Improve the efficiency of working with the DSA. (#1691)

* Improve the efficiency of working with the DSA.

Before, this often queried all of the items, files, or annotations in
the system.  Now it only queries what is required to perform the desired
task.

I've disabled a status check (by having it always return 0 values).  If
this is actually needed, we should add an endpoint to the DSA to do this
efficiently.  The listing endpoints do return counts in their headers.
The slow query is getting the count of images with annotations.

Signed-off-by: David Manthey <[email protected]>

* Find items regardless of folder; be faster for known folders

---------

Signed-off-by: David Manthey <[email protected]>

Author: manthey

monailabel/datastore/dsa.py

@@ -84,7 +84,7 @@ def get_label_by_image_id(self, image_id: str, tag: str) -> str:
     def get_annotations_by_image_id(self, image_id: str) -> Dict[str, Dict[str, List]]:
         image_id, name = self._name_to_id(image_id)
 
-        data = self.gc.get("annotation", parameters={"limit": 0})
+        data = self.gc.get(f"annotation/item/{image_id}", parameters={"limit": 0})
         result: Dict[str, Dict[str, List]] = {}
 
         # TODO(avirodov): probably can request only annotation for a given image_id, need to check how.
@@ -136,26 +136,40 @@ def get_image(self, image_id: str, params=None) -> Any:
     def _name_to_id(self, name):
         folders = self.folders if self.folders else self._get_all_folders()
         for folder in folders:
+            # First check if the name is directly present
+            data = self.gc.get("item", parameters={"folderId": folder, "name": name, "limit": 0})
+            for d in data:
+                if d.get("largeImage"):
+                    return d["_id"], d["name"]
+            # next check if the name is present in a stem form
             data = self.gc.get("item", parameters={"folderId": folder, "limit": 0})
             for d in data:
                 if d.get("largeImage") and d["name"] == name or Path(d["name"]).stem == name:
                     return d["_id"], d["name"]
-        return name
+        # Next check if the name is anywhere in the system
+        data = self.gc.get("item", parameters={"text": f'"{name}"' if '"' not in name else name, "limit": 0})
+        for d in data:
+            if d.get("largeImage") and d["name"] == name or Path(d["name"]).stem == name:
+                return d["_id"], d["name"]
+        # If we fail to find the item, the best we can do is return the name
+        return name, name
 
     def get_image_uri(self, image_id: str) -> str:
         try:
-            name = self.get_image_info(image_id)["name"]
+            info = self.get_image_info(image_id)
+            name = info["name"]
+            file_id = info.get("largeImage", {}).get("fileId")
         except girder_client.HttpError:
             image_id, name = self._name_to_id(image_id)
+            file_id = None
 
         if self.asset_store_path:
-            data = self.gc.get(f"item/{image_id}/files", parameters={"limit": 0})
-            assets = [d["assetstoreId"] for d in data]
-            for asset in assets:
-                files = self.gc.get(f"assetstore/{asset}/files", parameters={"limit": 0})
-                for f in files:
-                    if f["itemId"] == image_id:
-                        return str(os.path.join(self.asset_store_path, f["path"]))
+            if file_id is None:
+                data = self.gc.get(f"item/{image_id}/files", parameters={"limit": 0})
+                file_id = data[0]["_id"]
+            f = self.gc.get(f"resource/{file_id}?type=file")
+            if "path" in f and os.path.exists(os.path.join(self.asset_store_path, f["path"])):
+                return str(os.path.join(self.asset_store_path, f["path"]))
         else:
             cached = os.path.join(self.cache_path, name)
             if os.path.exists(cached):
@@ -243,9 +257,14 @@ def get_dataset_archive(self, limit_cases: Optional[int]) -> str:
         raise NotImplementedError
 
     def status(self) -> Dict[str, Any]:
+        # This is a very costly query, disable it for now
+        # return {
+        #     "total": len(self.list_images()),
+        #     "completed": len(self.get_labeled_images()),
+        # }
         return {
-            "total": len(self.list_images()),
-            "completed": len(self.get_labeled_images()),
+            "total": 0,
+            "completed": 0,
         }
 
     def json(self):