#14: fixed

src/digital_eval/evaluation.py

@@ -16,7 +16,7 @@
     floor
 )
 from multiprocessing import (
-    cpu_count, Queue, Lock, Manager, Array
+    cpu_count
 )
 from pathlib import (
     Path
@@ -285,6 +285,17 @@ def _map_page2013(elem: ET.Element) -> Tuple[str, int, int, int, int]:
     return (NOT_SET, min(_xs), min(_ys), max(_xs), max(_ys))
 
 
+def _get_line_pieces_from_piece(piece: Piece, lines: List[Piece] = None) -> List[Piece]:
+    if lines is None:
+        lines = []
+    if piece.level == PieceLevel.LINE and piece.transcription:
+        lines.append(piece)
+        return lines
+    for child in piece.pieces:
+        _get_line_pieces_from_piece(child, lines)
+    return lines
+
+
 def calculate_bounding_box(elements: List[ET.Element], map_func) -> Tuple[int, int, int, int]:
     """Review element's points to get points for
     minimum (top-left) and maximum (bottom-right)"""
@@ -346,11 +357,7 @@ def piece_to_text(file_path, frame=None, oneliner=True) -> Tuple[str | List[str]
         frame_piece = Piece()
         frame_piece.dimensions = frame
         filter_word_pieces(frame_piece, top_piece)
-        the_lines = [l
-                     for r in top_piece.pieces
-                     for l in r.pieces
-                     if l.transcription and l.level == PieceLevel.LINE]
-        # print('!!!!!!!!!', oneliner, len(top_piece.transcription), frame, )
+        the_lines = _get_line_pieces_from_piece(top_piece)
         if oneliner:
             return top_piece.transcription, len(the_lines)
         else:
@@ -583,7 +590,7 @@ def __init__(
         self.evaluation_map = {}
         self.text_mode = extras == EVAL_EXTRA_IGNORE_GEOMETRY
         self.is_sequential = False
-        self.metrics = []
+        self.metrics: List = []
         self.evaluation_report = {}
 
     def eval_all(self, entries: List[EvalEntry], sequential=False) -> None:
@@ -636,6 +643,7 @@ def eval_entry(self, entry: EvalEntry) -> EvalEntry:
 
         # evaluate metric copies
         _current_metrics = []
+
         for _m in self.metrics:
 
             path_g = entry.path_g
@@ -661,6 +669,7 @@ def eval_entry(self, entry: EvalEntry) -> EvalEntry:
 
             # read candidate data as text
             (txt_c, _) = to_text_func(path_c, coords, oneliner=True)
+
             if not txt_c:
                 print(f"[WARN ] candidate '{path_c}' contains no text")
 
@@ -720,7 +729,7 @@ def eval_map(self):
 
             # if more than one single evaluation item
             # calculate additional statistics to reflect
-            # impact of outlying data sets 
+            # impact of outlying data sets
             # take CA and number of GT into account
             # also calculate statistics (mean, std)
             if len(data_points) > 1:
@@ -753,13 +762,13 @@ def aggregate(self, by_type=False, by_metrics=[0, 1, 2, 3]):
         # aggregate on each directory
         for _metrics_index in by_metrics:
             for ee in self.evaluation_entries:
-                # if we do not have all these different metrics set, 
+                # if we do not have all these different metrics set,
                 # do of course not aggregate by non-existing index!
                 if _metrics_index >= len(self.evaluation_entries[0].metrics):
                     continue
                 path_key = f"{ee.metrics[_metrics_index].label}@{root_base}"
                 # ATTENZIONE! works only when forehand
-                # the *real* attribute has been accessed 
+                # the *real* attribute has been accessed
                 # *at least one time*
                 # kept this way for testing reasons
                 metric_value = ee.metrics[_metrics_index].value

src/digital_eval/metrics.py

@@ -410,7 +410,6 @@ def __init__(self, precision=2, normalization=UC_NORMALIZATION_NFKD, calc_func=a
         self._label = 'DictLT'
 
     def _forward(self):
-        # print('#####', len(self._data_reference), len(self._data_candidate))
         text: str = self._data_candidate
         text_list: List[str] = self._data_candidate.split()
         self._data_reference = text_list