Geo-spatial queries in MongoDB

My lightening talk from the Melbourne MongoDB meetup, on how there are multiple ways to query geo-spatial indexes

Transcript

1. geo-spatial queries in MongoDB page 1
2. First you need a geo-spacial index on your collection. db.places.ensureIndex({ location : "2d" }) Your values will need to be "within earth like bounds" ★ latitude -180 to 180 ★ longitude -90 to 90 You can specify other ranges incase you're not dealing with 'earth like bounds' db.places.ensureIndex( { location : "2d" } , { min : -500 , max : 500 } ) page 2
3. ings to be aware of e order in the array that you are indexing matters. longitude/latitude - otherwise you’re results are going to be wildly inaccurate. db.places.insert({ location : [30, 40] }) page 3
4. Find $near Query for things that are nearby to a longitude/latitude db.places.find({"location" : {"$near" : [32, 42]}}) Documents are returned in the order of distance from the location By default will return 100 documents, unless you specify a limit db.places.find({"location" : {"$near" : [32, 42]}}).limit(35) page 4
5. Find $within You can query for points within a shape ★ a circle ★ a square ★ polygon Your result-set is not sorted by distance, as a result this type of query is a bit quicker page 5
6. Find within circles You need to pass in the radius of the circle to de ne the size of it. db.places.find({"location" : { "$within" : {"$center" : [center, radius]}}}) page 6
7. Find within Polygons As of MongoDB v2+ you can query on polygons. myArea = [[ 10, 20 ], [ 10, 40 ], [ 30, 40 ], [ 30, 20 ]] db.places.find({ "location" : { "$within" : { "$polygon" : myArea } } }) If you’re not using MongoDB version 2, you probably should page 7
8. geoNear Returns the distance of each document from the location, as part of the resultset. Handy if you needed to show the distance from document to your location db.runCommand({ geoNear: 'locations', near: [32,42], maxDistance: 10 }) e maxDistance argument limits the distance from the central point. is example below quotes '10' degrees, realistically your numbers will be much smaller page 8
9. Your result-set will look something like this… { "ns" : "db.places", "near" : "1100010101001101111001111100010101", "results" : [ { "dis" : 1.924482540031314, "obj" : { "_id" : ObjectId("4f470841fb8d5f0b9a000023"), "location" : [ 33.81820081650795, 42.630697104587625 ], "name" : "My place" } }] continued… page 9
10. "stats" : { "time" : 0, "btreelocs" : 0, "nscanned" : 12, "objectsLoaded" : 1, "avgDistance" : 1.924482540031314, "maxDistance" : 1.924493274866918 }, "ok" : 1 Note the handy stats page 10
11. e distance is quoted in radians Which typically isn’t very useful. To the best of my knowledge you will need to calculate the distance in meters yourself. Question is how accurate you need your measurements to be. page 11
12. e earth isn't at, but MongoDB treats it as such (generally) Need to note that MongoDB treats your location queries as earth is a at two dimensional plane. It's a case of tradeoffs with accuracy and speed, as not only is the earth not at, it's also not perfectly spherical. If your calculations are near the poles, cover a lot of a area or need to be really accurate, you’ll need to do some math… http://en.wikipedia.org/wiki/Haversine_formula ere is no wrapping at the poles and international date line, although mongo will raise an error page 12
13. Geospatial Haystack Indexes Assuming you have a lot of data, you can split your geospatial indexes into smaller buckets. So that queries only utilise the indexes that are closest to them. eg: your search only uses the index for eastern Australia and not the whole world is is only appropriate if the documents are reasonably close together db.places.ensureIndex({ location : "geoHaystack", type : 1 }, { bucketSize : 10 }) “bucketSize” = degrees latitude/longitude page 13
14. A haystack index can only be used by a database command, a.k.a geoNear db.runCommand({ geoSearch : "places", near : [32, 42], maxDistance : 6, search : { type : "bar" }, limit : 30 }) page 14
15. Links e documentation is actually really good http://www.mongodb.org/display/DOCS/Geospatial+Indexing Presentations on geo-spatial http://www.slideshare.net/ggoodale/geospatial-indexing-and-querying-with-mongodb page 15
16. anks for listening… any questions? dougal@jtribe.com.au @mrdougal page 16