Running mongodb $geoWithin without a [long,lat] array - mongodb

Question

I have a mongodb $geoWithin query as followed
db.test.find(
{
'loc': {
$geoWithin: {
$geometry: {
type : "Polygon" ,
coordinates: [[list of co-ordinates]]
}
}
}
}
);
So here the query runs on the loc field which is a array of lng, lat values but fortunately in my data, lat and lng values are in 2 different fields like
{
lat:12,
long:122
}
In this case how can I run the above query?

Answer 1

The best thing you can do really is to tranform your documents to store the data better. By preference you should probably go for GeoJSON format. But more later.
Fortunately since $geoWithin does not actually "require" an index ( but it really still is the better option to have one ) you can actually do the transformation "on the fly" with the aggregation framework instead:
Transform "On the Fly"
Hoping you have at least MongoDB 2.6, there is $map:
db.collection.aggregate([
// Tranform to array
{ "$project": {
"location": {
"$map": {
"input": ["lng","lat"],
"as": "el",
"in": {
"$cond": [
{ "$eq": [ "$$el", "lng" ] },
"$long",
"$lat"
}
}
}
}
}},
// Then match
{ "$match": {
"location": {
"$geoWithin": {
"$geometry": {
"type": "Polygon" ,
"coordinates": [[list of co-ordinates]]
}
}
}
}}
])
MongoDB 3.2 has a much simpler syntax:
db.collection.aggregate([
// Tranform to array - pretty simple huh!
{ "$project": {
"location": ["$long","$lat"]
}},
// Then match
{ "$match": {
"location": {
"$geoWithin": {
"$geometry": {
"type": "Polygon" ,
"coordinates": [[list of co-ordinates]]
}
}
}
}}
])
Or if you still have MongoDB 2.4 - Upgrade! Okay, use this then:
db.collection.aggregate([
// Add an array field
{ "$project": {
"long": 1,
"lat": 1
"location": { "$const": [ "A", "B" ] }
}},
// Unwind it
{ "$unwind": "$location" },
// Group back and map it!
{ "$group": {
"_id": "$_id",
"location": {
"$push": {
"$cond": [
{ "$eq": [ "$location", "A" ] },
"$long",
"$lat"
]
}
}
}},
// Then match
{ "$match": {
"location": {
"$geoWithin": {
"$geometry": {
"type": "Polygon" ,
"coordinates": [[list of co-ordinates]]
}
}
}
}}
])
Tranform "Permanently"
But really the best case is to change the document structure permanantly. The modern way to do this is in bulk with something like:
var ops = [];
db.collection.find({}).forEach(function(doc) {
ops.push({
"updateOne": {
"filter": { "_id": doc._id },
"update": {
"$set": {
"location": {
"type": "Point",
"coordinates": [doc.long,doc.lat]
}
},
"$unset": { "long": "", "lat": "" }
}
}
});
// Send once in 1000 only
if ( ops.length % 1000 == 0 ) {
db.collection.bulkWrite(ops);
ops = [];
}
})
// Clear remaining queue
if ( ops.length > 0 )
db.collection.bulkWrite(ops);
But generally speaking, loop the source documents and update each one to create the new "location" field. Then of course "index" it:
db.collection.createIndex({ "location": "2dsphere" })
And now the documents look like that and actually have an index, you can use regular $geoWithin queries directly which will also work faster from the present indexed data.