKeyRelatedField(queryset=WorkerVersion.objects.all(), default=None)
+    transcriptions = TranscriptionBulkItemSerializer(many=True)
+
+    def validate(self, data):
+        element_ids = set(transcription['element_id'] for transcription in data['transcriptions'])
+        found_ids = set(Element.objects.filter(
+            id__in=element_ids,
+            corpus__in=Corpus.objects.writable(self.context['request'].user)
+        ).values_list('id', flat=True))
+
+        missing_ids = element_ids - found_ids
+        if not missing_ids:
+            return data
+
+        # Return an error message with list indexes just like DRF's ListField, for easier debugging
+        errors = {}
+        for i, transcription in enumerate(data['transcriptions']):
+            if transcription['element_id'] in missing_ids:
+                errors[str(i)] = {
+                    "element_id": [f'Element {transcription["element_id"]} was not found or cannot be written to.'],
+                }
+        raise ValidationError({'transcriptions': errors})
+
+    def create(self, validated_data):
+        transcriptions = [
+            Transcription(
+                worker_version=validated_data['worker_version'],
+                element_id=transcription['element_id'],
+                type=transcription['type'],
+                text=transcription['text'],
+                score=transcription['score'],
+            )
+            for transcription in validated_data['transcriptions']
+        ]
+        Transcription.objects.bulk_create(transcriptions)
+        return {"transcriptions": transcriptions}
+
+
 class ClassificationBulkSerializer(serializers.Serializer):
     """
     Single classification serializer for bulk insertion

arkindex/project/openapi/patch.yml

@@ -283,14 +283,6 @@ paths:
       description: Update the text of a manual transcription
     delete:
       description: Delete a manual transcription
-  /api/v1/transcription/bulk/:
-    post:
-      operationId: CreateTranscriptions
-    put:
-      operationId: UpdateTranscriptions
-      description: >-
-        Replace all existing transcriptions from a given recognizer on a page
-        with other transcriptions.
   /api/v1/metadata/{id}/:
     get:
       operationId: RetrieveMetaData