parse-genbank-location should warn about region/locality mix ups

[joverlee521]

Currently, parse-genbank-location strictly follows GenBank's documented pattern for geo_loc_name:

augur/augur/curate/parse_genbank_location.py

Lines 19 to 23 in 66e903a

	# Expected pattern for the location field is
	# "<country_value>[:<region>][, <locality>]"
	#
	# See GenBank docs for their "country" field:
	# https://www.ncbi.nlm.nih.gov/genbank/collab/country/

However, the GenBank records don't always follow this pattern as shown in nextstrain/rabies#10.

We've previously done this in ncov-ingest specifically for USA locations by checking for US state codes but we can do a more generalized check with something like pycountry. If there is a region/locality mix-up, the command should emit a warning with instructions on how to fix this with apply-geolocation-rules.

[genehack]

The alternative would be warning loudly and providing instructions on how to use the geo location file to override bad annotations?

I always worry that automatically fixing things like this will actually inject difficult-to-detect errors.

[joverlee521]

The alternative would be warning loudly and providing instructions on how to use the geo location file to override bad annotations?

I always worry that automatically fixing things like this will actually inject difficult-to-detect errors.

That's fair! We'd still have to use something like pycountry to detect these mix-ups to warn the users about them.