stash

backend/toolshed/api/files.py

@@ -53,10 +53,10 @@ def post_item_file(request, item_id):
 def get_staged_files(request, workflow_id):
     try:
         workflow = WorkflowInstance.objects.get(id=workflow_id, owner=request.user)
-        # Staged files are private working state the client already holds in full (name, size,
-        # mime_type, base64 data) from the moment it read them off disk/camera - the only thing
-        # it can't already know is whether/under what hash the upload was persisted, so that's
-        # all this returns, unlike the fuller FileSerializer representation item_files uses.
+        # Hash alone identifies a staged file (client and server hash content the same way, and
+        # bytes are fetchable from a hash-derived storage path) - useful mainly for discovering
+        # what another session/device already staged on this workflow, unlike the fuller
+        # FileSerializer representation item_files uses.
         return Response(list(workflow.staged_files.values_list('hash', flat=True)))
     except WorkflowInstance.DoesNotExist:
         return Response(status=status.HTTP_404_NOT_FOUND)
@@ -95,6 +95,7 @@ def delete_item_file(request, item_id, file_id, format=None):  # /item_files/
         item.files.remove(file_id)
         if file.connected_items.count() == 0 and file.profile_picture_users.count() == 0 \
                 and file.staged_by_workflows.count() == 0:
+            file.file.delete(save=False)
             file.delete()
         return Response(status=status.HTTP_204_NO_CONTENT)
     except InventoryItem.DoesNotExist:
@@ -123,6 +124,7 @@ def delete_staged_file(request, workflow_id, file_hash, format=None):  # /staged
         workflow.staged_files.remove(file)
         if file.connected_items.count() == 0 and file.profile_picture_users.count() == 0 \
                 and file.staged_by_workflows.count() == 0:
+            file.file.delete(save=False)
             file.delete()
         return Response(status=status.HTTP_204_NO_CONTENT)
     except WorkflowInstance.DoesNotExist:

backend/toolshed/api/inventory.py

@@ -126,6 +126,7 @@ class WorkflowInstanceViewSet(viewsets.ModelViewSet):
             for file in File.objects.filter(id__in=staged_file_ids):
                 if file.connected_items.count() == 0 and file.profile_picture_users.count() == 0 \
                         and file.staged_by_workflows.count() == 0:
+                    file.file.delete(save=False)
                     file.delete()
 
 

backend/toolshed/serializers.py

@@ -203,9 +203,11 @@ class InventoryItemSerializer(serializers.ModelSerializer):
 
 class WorkflowInstanceSerializer(serializers.ModelSerializer):
     owner = serializers.StringRelatedField(read_only=True)
-    # The client already holds the full file (name, size, mime_type, base64 data) for anything it
-    # staged itself - hash is the only thing it can't already know, so that's all this exposes,
-    # unlike InventoryItemSerializer.files which needs the fuller FileSerializer representation.
+    # Hash is enough to identify a staged file (the client computes the same SHA-256 the backend
+    # does, and can fetch bytes from a hash-derived storage path) - for anything staged by *this*
+    # session there's nothing more to say, and for a file staged elsewhere (another device/tab),
+    # hash is what lets this session recognize and fetch it. Unlike InventoryItemSerializer.files,
+    # no fuller FileSerializer representation is needed here.
     staged_files = serializers.SerializerMethodField()
 
     class Meta: