Fix JSON literal value version 1 check.

[dlongley]

So this fixes a bug where the JSON literal version 1 check was being performed on the unexpanded input instead of on the expanded output. IOW, it was looking for @value on the unexpanded input and then doing some version 1 check on it if @type was not @json. If the input uses an alias for @type (like type) this check will be erroneously triggered and throw an error. Example input:

{
  "@context": {
    "type": "@type"
  },
  "ex:foo": {
    "type": "@json",
    "@value": {test: 1}
  }
}

The conditional should have been checking for @type on the expanded output (expandedParent) instead -- a similar problem would have occurred if @value had been aliased.

However, in fixing this bug we start to fail some test:

#te029 Verifies that an exception is raised in Expansion when an invalid value object value is found

In looking at the input for that test, however, it seems like the test is wrong. The input is:

{
  "@type": true
}

Which is not a value object (there is no @value ... it is a top-level input). So I'm not quite sure what should be done here as I don't know what that test was suppose to test originally.

[davidlehn]

@dlongley That input you mention is for e028-in.jsonld, not e029-in.jsonld.

[dlongley]

@dlongley That input you mention is for e028-in.jsonld, not e029-in.jsonld.

Oh, well, something different is amiss I guess.

[davidlehn]

Not fine, test still fails. The expandedParent change isn't quite right. The @value checks need to be on the unexpanded data. Can fix by reverting the expandedParent change back to element except for that one that accesses @type. (Tested that with the new type and value alias tests.) Is that more correct?

[dlongley]

@davidlehn -- it seems like the check would just get skipped if you looked for @value on the unexpanded data and value alias were used.

[gkellogg]

It seems fine, but if you like, I'll take this over when I get on to JSON literals, in general.