Framing should use embedded contexts for compaction

[gkellogg]

As specified, framing uses a single context to perform compaction, however, there are cases where specific sub-frames may want to use their own context. For example, consider the following frame:

{
"@context": {
  "foaf": "http://xmlns.com/foaf/0.1/",
  "doap": "http://usefulinc.com/ns/doap#",
  "doap_developer": {"@id": "doap:developer", "@type": "@id"},
  "doap_id": "@id",
  "doap_name": "doap:name",
  "foaf_name": "foaf:name"
 },
  "doap:name": null,
  "doap_developer": {
  "@context": {
    "foaf_id": "@id",
    "foaf_name": "foaf:name"
  }
  }
}

This defines aliases for @id and foaf:name exclusive to the values of doap:developer. The desired output would use the embedded context definition for compacting the portions of that frame; example output from doing this might be the following:

{
  "@context": {
    "foaf": "http://xmlns.com/foaf/0.1/",
    "doap": "http://usefulinc.com/ns/doap#",
    "doap_developer": {"@id": "doap:developer", "@type": "@id"},
    "doap_id": "@id",
    "doap_name": "doap:name"
    "foaf_id": "@id",
    "foaf_name": "foaf:name"
  },
  "@graph": [{
    "doap_id": "http://rubygems.org/gems/json-ld",
    "doap_developer": {
      "doap_id": "http://greggkellogg.net/foaf#me",
      "foaf_name": "Gregg Kellogg"
    },
    "doap_name": "JSON::LD"
  }, {
    "doap_id": "http://rubygems.org/gems/rdf",
    "doap_developer": [
      {"foaf_id": "http://ar.to/#self", "foaf_name": "Arto Bendiken"},
      {"foaf_id": "http://bhuga.net/#ben", "foaf_name": "Ben Lavender"},
      {"foaf_id": "http://greggkellogg.net/foaf#me", "foaf_name": "Gregg Kellogg"}
    ],
    "doap_name": "RDF.rb"
  }, {
    "doap_id": "http://rubygems.org/gems/rdf-aggregate-repo",
    "doap_developer": {"foaf_id": "http://greggkellogg.net/foaf#me", "foaf_name": "Gregg Kellogg"},
    "doap_name": "RDF::AggregateRepo"
  }, {
    "doap_id": "http://rubygems.org/gems/rdf-json",
    "doap_developer": {"foaf_id": "http://ar.to/#self", "foaf_name": "Arto Bendiken"},
    "doap_name": "RDF::JSON"
  }, {
    "doap_id": "http://rubygems.org/gems/rdf-microdata",
    "doap_developer": {"foaf_id": "http://greggkellogg.net/foaf#me", "foaf_name": "Gregg Kellogg"},
    "doap_name": "RDF::Microdata"
  }, {
    "doap_id": "http://rubygems.org/gems/rdf-n3",
    "doap_developer": {"foaf_id": "http://greggkellogg.net/foaf#me", "foaf_name": "Gregg Kellogg"},
    "doap_name": "RDF::N3"
  }, {
    "doap_id": "http://rubygems.org/gems/rdf-rdfa",
    "doap_developer": {"foaf_id": "http://greggkellogg.net/foaf#me", "foaf_name": "Gregg Kellogg"},
    "doap_name": "RDF::RDFa"
  }, {
    "doap_id": "http://rubygems.org/gems/rdf-rdfxml",
    "doap_developer": {"foaf_id": "http://greggkellogg.net/foaf#me", "foaf_name": "Gregg Kellogg"},
    "doap_name": "RDF::RDFXML"
  }]
}

The use case for this is to allow an easy way to map JSON-LD with embedding to a CSV format. The CSV transformation would just take the property definitions and flatten into a single table, such as the following:

doap_id	doap_name	foaf_id	foaf_name
<http://rubygems.org/gems/json-ld>	JSON::LD	<http://greggkellogg.net/foaf#me>	Gregg Kellogg
<http://rubygems.org/gems/rdf>	RDF.rb	<http://ar.to/#self>	Arto Bendiken
<http://rubygems.org/gems/rdf>	RDF.rb	<http://bhuga.net/#ben>	Ben Lavender
<http://rubygems.org/gems/rdf>	RDF.rb	<http://greggkellogg.net/foaf#me>	Gregg Kellogg
<http://rubygems.org/gems/rdf-aggregate-repo>	RDF::AggregateRepo	<http://greggkellogg.net/foaf#me>	Gregg Kellogg
<http://rubygems.org/gems/rdf-json>	RDF::JSON	<http://ar.to/#self>	Arto Bendiken
<http://rubygems.org/gems/rdf-microdata>	RDF::Microdata	<http://greggkellogg.net/foaf#me>	Gregg Kellogg
<http://rubygems.org/gems/rdf-n3>	RDF::N3	<http://greggkellogg.net/foaf#me>	Gregg Kellogg
<http://rubygems.org/gems/rdf-rdfa>	RDF::RDFa	<http://greggkellogg.net/foaf#me>	Gregg Kellogg
<http://rubygems.org/gems/rdf-rdfxml>	RDF::RDFXML	<http://greggkellogg.net/foaf#me>	Gregg Kellogg

This is also consistent with how an HTML representation of SPARQL results might look from the following query:

PREFIX doap: <http://usefulinc.com/ns/doap#>
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
SELECT *
WHERE {
  ?doap_id a doap:Project; doap:name ?doap_name; doap:developer ?foaf_id .
  ?foaf_id foaf:name ?foaf_name .
}

[seralf]

Yes, i also think the csv representation whould be coherent to the html one, which uses tables.
On the other hand the process could be seen in the opposite direction: for example if the CSV-LD needs to introduce specific columns or rows in the format to track things like context, linking on other csv files or whatever, I think those features could be seen as an input to provide suggestions for the a review of the HTML result format.

[gkellogg]

You can see the CSV-LD proposal here, comments are welcome at public-csv-wg@w3.org and/or here on public-linked-json@w3.org. It's just a start, and it's not intended to allow any JSON-LD to be transformed to CSV, but it does use the framing mechanism to allow two-way transformation between CSV and JSON-LD. There is also a proposal for describing @context information in CSV but I don't consider this to be a primary use-case.

[gkellogg]

This is satisfied by PR #445 and issues #247, #262, and #369.