Parsing opf version 3

[gitting]

I'm just starting with ABS and so far liking it. For ebooks I lean towards providing metadata embedded in content.opf for epub, or in a folder as an external metadata.opf next to pdf.

The problem I observe is that for <package version="3.0" ... of metadata the parsing fails as is clear from the code, e.g.:

audiobookshelf/server/utils/parsers/parseOpfMetadata.js

Line 17 in a3cd9e4

function parseCreators(metadata) {

The code looks for opf:... attributes whereas v3.0 provides that info via <meta refines="#id-to-refine-here" property="prop-name-here">prop-value-here</meta> type elements.

I can elaborate further. Just wondering whether this problem was recognized as a cursory search didn't reveal anything.

[advplyr]

I've been building the parser based on examples that I have. I haven't looked much at the specs. Can you send me an example?

[gitting]

The following is a metadata element within a v3 package. That's what happens when a v2 epub is upgraded to v3 in Sigil. This form then passes when checked with epubcheck. If I mangle this form by manually rewriting with opf:... (for current ABS parsing to succeed), then epubcheck complains. Hence, ideally the parser should handle both forms.

<package version="3.0" unique-identifier="uuid_id" prefix="rendition: http://www.idpf.org/vocab/rendition/#" xmlns="http://www.idpf.org/2007/opf">
  <metadata xmlns:dcterms="http://purl.org/dc/terms/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:opf="http://www.idpf.org/2007/opf" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
    <dc:identifier id="uuid_id">urn:uuid:f98ccab9-5abe-4f57-a9ee-36463c0598e2</dc:identifier>
    <meta name="cover" content="cover"/>
    <dc:subject>War</dc:subject>

    <!-- titles -->
    <dc:title id="title1">The Breaking Wave</dc:title>
    <meta refines="#title1" property="title-type">main</meta>
    <dc:title id="title2">Requiem for a WREN</dc:title>
    <meta refines="#title2" property="title-type">subtitle</meta>

    <!-- author -->
    <dc:creator id="create1">Nevil Shute</dc:creator>
    <meta refines="#create1" property="role" scheme="marc:relators">aut</meta>
    <meta refines="#create1" property="file-as">Shute, Nevil</meta>

    <dc:language>en</dc:language>
    <meta property="dcterms:modified">2025-03-12T12:07:52Z</meta>
    <dc:date>1955</dc:date>
    <meta name="Sigil version" content="2.4.2"/>
  </metadata>
</package>

Here's the meta description in the spec.

[gitting]

I guess it could be something as straightforward as augmenting current parsers in case some property didn't get populated. E.g. for parseCreators something like:

function getRefine(metas, id, prop) {
    const ref = metas.find((m) => {
        const props = m['$']
        return (props['refines'] === `#${id}`) && (props['property'] === prop);
    })

    return ref ? ref['_'] : null
}

function parseCreators(metadata) {
    if (!metadata['dc:creator']?.length) return null

    return metadata['dc:creator'].map((c) => {
        if (typeof c !== 'object' || !c['$'] || !c['_']) return false
        const namespace =
            Object.keys(c['$']).find((key) => key.startsWith('xmlns:'))?.split(':')[1] || 'opf'

        const creator = {
            value: c['_'],
            role: c['$'][`${namespace}:role`] || null,
            fileAs: c['$'][`${namespace}:file-as`] || null
        }

        // in case v3 is to be probed
        const id = c['$']['id']

        if (id && !(creator.role && creator.fileAs)) {
            const metas = metadata['meta']
            creator.role = creator.role || getRefine(metas, id, 'role')
            creator.fileAs = creator.fileAs || getRefine(metas, id, 'file-as')
        }

        return creator
    })
}

[advplyr]

I got a draft started #4113