fix: PLT-802: _update_tasks_states job optimization

label_studio/core/settings/base.py

@@ -587,6 +587,7 @@
 DATA_MANAGER_ACTIONS = {}
 DATA_MANAGER_CUSTOM_FILTER_EXPRESSIONS = 'data_manager.functions.custom_filter_expressions'
 DATA_MANAGER_PREPROCESS_FILTER = 'data_manager.functions.preprocess_filter'
+BULK_UPDATE_IS_LABELED = 'tasks.functions.bulk_update_is_labeled_by_overlap'
 USER_LOGIN_FORM = 'users.forms.LoginForm'
 PROJECT_MIXIN = 'projects.mixins.ProjectMixin'
 TASK_MIXIN = 'tasks.mixins.TaskMixin'

label_studio/projects/functions/utils.py

@@ -47,7 +47,7 @@ def get_unique_ids_list(tasks_queryset):
 
     elif isinstance(tasks_queryset, QuerySet):
         # It's a Django QuerySet
-        return list(tasks_queryset.values_list('id', flat=True))
+        return list(tasks_queryset.values_list('id', flat=True).iterator(chunk_size=1000))
 
     else:
         raise ValueError(f'Unsupported type for tasks_queryset: {type(tasks_queryset)}')

label_studio/tasks/functions.py

@@ -230,3 +230,47 @@
             updated_count += len(batch_list)
 
     return updated_count
+
+
+def bulk_update_is_labeled_by_overlap(tasks_ids, project):
+    """Bulk update is_labeled status for tasks using overlap-based calculation only.
+
+    This function updates the is_labeled field for tasks based on overlap settings.
+    It only works when the project can use overlap (automatic assignment mode).
+    Processes tasks in batches to handle large datasets efficiently.
+
+    :param tasks_ids: List or QuerySet of task IDs
+    :param project: Project instance
+    :return: None
+    :raises ValueError: If project cannot use overlap
+    """
+    from django.db.models import F, Q
+    from django.db.models.lookups import GreaterThanOrEqual
+
+    if not tasks_ids:
+        return
+
+    if not project._can_use_overlap():
+        raise ValueError('Project cannot use overlap-based calculation')
+
+    # Convert to list if it's a QuerySet to enable slicing
+    if hasattr(tasks_ids, 'values_list'):
+        tasks_ids = list(tasks_ids)
+
+    # following definition of `completed_annotations` above, count cancelled annotations
+    # as completed if project is in IGNORE_SKIPPED mode
+    completed_annotations_f_expr = F('total_annotations')
+    if project.skip_queue == project.SkipQueue.IGNORE_SKIPPED:
+        completed_annotations_f_expr += F('cancelled_annotations')
+    finished_q = Q(GreaterThanOrEqual(completed_annotations_f_expr, F('overlap')))
+
+    # Process in batches
+    batch_size = settings.BATCH_SIZE
+    for i in range(0, len(tasks_ids), batch_size):
+        batch_ids = tasks_ids[i : i + batch_size]
+
+        # Update is_labeled=True for tasks that meet the overlap criteria
+        Task.objects.filter(id__in=batch_ids, project=project).filter(finished_q).update(is_labeled=True)
+
+        # Update is_labeled=False for tasks that don't meet the criteria
+        Task.objects.filter(id__in=batch_ids, project=project).exclude(finished_q).update(is_labeled=False)

label_studio/tasks/models.py

@@ -1376,7 +1376,7 @@
         task.save()
 
 
-def bulk_update_stats_project_tasks(tasks, project=None):
+def deprecated_bulk_update_stats_project_tasks(tasks, project=None):
     """bulk Task update accuracy
        ex: after change settings
        apply several update queries size of batch
@@ -1426,5 +1426,43 @@
                 )
 
 
+def bulk_update_stats_project_tasks(tasks, project=None):
+    """bulk Task update accuracy
+       ex: after change settings
+       apply several update queries size of batch
+       on updated Task objects
+       in single transaction as execute sql
+    :param tasks: QuerySet or list of Task objects
+    :param project: Project instance (optional, will be inferred from first task if not provided)
+    :return: None
+    """
+    from core.feature_flags import flag_set
+
+    bulk_update_is_labeled = load_func(settings.BULK_UPDATE_IS_LABELED)
+
+    if flag_set('fflag_fix_back_plt_802_update_is_labeled_20062025_short', user='auto'):
+        # New implementation using bulk_update_is_labeled
+        from projects.functions.utils import get_unique_ids_list
+
+        # Convert tasks to list of IDs
+        task_ids = get_unique_ids_list(tasks)
+
+        # Check if empty
+        if not task_ids:
+            return
+
+        # Get project if it's not in params
+        if project is None:
+            # Load first task to get project
+            first_task = Task.objects.get(id=task_ids[0])
+            project = first_task.project
+
+        # Call the loaded bulk_update_is_labeled function
+        bulk_update_is_labeled(task_ids, project)
+    else:
+        # Use deprecated implementation
+        return deprecated_bulk_update_stats_project_tasks(tasks, project)
+
+
 Q_finished_annotations = Q(was_cancelled=False) & Q(result__isnull=False)
 Q_task_finished_annotations = Q(annotations__was_cancelled=False) & Q(annotations__result__isnull=False)