Make RDF and XML attribute definition syntax consistent

[mark-dce]

Maybe this is just crazy extra work that no-one will use, but I find it pretty confusing that in AF9 there are totally different syntaxes to specify how you index object attributes depending on whether the object metadata is persisted as RDF or XML. It seems like you should be able to specify how you want your object indexed independently of how it's persisted.

Currently in AF9, you define an object using RDF persistence like this:

class Book < ActiveFedora::Base
  property :title, predicate: ::RDF::DC.title, multiple: false do |index|
    index.as :stored_searchable
  end
  property :author, predicate: ::RDF::DC.creator, multiple: false do |index|
    index.as :stored_searchable
  end
end

And in AF9, you would define the same attributes persisted to XML as follows:

class Codex < ActiveFedora::Base
  contains 'descMetadata', class_name: 'CodexMetadata'

  has_attributes :title, datastream: 'descMetadata', multiple: false
  has_attributes :author, datastream: 'descMetadata', multiple: false
end

class CodexMetadata < ActiveFedora::OmDatastream
  set_terminology do |t|
    t.root(path: "fields")
    t.title
    t.author
  end

  def self.xml_template
    Nokogiri::XML.parse("<fields/>")
  end
end

It seems awkward to specify the indexing as part of the persistence for XML only. Further, it seems like historical cruft that RDF attributes are defined using property and XML attributes are defined using \has_metadata`. Would it be possible to do something more like:

# A proposed revision of XML persistence
class Codex < ActiveFedora::Base
  contains 'descMetadata', class_name: 'CodexMetadata'

  property :title, delegate_to: 'descMetadata', multiple: false do |index|
    index.as :stored_searchable
  end  

  property :author, delegate_to: 'descMetadata', multiple: false do |index|
    index.as :stored_searchable
  end  
end

class CodexMetadata < ActiveFedora::OmDatastream
  set_terminology do |t|
    t.root(path: "fields")
    t.title
    t.author
  end

  def self.xml_template
    Nokogiri::XML.parse("<fields/>")
  end
end

In this example, the XML attachment is used to specify the persistence only and the indexing is handled identically regardless of whether RDF or XML is used to persist metadata. Rather than providing a predicate to the property you specify the delegate_to to identify which XML attachment the term is persisted to.

[awead]

This seems like a good suggestion. I'm sure it's possible to do, but I'm not sure of the work that would be required. We'd need to pull out the indexing bits from OM. Whereas you'd usually assign indexing behaviors in OM, like:

  set_terminology do |t|
    t.root(path: "fields")
    t.title(index_as: [: stored_searchable])
    t.author(index_as: [: stored_searchable])
  end

We'd want to pass those bits from the model to the OM datastream. It also be nice we could do both so as to support both varieties.

[mark-dce]

@awead I think @jcoyne has the new indexing work mostly done in https://github.com/projecthydra/active_fedora/tree/move_indexing_to_the_model . I've tested it manually and updated dive into Hydra accordingly. For now, I think we want to keep in the legacy option of specifying indexing from the OM datastream, but I'm proposing we deprecate that and remove it in Hydra 10.

Also, take a look at 31b9843 too. These two branches together make the abstractions for RDF and XML metadata much more similar, which is a big plus in training new adopters - and maybe keeping the codebase simpler overall.

[awead]

👍