Feature: Support non-namedNode subjects/objects in `match(Objects|Subjects)Of`

[jeswr]

So I can - e.g. match over literal types. I would expect the new type of the function to be matchSubjectsOf<T>(termWrapper: TermWrapperConstructor<T>, predicate?: string, object?: string | RDF.Term, graph?: string)

[langsamu]

@matthieubosquet this issue as well #14 (comment) highlight why I wanted Terms everywhere.

When we were designing that part and debating, I failed to articulate enough how Terms allow us to stay true to RDF/JS.
We went over to strings everywhere because we were focused on the few usage patterns that came to our minds then, and I lost track of the design goal of least interference.

Jesse is showing glimpses of usage patterns we were not expecting which, to my mind, prove we should have Terms everywhere.

[langsamu]

On the matter of string | Term:

I would very strongly recommend that we stay with a single type and not allow both of either strings or terms.

I understand that this is valid in both TS & JS, however I think that it is harder to implement and maintain and document and reason over. It is one of the reasons for JS bugs and unreadable TS type definitions.

If we allow either one of strings and terms then we will need to decide how to convert from strings to various term types. In my humble opinion, that would be a mistake as there is no idiomatic way to achieve it, only conventions and heuristics. Let's not do it.

Considering #12 (comment), I would tend to go with terms only.

[jeswr]

Considering #12 (comment), I would tend to go with terms only.

I am supportive of terms only