some fixes and added delete_repo.py

Author: Jonas Widmer

flow_preprocessor/preprocessing_logic/parse_textlines.py

@@ -303,7 +303,7 @@ def process_lines_from_xml_file(self) -> List[Line]:
         """
         line_document = None
         try:
-            line_document = self.xml_filename
+            line_document = self.get_image_file_name()
             line_list: List[Line] = []
             for text_line in self.root.findall(".//ns:TextLine", namespaces=self.xmlns):
                 line_number = self.get_line_id(text_line)
@@ -313,7 +313,7 @@ def process_lines_from_xml_file(self) -> List[Line]:
                 custom_attributes = self.get_custom_attribute(text_line)
                 if line_text == '' or line_coordinates == [] or line_baseline_points == []:
                     logger.warning(
-                        '%s - Skipping line %d in file %s as it is '
+                        '%s - Skipping line %s in file %s as it is '
                         'empty or has no coordinates or baseline points.',
                         self.__class__.__name__,
                         line_number,

flow_preprocessor/preprocessing_logic/preprocess.py

@@ -233,6 +233,7 @@ def preprocess_single_xml_file(self, xml_file: str) -> None:
             else:
                 gt_dict[line_name] = line.expand_abbreviations()
 
+            logger.info("Added to gt_dict: key = %s, value = %s", line_name, gt_dict[line_name])
             image_path = os.path.join(self.in_path, page.image_file_name)
 
             if not self.crop:
@@ -274,12 +275,11 @@ def _save_gt_dict(gt_dict: Dict[str, str], out_path: str) -> None:
         :param gt_dict: Dictionary containing line names and texts.
         :param out_path: The output path where the file will be saved.
         """
-        file_path = os.path.join(out_path, 'gt.txt')
+        file_path = os.path.join(out_path, "gt.txt")
 
-        with open(file_path, "a", encoding='utf-8') as txt_file:
+        with open(file_path, "a", encoding="utf-8") as txt_file:
             for line_name, line_text in gt_dict.items():
-                escaped_text = json.dumps(line_text)
-                txt_file.write(f"{line_name}\t{escaped_text}\n")
+                txt_file.write(f"{line_name}\t{line_text}\n")
 
     @staticmethod
     def _get_gt_dict(out_path: str) -> Dict[str, str]:

flow_preprocessor/preprocessing_logic/process_images.py

@@ -122,7 +122,7 @@ def extract_line_from_image(self,
             raise ImageProcessException(f'The image cannot be opened and identified for file {in_path}, {e}') from e
         except ValueError as e:
             logger.error(
-                '%s - Wrong value provided for file %s on line %d, %s',
+                '%s - Wrong value provided for file %s on line %s, %s',
                 self.__class__.__name__,
                 in_path,
                 line_number,
@@ -174,7 +174,7 @@ def crop_line_from_image(self,
             cropped_image.paste(cutout, (0, 0), mask=cutout)
             cropped_image = cropped_image.convert('RGB')
             logger.info(
-                '%s - Successfully extracted line %d for image %s',
+                '%s - Successfully extracted line %s for image %s',
                 self.__class__.__name__,
                 line_number,
                 image_path,

flow_preprocessor/preprocessing_logic/status.py

@@ -6,7 +6,7 @@
 # IMPORT STATEMENTS
 # ===============================================================================
 from datetime import datetime
-from typing import List, Any
+from typing import List, Union
 from flow_preprocessor.preprocessing_logic.models import PreprocessState, StateEnum
 from flow_preprocessor.exceptions.exceptions import ImageFetchException
 
@@ -92,7 +92,7 @@ async def update_progress(self,
 
         return PreprocessState(**self.state.model_dump(by_alias=True))
 
-    def update_image_list(self, new_line_images: Any[str, List]) -> PreprocessState:
+    def update_image_list(self, new_line_images: Union[str, List]) -> PreprocessState:
         """
         Update the image list.
 

flow_preprocessor/utils/delete_repo.py

@@ -0,0 +1,24 @@
+"""
+Simple Util to delete a repository.
+"""
+
+import shutil
+import os
+from typing import Tuple
+
+
+async def deleteRepo(repo_name: str) -> Tuple[bool, str]:
+    """
+    Delete a repository locally.
+    """
+    dir_to_delete = os.path.join('data', repo_name.replace('/', '___'))
+    if os.path.exists(dir_to_delete):
+        shutil.rmtree(dir_to_delete)
+        success = True
+    else:
+        return False, f"Directory for {repo_name} doesn't exist"
+
+    if success:
+        return True, f"Successfully deleted {repo_name}"
+    else:
+        return False, f"Failed to delete {repo_name}"