Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

The generated nav.xhtml no longer contains an empty <nav epub:type="landmarks"> wrapper. epub-builder emits the wrapper even when there are no landmark entries, which fails epubcheck RSC-005 ("element \"nav\" incomplete; missing required element \"ol\""). The writer's post-processing pass now strips a landmarks nav that has no entries; a landmarks nav with real entries is kept verbatim. Built EPUBs that previously carried this epubcheck error are now clean, and the test suite no longer allowlists it.
The generated toc.ncx no longer trips epubcheck RSC-005 "different playOrder values for navPoint/navTarget/pageTarget that refer to same target". epub-builder gives every navPoint a fresh sequential playOrder, so a file appearing in the TOC more than once (e.g. as both a nested child and a top-level entry — common in real books) got distinct values where the NCX spec requires one shared value. The writer now renumbers playOrder in document order, 1-based, with repeat targets reusing the first occurrence's value. The test suite no longer allowlists this error either.

LangelicEpub.Document gains a page_progression_direction field ("rtl", "ltr", or nil). When set, build/1 writes the OPF <spine page-progression-direction> attribute so right-to-left target languages (Arabic, Hebrew, …) paginate correctly in real readers. For "rtl", the generated nav.xhtml root <html> also gets dir="rtl" and the document language so table-of-contents labels render in the correct direction.
LangelicEpub.Error may now have kind: :invalid_page_direction when page_progression_direction is anything other than "rtl", "ltr", or nil. The value is rejected at build time rather than silently dropped.

When page_progression_direction is nil, the built OPF now omits the <spine> page-progression-direction attribute entirely. Previously the underlying epub-builder unconditionally emitted page-progression-direction="ltr". An omitted direction is semantically identical (readers default to ltr) but no longer hard-codes a direction the caller never asked for. Existing nil builds change bytes but not rendering.

The reader continues to not surface a source EPUB's spine direction: parse/1 always returns page_progression_direction: nil. Direction is a build-time decision derived from the target language, never round-tripped from the source (a rtl Japanese source rebuilt into English must shed it).

Dependency updates: rustler 0.37 → 0.38 (Elixir + crate), zip 7 → 8, quick-xml 0.39 → 0.40, plus uuid, credo, and ex_doc bumps. The quick-xml Attribute::unescape_value API was deprecated; migrated to normalized_value(XmlVersion::Implicit1_0), which is behavior-identical (same XML 1.0 implicit version, depth, and predefined-entity resolver). No user-facing API change.
Reader no longer depends on the iepub crate. All structural data (manifest, spine, metadata, cover, TOC) now comes from a pure zip + quick_xml stack. This removes a large transitive dependency surface and lets us tolerate non-OCF-conformant mimetype files (trailing \n, \r\n, space, or a leading UTF-8 BOM) — common in Calibre exports and other real-world EPUBs that other readers accept silently. Genuine non-EPUB content is still rejected with a new :invalid_mimetype error kind.

LangelicEpub.Error may now have kind: :invalid_mimetype when the mimetype zip entry is missing or its content (after trimming a UTF-8 BOM and whitespace) is not application/epub+zip.
OTP 29 support: rustler 0.38 builds against OTP 29's NIF interface, and CI now tests on OTP 29 / Elixir 1.20 in addition to OTP 26/27. The precompiled NIF 2.16 artifact forward-loads on OTP 29's newer NIF ABI, so no new artifact is shipped (a 2.17 artifact would needlessly drop OTP 26 compatibility).

Dublin Core metadata is now matched by its namespace URI rather than a hard-coded dc: prefix (the OPF is parsed with quick-xml's NsReader), so dc:-equivalent elements bound to a non-standard prefix or a default namespace are recovered instead of silently dropped. Parsing the OPF also now rejects elements that declare more than 256 namespace bindings, bounding a malformed or hostile package.

LangelicEpub.parse/1 — parse EPUB 2 and EPUB 3 bytes into a %LangelicEpub.Document{} with spine, assets, table of contents, and metadata (including fields like <dc:language>, <dc:rights>, and multiple <dc:creator> entries that the underlying iepub crate does not expose natively).
LangelicEpub.build/1 — emit EPUB 3 bytes from a %LangelicEpub.Document{}, with a backward-compatible toc.ncx alongside the EPUB 3 nav.xhtml so EPUB 2-only readers still navigate correctly.
Validation of required fields (title, identifier, language) and spine/asset ID uniqueness at build time. UTF-8 is enforced on chapter data to prevent silent corruption in downstream readers.
Rust panics inside the NIF are caught (std::panic::catch_unwind) and returned as {:error, %LangelicEpub.Error{kind: :panic, ...}} so a malformed input cannot crash the BEAM scheduler.
Precompiled NIFs published via GitHub Releases for aarch64-apple-darwin, x86_64-apple-darwin, aarch64-unknown-linux-gnu, x86_64-unknown-linux-gnu, and x86_64-unknown-linux-musl.