Points Closest to the Origin
As we've established in From the Viewport to the Query , any rectangle or bounding box can be targeted using the geo_bounding_box query.
Similarly, any polygon (incl. polygon circles) can be targeted through the geo_polygon query.
But what if we want all points that lie within a certain distance from an origin point? Think the nearest restaurants, the closest Uber driver, etc. That's what the geo_distance query is for.
I want to retrieve restaurants that serve pizza and lie within 0.5 miles of an origin geo point. I also want to give precedence to those that are closer to the origin:
Imaginary Restaurants in Midtown Manhattan. Map courtesy of geojson.io
PUT restaurants
{
"mappings": {
"properties": {
"location": {
"type": "geo_point"
},
"menu": {
"type": "text",
"fields": {
"keyword": {
"type": "keyword"
}
}
}
}
}
}POST restaurants/_doc
{
"name": "French Eatery",
"location": [-74.00620, 40.73981],
"menu": [
"european",
"french",
"pizza"
]
}
POST restaurants/_doc
{
"name": "Flatiron Streetfood",
"location": [-73.99500, 40.74039],
"menu": [
"pizza",
"kebab"
]
} We'll first apply a simple match query to target pizza restaurants:
{
"match": {
"menu": "pizza"
}
} Next, we'll be utilizing a geo_distance query but will need to keep in mind that it's a purely boolean, constant-score query — meaning it'll assign a score of 1 to any docs it matches and filter out those that are too far away:
{
"geo_distance": {
"distance": "0.5mi",
"location": [-73.9982, 40.7388]
}
} So on top of that, we'll apply a decay function_score query which is a fancy way of saying "the further one moves from the origin, the lower the location's score":
{
"function_score": {
"functions": [
{
"exp": {
"location": {
"origin": [-73.9982, 40.7388],
"scale": "0.5mi"
}
}
}
]
}
} Let's envelop them all in a must query:
POST restaurants/_search
{
"query": {
"bool": {
"must": [
{
"match": {
"menu": "pizza"
}
},
{
"geo_distance": {
"distance": "0.5mi",
"location": [-73.9982, 40.7388]
}
},
{
"function_score": {
"functions": [
{
"exp": {
"location": {
"origin": [-73.9982, 40.7388],
"scale": "0.5mi"
}
}
}
]
}
}
]
}
}
} There's a whole bunch of decay functions to choose from but I went with the exponential exp which has the smoothest decay curve.