根据条件在MongoDB中进行投影

Question

My document structure in c#:

 public class HashTableDocument : Model
    {
        public int Id { get; set; }
        public Dictionary<string, HashSet<int>> items= new Dictionary<string, HashSet<int>>();
    }

in Mongo:

{
    "_id" : 218,
    "items" : {
        "1" : [ 
            52711, 
            201610, 

        ],
        "2" : [ 
            246421, 
            390200
        ],
        "3" : [ 
            105628, 
            768519
        ],
        "26" : [ 
            17435, 
            22252, 
            61389, 
            65184, 
            72859, 
            81421, 
            931469, 
            933505, 
            938377, 
            959836
        ],
        "27" : [ 
            26917, 
            38706, 
            53862, 
            111816, 
            827294, 
            858348, 
            870334
        ]
    }
}

I want to be able to pass in any List ('x') of Integers to Mongo. And project on only those key value pairs, if values contains any of the integer in given list ('x').

For example, in above document. if i pass List = { 52711, 105628, 17435, 81421} to Mongo then It should return

{
    "_id" : 218,
    "items" : {
        "1" : [ 
            52711, 
            201610, 

        ],
        "3" : [ 
            105628, 
            768519
        ],
        "26" : [ 
            17435, 
            22252, 
            61389, 
            65184, 
            72859, 
            81421, 
            931469, 
            933505, 
            938377, 
            959836
        ],
     }
}

because each one of those key's value contains at least one element in it's list.

Answer 1

I don't know the C# syntax, but here's how to do it using the aggregation framework. Note that this uses the $objectToArray expression, introduced in version 3.4.4.

> db.test.aggregate([{
  $project: {
    x: {
      $filter: {
        input: {$objectToArray: "$items"},
        cond: {
          $gt: [
            {
              $size: {
                $filter: {
                  input: "$$this.v",
                  as: "int",
                  cond: {$in: ["$$int", [52711, 105628, 17435, 81421]]}
                }
              }
            },
            0
          ]
        }
      }
    }
  }
}])
{
  "result": [
    {
      "_id": 218,
      "items": [
        {
          "k": "1",
          "v": [
            52711,
            201610
          ]
        },
        {
          "k": "3",
          "v": [
            105628,
            768519
          ]
        },
        {
          "k": "26",
          "v": [
            17435,
            22252,
            61389,
            65184,
            72859,
            81421,
            931469,
            933505,
            938377,
            959836
          ]
        }
      ]
    }
  ],
  "ok": 1
}

However, it's generally not easy to do such computation when you have a structure like you do. This aggregation cannot use any indexes to restrict its search. Have you considered using the following schema instead?

{
  "_id": 218,
  "items": [
    {k: "1", v: [52711, 201610]},
    {k: "2", v: [246421, 390200]},
    {k: "3", v: [105628, 768519]},
    {k: "26", v: [17435, 22252, 61389, 65184, 72859, 81421, 931469, 933505, 938377, 959836]},
    {k: "27", v: [26917, 38706, 53862, 111816, 827294, 858348, 870334]},
  ]
}

Then your problem becomes much simpler, and you can do the following instead:

db.test.aggregate([
  {$match: {"items.v": {$in: [52711, 105628, 17435, 81421]}}},
  {
    $project: {
      items: {
        $filter: {
          input: "$items",
          cond: {
            $size: {
              $setIntersection:
                  [[52711, 105628, 17435, 81421], "$$this.v"]
            }
          }
        }
      }
    }
  }
])

And if you created an index on the field "items.v", the initial $match stage could leverage that index to do a more efficient query.