perf(language-server): parse the document once per selection range request

Selection ranges built the model again for every requested position when the document had no version, so a request with many positions parsed the whole file once per position. The model is now taken once per request, which roughly halves the time of the selection range parity tests.

Author: d-biehl

packages/language_server/src/robotcode/language_server/robotframework/parts/selection_range.py

@@ -34,10 +34,11 @@ def __init__(self, parent: "RobotLanguageServerProtocol") -> None:
     def collect(self, sender: Any, document: TextDocument, positions: List[Position]) -> Optional[List[SelectionRange]]:
         namespace = self.parent.documents_cache.get_namespace(document)
         semantic_model = namespace.semantic_model
+        model = self.parent.documents_cache.get_model(document)
 
         results: List[SelectionRange] = []
         for position in positions:
-            nodes = get_nodes_at_position(self.parent.documents_cache.get_model(document), position)
+            nodes = get_nodes_at_position(model, position)
 
             if not nodes:
                 break