I have a document structure that is deeply nested, like this:

{id: 1, 
 forecasts: [ { 
             forecast_id: 123, 
             name: "Forecast 1", 
             levels: [ 
                { level: "proven", 
                  configs: [
                            { 
                              config: "Custom 1",
                              variables: [{ x: 1, y:2, z:3}]
                            }, 
                            { 
                              config: "Custom 2",
                              variables: [{ x: 10, y:20, z:30}]
                            }, 
                    ]
                }, 
                { level: "likely", 
                  configs: [
                            { 
                              config: "Custom 1",
                              variables: [{ x: 1, y:2, z:3}]
                            }, 
                            { 
                              config: "Custom 2",
                              variables: [{ x: 10, y:20, z:30}]
                            }, 
                    ]
                }
            ]
        }, 
    ]

}

I'm trying to update the collection to insert a new config, that looks like this:

newdata =  {
  config: "Custom 1", 
  variables: [{ x: 111, y:2222, z:3333}]
}

I'm trying something like this in mongo (in Python):

db.myCollection.update({"id": 1, 
                        "forecasts.forecast-id": 123, 
                        "forecasts.levels.level": "proven", 
                        "forecasts.levels.configs.config": "Custom 1"
                         },
                         {"$set": {"forecasts.$.levels.$.configs.$": newData}}
                      )

I'm getting "Cannot apply the positional operator without a corresponding query field containing an array" error though. What is the proper way to do this in mongo? This is mongo v2.4.1.

Solution 1

Unfortunately, you can't use the $ operator more than once per key, so you have to use numeric values for the rest. As in:

db.myCollection.update({
    "id": 1, 
    "forecasts.forecast-id": 123, 
    "forecasts.levels.level": "proven", 
    "forecasts.levels.configs.config": "Custom 1"
  },
  {"$set": {"forecasts.$.levels.0.configs.0": newData}}
)

MongoDB's support for updating nested arrays is poor. So you're best off avoiding their use if you need to update the data frequently, and consider using multiple collections instead.

One possibility: make forecasts its own collection, and assuming you have a fixed set of level values, make level an object instead of an array:

{
  _id: 123,
  parentId: 1,
  name: "Forecast 1", 
  levels: {
    proven: { 
      configs: [
        { 
          config: "Custom 1",
          variables: [{ x: 1, y:2, z:3}]
        }, 
        { 
          config: "Custom 2",
          variables: [{ x: 10, y:20, z:30}]
        }, 
      ]
    },
    likely: {
      configs: [
        { 
          config: "Custom 1",
          variables: [{ x: 1, y:2, z:3}]
        }, 
        { 
          config: "Custom 2",
          variables: [{ x: 10, y:20, z:30}]
        }, 
      ]
    }
  }
}

Then you can update it using:

db.myCollection.update({
    _id: 123,
    'levels.proven.configs.config': 'Custom 1'
  },
  { $set: { 'levels.proven.configs.$': newData }}
)

Solution 2

Managed to solve it with using mongoose:

All you need to know is the '_id's of all of the sub-document in the chain (mongoose automatically create '_id' for each sub-document).

for example -

  SchemaName.findById(_id, function (e, data) {
      if (e) console.log(e);
      data.sub1.id(_id1).sub2.id(_id2).field = req.body.something;

      // or if you want to change more then one field -
      //=> var t = data.sub1.id(_id1).sub2.id(_id2);
      //=> t.field = req.body.something;

      data.save();
  });

More about the sub-document _id method in mongoose documentation.

explanation:_id is for the SchemaName, _id1 for sub1 and _id2 for sub2 - you can keep chaining like that.

*You don't have to use findById method, but it's seem to me the most convenient as you need to know the rest of the '_id's anyway.

Solution 3

MongoDB has introduced ArrayFilters to tackle this issue in Version 3.5.2 and later.

New in version 3.6.

Starting in MongoDB 3.6, when updating an array field, you can specify arrayFilters that determine which array elements to update.

[https://docs.mongodb.com/manual/reference/method/db.collection.update/#specify-arrayfilters-for-an-array-update-operations][1]

Let's say the Schema design as follows :

var ProfileSchema = new Schema({
    name: String,
    albums: [{
        tour_name: String,
        images: [{
            title: String,
            image: String
        }]
    }]
});

And Document created looks like this :

{
   "_id": "1",
   "albums": [{
            "images": [
               {
                  "title": "t1",
                  "url": "url1"
               },
               {
                  "title": "t2",
                  "url": "url2"
               }
            ],
            "tour_name": "london-trip"
         },
         {
            "images": [.........]: 
         }]
}

Say I want to update the "url" of an image. Given - "document id", "tour_name" and "title"

For this the update query :

Profiles.update({_id : req.body.id},
    {
        $set: {

            'albums.$[i].images.$[j].title': req.body.new_name
        }
    },
    {
        arrayFilters: [
            {
                "i.tour_name": req.body.tour_name, "j.image": req.body.new_name   // tour_name -  current tour name,  new_name - new tour name 
            }]
    })
    .then(function (resp) {
        console.log(resp)
        res.json({status: 'success', resp});
    }).catch(function (err) {
    console.log(err);
    res.status(500).json('Failed');
})

Solution 4

This is a very OLD bug in MongoDB

https://jira.mongodb.org/browse/SERVER-831

Solution 5

I was facing same kind of problem today, and after lot of exploring on google/stackoverflow/github, I figured arrayFilters are the best solution to this problem. Which would work with mongo 3.6 and above. This link finally saved my day: https://thecodebarbarian.com/a-nodejs-perspective-on-mongodb-36-array-filters.html

const OrganizationInformationSchema = mongoose.Schema({
user: {
    _id: String,
    name: String
},
organizations: [{
    name: {
        type: String,
        unique: true,
        sparse: true
    },
    rosters: [{
        name: {
            type: String
        },
        designation: {
            type: String
        }
    }]
}]
}, {
    timestamps: true
});

And using mongoose in express, updating the name of roster of given id.

const mongoose = require('mongoose');
const ControllerModel = require('../models/organizations.model.js');
module.exports = {
// Find one record from database and update.
findOneRosterAndUpdate: (req, res, next) => {
    ControllerModel.updateOne({}, {
        $set: {
            "organizations.$[].rosters.$[i].name": req.body.name
        }
    }, {
        arrayFilters: [
            { "i._id": mongoose.Types.ObjectId(req.params.id) }
        ]
    }).then(response => {
        res.send(response);
    }).catch(err => {
        res.status(500).send({
            message: "Failed! record cannot be updated.",
            err
        });
    });
}
}

Solution 6

It's fixed. https://jira.mongodb.org/browse/SERVER-831

But this feature is available starting with the MongoDB 3.5.12 development version.

Note: This question asked on Aug 11 2013 and it's resolved on Aug 11 2017

Solution 7

Given how MongoDB doesn't appear to provide a good mechanism for this, I find it prudent to use mongoose to simply extract the element from the mongo collection using .findOne(...), run a for-loop search on its relevant subelements (seeking by say ObjectID), modify that JSON, then do Schema.markModified('your.subdocument'); Schema.save(); It's probably not efficient, but it is very simple and works fine.

Solution 8

I searched about this for about 5 hours and finally found the best and easiest solution: HOW TO UPDATE NESTED SUB-DOCUMENTS IN MONGO DB

{id: 1, 
forecasts: [ { 
         forecast_id: 123, 
         name: "Forecast 1", 
         levels: [ 
            { 
                levelid:1221
                levelname: "proven", 
                configs: [
                        { 
                          config: "Custom 1",
                          variables: [{ x: 1, y:2, z:3}]
                        }, 
                        { 
                          config: "Custom 2",
                          variables: [{ x: 10, y:20, z:30}]
                        }, 
                ]
            }, 
            { 
                levelid:1221
                levelname: "likely", 
                configs: [
                        { 
                          config: "Custom 1",
                          variables: [{ x: 1, y:2, z:3}]
                        }, 
                        { 
                          config: "Custom 2",
                          variables: [{ x: 10, y:20, z:30}]
                        }, 
                ]
            }
        ]
    }, 
]}

Query:

db.weather.updateOne({
                "_id": ObjectId("1"), //this is level O select
                "forecasts": {
                    "$elemMatch": {
                        "forecast_id": ObjectId("123"), //this is level one select
                        "levels.levelid": ObjectId("1221") // this is level to select
                    }
                }
            },
                {
                    "$set": {
                        "forecasts.$[outer].levels.$[inner].levelname": "New proven",
                    }
                },
                {
                    "arrayFilters": [
                        { "outer.forecast_id": ObjectId("123") }, 
                        { "inner.levelid": ObjectId("1221") }
                    ]
                }).then((result) => {
                    resolve(result);
                }, (err) => {
                    reject(err);
                });

Solution 9

Sharing my lessons learned. I faced the same requirement recently where i need to update a nested array item. My structure is as follows

  {
    "main": {
      "id": "ID_001",
      "name": "Fred flinstone Inc"
    },
    "types": [
      {
        "typeId": "TYPE1",
        "locations": [
          {
            "name": "Sydney",
            "units": [
              {
                "unitId": "PHG_BTG1"
              }
            ]
          },
          {
            "name": "Brisbane",
            "units": [
              {
                "unitId": "PHG_KTN1"
              },
              {
                "unitId": "PHG_KTN2"
              }
            ]
          }
        ]
      }
    ]
  }

My requirement is to add some fields in a specific units[]. My solution is first to find the index of the nested array item (say foundUnitIdx) The two techniques I used are

  1. use the $set keyword
  2. specify the dynamic field in $set using the [] syntax

                query = {
                    "locations.units.unitId": "PHG_KTN2"
                };
                var updateItem = {
                    $set: {
                        ["locations.$.units."+ foundUnitIdx]: unitItem
                    }
                };
                var result = collection.update(
                    query,
                    updateItem,
                    {
                        upsert: true
                    }
                );
    

Hope this helps others. :)

Solution 10

EASY SOLUTION FOR Mongodb 3.2+ https://docs.mongodb.com/manual/reference/method/db.collection.replaceOne/

I had a similar situation and solved it like this. I was using mongoose, but it should still work in vanilla MongoDB. Hope it's useful to someone.

const MyModel = require('./model.js')
const query = {id: 1}

// First get the doc
MyModel.findOne(query, (error, doc) => {

    // Do some mutations
    doc.foo.bar.etc = 'some new value'

    // Pass in the mutated doc and replace
    MyModel.replaceOne(query, doc, (error, newDoc) => {
         console.log('It worked!')
    })
}

Depending on your use case, you might be able to skip the initial findOne()

Solution 11

Okkk.we can update our nested subdocument in mongodb.this is our schema.

var Post = new mongoose.Schema({
    name:String,
    post:[{
        like:String,
        comment:[{
            date:String,
            username:String,
            detail:{
                time:String,
                day:String
            }
        }]
    }]
})

solution for this schema

  Test.update({"post._id":"58206a6aa7b5b99e32b7eb58"},
    {$set:{"post.$.comment.0.detail.time":"aajtk"}},
          function(err,data){
//data is updated
})