Dan Moore!

Set up your own geocode service

Update, 2/9/06: this post only outlines how to set up a geocode engine for the United States. I don't know how to do it for any other countries.

Geocoder.us provides you with a REST based geocoding service, but their commercial services are not free. Luckily, the data they use is public domain, and there are some helpful perl modules which make setting up your own service a snap. This post steps you through setting up your own geocoding service (for the USA), based on public domain census data. You end up with a Google map of any address in the USA, but of course the lat/long you find could be used with any mapping service.

First, get the data.

$ wget -r -np -w 5 --random-wait ftp://www2.census.gov/geo/tiger/tiger2004se/ 

If you only want the data for one state, put the two digit state code at the end of the ftp:// url above (eg ftp://www2.census.gov/geo/tiger/tiger2004se/CO/ for Colorado's data).

Second, install the needed perl modules. (I did this on cygwin and linux, and it was a snap both times. See this page for instructions on installing to a nonstandard location with the CPAN module and don't forget to set your PERL5LIB variable.)

$ perl -MCPAN -e shell
cpan> install S/SD/SDERLE/Geo-Coder-US-1.00.tar.gz
cpan> install S/SM/SMPETERS/Archive-Zip-1.16.tar.gz

Third, import the tiger data (this code comes from the Geo::Coder::US perldoc, and took 4.5 hours to execute on a 2.6ghz pentium4 with 1 gig of memory). Note that if you install via the CPAN module as shown above, the import_tiger_zip.pl file is under ~/.cpan/:

$ find www2.census.gov/geo/tiger/tiger2004se/CO/ -name \*.zip 
  | xargs -n1 perl /path/to/import_tiger_zip.pl geocoder.db

Now you're ready to find the lat/long of an address. Find one that you'd like to map, like say, the Colorado Dept of Revenue: 1375 Sherman St, Denver, CO.

$ perl -MGeo::Coder::US -e 'Geo::Coder::US->set_db( "geocoder.db" );
my($res) = Geo::Coder::US->geocode("1375 Sherman St, Denver, CO" ); 
print "$res->{lat}, $ res->{long}\n\n";'

39.691702, -104.985361

And then you can map it with Google maps.

Now, why wouldn't you just use Yahoo!'s service (which provides geocoding and mapping APIs)? Perhaps you like Google's maps better. Perhaps you don't want to use a mapping service at all, you just want to find lat/longs without reaching out over the network.