Mongoose - finding subdocuments by criteria

In Mongoose, you can also use the elegant .populate() function like this:

parents
.find({})
.populate({
  path: 'children',
  match: { age: { $gte: 18 }},
  select: 'name age -_id'
})
.exec()

You can use $elemMatch as a query-projection operator in the most recent MongoDB versions. From the mongo shell:

db.parents.find(
    {'children.age': {$gte: 18}},
    {children:{$elemMatch:{age: {$gte: 18}}}})

This filters younger children's documents out of the children array:

{ "_id" : ..., "children" : [ { "name" : "Margaret", "age" : 20 } ] }
{ "_id" : ..., "children" : [ { "name" : "John", "age" : 22 } ] }

As you can see, children are still grouped inside their parent documents. MongoDB queries return documents from collections. You can use the aggregation framework's $unwind method to split them into separate documents:

> db.parents.aggregate({
    $match: {'children.age': {$gte: 18}}
}, {
    $unwind: '$children'
}, {
    $match: {'children.age': {$gte: 18}}
}, {
    $project: {
        name: '$children.name',
        age:'$children.age'
    }
})
{
    "result" : [
        {
            "_id" : ObjectId("51a7bf04dacca8ba98434eb5"),
            "name" : "Margaret",
            "age" : 20
        },
        {
            "_id" : ObjectId("51a7bf04dacca8ba98434eb6"),
            "name" : "John",
            "age" : 22
        }
    ],
    "ok" : 1
}

I repeat the $match clause for performance: the first time through it eliminates parents with no children at least 18 years old, so the $unwind only considers useful documents. The second $match removes $unwind output that doesn't match, and the $project hoists children's info from subdocuments to the top level.