Fix Context Processing Algorithm 5.2.4

[skodapetr]

The specification https://www.w3.org/TR/json-ld11-api/#algorithm step 5.2.4 states:

If context was previously dereferenced, then the processor MUST NOT do a further dereference, and context is set to the previously established internal representation: set context document to the previously dereferenced document, and set loaded context to the value of the @context entry from the document in context document.

The current implementation is not according the specification as caching is done on level of context not context document as stated by the specification. This becomes an issue if the same document, remote context, is loaded multiple times.

In the current implementation, where the whole context in cached I got error on following example:

{
  "@context": [
    {
      "@version": 1.1,
      "Pracovní místo": {
        "@id": "https://slovník.gov.cz/datový/pracovní-místa/pojem/pracovní-místo",
        "@context": {
          "kvalifikace": {
            "@id": "https://slovník.gov.cz/datový/pracovní-místa/pojem/požadovaná-kvalifikace",
            "@context":  "http://example.com/práce.json"
          }
        }
      }
    },
    {
      "@version": 1.1,
      "věda": "https://slovník.gov.cz/generický/věda-a-výzkum/pojem/",
      "Výzkumné pracoviště": {
        "@id": "věda:ResearchOrganization",
        "@context": [
         "http://example.com/práce.json",
          {
            "orjk": {
              "@id": "věda:orjk"
            }
          }          
        ]
      }
    }
  ],
  "typ": "Pracoviště",
  "iri": "https://data.mff.cuni.cz/zdroj/číselník/organizační-struktura/oddělení/204"
}

where it fail to resolve věda:orjk.

The reason is that before loading the context with this class the remote context from http://example.com/práce.json is loaded from the cache. The issue as that the context was resolved using the first context, the one with Pracovní místo and thus there is no definition of věda.

The solution is to simple cache the JSON documents, as I believe the specification states. Then then loading the remote context from http://example.com/práce.json using the cache, it is properly resolved with the current parent. As a result the term věda is available for resolution of věda:orjk.

[filip26]

Hi,
thank you for the PR. The motivation for caching processed contexts is to save time and do not repeat the same computation multiple times. As far as I understand the issue you ran into, the current caching implementation introduces an unwanted side effect - a bug.

Feel free to open issues, or other PRs.

[jakubklimek]

@filip26 thanks for merging this. Would you consider releasing this bugfix also to Maven?

[filip26]

Good question, there have been more changes since 1.0.0. Let's plan 1.0.3 release on Friday 4/23.