Find Documents Where a Field Compares with Another in an Array

Question

Let's say I have a collection of documents that look like this:

{
    "_id" : ObjectId("5afa6df3a24cdb1652632ef5"),
    "createdBy" : {
        "_id" : "59232a1a41aa651ddff0939f"
    },
    "owner" : {
        "_id" : "5abc4dc0f47f732c96d84aac"
    },
    "acl" : [
        {
            "profile" : {
                "_id" : "59232a1a41aa651ddff0939f"
            }
        },
        {
            "profile" : {
                "_id" : "5abc4dc0f47f732c96d84aac"
            }
        }
    ]
}

I want to find all documents where createdBy._id != owner._id, AND where the createdBy._id appears in one of the entries in the acl array. Eventually, I will want to update all such documents to set the owner._id field to equal the createdBy._id field. For now, I'm just trying to figure out how to query the subset of documents I want to update.

So far, I have come up with this:

db.boards.find({
  $where: "this.createdBy._id != this.owner._id", 
  $where: function() {
    return this.acl.some(
      function(e) => {
        e.profile._id === this.createdBy._id
      }, this);
  }
)

(I have used ES5 syntax just in case ES6 isn't ok)

But when I run this query, I get the following error:

Error: error: { "ok" : 0, "errmsg" : "TypeError: e.profile is undefined :\n_funcs2/<@:2:36\n_funcs2@:2:12\n", "code" : 139 }

How do I perform this query / what is going on here? I would have expected my query to work, based on the docs I've read. Above, e should be an element of the acl array, so I expect it to have a field profile, but that doesn't seem to be the case.

Note, I'm using Mongo 3.2, so I can't use $expr, which I've seen some resources suggest is a possibility.

Resolution

It turns out that I had made an incorrect assumption about the schema of this collection. The reason I was running into the above error is because some documents have an acl array with an element that doesn't have a profile field. The below query checks for this case. It also has a single $where, because the way I had written it originally (with two) seemed to end up giving me an OR of the conditions instead of an AND.

db.boards.find({
  $where: function() {
    return this.acl.some(
      function(e) => {
        e.profile !== undefined && e.profile._id === this.createdBy._id && this.createdBy._id != this.owner._id
      }, this);
  }
)

Answer 1

You can still use aggregate() here with MongoDB 3.2, but just using $redact instead:

db.boards.aggregate([
  { "$redact": {
    "$cond": {
      "if": {
        "$and": [
          { "$ne": [ "$createdBy._id", "$owner._id" ] },
          { "$setIsSubset": [["$createdBy._id"], "$acl.profile._id"] }
        ]
      },
      "then": "$$KEEP",
      "else": "$$PRUNE"
    }
  }}
])

Or with $where for the MongoDB 3.2 shell, you just need to keep a scoped copy of this, and your syntax was a bit off:

db.boards.find({
  "$where": function() {
    var self = this;
    return (this.createdBy._id != this.owner._id)
      && this.acl.some(function(e) {
        return e.profile._id === self.createdBy._id
     })
  }
})

Or in an ES6 compatible environment then:

db.boards.find({
  "$where": function() {
    return (this.createdBy._id != this.owner._id)
      && this.acl.some(e => e.profile._id === this.createdBy._id)
  }
})

The aggregate is the most performant option of the two and should always be preferable to using JavaScript evalulation

And for what it's worth, the newer syntax with $expr would be:

db.boards.find({
  "$expr": {
    "$and": [
      { "$ne": [ "$createdBy._id", "$owner._id" ] },
      { "$in": [ "$createdBy._id", "$acl.profile._id"] }
    ]
  }
})

Using $in in preference to $setIsSubset where the syntax is a little shorter.

---

NOTE The only reason the JavaScript comparison here works is because you have mistakenly stored ObjectId values as "strings" in those fields. Where there is a "real" ObjectId just like in the _id field, the comparison needs to take the "string" from valueOf() in order to compare:

    return (this.createdBy._id.valueOf() != this.owner._id.valueOf())
      && this.acl.some(e => e.profile._id.valueOf() === this.createdBy._id.valueOf())

Without that it's actually an "Object Comparison" with JavaScript and { a: 1 } === { a: 1 } is actually false. So avoiding that complexity is another reason there are native operators for this instead.