Bug in scoped contexts

[jakubklimek]

There is probably a bug in handling scoped contexts. This is the same as json-ld/json-ld.org#734
In the snippet below, začátek and konec (beginning and end in Czech) have equivalent contexts. However, when used in data, they behave differently. Namely the context in konec is ignored, while when začátek is used, the context is applied. When the začátek definition is removed from the context, konec starts working. This can be tested in json-ld playground.

{
    "@context": {
        "typ": "@type",
        "time": "http://www.w3.org/2006/time#",
        "Časový interval": {
            "@id": "time:ProperInterval",
            "@context": {
                "začátek": {
                    "@id": "time:hasBeginning",
                    "@context": {
                        "time": "http://www.w3.org/2006/time#",
                        "xsd": "http://www.w3.org/2001/XMLSchema#",
                        "časy": "https://slovník.gov.cz/generický/čas/pojem/",
                        "Časový okamžik": {
                            "@id": "time:Instant",
                            "@context": {
                                "datum": {
                                    "@id": "time:inXSDDate",
                                    "@type": "xsd:date"
                                },
                                "datum_a_čas": {
                                    "@id": "time:inXSDDateTimeStamp",
                                    "@type": "xsd:dateTimeStamp"
                                },
                                "nespecifikovaný": {
                                    "@id": "časy:je-nespecifikovaný",
                                    "@type": "xsd:boolean"
                                }
                            }
                        }
                    }

                },
                "konec": {
                    "@id": "time:hasEnd",
                    "@context": {
                        "time": "http://www.w3.org/2006/time#",
                        "xsd": "http://www.w3.org/2001/XMLSchema#",
                        "časy": "https://slovník.gov.cz/generický/čas/pojem/",
                        "Časový okamžik": {
                            "@id": "time:Instant",
                            "@context": {
                                "datum": {
                                    "@id": "time:inXSDDate",
                                    "@type": "xsd:date"
                                },
                                "datum_a_čas": {
                                    "@id": "time:inXSDDateTimeStamp",
                                    "@type": "xsd:dateTimeStamp"
                                },
                                "nespecifikovaný": {
                                    "@id": "časy:je-nespecifikovaný",
                                    "@type": "xsd:boolean"
                                }
                            }
                        }
                    }

                }
            }
        }
    },
    "typ": "Časový interval",
    "konec": {
        "typ": "Časový okamžik",
        "datum_a_čas": "2019-01-14T09:30:00+02:00"
    }
}

[RinkeHoekstra]

I have a similar issue with scoped contexts and IRI expansion using scoped @base or @vocab definitions. Only the first scope context is used, the other is ignored. Adding this here, and linking to json-ld/json-ld.org#734 .

To test: two examples. One using @base and one using @vocab:

Using @base

{
  "@context": {
    "@base": "http://example.com/entity/test/",
    "@vocab": "http://example.com/schema/test/",
    "b": {
      "@context": {
        "@base": "http://example.com/test/"
      },
      "@type": "@id"
    },
    "a": {
      "@context": {
        "@base": "http://example.com/test/"
      },
      "@type": "@id"
    }
  },
  "@id": "justATest",
  "b": "1234",
  "a": "5678"
}

This should give

Subject	Predicate	Object
http://example.com/entity/test/justATest	http://example.com/schema/test/b	http://example.com/test/1234
http://example.com/entity/test/justATest	http://example.com/schema/test/a	http://example.com/test/5678

But instead gives

Subject	Predicate	Object
http://example.com/entity/test/justATest	http://example.com/schema/test/b	http://example.com/test/1234
http://example.com/entity/test/justATest	http://example.com/schema/test/a	5678

Using @vocab

Different context, now using the @vocab mechanism... same results:

{
  "@context": {
    "@base": "http://example.com/entity/test/",
    "@vocab": "http://example.com/schema/test/",
    "b": {
      "@context": {
        "@vocab": "http://example.com/test/"
      },
      "@type": "@vocab"
    },
    "a": {
      "@context": {
        "@vocab": "http://example.com/test/"
      },
      "@type": "@vocab"
    }
  },
  "@id": "justATest",
  "b": "1234",
  "a": "5678"
}

Results:

Subject	Predicate	Object
http://example.com/entity/test/justATest	http://example.com/schema/test/a	5678
http://example.com/entity/test/justATest	http://example.com/schema/test/b	http://example.com/test/1234

Change the order in the @context

Just including the example from the @base method... This context defines a first, and then b. Now the effect is the other way around. The value for a is properly expanded, and the one for b isn't.

{
  "@context": {
    "@base": "http://example.com/entity/test/",
    "@vocab": "http://example.com/schema/test/",
    "a": {
      "@context": {
        "@base": "http://example.com/test/"
      },
      "@type": "@id"
    },
    "b": {
      "@context": {
        "@base": "http://example.com/test/"
      },
      "@type": "@id"
    }
  },
  "@id": "justATest",
  "b": "1234",
  "a": "5678"
}

Results:

Subject	Predicate	Object
http://example.com/entity/test/justATest	http://example.com/schema/test/a	http://example.com/test/5678
http://example.com/entity/test/justATest	http://example.com/schema/test/b	1234