Docker stuff and folder watcher

src/SuwayomiMover.py

@@ -161,6 +161,60 @@ def _al_id_from_metadata(md: dict) -> "int | None":
     return None
 
 
+def _chapter_image_size(chapter_dir: Path) -> int:
+    """Returns the total file size of all images in a chapter folder."""
+    return sum(
+        f.stat().st_size
+        for f in chapter_dir.iterdir()
+        if f.is_file() and f.suffix.lower() in _IMAGE_EXTS
+    )
+
+
+def _deduplicate_chapters(
+    chapter_items: list[tuple[Path, dict, str]],
+) -> tuple[list[tuple[Path, dict, str]], list[Path]]:
+    """
+    When multiple chapter folders share the exact same chapter number
+    (e.g. two folders for chapter "2" — not "2" vs "2.2"), keeps only the
+    one with the highest total image file size, which is a reliable proxy
+    for image quality.
+
+    Chapter number comes from ComicInfo.xml <Number>; comparison is an exact
+    string match so "2" and "2.2" are never considered duplicates.
+
+    Returns
+    -------
+    kept     : deduplicated chapter_items list (original sort order preserved)
+    rejected : Path list of lower-quality duplicate folders to be removed
+    """
+    best:      dict[str, tuple[Path, dict, str]] = {}
+    best_size: dict[str, int]                    = {}
+    rejected:  list[Path]                        = []
+
+    for item in chapter_items:
+        chapter_dir, fields, chapter_num = item
+        size = _chapter_image_size(chapter_dir)
+
+        if chapter_num not in best:
+            best[chapter_num]      = item
+            best_size[chapter_num] = size
+        elif size > best_size[chapter_num]:
+            prev_dir = best[chapter_num][0]
+            print(f"  [dup] ch.{chapter_num}: replacing {prev_dir.name!r} "
+                  f"({best_size[chapter_num]:,}B) with {chapter_dir.name!r} "
+                  f"({size:,}B) — higher quality")
+            rejected.append(prev_dir)
+            best[chapter_num]      = item
+            best_size[chapter_num] = size
+        else:
+            print(f"  [dup] ch.{chapter_num}: skipping {chapter_dir.name!r} "
+                  f"({size:,}B), keeping {best[chapter_num][0].name!r} "
+                  f"({best_size[chapter_num]:,}B)")
+            rejected.append(chapter_dir)
+
+    return list(best.values()), rejected
+
+
 def _extract_chapter_num(folder_name: str) -> "str | None":
     """
     Fallback: extracts chapter number from the folder name.
@@ -325,6 +379,11 @@ class SuwayomiMover:
                 continue
             chapter_items.append((chapter_dir, fields, chapter_num))
 
+        chapter_items, rejected_dirs = _deduplicate_chapters(chapter_items)
+        if self._delete_source:
+            for d in rejected_dirs:
+                shutil.rmtree(d, ignore_errors=True)
+
         # <Series> from the first chapter's XML → strip source labels → clean title
         # for the MangaBaka search.  Folder name is the last resort.
         raw_series = manga_title
@@ -436,7 +495,7 @@ if __name__ == "__main__":
     )
 
     # Process a single series
-    result = mover.process_series("Yofukashi no Uta_ Rakuen-hen")
+    result = mover.process_series("Yofukashi no Uta")
     ok     = sum(1 for c in result["chapters"] if c["ok"])
     failed = sum(1 for c in result["chapters"] if not c["ok"])
     print(f"\nDone: {ok} ok, {failed} failed")