[languages] Non-default language canonicals, more test cases

scripts/geodata/language_id/disambiguation.py

@@ -195,7 +195,8 @@ def disambiguate_language(text, languages):
                 elif not seen_languages and len(potentials) == 1 and len(t[0][1]) > 1:
                     possible_lang = lang if possible_lang is None or possible_lang == lang else None
 
-            if seen_languages and valid and not any((l in seen_languages for l in valid)):
+            if seen_languages and valid and not any((l in seen_languages for l in valid)) and \
+               (not any((valid_languages.get(l) for l in valid)) or any((valid_languages.get(l) for l in seen_languages))):
                 return AMBIGUOUS_LANGUAGE
 
             if len(valid) == 1:

scripts/geodata/tests/test_disambiguation.py

@@ -29,13 +29,20 @@ country_test_cases = [
     ('No 2 School House', 'us', UNKNOWN_LANGUAGE),
     ('E Thetford Rd', 'us', 'en'),
     ('El Camino', 'us', 'es'),
+    ('The El Camino', 'us', 'en'),
+    ('Via Antiqua Street', 'us', 'en'),
+    ('Salt Evaporator Plan Road', 'us', 'en'),
+    ('Calle Las Brisas North', 'us', 'en'),
+    ('Chateau Estates', 'us', 'en'),
+    ('Grand Boulevard', 'us', 'en'),
     ('Rue Louis Phillippe', 'us', 'fr'),
-    ('Calle Street', 'us', AMBIGUOUS_LANGUAGE),
+    ('Calle Street', 'us', 'en'),
     ('Del Rio Avenue', 'us', 'en'),
     ('South Signal Butte Road', 'us', 'en'),
     ('Chief All Over', 'us', UNKNOWN_LANGUAGE),
     ('South Alameda Street', 'us', 'en'),
     ('The Alameda', 'us', 'en'),
+    ('Rincon Road', 'us', 'en'),
 
     # Avenue + stopword
     ('Avenue du Bourget-du-Lac', 'je', 'fr'),