Mongodb select random query from collection except few ids - node.js

Question

I have a "User" and "Chat" collection
The User scheme is like:
username: {
type: String,
unique: true, required: true
}
And the Chat scheme:
from: {
type: Schema.Types.ObjectId,
required: true
},
to:{
type: Schema.Types.ObjectId,
required: true
},
content:{
type:String,
required:true
}
My purpose is to start a chat between an user and another random selected user.
My code so far to select random user:
User.aggregate([
{ "$match": { "_id": { "$ne": mongoose.Types.ObjectId(from_id) }}},
{$sample: {size: 1}}]
I'm using "$ne": mongoose.Types.ObjectId(from_id) so it won't select the same user.
What I want to achieve is choosing an user except same user (done) and also except someone that the first user has already chat with.
Like if there is already a chat between user1 and user2 then the code must choose another user except user2.

Answer 1

As an alternative to your approach:
const count = model.estimatedDocumentCount()
if (count === 2) {
throe new Error('Not possible')
}
let newUser = null
let i = 0
while(true) {
const random = Math.random() * count
newUser = await model.findOne({_id: {$ne: fromId}}).skip(random).exec()
if (!(await chatModel.findOne({$or: [{from: fromId, to: newUser._id}, {to: fromId, from: newUser._id}]}, {_id: 1}).exec())) {
break
}
if (i === 1000) {
throw new Error('Ah well not today') // can come up with some alternative strategy
}
++i
}
Use a compound index like ({from: 1, to: 1}) on Chat collection.
Should work nicely on large collection (because probability of hitting the same is low), as well as, small because collection should fit into memory and be easy to read from disk.