Count links in Arrays in MongoDB collection - json

Question

I have a collection with objects, which are linking to other objects in the array:
{
"_id" : ObjectId("53f75bedc5489f86666d305e"),
"id" : "2",
"links_to" : [
{
"id" : 1,
"label" : null,
},
{
"id" : 3,
"label" : null,
},
{
"id" : 60,
"label" : null,
},
{
"id" : 23,
"label" : null,
},
},
{
"_id" : ObjectId("53f75bedc5489f86666d305e"),
"id" : "3",
"links_to" : [
{
"id" : 4,
"label" : null,
},
{
"id" : 8,
"label" : null,
},
{
"id" : 23,
"label" : null,
},
{
"id" : 2,
"label" : null,
},
},
...
Now I would like to write a query, which gives as an output for each id the number of links. Eg.:
{"id": 1, "numberOfLinks": 21},
{"id": 2, "numberOfLinks": 15},
...
Thanks in advance.

Answer 1

The best approach is to keep the count on the document and update it when you either $push or $pull elements of the array using the $inc operator. In this way the field is maintained on the document itself:
{
"links_to": [],
"linkCount": 0
}
When you "push"
db.collecction.update(
{},
{ "$push": { "links_to": newLink }, "$inc": { "linkCount": 1 } }
)
And "pull":
db.collecction.update(
{},
{ "$pull": { "links_to": newLink }, "$inc": { "linkCount": -1 } }
)
Without doing this, you can use the $size operator from the aggregation framework in Mondern MongoDB to get the array length:
db.collection.aggregate([
{ "$project": {
"numberOfLinks": { "$size": "$link_count" }
}}
])
Or in versions prior to MongoDB 2.6 you can count the array members after $unwind and $group:
db.collection.aggregate([
{ "$unwind": "$link_count" },
{ "$group": {
"_id": "$id",
"numberOfLinks": { "$sum": 1 }
}}
])
So usually unless you want something specifically "dynamic" then just maintain the count on the document. This avoids the overhead of calculation when you query.

Answer 2

Actually this is fairly simple to achieve using aggregation:
db.foo.aggregate([
{$unwind: "$links_to" },
{$group: { _id: {"lti":"$links_to.id"}, numberOfLinks: {$sum: 1} } },
{$project: { _id:0, id: "$_id.lti", numberOfLinks: "$numberOfLinks" } }
])
produces the desired output, though in reversed order of fields, at least in the shell output:
{ "numberOfLinks" : 3, "id" : 3 }
{ "numberOfLinks" : 3, "id" : 2 }
{ "numberOfLinks" : 1, "id" : 5 }
{ "numberOfLinks" : 2, "id" : 4 }
{ "numberOfLinks" : 3, "id" : 1 }
If you can live with an output like:
{ "_id" : { "linksToId" : 3 }, "numberOfLinks" : 3 }
{ "_id" : { "linksToId" : 2 }, "numberOfLinks" : 3 }
{ "_id" : { "linksToId" : 5 }, "numberOfLinks" : 1 }
{ "_id" : { "linksToId" : 4 }, "numberOfLinks" : 2 }
{ "_id" : { "linksToId" : 1 }, "numberOfLinks" : 3 }
you can skip the $project step of the aggregation pipeline.
This is extremely efficient. I did a test basically doing the same thing over a collection of 5M documents with roughly 17M relations. Takes 18 seconds on a not exactly high performance server.