CODEWITHSUNDEEP
elasticsearchfilterdistance
woshitom
woshitom

Reputation: 5131

Elasticsearch custom geo distance filter

From an Elasticsearch query I'd like to retrieve all the points within a variable distance. Let say I have 2 shops, one is willing to deliver at maximum 3 km and the other one at maximum 5 km:

PUT /my_shops/_doc/1
{
    "location": {
      "lat": 40.12,
      "lon": -71.34
    },
    "max_delivery_distance": 3000
}

PUT /my_shops/_doc/2
{
    "location": {
      "lat": 41.12,
      "lon": -72.34
    },
    "max_delivery_distance": 5000
}

For a given location I'd like to know which shops are able to deliver. IE query should return shop1 if given location is within 3km and shop2 if given location is within 5km

GET /my_shops/_search
{
  "query": {
    "bool": {
      "must": {
        "match_all": {}
      },
      "filter": {
        "geo_distance": {
          "distance": max_delivery_distance,
          "location": {
            "lat": 40,
            "lon": -70
          }
        }
      }
    }
  }
}

Upvotes: 0

Views: 525

Answers (2)

Val
Val

Reputation: 217254

There's another way to solve this without scripting (big performance hogger !!) and let ES sort it out using native Geo shapes.

I would model each document as a circle, with a center location and a (delivery) radius. First, your index mapping should look like this:

PUT /my_shops
{
  "mappings": {
    "properties": {
      "delivery_area": {
        "type": "geo_shape",
        "strategy": "recursive"
      }
    }
  }
}

Then, your documents then need to have the following form:

PUT /my_shops/_doc/1
{
  "delivery_area" : {
    "type" : "circle",
    "coordinates" : [-71.34, 40.12],
    "radius" : "3000m"
  }
}

PUT /my_shops/_doc/2
{
  "delivery_area" : {
    "type" : "circle",
    "coordinates" : [-72.34, 41.12],
    "radius" : "5000m"
  }
}

And finally the query simply becomes a geo_shape query looking at intersections between a delivery point and the delivery area of each shop.

GET /my_shops/_search
{
  "query": {
    "bool": {
      "filter": {
        "geo_shape": {
          "delivery_area": {
            "shape": {
              "type": "point",
              "coordinates": [ -70, 40 ]
            },
            "relation": "contains"
          }
        }
      }
    }
  }
}

That's it! No scripting, just geo operations.

Upvotes: 2

Briomkez
Briomkez

Reputation: 577

I think that you need to work with a script to use another field as parameter. After some research I come to this answer:

GET my_shops/_search
{
  "query": {
    "script": {
      "script": {
        "params": {
          "location": {
            "lat": 40,
            "lon": -70
          }
        },
        "source": """
      return doc['location'].arcDistance(params.location.lat, params.location.lon)/1000 <= doc['max_delivery_distance'].value"""
      }
    }
  }
}

Basically, we exploit the fact that the classes related to the GEO points are whitelisted in painless https://github.com/elastic/elasticsearch/pull/40180/ and that scripts accepts additional parameters (your fixed location).

According to the documentation of arcDistance we retrieve the size in meters, so you need to convert this value into km by dividing by 1000.

Additional Note

I assume that location and max_delivery_distance are always (for each document) defined. If it is not the case, you need to cover this case.

Reference

Upvotes: 1

Related Questions