Expose EPUB ISBNs via metadata.isbns

[raphi011]

Summary

The EPUB parser now exposes the publication's ISBNs through a new metadata.isbns property. A package's primary identifier is usually a UUID, so an ISBN — when present — lives in a secondary dc:identifier. EPUBMetadataParser scans every dc:identifier and collects those recognizable as ISBNs.

Detection

A dc:identifier is treated as an ISBN when it carries any of:

• opf:scheme="ISBN" (EPUB 2)
• an identifier-type ONIX Code List 5 refinement — 15 (ISBN-13) / 02 (ISBN-10) (EPUB 3)
• a urn:isbn: URN value
• Calibre's bare isbn: value prefix

Value handling

Values are returned as declared — only the scheme prefix (urn:isbn: / isbn:) and separators (hyphens, whitespace) are stripped. There is no normalization: an ISBN-10 is returned as ISBN-10, not converted to ISBN-13, and no check-digit synthesis happens, so the toolkit never invents a value the publisher didn't write.

A publication may declare more than one ISBN — the ISBN-10 and ISBN-13 of the same book, or distinct ISBNs per format — so isbns is a list, returned in document order with duplicates removed.

API

public extension Metadata {
    /// ISBNs declared by the publication, as found in the OPF.
    var isbns: [String] { ... }
}

Surfaced following the existing mediaOverlay extension pattern: the values are stashed in otherMetadata and read back through a typed accessor, so there is no change to the public Metadata type — only an additive computed property on the Shared EPUB extension.

Testing

EPUBMetadataParserTests cases + OPF fixtures covering: EPUB 2 (opf:scheme), EPUB 3 (identifier-type for both ISBN-13 and ISBN-10), urn:isbn:, Calibre's isbn: prefix, hyphen stripping, ISBN-10 returned verbatim (no conversion), multiple ISBNs in document order, duplicate-across-schemes dedup, and the no-ISBN (UUID-only) case. make format reports no changes.

Notes

• No JavaScript bundle changes.

[raphi011]

Superseded by the upstream PR readium#837.