Overhauling implementatation of ActiveFedora::Rdf

[no-reply]

** THIS BREAKS BACKWARD COMPATIBILITY **

This is a major rewrite of RDF Datastreams and the underlying node
support, tightly coupling the ActiveFedora::Rdf to rdf.rb. It replaces
the old RdfNode and related modules with a Resource class that
subclasses RDF::Graph.

The commit adds support for RDF Resources (nodes) persisted to an
RDF::Repository, which allows many ActiveFedora::Base objects to share
data without duplicating statements in each of their datastreams. It
will handle deep nodes in graphs gracefully, even if they are not
expected in the model. The other key features are some powerful
convenience methods for handling Linked Data; namely, #fetch (which
pulls data from an upstream URI) and #rdf_label (which is a way of
getting a human-readable label from common fields).

It also improves handling of typed data. Primarily by virtue of using
df.rb. Saving a Date object (for example) will serialize correctly
typed data and return a Date object when parsed.

A demo of the features included in the commit is here:
https://gist.github.com/no-reply/9519740

The commit breaks backwards compatability in several ways. Full
documentation of compatibility breakage and deprecation is here:
https://gist.github.com/terrellt/75a2fc84fd3d274a1bbd

• New object types now subclass ActiveFedora::Rdf::Resource:

  class Component < ActiveFedora::Rdf::Resource
  end

  instead of:

  class Component
  include ActiveFedora::RdfObject
  end

• Parameters for new Resources are changed.

• A resource is always returned for URI subjects, even when no
  class can be identified. This allows better handling of nested data
  from sources whose structure is unknown; properties won't be
  registered, but other Resource methods will be available. You can get
  the string version by calling #rdf_subject.

• Typed literals are supported. This includes changes to
  serialization of RDF! Passing (e.g.) a Ruby Date object as the value
  of a property/triple will yield a typed literal in RDF serializations
  and a Ruby Date object will be returned when fetching that data.

It also deprecates a few things for future removal.

• rdf_type is deprecated in favor of a general configuration syntax
  for Resources.
• map_predicates is superceded by property, which registers
  properties in a simpler way (without need for a block) using a pattern
  borrowed from some non-core ruby-rdf libraries.

This work depends on the linkeddata gem for support loading arbitrary
RDF serialations (see: ActiveFedora::Rdf::Resource#fetch).

[no-reply]

This is perhaps a few hours premature. Our cCLA is on the way, but it may not be on file yet.

Thought it wouldn't hurt for the PR to get some eyeballs today, since it's a pretty major commit.

[jcoyne]

Since this is such a big commit, with a massive impact, I propose we have at least 3 👍 before we merge.

[no-reply]

I'm in support of that @jcoyne.

This is squashed, in line with the contribution guidelines. The history is still available at https://github.com/no-reply/active_fedora/tree/rdf, if that helps people review the content.

[mjgiarlo]

I will be pointing my local ScholarSphere installation at this branch this afternoon to see what breaks and compare that to the documentation above. Stay tuned.

[jcoyne]

@no-reply can you rebase please ;)

[jcoyne]

Please use the ruby 1.9 hash syntax (http://guides.rubyonrails.org/contributing_to_ruby_on_rails.html#follow-the-coding-conventions)

[jcoyne]

klass = property_config.fetch(:class_name, ActiveFedora::Rdf::Resource)

[mjgiarlo]

I agree with @jcoyne's suggestions. ;) Once those are incorporated, I give this a hearty 👍, and will commit to testing a release candidate in a timely fashion.

It'd be good to hear from the UCSD contingent too. @escowles?

[escowles]

I agree with @mjgiarlo ;) -- I have started a branch to use this in damspas (https://github.com/ucsdlib/damspas/tree/feature/af7) and will spend some quality time on this, probably at c4l and definitely the following week.

[no-reply]

This has been rebased (and unceremoniously force-pushed) for further comment and/or merge.