In general I think of single-precision floats as being precise to approximately a city block. Double-precision floats should be accurate to within inches.
Calculations with decimals are slower so I don't recommend them unless you need some really extreme precision (in which case you should probably be using a proper geospatial datastore).
Geocoder is lighter, easier to use, and has features missing from GeoKit like built-in result caching. It's also designed to be compatible with more storage engines (like MongoDB) and environments (JRuby, Rack, SQLite, etc).
GeoKit has been around longer but Geocoder is also production quality.
Geocoder currently supports Google, Yahoo, Bing, Yandex and Geocoder.ca APIs for street address geocoding. Geocoder.ca is free but only provides data for the US and Canada. Easy enough to configure:
ruby
Geocoder::Configuration.lookup = :geocoder_ca
The Geocoder README has a table comparing the features of the various APIs.
Check out MySQL spatial extensions and PostGIS.
In general I think of single-precision floats as being precise to approximately a city block. Double-precision floats should be accurate to within inches.
Calculations with decimals are slower so I don't recommend them unless you need some really extreme precision (in which case you should probably be using a proper geospatial datastore).
Geocoder is lighter, easier to use, and has features missing from GeoKit like built-in result caching. It's also designed to be compatible with more storage engines (like MongoDB) and environments (JRuby, Rack, SQLite, etc).
GeoKit has been around longer but Geocoder is also production quality.
Geocoder currently supports Google, Yahoo, Bing, Yandex and Geocoder.ca APIs for street address geocoding. Geocoder.ca is free but only provides data for the US and Canada. Easy enough to configure:
The Geocoder README has a table comparing the features of the various APIs.