Comic archive checker: files for "How to Translate a CBZ or CBR Comic Archive"
https://www.booktranslator.app/blog/translate-cbz-comics

FILES

check_comic_archive.py
  Reads a .cbz file (or a folder of extracted pages) and prints the page order,
  sorting problems, junk entries, non-page images, spreads, tall strips, and
  the ComicInfo.xml reading direction. With --repack it writes a new CBZ with
  zero-padded page names. Python 3.8+ standard library only; runs offline.

make_test_comic.py
  Builds test-comic.cbz, the sample archive used in the article: 12 numbered
  pages without zero padding, an unnumbered cover.png, one 800x600 spread
  (page 7), an AVIF file named 13.avif, macOS junk entries, and a
  ComicInfo.xml with Manga=YesAndRightToLeft and a 13-row page table.
  The archive is byte-identical between runs on the same Python build. With
  Python 3.14.3 on macOS it is
  SHA-256 d386bba72097df250680a5e51d2cfa6a7c65b9e1610a34aa6b114fe9db1aeeec
  (a Python linked against a different zlib, such as zlib-ng, can differ).

REPRODUCE

  python3 make_test_comic.py
  python3 check_comic_archive.py test-comic.cbz
  python3 check_comic_archive.py test-comic.cbz --first cover.png --repack fixed.cbz
  python3 check_comic_archive.py fixed.cbz

The first check reports four problems: a plain character sort disagrees with
natural order at position 2, cover.png sorts to position 13, page numbers mix
1 and 2 digits, and 13.avif is not a page format. After the repack, the pages
are 0001.png..0013.png with the cover first, the junk entries are gone, and
only the AVIF warning remains.

HOW THE SORT WAS CHECKED

natural_key() approximates JavaScript's
Intl.Collator("en", {numeric: true, sensitivity: "base"}), the comparison
BookTranslator uses to order .cbz pages. On 2026-10-30 its order matched
Node.js 25.2.1 on 18 hand-picked ASCII names (chapter folders, mixed case,
spaces, hyphens and underscores) and 400 random names drawn from
"ab AB_-./0123456789". It does not fold accents or kana: names such as
"Écran 1" or katakana/hiragana mixes can sort differently. Zero-padded ASCII
names avoid the question for every tool.

LIMITS

- RAR (CBR), 7z (CB7) and tar (CBT) archives are detected, not opened.
  Extract them first and run the script on the folder.
- --first changes page positions; a ComicInfo.xml <Pages> table is copied
  as it is, so check its Image numbers after reordering.
- Image sizes come from file headers. Unusual or damaged headers report no size.
- The checker cannot see text in images or judge translation quality.