Schuyler Erle > Geo-Coder-US-1.00 > Geo::Coder::US

Download:
Geo-Coder-US-1.00.tar.gz

Dependencies

Annotate this POD (1)

CPAN RT

New  1
Open  0
View/Report Bugs
Module Version: 1.00   Source  

NAME ^

Geo::Coder::US - Geocode (estimate latitude and longitude for) any US address

SYNOPSIS ^

  use Geo::Coder::US;

  Geo::Coder::US->set_db( "geocoder.db" );

  my @matches = Geo::Coder::US->geocode(
                    "1600 Pennsylvania Ave., Washington, DC" );

  my @matches = Geo::Coder::US->geocode(
                    "42nd & Broadway New York NY" )

  my ($ora) = Geo::Coder::US->geocode(
                    "1005 Gravenstein Hwy N, 95472" );

  print "O'Reilly is located at $ora->{lat} degrees north, "
                               "$ora->{long} degrees east.\n";

DESCRIPTION ^

Geo::Coder::US provides a complete facility for geocoding US addresses, that is, estimating the latitude and longitude of any street address or intersection in the United States, using the TIGER/Line data set from the US Census Bureau. Geo::Coder::US uses Geo::TigerLine to parse this data, and DB_File to store a highly compressed distillation of it, and Geo::StreetAddress::US to parse addresses into normalized components suitable for looking up in its database.

You can find a live demo of this code at http://geocoder.us/. The demo.cgi script is included in eg/ directory distributed with this module, along with a whole bunch of other goodies. See Geo::Coder::US::Import for how to build your own Geo::Coder::US database.

Consider using a web service to access this geocoder over the Internet, rather than going to all the trouble of building a database yourself. See eg/soap-client.pl, eg/xmlrpc-client.pl, and eg/rest-client.pl for different examples of working clients for the rpc.geocoder.us geocoder web service.

METHODS ^

In general, the only methods you are likely to need to call on Geo::Coder::US are set_db() and geocode(). The following documentation is included for completeness's sake, and for the benefit of developers interested in using bits of the module's internals.

Note: Calling conventions for address and intersection specifiers are discussed in the following section on CALLING CONVENTIONS.

Geo::Coder::US->geocode( $string )

Given a string containing a street address or intersection, return a list of specifiers including latitude and longitude for all matching entities in the database. To keep from churning over the entire database, the given address string must contain either a city and state, or a ZIP code (or both), or geocode() will return undef.

geocode() will attempt to normalize directional prefixes and suffixes, street types, and state abbreviations, as well as substitute TIGER/Line's idea of the "primary street name", if an alternate street name was provided instead.

If geocode() can parse the address, but not find a match in the database, it will return a hashref containing the parsed and normalized address or intersection, but without the "lat" and "long" keys specifying the location. If geocode() cannot even parse the address, it will return undef. Be sure to check for the existence of "lat" and "long" keys in the hashes returned from geocode() before attempting to use the values! This serves to distinguish between addresses that cannot be found versus addresses that are completely unparseable.

geocode() attempts to be as forgiving as possible when geocoding an address. If you say "Mission Ave" and all it knows about is "Mission St", then "Mission St" is what you'll get back. If you leave off directional identifiers, geocode() will return address geocoded in all the variants it can find, i.e. both "N Main St" and "S Main St".

Don't be surprised if geocoding an intersection returns more than one lat/long pair for a single intersection. If one of the streets curves greatly or doglegs even slightly, this will be the likely outcome.

geocode() is probably the method you want to use. See more in the following section on the structure of the returned address and intersection specifiers.

Geo::Coder::US->geocode_address( $string )

Works exactly like geocode(), but only parses addresses.

Geo::Coder::US->geocode_intersection( $string )

Works exactly like geocode(), but only parses intersections.

Geo::Coder::US->filter_ranges( $spec, @candidates )

Filters a list of address specifiers (presumably from the database) against a query specifier, filtering by prefix, type, suffix, or primary name if possible. Returns a list of matching specifiers. filter_ranges() will ignore a filtering step if it would result in no specifiers being returned. You probably won't need to use this.

Geo::Coder::US->find_ranges( $address_spec )

Given a normalized address specifier, return all the address ranges in the database that appear to cover that address. find_ranges() ignores prefix, suffix, and type fields in the specifier for search purposes, and then filters against them ex post facto. The intention for find_ranges() to find the closest match possible in preference to returning nothing. You probably want to use lookup_ranges() instead, which will call find_ranges() for you.

Geo::Coder::US->lookup_ranges( $address_spec, @ranges )

Given an address specifier and (optionally) some address ranges from the database, interpolate the street address into the street segment referred to by the address range, and return a latitude and longitude for the given address within each of the given ranges. If @ranges is not given, lookup_ranges() calls find_ranges() with the given address specifier, and uses those returned. You probably want to just use geocode() instead, which also parses an address string and determines whether it's a proper address or an intersection automatically.

Geo::Coder::US->find_segments( $intersection_spec )

Given a normalized intersection specifier, find all of the street segments in the database matching the two given streets in the given locale or ZIP code. find_segments() ignores prefix, suffix, and type fields in the specifier for search purposes, and then filters against them ex post facto. The intention for find_segments() to find the closest match possible in preference to returning nothing. You probably want to use lookup_intersection() instead, which will call find_segments() for you.

Geo::Coder::US->lookup_intersection( $intersection_spec )

Given an intersection specifier, return all of the intersections in the database between the two streets specified, plus a latitude and longitude for each intersection. You probably want to just use geocode() instead, which also parses an address string and determines whether it's a proper address or an intersection automatically.

CALLING CONVENTIONS ^

Most Geo::Coder::US methods take a reference to a hash containing address or intersection information as one of their arguments. This "address specifier" hash may contain any of the following fields for a given address:

ADDRESS SPECIFIER

number

House or street number.

prefix

Directional prefix for the street, such as N, NE, E, etc. A given prefix should be one to two characters long.

street

Name of the street, without directional or type qualifiers.

type

Abbreviated street type, e.g. Rd, St, Ave, etc. See the USPS official type abbreviations at http://www.usps.com/ncsc/lookups/abbr_suffix.txt for a list of abbreviations used.

suffix

Directional suffix for the street, as above.

city

Name of the city, town, or other locale that the address is situated in.

state

The state which the address is situated in, given as its two-letter postal abbreviation. See http://www.usps.com/ncsc/lookups/abbr_state.txt for a list of abbreviations used.

zip

Five digit ZIP postal code for the address, including leading zero, if needed.

lat

The latitude of the address, as returned by geocode() et al. If you provide this to as part of an argument to a Geo::Coder::US method, it will be ignored.

long

The longitude of the address, as returned by geocode() et al. If you provide this to as part of an argument to a Geo::Coder::US method, it will be ignored.

INTERSECTION SPECIFIER

prefix1, prefix2

Directional prefixes for the streets in question.

street1, street2

Names of the streets in question.

type1, type2

Street types for the streets in question.

suffix1, suffix2

Directional suffixes for the streets in question.

city

City or locale containing the intersection, as above.

state

State abbreviation, as above.

zip

Five digit ZIP code, as above.

lat, long

A single latitude and longitude for the intersection, as specified above. If you provide these values as part of an argument to a Geo::Coder::US method, they will be ignored.

BUGS, CAVEATS, MISCELLANY ^

The TIGER/Line data is notoriously buggy and inaccurate, but it seems to work reasonably well for urban areas. Geo::Coder::US uses interpolation to estimate the position of a particular address within a block, which means that it will necessarily be slightly inaccurate. Hey, it's only 14 meters off for my house, which is better than the 300 meter error given by another prominent geocoder, and definitely close enough for navigation.

In rural areas, TIGER/Line doesn't give names for lots of putative roads, even if the roads have names. Maybe the sign blew down the day before the Census agents got there, assuming there was ever a sign. What can you do? Similarly, lots of rural areas have official county subdivision names that an ordinary user would never think to give. Probably the right thing to do is map in names from a ZIP code database, but that data's not in TIGER/Line. What can you do? In general, you should expect the geocoder to be a lot more accurate in urban versus rural areas.

There may be many kinds of US street addresses which Geo::Coder::US can't parse. In particular, Geo::Coder::US strips out letters and dashes from house numbers, which may cause ambiguous results in certain parts of the country (particularly rural Michigan and Illinois, I think). Mea culpa. Send patches.

The full TIGER/Line data set is one heck of a lot of data -- about four gigabytes compressed, and over 24 gigs uncompressed. The BerkeleyDB database covering the whole US runs to 750+ megabytes uncompressed, or about 305 megs compressed. Unfortunately, I am not at present able to offer copies for download.

It would be nice to see a version of this for other countries, e.g. Geo::Coder::CA, Geo::Coder::DK, Geo::Coder::UK, with the same methods. Contact your local legislator about why the public geographic data for your country isn't freely available like it is in the US. If street address data is freely available for your country of choice, what are you waiting for?

TODO ^

Reverse geocoding methods, to retrieve the nearest street address from a given lat/long. This would probably necessitate using a R-tree or some other spatial indexing algorithm.

A metaphone index, for doing fuzzy matching on misspelled street and place names.

SEE ALSO ^

DB_File(3pm), Geo::TigerLine(3pm), Geo::Coder::US::Import(3pm)

TIGER/Line is a registered trademark of the US Census Bureau. Find out more, and get the latest TIGER/Line files from http://www.census.gov/geo/www/tiger/. Actually, the best place to download the data from is http://www2.census.gov/geo/tiger/tiger2004fe/.

You can find a live demo of this code at http://geocoder.us/. Our consultancy, Locative Technologies, offers service and support for this software on a contractual basis.

AUTHORS ^

Schuyler Erle <schuyler@nocat.net>

Jo Walsh <jo@frot.org>

Geo::Coder::US incorporates a patch submitted by John P. Linderman. Submit a useful patch and get your name added here, too!

COPYRIGHT AND LICENSE ^

Copyright (C) 2004 by Schuyler Erle and Jo Walsh

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself, either Perl version 5.8.3 or, at your option, any later version of Perl 5 you may have available.

syntax highlighting: