termFromId and termToId do not roundtrip on quoted literals

[jeswr]

As demonstrated by this failing test doing a round-trip of termFromId and termToId causes Literal TermTypes to become NamedNode term types because the internal id without literals gets used in

N3.js/src/N3DataFactory.js

Line 236 in 520054a

return term.id;

and the result is that termFromId interprets it as a namedNode because it does not contain quotes.

We need to either (a) add quotes to literals missing them when Literals are created, or (b) remove

N3.js/src/N3DataFactory.js

Line 236 in 520054a

return term.id;

which is the optimisation causing the bug.

Note that NamedNodes can also become literals if we create a NamedNode of the form new NamedNode('"s"')

[RubenVerborgh]

I think the problem is in the test: new Literal('abc') is not meaningful, because the (internal) Literal constructor (which is the Term constructor) requires a correct id as argument. This can also be seen here: main...jeswr:N3.js:bug/literal-roundtrip-failure#diff-8dc308ca806460d3608d347268016fb538008d8f9926902f331cc3553ababbfaR228

Note that this differs from the exported literal function (with which the test would succeed):

N3.js/src/N3DataFactory.js

Lines 339 to 342 in 520054a

	function literal(value, languageOrDataType) {
	// Create a language-tagged string
	if (typeof languageOrDataType === 'string')
	return new Literal(`"${value}"@${languageOrDataType.toLowerCase()}`);

[jeswr]

because the (internal) Literal constructor

These constructors are actually exported (c.f.

N3.js/src/index.js

Lines 38 to 39 in 520054a

	NamedNode,
	Literal,

). Perhaps it then worth removing these exports in the next mver (which I believe will be required in #311 anyway since there are a couple of minor breaking changes required to migrate from RDF* -> RDF-star)?

[RubenVerborgh]

Good idea, following up in #313