Updates the definition of graph isomorphism

[hartig]

.. as proposed in #128, including the additional change proposed in #128 (comment)

Notice that I also removed the "that is a node of G" part of bullet point 2 and 3. This change is needed for the new recursive part of the definition to be correct (i.e., when applying M to IRIs or literals inside triple terms).

Additionally, I also changed the last (now fifth) bullet point by putting M(p) in the second triple mentioned in this bullet point. While this change is not strictly necessary (because p is an IRI and M maps each IRI to itself), using M here looks a bit more consistent.

---

Preview | Diff

[pchampin]

LGTM.

I notice that M was also applied to the predicate of (asserted) triples and triple terms, whereas the original definition kept p as is. Since p is always an IRI, this is absolutely equivalent, but more verbose...

I have a slight preference for the terse version (p instead of M(p)), but I can live with both.

[hartig]

I have a slight preference for the terse version (p instead of M(p)), but I can live with both.

I'm fine with changing it back to p if others prefer that as well. However, in this case, I would also change the M(p) in the new fourth bullet point to p. Just to be consistent.

[afs]

I prefer M(p) because it extends to generalized RDF (and symmetric RDF if we have to add blank node predicates for sementics).

Also, it looks clearer to me because it is mapping p even if it only ever results in p.

[pfps]

Looks good to me

[hartig]

Given the approvals of all other editors, and also of @pfps who created the issue that this PR addresses, I will merge the PR now.