Correct `canonize` documentation

[dpetran]

The docs on the README for canonize state:

const canonized = await jsonld.canonize(doc, {
  algorithm: 'URDNA2015',
  format: 'application/n-quads'
});

In order to properly accept n-quads the option needs to be:

const canonized = await jsonld.canonize(doc, {
  algorithm: 'URDNA2015',
  inputFormat: 'application/n-quads'
});

[davidlehn]

Something is a bit off here. I think the intent in those docs is that format was for the output format, and the input doc is assumed to be JSON-LD. With the inputFormat change here, that would run an N-Quads parser on the input instead of running toRdf. Is that what you were trying to do? Maybe just need to be clearer in examples that JSON-LD is the assumed input?

Looking here and in rdf-canonize, I'm not sure if format is even used currently? It may be left over from legacy code, or was put in with the intent to allow other formats than N-Quads. Due to how rdf-canonize is currently used, I'm not sure what other output would make sense. But if we're accepting format, is should probably at least test that it is the supported application/n-quads value.

[dpetran]

Ah, I was passing it an rdf document so that's how I assumed it was intended to be called (also, that's what the spec says the algorithm takes). I didn't continue reading the source past the inputFormatcheck, so I didn't think I could pass json-ld directly.

Does output format make sense for this function? I thought canonize always returned a string of nquads.

[dpetran]

I see now that in the README doc is a json-ld document. In that case we can close this PR - I think, though, that the options can be removed from this example call, since it handles json-ld just fine.

[gkellogg]

I think rdflib.js uses a plugin architecture for parsers, and should be able to select an appropriate parser based on the MIME type. If you're going to do something with format, it would probably be good to not bake in too many formats, and maybe better leverage rdflib.js for finding parsers. Ideally, it should play well if the input is Turtle, TriG, or even RDFa, given loaded libraries that support it.