Mongodb geospatial query performance - mongodb

I am writing a nodejs app involving geocodes ( ~50million places) stored in mongodb (using mongo native driver). For development I am testing with sample data (airports of the world ~45k locations http://www.ourairports.com/data/airports.csv ). I am following GeoJSON format and building 2dsphere index on location field(named "geometry"). Then I am doing a $geowithin query to find locations.
Evrything works as expected. The issue is that when I am querying for a polygon containing US, the query is taking around 4-6sec for returning 22845 locations.
db. airports.ensureIndex({ geometry: '2dsphere' })
db.airports.find({ geometry: { '$geoWithin': { '$geometry': { type: 'Polygon', coordinates: [ [ [ -127.32917843921399, 75.11297289119061 ], [ -71.32126356078601, 75.11297289119061 ], [ -71.32126356078601, 12.305525108809391 ], [ -127.32917843921399, 12.305525108809391 ], [ -127.32917843921399, 75.11297289119061 ] ] ] } } } })
I am sure that this cant be correct. This way I cant imagine response times with 50million points.
Would appreciate if someone can point me in direction to improve the performance of this.

Related

Does precision influence performance with MongoDB GeoSpatial Queries?

Consider a query of:
db.stuff.find({ Location:
{ $geoWithin:
{ $centerSphere: [ [ -73.93414657, 40.82302903 ], 5 / 3963.2 ] } } })
Would it be reasonable to expect a performance boost if the coordinates were less precise? If so, by how much?
db.stuff.find({ Location:
{ $geoWithin:
{ $centerSphere: [ [ -73.93414, 40.82302 ], 5 / 3963.2 ] } } })
Bonus points if you have links to documentation on performance tuning mongo geospatial queries.
Unfortunately no.
And you can test this yourself by looking at $executionStats.
Be aware that you should change the coordinated on every call to avoid metric distortion due to caching.
More info about mongodb's spatial B-tree indexing here.

MongoDB $near query accuracy issue

In my application I would like to query items that are close by (e.g. within 5km) to a coordinate and I tried to use $near to achieve that. With a quick look I thought it worked but after I tested it further it seems the query is somewhat inaccurate. Here is my setup:
I selected 2 coordinates that are a bit less than 5km apart from each other:
61.4644750214197, 23.8426943813556
61.497133399999996, 23.778528100000003
(At least according to tools like this, this or this the distance between those coordinates should be about ~4,99km)
I added one of the coordinates into empty "items" collection:
db.items.insert({
"geo" : {
"type" : "Point",
"coordinates" : [
61.4644750214197,
23.8426943813556
]
}
});
I added "2dsphere" index to the collection make geospatial queries possible:
db.items.createIndex( { geo : "2dsphere" } )
Finally, I used the other coordinate with $near query:
db.items.find({geo: {
$near: {
$geometry: {
type: "Point" ,
coordinates: [ 61.497133399999996, 23.778528100000003 ]
},
$maxDistance: 5000 // according to docs with '2dsphere' index and GeoJSON this is is meters
}
}}).count()
I expected the result to be 1 but instead it is 0. Even if I set $maxDistance to 7000 the result is still 0, but if I set it to 8000 the result will be 1.
Am I doing something wrong or are MongoDB geospatial queries (or just $near query?) that inaccurate? If so, is there a better way to get accurate results for this kind of query? This is my first time dealing with geospatial queries in MongoDB so there is probably a trivial explanation for my problem.
EDIT:
Basically I was dreaming of a functionality to show all items in map within X kilometres from users current location and X could be determined by user. It would be awkward if an item within 5km would not be visible even when the user wants to filter items within 7km.
I have tried most of the options for doing this query, like $centerSphere, $nearSphere and geoNear with similar results. They all seem to claim that the distance between my earlier mentioned coordinates is somewhere between 7-8km. I'm starting to think either 1. I'm missing some key peace of information about how distances work in general or 2. it simply is not possible to solve my problem with mongodb. Below are my queries for the other options:
$centerSphere (0 results with 5, 6 and 7km but 1 result with 8km):
db.items.find( { geo: {
$geoWithin: { $centerSphere: [ [ 61.497133399999996, 23.778528100000003 ], 5/6378.1 ]
}
}}).count()
geoNear (0 results with maxDistance 5000, 6000 and 7000 but 1 result with 8000):
db.runCommand(
{
geoNear: "items",
near: { type: "Point", coordinates: [ 61.497133399999996, 23.778528100000003 ] },
spherical: true,
maxDistance: 5000
}
)
I understand I am late to the party, but for all those who are facing similar issue
The problem here is that when you store that data into "coordinates", it must be in the [longitude, latitude] order because this is how mongodb works. https://docs.mongodb.com/manual/geospatial-queries/#spherical I just ran your example with reversed order of coordinates and it worked as expected.
db.items.insert({
"geo" : {
"type" : "Point",
"coordinates" : [
23.8426943813556,
61.4644750214197
]
}
});
And then i ran
db.items.find({geo: {
$near: {
$geometry: {
type: "Point" ,
coordinates: [ 23.778528100000003 , 61.497133399999996]
},
$maxDistance: 5000
}
}}).count()
The count here is 1:
Hope it helps

MongoDB Geospacial Query on MultiPolygon

I've been reading through mongo's docs on geospacial querying, and have things working well for singl Polygon types but am having trouble with MultiPolygon. What I want to do is essentially this:
Given a MultiPolygon outlining areas of exclusion:
{
"type" : "MultiPolygon",
"coordinates" : [
[
[
[
-117.873730659485,
33.6152089844919
],
[
-117.873065471649,
33.615048159758
],
[
-117.873044013977,
33.614690770386
],
[
-117.873666286469,
33.6146729008785
],
[
-117.873730659485,
33.6152089844919
]
]
]
]
}
I simply want to be able to pass in a Point to see if it is excluded. I've tried $geoIntersects just to see if it even can determine if a Point is included or not, but that doesn't work. In the end, I want to check that a point is not included within the exclusion list, but the query is simpler without the additional $not operator... Here's what I've been trying:
var geoPoint = {type: 'Point', coordinates: [-117.8731230, 33.6150696]};
db.myCollection.aggregate([
{$match: {'exclusionsPolygons': {$geoIntersects: {$geometry: geoPoint}}}}
]);
Note that if I do the same exact thing with a GeoJSON type of Polygon then it works just fine:
Given this single polygon:
{
"type" : "Polygon",
"coordinates" : [
[
[
-117.8711744,
33.6129677
],
[
-117.8751744,
33.6129677
],
[
-117.874444839148,
33.6162171973226
],
[
-117.87287399259,
33.6172714730352
],
[
-117.871410434393,
33.6165209730032
],
[
-117.8711744,
33.6129677
]
]
]
}
This query works just find and returns the item(s) whose singular polygon contains the point:
var geoPoint = {type: 'Point', coordinates: [-117.8731230, 33.6150696]};
db.myCollection.aggregate([
{$match: {'singularPolygon': {$geoIntersects: {$geometry: geoPoint}}}}
]);
After some tinkering, it turns out the result set was right and I was wrong...
I was using the areas of interest on the map to get addresses to try to query against. One such place was, I thought, in an exclusion polygon:
However, once I made the polygon larger the result set started coming back as I expected it to... So, I reset the polygon and double-checked the map content, finding that if I zoom in further the area of interest was actually excluded from the polygon as there are multiple areas of interest contained:
Whoops - my bad :)

Query polygons $geowithin $box mongodb doesn't return anything

Guys I am trying to query all the polygons inside by bounding box but it simply returns 0.. It should be returning many polygons!
Alternatively I try to query a much larger bounding box and nothing happens!
My query is:
{
geometry:
{ $geoWithin:
{ $box:
[ [-73.995762,40.764826], [-73.934034,40.802038] ]
}
}
}
Notice that the very same query returns a valid result for geometries of type Point
to query all the polygons inside by bounding box but it simply returns 0
The $box operator for $geoWithin only supports documents based on grid coordinates and does not support GeoJSON shapes format.
the very same query returns a valid result for geometries of type Point
Depending on your documents structure, and how you are querying them, this is probably treated as being grid coordinates i.e. {geometry: [<long>, <lat>]}
This may not have worked for your polygons document because GeoJSON Polygons require an extra array wrapper. i.e. [[ [<long>, <lat>] ]] invalidating the grid coordinates format.
If your documents are in GeoJSON format, and you would like to select an area, you could utilise $geometry instead.
db.places.find(
{
'geometry': {
$geoWithin: {
$geometry: {
type : "Polygon" ,
coordinates: [ [ [ 0, 0 ], [ 3, 6 ], [ 6, 1 ], [ 0, 0 ] ] ]
}
}
}
}
)
Worth noting that MongoDB Compass geospatial visualisation currently (v1.6) does not support GeoJSON yet.
Here goes the answer:
$box do work with GeoJSON, but not with polygons! You need to generate a geometry with the $box instead.
Also, it always has to has the start point and point, so a 4-point polygon will have 5 coordinate tuples
BUT it gets trickier, it has to follow the right hand rule (anticlockwise).
TESTING GEOJSON WITH $BOX AND POINTS
my query:
{
geometry: {
$geoWithin: {
$box: [
[ -71.934034, 38.764826 ],
[ -75.995762, 43.802038 ]
]
}
}
}

MongoDB: Select all polygons within the radius of a point

I have two collections in my database: Post and Gallery. A Post has a single GeoJSON Point location, and a gallery is a collection of posts. The gallery's location is a GeoJSON Polygon bounding the gallery's posts (using quickhull algorithm). I now need to query for all galleries within x miles of a certain point, however I'm not getting any results even if I query from right next to my polygon.
I would like the behavior to be exactly the same as the following:
db.posts.find({
'location': {
$geoWithin: {
$centerSphere: [[-70, 30], 1000/3959]
}
}
});
Here, all posts within a the radius are returned. However, when I run the same type of function in this way, I am returned nothing, which is not correct:
db.galleries.find({
'location': {
$geoWithin: {
$centerSphere: [[-70, 30], 1000/3959]
}
}
});
One of my galleries has the following location (is 2dsphere index with 2dsphereIndexVersion = 2):
"location": {
"type": "Polygon",
"coordinates": [
[
[
-73.986882,
40.682829
],
[
-73.971089,
40.6672045
],
[
-73.955296,
40.65158
],
[
-73.986882,
40.682829
]
]
]
}
How do I query for location polygons that at least intersect with my radius?
I had the same problem and tried the exact same query that you initially tried. I'm not sure why it didn't work, but I was eventually able to get $near to do the job. Keep in mind that $near uses meters when calculating $maxDistance, so I had to convert my 10 mile desired distance to meters by multiplying by 1609.34. Here's the query that I ended up using:
db.maTowns.find(
{
'geometry': {
$near: {
$geometry: {
'type': "Point",
'coordinates': [ -71, 42 ]
},
$maxDistance: 10 * 1609.34
}
}
}
)
Select all polygons within the radius of a point
With the recent release of MongoDB version 3.6.0-rc0, you can now query GeoJSON LineStrings and Polygons with $geoWithin geospatial operator $centerSphere.
See also SERVER-27968 for more information about the change. Note that this change is pending to be backported.
Also maybe related for $geoIntersects and $centerSphere is ticket SERVER-30390