TWiki > Main > GeoPosition	TWiki webs: Main | TWiki | Know | Sandbox
	Main . { Changes | Index? | Search | Go }

Plugin: geoPosition

The geoPosition Plugin enables local searches. The plugin parses geographical meta tags (geo.position, DC.coverage.spatial and ICBM). If there are no coordinates in the document, coordinates can be loaded from conf/geodata.txt file.

Download at http://nutch.eventax.com/

Query Syntax

restaurant position:n52e10.0r10

hotel position:s-52e10.0r10

party position:n52.023w10.0r100

politics position:n52e10.0r100

Possible indentifiers are n, s, e, w and r.

Use either n or s. Values between -90 degrees and +90 degrees are useful.

Use either e or w. Values between -180 degrees and +180 degrees are useful.

Use r in kilometers. Values bigger then 3000 km might not work.

Config File Options

geoPosition.step

The accuracy positions are stored can be changed in the config file. Default is 1000m.

geoPosition.Domain2PositionFile

Default filename: conf/geodata.txt

The file consists of: URL-Prefix, North and East, tab-separated. Use negative values for south and west coordinates.

Example:

http://www.berlin.de 52.1234 9.9876

http://www.germany.de/berlin 52.1234 9.9876

Internal Documentation

Our Earth

The plugin further assumes that that the sea level is the same throughout the whole world. This should not make the error significantly larger.

Coordinate system

To avoid cpu-consuming calculations of sine, cosine and tangent, all the geographic coordinates are transformed before indexing to a 3-D-System with x, y, z.

The Point of origin is the middle of the earth.

x is the line through the Greenwich meridian and the equator.

z is the line through the north pole.

y is 90 degrees to x and 90 degrees to z.

Storeing

Searching

Searches are done by putting a cube around the point of search and searching all stuff between min and max values in each direction. In a 2D view this means, all stuff within a square instead of a circle is fetched. This means a maximum fault of nearly 40% in the distance, hits retrieved from. But it is fast. Distance ranking should minimize this problem in future.

Running search engines using this plugin

• This plugin is used for local searches at http://www.umkreisfinder.de/.

To do

• Checking the existing implementation.
• Speeding up all the stuff
  • Caching Range Queries
  • Using Field Cache http://jakarta.apache.org/lucene/docs/api/org/apache/lucene/search/FieldCache.html#getInts(org.apache.lucene.index.IndexReader,%20java.lang.String)
  • Using Hit Collector http://jakarta.apache.org/lucene/docs/api/org/apache/lucene/search/HitCollector.html

• Enabling users to add locations in well know ways, e.g. zip:12345.
• Also using distance for ranking.
• Using whois informations to get positions, if useful.
• Parsing addresses on websites to get positions.

-- MatthiasJaekle - 14 Oct 2004