OAK Adapters vs. Invalid URI/CURIEs about ontology-access-kit HOT 5 OPEN

Mid term solution: https://github.com/INCATools/semantic-sql/pull/77/files (this means these specific IRIs wont show up as invalid after the next OBO OAK release).

Please bring this issue to @cmungall attention.

from ontology-access-kit.

See also #688

In general the provider of upstream sources should try and provide prefix mappings such that the number of uncontracted URIs is minimal

I realize that this is harder than it should be for sql sources, due to the way semsql prefixes work. I apologize for this, and there will be a fix in place when we move away from rdftab. See INCATools/semantic-sql#41.

Even with this fix in place, there will inevitably be cases where URIs cannot be contracted (again, the number now is artifically high). In these cases we need to decide on the correct behavior:

1. throw error
2. treat URIs as if they are CURIEs (potentially renaming CURIE type to IDENTIFIER)
3. Quote URIs to discriminate them from CURIEs (note: this is currently the behavior of the semsql adapter, but it's not the behavior of the rdf adapters, this inconsistency is confusing for users)

I tend towards 3. I think this was @joeflack4's instincts with #688. However, I would do it this way:

1. rename CURIE to IDENTIFIER, such that this is used universally in signatures. This will be a large diff, but it will not affect runtime
2. Formally define this as IDENTIFIER = CURIE | QUOTED_URI. Again, this is typing information that does not affect runtime, but it helps make the intentions clear

We would also need a PR to change the default behavior of the rdf adapter

from ontology-access-kit.

On the typing front, I approve of the IDENTIFIER proposal.

from ontology-access-kit.

For completeness, there is also this option: ignore triples with entities that cannot be contracted and print a warning during semsql.

For programmatic usage its always going to be a bit annoying to have to distinguish between CURIE and "something else" based on some syntactic convention but I guess the "quote" solution is alright. Its not much different tbh the the <> situation we have right now, but at least we don't have to use a.replace('<','') anymore.

Maybe not a decision for Chrismtas period

from ontology-access-kit.