Report queries not handling empty labels on ERROR level

[psiotwo]

Standard set of report queries for the ERROR level does not seem to handle cases when a label is empty "".

There are rules e.g. to detect

• missing label
• label whitespace

But there does not seem to be check that a label is an empty string "" or an empty lang literal, e.g. ""@en. Should not this be covered on the ERROR level?

Also, this causes confusion in combination with the duplicate label query because it reports all concepts with empty labels as potential duplicates.

[matentzn]

There are maaaany checks we could add to ROBOT. The problem is that with every check, runtime increases. So we basically always sit together and decide:

1. How prevalent is this error
2. How damaging is this error
3. How long is the runtime of the check

Only if 1 is high or 2 is very high while 2 is reasonable, we start adding checks.. I have never seen this happen! https://api.triplydb.com/s/txFJGiVyt

[matentzn]

That said, http://robot.obolibrary.org/report_queries/missing_label could probably be extended to support "" and ""@en, if you like to give it a shot!

[psiotwo]

Clear @matentzn - I think this check would make sense because not having a label and having an empty label is often just a matter of inconsistent serialisation of the particular tool and does not bear any semantics (although it could, of course).

So having a check for missing_label not including empty labels seems a bug to me w.r.t. this check.

But I understand your reasoning and your proposal for missing_label was one option I was considering. I will prepare some PR for this and we can discuss it through there ...

[jamesaoverton]

Thanks @psiotwo! I think this will be a good addition, but yes let's ensure that there isn't a big performance hit.

[psiotwo]

Actually, is there any established way to test (both unit/performance tests) report queries? I must be overlooking st. in the codebase ...

[jamesaoverton]

No, you're not missing anything, there's no code here for that.

@matentzn runs a million ontology projects and the OBO Dashboard, so maybe he can suggest a project that takes 5-10 minutes to run robot report, and we can test against that.

[matentzn]

Actually, it would not hurt to add a few unit tests here as well. But yeah, we develop a query, make sure it works on a corpus, then push it to robot.

[psiotwo]

Closing via #1018 .