[BUGFIX] Prevent corrupt writes to download archive (#983)

Attempts to make writes to the download archive safer (https://github.com/jmbannon/ytdl-sub/issues/982).

`ytdl-sub` will now *copy* the download archive from the working directory to the output directory with a temp name, then perform a *move* to store it with its final expected name. This will drastically lower the window of time where  the process could die mid-write and corrupt it on the next read.

src/ytdl_sub/subscriptions/subscription_download.py

@@ -161,7 +161,7 @@ class SubscriptionDownload(BaseSubscription, ABC):
                 )
 
             self.download_archive.save_download_mappings()
-            FileHandler.delete(self.download_archive.working_file_path)
+            FileHandler.delete(self.download_archive.working_ytdl_file_path)
 
     @contextlib.contextmanager
     def _remove_empty_directories_in_output_directory(self):

src/ytdl_sub/subscriptions/subscription_ytdl_options.py

@@ -89,7 +89,9 @@ class SubscriptionYTDLOptions:
         ytdl_options = {}
 
         if self._preset.output_options.maintain_download_archive:
-            ytdl_options["download_archive"] = self._enhanced_download_archive.working_file_path
+            ytdl_options["download_archive"] = (
+                self._enhanced_download_archive.working_ytdl_file_path
+            )
         if self._preset.output_options.keep_max_files:
             keep_max_files = int(
                 self._overrides.apply_formatter(self._preset.output_options.keep_max_files)

src/ytdl_sub/utils/file_handler.py

@@ -385,7 +385,11 @@ class FileHandler:
         dst_file_path
             Destination file
         """
-        shutil.copyfile(src=src_file_path, dst=dst_file_path)
+        # Perform the copy by first writing to a temp file, then moving it.
+        # This tries to prevent corrupted writes if the processed dies mid-write,
+        atomic_dst = f"{dst_file_path}-ytdl-sub-incomplete"
+        shutil.copyfile(src=src_file_path, dst=atomic_dst)
+        shutil.move(src=atomic_dst, dst=dst_file_path)
 
     @classmethod
     def move(cls, src_file_path: Union[str, Path], dst_file_path: Union[str, Path]):

src/ytdl_sub/ytdl_additions/enhanced_download_archive.py

@@ -506,10 +506,19 @@ class EnhancedDownloadArchive:
         """
         Returns
         -------
-        The download mapping's file path in the working directory.
+        The download mapping's file path in the working directory for ytdl usage
         """
         return str(Path(self.working_directory) / self.file_name)
 
+    @property
+    def working_ytdl_file_path(self) -> str:
+        """
+        Returns
+        -------
+        The download mapping's file path in the working directory for ytdl usage
+        """
+        return f"{self.working_file_path}-ytdl-archive"
+
     @property
     def mapping(self) -> DownloadMappings:
         """
@@ -541,7 +550,7 @@ class EnhancedDownloadArchive:
             return self
 
         # Otherwise, create a ytdl download archive file in the working directory.
-        self.mapping.to_download_archive().to_file(self.working_file_path)
+        self.mapping.to_download_archive().to_file(self.working_ytdl_file_path)
 
         return self
 
@@ -603,15 +612,18 @@ class EnhancedDownloadArchive:
         if self._migrated_file_name:
             self._download_mapping.to_file(output_json_file=self.working_file_path)
             self.save_file_to_output_directory(
-                file_name=self.file_name, output_file_name=self._migrated_file_name
+                file_name=self.file_name, output_file_name=self._migrated_file_name, copy_file=True
             )
+            FileHandler.delete(file_path=self.working_file_path)
+
             # and delete the old one if the name differs
             if self._file_name != self._migrated_file_name:
                 self.delete_file_from_output_directory(file_name=self.file_name)
         # Otherwise, only save if there are changes to the transaction log
         elif not self.get_file_handler_transaction_log().is_empty:
             self._download_mapping.to_file(output_json_file=self.working_file_path)
-            self.save_file_to_output_directory(file_name=self.file_name)
+            self.save_file_to_output_directory(file_name=self.file_name, copy_file=True)
+            FileHandler.delete(file_path=self.working_file_path)
         return self
 
     def delete_file_from_output_directory(self, file_name: str):