feature(dataset indices): create necessary indices on startup

deepdataspace/task/__init__.py

@@ -13,6 +13,10 @@
 from celery.signals import worker_ready
 
 from deepdataspace import environs
+from deepdataspace.model.dataset import DataSet
+from deepdataspace.model.dataset import DatasetStatus
+from deepdataspace.model.image import Image
+from deepdataspace.task.celery import app
 from deepdataspace.task.celery import app
 from deepdataspace.task.ping import ping
 
@@ -71,8 +75,38 @@ def import_and_process_data_dir(data_dir: str, enforce: bool = False, auto_trigg
             logger.error(str(err))
 
 
+@app.task
+def create_dataset_index():
+    """
+    Create index for all datasets.
+    """
+
+    logger.info(f"create_dataset_index starts")
+
+    total = DataSet.count_num({})
+    datasets = DataSet.find_many({})
+    for idx, dataset in enumerate(datasets):
+        logger.info(f"[{idx + 1}/{total}] creating index for dataset {dataset.id}")
+        if dataset.status != DatasetStatus.Ready:
+            logger.info(f"dataset status is not {DatasetStatus.Ready}, skip it")
+        else:
+            ImageModel = Image(dataset.id)
+            ImageModel.get_collection().create_index([
+                ("objects.category_id", 1),
+            ])
+            logger.info(f"dataset created index on field: objects.category_id")
+
+            ImageModel.get_collection().create_index([
+                ("idx", 1)
+            ])
+            logger.info(f"dataset created index on field: idx")
+
+
 @worker_ready.connect
 def startup_jobs(sender=None, headers=None, body=None, **kwargs):
+    # ensure necessary indices
+    create_dataset_index.apply_async()
+
     # import all datasets
     data_dir = environs.DATASET_DIR
     import_and_process_data_dir.apply_async(args=(data_dir, False, True))