$geowithin with Foriegn Collection on $lookup - mongodb

Question

I have two collections Members and MobileUserLocations - where each users locations is saved(Can be multiple) as userId as the foreign field.
Members:
{
_id: ObjectId("591553ffa4233a181506880c"),
userName: "Test user"
}
MobileUserLocations:
{ _id: ObjectId("59156070a4233a1815068b6b"),
userId: ObjectId("591553ffa4233a181506880c"),
location: {type: "Point", coordinates: [76.9121, 10.2232]]},
updatedOn: 2017-05-12T07:12:48.626Z,
status: 1
},
{ _id: ObjectId("59156070a4233a1815068b6b"),
userId: ObjectId("591553ffa4233a181506880c"),
location: {type: "Point", coordinates: [76.8121, 10.1232]]},
updatedOn: 2017-05-12T07:12:48.626Z,
status: 1
}
I want to get the Members who are within a radius - say 5km with reference to a particular geo point - say: [10.0132295, 76.3630502] (lat,lng format).
I tried this:
collection.aggregate([
{$match: {_id: { $ne: options.accessToken.userId }},
{ "$lookup": {
"localField": "_id",
"from": "MobileUserLocations",
"foreignField": "userId",
"as": "userLocInfo"
}
},
{
$project: {
_id: 1,
userLocInfo: {
"$filter": {
"input": "$userLocInfo",
"as": "userLoc",
"cond": {
"$eq": [ "$$userLoc.status", -1],
"$$userLoc.location": {"$geoWithin": {"$centerSphere": [[76.3630502, 10.0132295], 5 / 3963.2]}}
}
}
}
}
},
{$unwind: "$userLocInfo"}
]
But not getting. If I am removing the $geowithin from the filter cond, it is getting, otherwise not getting. But if I am individullay querying the collections, I am getting the result.
Can anyone know the issue?

Answer 1

That does not work because $geoWithin is not a "logical operator", but it's a "query operator" and can only be used in an aggregation pipeline using $match. Fortunately for you, that is really what you want. Though you don't yet see why:
collection.aggregate([
{ "$match": {
"_id": { "$ne": options.accessToken.userId }
}},
{ "$lookup": {
"localField": "_id",
"from": "MobileUserLocations",
"foreignField": "userId",
"as": "userLocInfo"
}},
{ "$unwind": "$userLocInfo" },
{ "$match": {
"userLocInfo.status": -1,
"userLocInfo.updatedOn": "2017-05-12T12:11:04.183Z",
"userLocInfo.location": {
"$geoWithin": {
"$centerSphere": [[76.3630502, 10.0132295], 5 / 3963.2]
}
}
}}
])
There's a really good reason for that aside from it's the only way it works. To understand, look at the "explain" output:
{
"$lookup" : {
"from" : "MobileUserLocations",
"as" : "userLocInfo",
"localField" : "_id",
"foreignField" : "userId",
"unwinding" : {
"preserveNullAndEmptyArrays" : false
},
"matching" : {
"$and" : [
{
"status" : {
"$eq" : -1.0
}
},
{
"updatedOn" : {
"$eq" : "2017-05-12T12:11:04.183Z"
}
},
{
"location" : {
"$geoWithin" : {
"$centerSphere" : [
[
76.3630502,
10.0132295
],
0.00126160678239806
]
}
}
}
]
}
}
}
What that shows you is that both the $unwind and following $match get absorbed into the $lookup stage itself. This means that the $geoWithin and other conditions are actually executed on the foreign collection "before" the results are returned.
This is how $lookup deals with resulting joins that can possibly breach the 16MB limit. It's also the most efficient way you can presently "filter" results of the join.
So that's what you really want to do here instead.
Based on the data in your question, this statement:
db.members.aggregate([
{ "$lookup": {
"localField": "_id",
"from": "MobileUserLocations",
"foreignField": "userId",
"as": "userLocInfo"
}},
{ "$unwind": "$userLocInfo" },
{ "$match": {
"userLocInfo.location": {
"$geoWithin": {
"$centerSphere": [[76.9121, 10.2232], 5 / 3963.2]
}
}
}}
])
Filters out the one location in $lookup that matches the constraint:
/* 1 */
{
"_id" : ObjectId("591553ffa4233a181506880c"),
"userName" : "Test user",
"userLocInfo" : {
"_id" : ObjectId("59c3c37359f55d64d6e30297"),
"userId" : ObjectId("591553ffa4233a181506880c"),
"location" : {
"type" : "Point",
"coordinates" : [
76.9121,
10.2232
]
},
"updatedOn" : ISODate("2017-05-12T07:12:48.626Z"),
"status" : 1.0
}
}