dimanche 30 décembre 2012

A geocoding client API in GDAL/OGR

A new API has been added to GDAL/OGR to use geocoding services, such as OpenStreetMap Nominatim, MapQuest Nominatim, Yahoo! PlaceFinder, GeoNames or Bing.

From C, this is as simple as :
/* Create a session with default options */
OGRGeocodingSessionH hSession = OGRGeocodeCreateSession(NULL);

/* Now query the service */
OGRLayerH hLayer = OGRGeocode(hSession, "Paris", NULL, NULL);

/* Use the result layer (the first feature is generally the most relevant) */

/* Cleanup */
OGRGeocodeFreeResult(hLayer);
OGRGeocodeDestroySession(hSession);
More interesting, you can use that capability in SQL with the ogr_geocode() function that has been added to the SQLite SQL dialect :
SELECT ST_Centroid(ogr_geocode('Paris'))

returns:

OGRFeature(SELECT):0
POINT (2.342878767069653 48.85661793020374)
or a bit more evolved, to add geocoded information to a CSV file :
ogrinfo cities.csv -dialect sqlite -sql \
   "SELECT *, ogr_geocode(city, 'country') AS country,
    ST_Centroid(ogr_geocode(city)) FROM cities" \
    --config OGR_GEOCODE_LANGUAGE en

returns:

OGRFeature(SELECT):0
id (Real) = 1
city (String) = Paris
country (String) = France
POINT (2.342878767069653 48.85661793020374)

OGRFeature(SELECT):1
id (Real) = 2
city (String) = London
country (String) = United Kingdom
POINT (-0.109369427546499 51.500506667319407)

[...]

OGRFeature(SELECT):6
id (Real) = 7
city (String) = Beijing
country (String) = People's Republic of China
POINT (116.391195 39.9064702)

For better efficiency, the results of geocoding queries are cached into a local datasource (by default, a simple SQLite database in the working directory, but the location can be overriden by a configuration option, and other formats - CSV or PostgreSQL - can be selected).

Reverse geocoding is also possible with OGRGeocodeReverse() or the ogr_geocode_reverse() SQL function :
SELECT ogr_geocode_reverse(-100,45,'country','zoom=3') AS country

returns :
OGRFeature(SELECT):0
  country (String) = United States of America
Note: the online services that you may use have generally Term of Uses that disallow bulk geocoding, so use the above capability with care.