Metadata injection in epub3-to-epub3 does not update dc:identifier and title in content files

[martinpub]

Currently, the metadata injection is able to update dc:identifier and dc:title in package.opf, however, it is not doing anything to content files with regards to the corresponding <title> and <meta name="dc:identifier" content="xxx"/> values. This causes a mismatch in package.opf metadata and content file metadata, and it also causes non-valid EPUB 3 files according to the Nordic Guidelines.

Could this be easily added @bertfrees? I thought we discussed it at the spec stage, but perhaps we didn't. Also, it apparently hasn't been tested enough until now.

[bertfrees]

This is because the EPUB specification doesn't say anything about metadata in content documents.

It could be a bit tricky to implement this because you need to detect that the metadata fields in content and package documents indeed correspond. For example we could say a title element in a content document corresponds with a dc:title element in the package document if

• there is exactly one title element (not more than one is allowed in HTML)
• there is exactly one dc:title element
• their text matches

and something similar for dc:identifier vs. <meta name="dc:identifier">. We would then only update metadata fields in content documents that correspond to metadata fields in the package document that were updated (and maybe it should only be done when enabled with an option).

Alternatively, we could have an option similar to update-lang-attributes that synchronizes the metadata in the content documents with the package document:

• derive <meta name="dc:identifier"> element from the dc:identifier element marked as the "unique identifier", and remove any existing <meta name="dc:identifier"> elements
• derive title element from the dc:title element if there is exactly one (or from the first)

[martinpub]

Thanks @bertfrees, yes I see the implementation design issues here. I like the second option for its simplicity and similarity as a kind of "force" option like "update-lang-attributes" toggle. It could be named "update-contentdoc-meta"?

[bertfrees]

Yes I think I agree.

[martinpub]

That's great. Could you prioritise this over #18? This one is earlier in our production line. Also an estimate for when it could be done is very much appreciated, even though I know it's not always possible.

[martinpub]

Hi @bertfrees, any progress here? Just checking in.

[bertfrees]

Not yet.

[bertfrees]

I have implemented this. It's available on the master branch of daisy/pipeline-modules: daisy/pipeline-modules@78a2e6d.

[martinpub]

Great, thank you very much @bertfrees!

@kalaspuffar can I kindly request a cherry-pick to the fork?

[kalaspuffar]

Hi @bertfrees and @martinpub

Of course, I could help with this, but wouldn't it be safer if Bert creates a PR with this change, as usual, ensuring that all code will be merged.

Best regards
Daniel

[martinpub]

Sure @kalaspuffar. @bertfrees is that possible?