Is there a way to query an embedded document in an embedded document? - mongodb

Question

I have a weird mongodb document, but still need to query on it. Is it possible?
For example: I need every player within a certain radius.
{
"_id" : ObjectId("55d89c63c746230c200c528e"),
"speler_id" : 12,
"naam" : "Arjen Robben",
"seconds" : [
[
{
"locatie" : [
8.7173307286181370,
33.2784843816214250
],
"timestamp" : ISODate("1970-01-01T19:00:01.000Z")
},
{
"locatie" : [
-45.8853075448968970,
138.1526615469845800
],
"timestamp" : ISODate("1970-01-01T19:00:02.000Z")
},
{
"locatie" : [
80.5503710377444690,
10.0500048843973580
],
"timestamp" : ISODate("1970-01-01T19:00:03.000Z")
}
]
]
}

Answer 1

Well you can always use $geoWithin with $center or $centerSphere ( depending on whether these are global geometry coordinates or just a flat plane, for distance caluation purposes ) after processing with $unwind in the aggregation framework:
db.collection.aggregate([
{ "$unwind": "$seconds" },
{ "$unwind": "$seconds" },
{ "$match": {
"seconds.locatie": {
"$geoWithin": {
"$centerSphere": [
[
8.7173307286181370,
33.2784843816214250
],
100
]
}
}
}}
])
Which on the presented data would return:
{
"_id" : ObjectId("55d89c63c746230c200c528e"),
"speler_id" : 12,
"naam" : "Arjen Robben",
"seconds" : {
"locatie" : [
8.717330728618137,
33.278484381621425
],
"timestamp" : ISODate("1970-01-01T19:00:01Z")
}
}
{
"_id" : ObjectId("55d89c63c746230c200c528e"),
"speler_id" : 12,
"naam" : "Arjen Robben",
"seconds" : {
"locatie" : [
80.55037103774447,
10.050004884397358
],
"timestamp" : ISODate("1970-01-01T19:00:03Z")
}
}
Since $geoWithin does not "require" a geospatial index, then this is fine to use at later aggregation stages than the initial match. The $centerSphere in this case defines a point to query from and the "radius" extending from that point. This is just really a geometery "shortcut" as you can alternately provdide a GeoJSON polygon of your own definition.
But it's not really great. And mostly because it will not be able to use an index and therefore is pretty much brute force over the whole collection, and in that you cannot do nice things like return the distance from the queried point, like you can do with $geoNear.
Therefore while you can do things like this, most geoSpatial queries with MongoDB are best left to keeping that location data at the top level of the document, rather than embedded within arrays. So such modelling usually means having separate collection objects rather than embedded ones for the best results.
If you want an aggregated array in your response, then it is better to do this in aggregation after the intial geospatial query is made.