Reverse- geocoding: How to determine the city closest to a (lat,lon) with BigQuery SQL?

Question

I have a huge collection of points - and I want to determine the closest city to each point. How can I do this with BigQuery?

Answer 1

I have a huge collection of points ...

Felipe's solution is perfect in many ways but I found that in cases when you really have not just few points to search for closest city and you cannot limit yourself with 60 miles distance below solution works much better

#standardSQL
WITH a AS (
  # a table with points around the world
  SELECT ST_GEOGPOINT(lon,lat) my_point
  FROM `fh-bigquery.geocode.201806_geolite2_latlon_redux`  
), b AS (
  # any table with cities world locations
  SELECT *, ST_GEOGPOINT(lon,lat) latlon_geo, ST_ASTEXT(ST_GEOGPOINT(lon,lat)) hsh 
  FROM `fh-bigquery.geocode.201806_geolite2_latlon_redux` 
)
SELECT AS VALUE 
  ARRAY_AGG(
    STRUCT(my_point, city_name, subdivision_1_name, country_name, continent_name) 
    LIMIT 1
  )[OFFSET(0)]
FROM (
  SELECT my_point, ST_ASTEXT(closest) hsh 
  FROM a, (SELECT ST_UNION_AGG(latlon_geo) arr FROM b),
  UNNEST([ST_CLOSESTPOINT(arr, my_point)]) closest
)
JOIN b 
USING(hsh)
GROUP BY ST_ASTEXT(my_point)

Note:

• I am using ST_CLOSESTPOINT function
• To mimic case of not just few points ... I am using same table as in in b so there are 100K points to search closest city for and also no limits on how close or far lookup city can be (for this scenario - query in original answer will end up with famous Query exceeded resource limits - while otherwise it shows much better if not best performance as it is truly stated in that answer)