Is it possible to change the standard prefixes?

[wu-lee]

This is a question, essentially a request for documentation, if possible, or a feature request if not.

Is it possible to change the prefixes used for the standard vocabularies defined here, such that RDF::URI#qname uses them when shorting URIs (and #pname, RDF::Vocab.exand_pname etc.)?

Using my case as an example, I would like to use the prefix dcterms instead of dc, and dc instead of dc11. This is because I want to work with TTL files with a pre-existing choice of prefixes which differ from the ones used in RDF::Vocab. (Both read and write them.)

I have found the RDF::Vocabulary.register method, which seems to allow adding new prefixes, and RDF::Vocabulary.limit, which are candidates when combined, but I am uncertain if these are designed to do this job reliably.

[wu-lee]

Empirically, using #limit and #register together would be unweildy, but in any case seems not to do the job:

require 'rdf'
require 'rdf/vocab'
puts RDF::URI.new("http://purl.org/dc/terms/foo").pname # => "dc:foo"

vocabs = RDF::Vocabulary.vocab_map.keys
vocabs.delete(:dc)
vocabs.delete(:dc11)
RDF::Vocabulary.limit_vocabs(*vocabs)
RDF::Vocabulary.register(:dcterms, RDF::Vocab::DC)

puts RDF::URI.new("http://purl.org/dc/terms/foo").pname # => http://purl.org/dc/terms/foo

(Using Ruby 2.7.0, rdf-3.1.15 and rdf-vocab-3.1.14)

[wu-lee]

This way works, but only partially, and I'm unsure if monkeying with #vocab_map directly is going to break something somewhere else too:

require 'rdf'
require 'rdf/vocab'

puts RDF::Vocabulary.expand_pname('dc:creator') 
# => http://purl.org/dc/terms/creator
puts RDF::URI.new("http://purl.org/dc/terms/creator").pname 
# => dc:creator
puts RDF::Vocabulary.expand_pname('dcterms:creator') 
# => dcterms:creator
puts RDF::URI.new("http://purl.org/dc/elements/1.1/creator").pname 
# => dc11:creator

RDF::Vocabulary.vocab_map.delete(:dc)
RDF::Vocabulary.vocab_map.delete(:dc11)

RDF::Vocabulary.register(:dc, RDF::Vocab::DC11)
RDF::Vocabulary.register(:dcterms, RDF::Vocab::DC)

puts RDF::Vocabulary.expand_pname('dc:creator')
# => http://purl.org/dc/elements/1.1/creator

puts RDF::URI.new("http://purl.org/dc/terms/creator").pname
# => dc:creator

puts RDF::Vocabulary.expand_pname('dcterms:creator') 
# => http://purl.org/dc/terms/creator

# All OK so far, but this shortening transform now goes back to `dc11`
puts RDF::URI.new("http://purl.org/dc/elements/1.1/creator").pname 
# => dc11:creator  


# Setting dcterms to like this gets the wrong shortening, and results
# in exceptions when expanding:
RDF::Vocabulary.register(:dcterms, "http://purl.org/dc/terms/")

puts RDF::URI.new("http://purl.org/dc/terms/creator").pname  
# => dc:creator
puts RDF::Vocabulary.expand_pname('dcterms:creator') 
# => ERROR: uninitialized constant RDF::DCTERMS (NameError)

# Whereas setting it like this allows transforms to succeed both ways, but 
# results in #pname reverting to the wrong prefix 
# (`dc:creator`, instead of `dcterms:creator`)
RDF::Vocabulary.register(:dcterms, RDF::Vocab::DC, uri: "http://purl.org/dc/terms/")

puts RDF::URI.new("http://purl.org/dc/terms/creator").pname  
# => dc:creator
puts RDF::Vocabulary.expand_pname('dcterms:creator') 
# => http://purl.org/dc/terms/creator

[gkellogg]

Not well documented in RDF::URI#pname, but it is somewhat better in RDF::URI#qname, which it depends on. You can pass a prefixes keyword parameter with a map of prefix name to URIl, which will be used for creating PNames.

RDF::URI('http://purl.org/dc/terms/creator').
  pname(prefixes: {dcterms: 'http://purl.org/dc/terms/'})
  # => "dcterms:creator"

The prefixes used by RDF::Vocabulary come from the class names of the vocabularies themselves. We could maybe add some registration mechanism to allow alternative prefixes to be used, instead, but that probably wouldn't be much more useful.

I can improve the method documentation for URI#qname and URI#pname with some examples. If you'd like to see something in the RDF::URI class documentation, let me know.

[gkellogg]

Fixed in 54ae4d4.

[wu-lee]

Ah - I've been using this documentation, which doesn't mention that parameter in #qname or #pname:

https://www.rubydoc.info/github/ruby-rdf/rdf/RDF/URI#qname-instance_method

Also, it seems this parameter isn't supported in the version I'm using (rdf-3.1.15). But deleting Gemfile.lock and re-running bundle install gets a new version, which does :)

I'm pleased to say that in my specific case right now, this is sufficient. However, it's often not feasible to go and alter all the usages of #pname to pass the prefix: parameter when I'm using the RDF gems or some other pre-existing code to handle some of the work. I may hit that problem later.

So that'd be a feature request: please allow arbitrary prefixes to be configured?

In any case, what might also be worth documenting is whether it's okay to re-register and/or delete things in #vocab_map, or if these should be considered append-only or read-only. Also, it'd be helpful to explicitly say that making changes to #vocab_map affect RDF::URI (and wherever else, or not, as is the case).

[gkellogg]

Rubydoc doesn't update itself very frequently. The latest version of the documents can usually be found at http://rdf.greggkellogg.net/yard/index.html. The prefixes argument as first released in 3.2, so it wouldn't have shown up there, I'm afraid.

Configuration of custom prefixes would probably look like either of the following:

RDF::Vocab::DC.__prefix__ =  :dcterms

This would set a class variable for the prefix to use, and RDF::Vocabulary.__prefix__ would turn into the following::

      def __prefix__
        instance_variable_defined?(:@__prefix__) ?
          @__prefix__ : 
          __name__.split('::').last.downcase.to_sym
      end

I'm going to move this issue to the RDF.rb repository.

[gkellogg]

@wu-lee See if that change works for you, and it will be in the next release.