Remove duplicates based on a key and referenced Objects in Mongodb? - mongodb

Question

I have MongoDB models of Actor and Movies. The Mongoose schema of both the models is as following :
var ActorsSchema = new Schema({
id : {
type : Number
},
known_for:[{
type: Schema.Types.ObjectId,
ref: 'Movie'
}]
})
var MovieSchema = new Schema({
genres: [{
type: Schema.Types.ObjectId,
ref: 'Genre'
}],
id: {
type: Number
}
});
known_for attribute in the actor model contains the reference to a list of movies in which that actor has starred.
I want to delete duplicate Actor records which would be determined using the id fieled (not the _id). But what I also want to do is delete the movies referenced in the deleted actor's record in the known_for field to also be deleted and I want to do that from the Mongo interface as the number of records in these documents is very large and performing this function programmatically would be time inefficient.
I have looked in to a related question but it does not apply to models who reference other models as there fields.

Answer 1

Consider using the aggregation framework to identify the duplicate documents, get a list of the duplicate _ids for the actors collection alongside the arrays of movie ids and issue remove and update commands with the ids array as the query.
For testing purposes, suppose you have the following data in your collections (with minimum test cases, for demonstration purposes of course):
db.movies.insert([
{
"_id" : ObjectId("5543e79e42063d2be5d2ea84"),
"id" : 1,
"genres" : []
},
{
"_id" : ObjectId("5543e79e42063d2be5d2ea85"),
"id" : 2,
"genres" : []
},
{
"_id" : ObjectId("5543e79e42063d2be5d2ea86"),
"id" : 3,
"genres" : []
}
]);
db.actors.insert([
{ id: 1, known_for: [ObjectId("5543e79e42063d2be5d2ea84")] },
{ id: 1, known_for: [ObjectId("5543e79e42063d2be5d2ea84")] },
{ id: 2, known_for: [ObjectId("5543e79e42063d2be5d2ea84"), ObjectId("5543e79e42063d2be5d2ea85")] },
{ id: 3, known_for: [ObjectId("5543e79e42063d2be5d2ea85"), ObjectId("5543e79e42063d2be5d2ea86")] }
]);
Now for the magical part. The aggregation pipeline groups the actors documents by id, calculates the grouped count, creates two array fields which hold the actor _id duplicates and the movies object ids. The pipeline outputs the results to a collection dupes that will be used later on to remove the duplicates:
db.actors.aggregate([
{
"$group": {
"_id": "$id",
"duplicates": { "$addToSet": "$_id" },
"movies": { "$addToSet": "$known_for"},
"count": { "$sum": 1 }
}
},
{
"$match": {
"count": { "$gt": 1 }
}
},
{
"$out": "dupes"
}
])
Querying the dupes collection will give the result:
/* 1 */
{
"_id" : 1.0000000000000000,
"duplicates" : [
ObjectId("5543fc8e42063d2be5d2eaa2"),
ObjectId("5543fc8e42063d2be5d2eaa1")
],
"movies" : [
[
ObjectId("5543e79e42063d2be5d2ea84")
]
],
"count" : 2
}
Now for the fun part. Use the dupes collection to then remove the dupes from the actors collection. As you have noticed from the dupes collection, the movies field is an array of arrays so you will need to flatten it and use the flattened array to then remove the movies and pull the orphaned movie references from the actors collection:
db.dupes.find({}).find({}).forEach( function (doc) {
var movie_dupes = [];
db.actors.remove({ "_id": { "$in": doc.duplicates } });
doc.movies.forEach( function (arr){
arr.forEach(function (id){
movie_dupes.push(id)
});
});
db.movies.remove({ "_id": { "$in": movie_dupes } });
db.actors.update({ "known_for": { "$in": movie_dupes } }, { "$pull": { "known_for": { "$in": movie_dupes } } }, { "multi": true });
});
Logs to console:
Removed 2 record(s) in 38ms
Removed 1 record(s) in 2ms
Updated 1 existing record(s) in 1ms
Now to verify whether our duplicates have been obliterated:
db.actors.find()
/* 1 */
{
"_id" : ObjectId("5543fc8e42063d2be5d2eaa3"),
"id" : 2,
"known_for" : [
ObjectId("5543e79e42063d2be5d2ea85")
]
}
/* 2 */
{
"_id" : ObjectId("5543fc8e42063d2be5d2eaa4"),
"id" : 3,
"known_for" : [
ObjectId("5543e79e42063d2be5d2ea85"),
ObjectId("5543e79e42063d2be5d2ea86")
]
}
Actor with id 1 (which was a duplicate) was indeed removed.
db.movies.find()
/* 1 */
{
"_id" : ObjectId("5543e79e42063d2be5d2ea85"),
"id" : 2,
"genres" : []
}
/* 2 */
{
"_id" : ObjectId("5543e79e42063d2be5d2ea86"),
"id" : 3,
"genres" : []
}
Movie with ObjectId("5543e79e42063d2be5d2ea84") was removed.