MongoDB projection on specific nested properties

Using MongoDB 3.4.4 and newer versions:

db.collection.aggregate([
    { "$addFields": { 
        "data.visits.daily": {
            "$arrayToObject": {
                "$filter": {
                    "input": { "$objectToArray": "$data.visits.daily" },
                    "as": "el",
                    "cond": {
                        "$and": [
                            { "$gte": ["$$el.k", "2018-09-06"] },
                            { "$lte": ["$$el.k", "2018-09-07"] },
                        ]
                    }
                }
            }
        }
    } }
])

The above pipeline will yield the final output

{
    "data" : {
        "visits" : {
            "daily" : {
                "2018-09-06" : 2969,
                "2018-09-07" : 2624
            }
        }
    }
}

Explanations

The pipeline can be decomposed to show each individual operator's results.

$objectToArray

$objectToArray enables you to transform the document with dynamic keys into an array that contains a element for each field/value pair in the original document. Each element in the return array is a document that contains two fields k and v.

Running the pipeline with just the operator in a $project stage

db.collection.aggregate([
    { "$project": {
        "keys": { "$objectToArray": "$data.visits.daily" }
    } }
])

yields

{
    "_id" : ObjectId("5bab6d09b1951fef20a5dce4"),
    "keys" : [ 
        {
            "k" : "2018-09-05",
            "v" : 3586
        }, 
        {
            "k" : "2018-09-06",
            "v" : 2969
        }, 
        {
            "k" : "2018-09-07",
            "v" : 2624
        }, 
        {
            "k" : "2018-09-08",
            "v" : 2803
        }, 
        {
            "k" : "2018-09-09",
            "v" : 3439
        }, 
        {
            "k" : "2018-09-10",
            "v" : 3655
        }
    ]
}

$filter

The $filter operator acts as a filtering mechanism for the array produced by the $objectToArray operator, works by selecting a subset of the array to return based on the specified condition which becomes your query.

Consider the following pipeline which returns an array of the key/value pair that matches the condition "2018-09-06" <= key <= "2018-09-07"

db.collection.aggregate([
    { "$project": {
        "keys": { 
            "$filter": {
                "input": { "$objectToArray": "$data.visits.daily" },
                "as": "el",
                "cond": {
                    "$and": [
                        { "$gte": ["$$el.k", "2018-09-06"] },
                        { "$lte": ["$$el.k", "2018-09-07"] },
                    ]
                }
            }  
        }
    } }
])

which yields

{
    "_id" : ObjectId("5bab6d09b1951fef20a5dce4"),
    "keys" : [ 
        {
            "k" : "2018-09-06",
            "v" : 2969
        }, 
        {
            "k" : "2018-09-07",
            "v" : 2624
        }
    ]
}

$arrayToObject

This will transform the filtered array above from

[ 
    {
        "k" : "2018-09-06",
        "v" : 2969
    }, 
    {
        "k" : "2018-09-07",
        "v" : 2624
    }
]

to the original document with the dynamic key

{
    "2018-09-06" : 2969,
    "2018-09-07" : 2624
}

so running the pipeline

db.collection.aggregate([
    { "$project": { 
        "keys": {
            "$arrayToObject": {
                "$filter": {
                    "input": { "$objectToArray": "$data.visits.daily" },
                    "as": "el",
                    "cond": {
                        "$and": [
                            { "$gte": ["$$el.k", "2018-09-06"] },
                            { "$lte": ["$$el.k", "2018-09-07"] },
                        ]
                    }
                }
            }
        }
    } }
])

will produce

{
    "_id" : ObjectId("5bab6d09b1951fef20a5dce4"),
    "keys" : {
        "2018-09-06" : 2969,
        "2018-09-07" : 2624
    }
}

But of course you would want to preserve the original schema i.e. the current fields so you would need to use $addFields instead of the $project pipeline used for illustrated.

$addFields

This is is equivalent to a $project stage that explicitly specifies all existing fields in the input documents and adds the new fields. Specifying an existing field name in an $addFields operation causes the original field to be replaced and you would need to use dot notation to to update the embedded data.visits.daily field with the dynamic keys.