refactor: clean up splitting between data and expected

Author: Ciznia

src/analyzer/data_loader.py

@@ -2,6 +2,8 @@
 
 from chess_utils import fen_to_tensor
 
+from .labels import label_to_vector
+
 
 def load_chessfile_predict(filepath: str) -> List[str]:
     with open(filepath, "r", encoding="utf-8") as f:
@@ -11,7 +13,6 @@ def load_chessfile_predict(filepath: str) -> List[str]:
 def load_chessfile_train(
     filepath: str, encoding: str = "simple"
 ) -> List[Tuple[List[float], List[float]]]:
-    from .labels import label_to_vector
 
     dataset: List[Tuple[List[float], List[float]]] = []
 
@@ -21,18 +22,12 @@ def load_chessfile_train(
             if not line:
                 continue
 
-            parts = line.rsplit(maxsplit=2)
-            if len(parts) < 2:
-                print(f"Warning: line {line_num} has no label, skipping")
+            parts = line.split()
+            if len(parts) < 7 or len(parts) > 8:
+                print(f"Warning: line {line_num} has invalid format, skipping")
                 continue
-
-            if len(parts) == 3 and parts[1] in ("Check", "Checkmate"):
-                fen = parts[0]
-                label_str = f"{parts[1]} {parts[2]}"
-            else:
-                fen = " ".join(parts[:-1])
-                label_str = parts[-1]
-
+            fen = " ".join(parts[:6])
+            label_str = " ".join(parts[6:])
             try:
                 x = fen_to_tensor(fen, encoding=encoding)
                 y = label_to_vector(label_str)