Using ElasticSearch for Spatial Search tutorial

In this post I will demonstrate using ElasticSearch to spatially query records and  filter them by attributes.

ElasticSearch is built on top of Lucence which in version 4.0 supports Spatial query features, for those interested here is an example and link to javadocs.

You will need to install ElasticSearch, read about installing it here, also install marvel plugin for configuration and testing.

For this demonstration I have used data from Geofabrik. I have used point.shp file from Great Britan but you can use data from any other country as well.

To extract the data,I wrote a quick python script using pyshp library

One can easily install it by typing sudo easy_install pyshp in shell

Mapping type setup

Before uploading the records we need to create an Index in elasticsearch and apply schema mapping where we will instruct elasticsearch to create a geo index on location attribute.

We can do this easily by issuing a curl request

curl -XPUT http://localhost:9200/places -d '
{
    "mappings": {
        "place": {
            "properties": {
                "id": {"type": "double"},
                "name": {"type": "string"},
                "type": {"type": "string"},
                "location": {"type": "geo_point"}
            }
        }
    }
}

Once the schema for index places has been set up, its now time to add records to the index.

In the python script below modify the name of the shape file path and execute.

import shapefile
import urllib2
import json
sf = shapefile.Reader("points")
sr = sf.shapeRecords()

for r in sr:
    try :
        if r.record[2].strip() and r.record[3].strip():
            req = urllib2.Request('http://localhost:9200/places/place/')
            req.add_header('Content-Type', 'application/json')
            data = {'id': r.record[0].strip(),'name':r.record[2].strip(),'type':r.record[3].strip(),'location':{'lat':r.shape.points[0][1],'lon':r.shape.points[0][0]}}
            response = urllib2.urlopen(req, json.dumps(data))
            print r.record[2]
    except Exception,e:
        print e
        #print "ERROR ",r.record[0],r.record[2],r.record[3] , r.shape.points[0][0], r.shape.points[0][1]
        pass

The script inserts all records which have a valid name and type column into index.

Verify

Once complete we can quickly verify the number of records in the index by issuing.

GET places/_count from marvel or from a browser http://localhost:9200/places/_count

Spatial Search

In Marvel/Sense

GET places/_search
{
  "sort" : [
      {
          "_geo_distance" : {
              "location" : {
                    "lat": 51.5286416,
                "lon": -0.10159870000006777
              },
              "order" : "asc",
              "unit" : "km"
          }
      }
  ],
  "query": {
    "filtered" : {
        "query" : {
            "match_all" : {}
        },
        "filter" : {
            "geo_distance" : {
                "distance" : "20km",
                "location" : {
                   "lat": 51.5286416,
                   "lon": -0.10159870000006777
                }
            }
        }
    }
  }
}

or by curl

curl -XGET 'http://localhost:9200/places/place/_search?pretty=true' -d '
{
  "sort" : [
      {
          "_geo_distance" : {
              "location" : {
                    "lat": 51.5286416,
                   "lon": -0.10159870000006777
              },
              "order" : "asc",
              "unit" : "km"
          }
      }
  ],
  "query": {
    "filtered" : {
        "query" : {
            "match_all" : {}
        },
        "filter" : {
            "geo_distance" : {
                "distance" : "20km",
                "location" : {
                    "lat": 51.5286416,
                   "lon": -0.10159870000006777
                }
            }
        }
    }
  }
}'

To search by Geodistance as well as a term filter, modify the query to.

GET places / _search ? size = 100 & from = 0 {
    "sort": [{
        "_geo_distance": {
            "location": {
                "lat": 51.5286416,
                "lon": -0.10159870000006777
            },
            "order": "asc",
            "unit": "km"
        }
    }],
    "query": {
        "filtered": {
            "query": {
                "bool": {
                    "should": [{
                        "term": {
                            "type": "pub"
                        }
                    }]
                }
            },
            "filter": {
                "geo_distance": {
                    "distance": "1km",
                    "location": {
                        "lat": 51.5286416,
                        "lon": -0.10159870000006777
                    }
                }
            }
        }

    }
}

For better visualisation I have created a nice webapp here.

Web App

Hope this helps

admin

Using ElasticSearch for Spatial Search tutorial  by  admin