I would like to find a way to specify a custom FTS5 rankingin such a way that empty values receive a higher rank. It would be easier to explain the question through the following setup.

sqlite> CREATE TABLE locations(rowid INTEGER PRIMARY KEY, country TEXT NOT NULL, city TEXT NOT NULL);
sqlite> 
sqlite> INSERT INTO locations (country, city) VALUES ("Mexico", "Mexico City");
sqlite> INSERT INTO locations (country, city) VALUES ("Mexico", "Cancun");
sqlite> INSERT INTO locations (country, city) VALUES ("Mexico", "Acapulco");
sqlite> INSERT INTO locations (country, city) VALUES ("Mexico", "");
sqlite> 
sqlite> CREATE VIRTUAL TABLE locations_search USING fts5(country, city, content='locations', prefix='3 4 5');
sqlite> INSERT INTO locations_search(locations_search, rank) VALUES('rank', 'bm25(1.0, 2.0)');
sqlite> INSERT INTO locations_search(rowid, country, city) SELECT rowid, country, city FROM locations;

Now if I perform the following query, the result looks something like this:

sqlite> SELECT country, city, rank FROM locations_search WHERE locations_search MATCH 'Mex*' ORDER BY rank;
Mexico|Mexico City|-1.52475247524752e-06
Mexico||-1.21259842519685e-06
Mexico|Cancun|-9.4478527607362e-07
Mexico|Acapulco|-9.4478527607362e-07

How can I write a query that ranks the second result higher, so country "Mexico" without a city should be the top ranked result, followed by "Mexico, Mexico City"?