kindle_unpack 0.1.1

Pure-Dart library for reading Amazon Kindle ebook files (MOBI / AZW / AZW3 / KF8) and re-emitting them as EPUB 3. No DRM removal.

Changelog

0.1.1

• Fix toEpub() so the OPF declares an EPUB 3 nav document (<item ... properties="nav"/> pointing at a generated nav.xhtml). Without it, strict EPUB 3 parsers like epubx failed to locate the TOC and threw EPUB parsing error: TOC item, not found in EPUB manifest. The packager still emits the NCX for EPUB 2 reader compatibility.

0.1.0

First release. Reads MOBI / AZW / AZW3 / KF8 files end-to-end and re-emits them as EPUB 3.

• PDB container — Palm Database header + record list parser.
• PalmDOC + MOBI headers — fixed + variable-length headers, with the field offsets that real-world files actually use (the MobileRead / KindleUnpack layout, not the offsets some other references list).
• EXTH metadata — typed accessors for title, authors, publisher, description, ISBN, subjects, ASIN, language, cover/thumbnail offset, KF8 boundary record, etc.
• PalmDOC decompression — strict LZ77-style byte-coded interpreter.
• HUFF/CDIC decompression — Huffman + dictionary-coded compression with recursive expansion of non-precoded entries.
• Per-record trailing-data stripping — handles the multibyte-overlap indicator and bit-N length-prefixed trailers from MOBI offset 0xF2.
• Image extraction — JPEG / PNG / GIF / BMP / SVG with cover and thumbnail resolution via EXTH 201/202.
• KF8 / AZW3 detection — distinguishes standalone KF8, Mobi-7-only, and dual (combo) files via the EXTH 121 boundary plus the BOUNDARY sentinel.
• FDST flow splitting — slices the rawML into HTML / CSS / SVG / auxiliary flows.
• INDX parser — main + entry blocks + CTOC strings + the byte-length and small-count tag encodings.
• Skeleton + fragment splicing — reconstructs individual XHTML files from KF8's collapsed primary flow, with the partial-tag boundary heuristic for malformed splice points.
• RESC manifest — extracts the OPF XML from the RESC record.
• FONT records — parses the FONT header, deobfuscates the XOR-mangled prefix, decompresses zlib payloads, and sniffs TTF / TTC / OTF.
• EPUB 3 packaging — KindleBook.fromBytes(bytes).toEpub() produces a valid EPUB zip (mimetype-first uncompressed, OPF + NCX + parts + images + CSS + fonts).

Real-world fixtures used during development:

• Pride and Prejudice (Project Gutenberg #1342) — MOBI v6, PalmDOC.
• A DRM-free AZW3 — KF8 v8, HUFF/CDIC, 70 XHTML parts, 8 images.

Known gaps for follow-up versions:

• The OPF and NCX are synthesised from EXTH metadata, not lifted from the file's own RESC manifest / NCX index. Output is valid EPUB but doesn't preserve original spine ordering refinements.
• No DRM removal — encrypted files are detected and rejected, never decrypted.
• Topaz (.tpz / .azw1) and Kindle Print Replica (.kpf) are out of scope.
• Flutter web isn't supported (the HUFF/CDIC decoder uses 64-bit shifts and the FONT decoder uses dart:io's ZLibCodec).