Repository#supports? for :literal_equality

[no-reply]

:literal_equality might be the wrong term.

Most triplestores inline literals by their value, rather than storing a value + a string expression, and fail some literal equality specs as a result. We should make conformance for those specs optional for RDF::Repository impls, and add a #supports? flag for it. We could then use that flag to turn off the handful of literal equality specs that test, e.g. '01' != '1' for repositories that don't want or need support for that "feature".

[no-reply]

@bendiken: @gkellogg and I discussed this via skype today. Your feedback is welcome. I'll probably get to this soon, since it's relevant for my Blazegraph work.

[no-reply]

Some baseline tests for literal equality support are in a branch at https://github.com/ruby-rdf/rdf-spec/compare/feature/equality.

Interestingly, RDF::Repository::Implementation fails several of them, due to reliance on Array#include?. We should have some discussion on the level of support expected here; I'll wait on some resolution there before making the existing 01 != 1 tests conditional on :supports?(:literal_equality).

Also related is the ongoing equality work connected to #260.

[gkellogg]

I think we need to allow both interpretations, and have implementations that go both ways, with tests conditional on supports? :literal_equality.

Perhaps a repository that does not support literal_equality always canonicalizes literals on insert, which is basically what this is saying.

[no-reply]

Closed via: #281.