Add diff tool about kgx

[jmcmurry]

My orthogonal drive-by thoughts. I'm a little out of the loop so my concerns may already have been accounted for.

Such diffs would be great; to be informative requires that there be effective collapsing of functionally equivalent IDs (both for nodes and for edges).

Nodes:

• Diffs will be artificially noisy if a given node N (whose ID is syntactically consistently across KGs) has the predicate Y (whose ID is syntactically different across KGs).
• Diffs will be artificially sparse in cases where a given node N has IDs syntactically different across KGs.

Edges:

• Diffs will be artificially sparse in cases where gene-to-disease are not equated with disease-to-gene.

Thus if we can agree on a common identifier syntax, that helps. However, we still need the ability to collapse conceptual equivalents such as between MONDO and Orphanet; no syntactic sugar can fix that.

Feel free to talk me down; I realize this equivalence business is non-trivial and something we are trying to tackle in Data Commons too.

[yy20716]

Chris, I was trying to handle this issue but let me ask you some questions first if you don't mind.

• Could you please clarify what type means in your description? I guess it would not be rdf:type, so I checked the kg standard specification. It seems that the type meant category, but was not sure about this.
• As far as I know the internal graph generated by the current rdf_transformer does not have any corresponding fields yet (i.e. even category one). Actually, the current codes do not add or asssign any node attributes but only edge ones. I guess this happens because the specification and the codes are not finalized yet, but I just wonder how to proceed this task. Any suggestions would be helpful. Thank you.