Cu 8692zguyq no preferred name

[mart-r]

The CAT.add_and_train_concept (by using CAT.add_concept internally) is able to add new concepts without registering a preferred name.

From my testing, this most likely has to do with the fact that the name specified is too short or empty. The minimum length of names is controlled by config.cdb_maker.min_letters_required. If the one (and only) name tied to a CUI has fewer characters than that (defaults to 2), then the medcat.preprocessing.cleaners.prepare_name method will return an empty dict and CDB.add_concept method cannot add a preferred name.

What I've done so far is the following:

• Add a few tests for prepare_name
  • Mostly to better understand how it works and what it does
• Add a warning in CDB.add_concept regarding no preferred name
• Add a warning in CAT.add_and_train_concept if names as returned to prepare_name is empty
  • Because in this case no preferred name will be able to be set
• Add tests to make sure the warnings work for short and empty names
  • As well as to make sure they don't if the name is sufficiently long

EDIT:
Had to run GHA a total of 13 times. For whatever reason, stuff kept being cancelled. I hope #368 will fix this for the future. GHA for that one was all good first time around.

[tomolopolis]

Task linked: CU-8692zguyq CAT.add_and_train_concept can add a concept with no preferred name

[tomolopolis]

lgtm - minor on the unittest asserts

[tomolopolis]

minor: there's an assertLogs you can use here instead of patching the method.

[mart-r]

Just tried that. And it works fine for 3.10+, but fails in 3.8 and 3.9 since these do not have the assertNoLogs method yet. The method was added in 3.10.
https://docs.python.org/3/library/unittest.html#unittest.TestCase.assertNoLogs

But I think I should be able to engineer a workaround.