unify definitions of graph/dataset isomorphism #159
Conversation
makes it easier to reuse to define dataset isomorphism
| @@ -1053,17 +1052,11 @@ <h3>RDF Dataset Comparison</h3> | |||
| graph <var>NG1</var> and the RDF dataset <var>D2</var> with default graph | |||
| <var>DG2</var> and any named graph <var>NG2</var>) | |||
| are <dfn data-lt="dataset isomorphism" class="export">dataset-isomorphic</dfn> if and only if | |||
| there is a bijection <var>M</var> between the nodes, triples and graphs in | |||
| <var>D1</var> and those in <var>D2</var> such that all of the following properties hold:</p> | |||
| there is a bijection <var>M</var> from [=blank nodes=] to [=blank nodes=], | |||
There was a problem hiding this comment.
@gkellogg I don't understand your comment.
That being said, I reused the language about NG1 and NG2 as is, but I find it quite confusing. I'll change it into something that I find more understandable. Hopefully it will solve your comment above as well :)
There was a problem hiding this comment.
My reading of this is that it covers isomorphism of terms within the default and named graphs, but not on the graph names, themselves. The graph names also need to be part of M.
There was a problem hiding this comment.
The previous wording was indeed ambiguous, as it said "bijection M between nodes, triples and graphs", sounding like graph names were excluded. Actually, there were not excluded, as later it said:
<n, G> is in NG1 if and only if <M(n), M(G)> is in NG2.
so graph named are also mapped via M.
And the proposed wording in this PR (as well as Olaf's counterproposal in #160) is described as mapping from terms to terms, regardless on where those terms are used. So hopefully, this should be even clearer that graph names are also mapped.
|
While I agree with the general direction and the need to unify the definitions, I don't find this proposal particularly clean from a formal perspective; especially the part about extending M. The other major issue (which you have just mentioned yourself and which was there already before this PR) is the one about NG1 and NG2. Give me a few minutes to make another proposal. |
address #157
Preview | Diff