MongoDB aggregate by field exists

Question

I have a hard time believing this question hasn\'t been asked and answered somewhere already, but I can\'t find any trace of it.

I have a MongoDB aggregation query t

Answer 1

I solved the same problem just last night, this way:

> db.test.aggregate({$group:{_id:{$gt:["$field", null]}, count:{$sum:1}}})
{ "_id" : true, "count" : 2 }
{ "_id" : false, "count" : 2 }

See http://docs.mongodb.org/manual/reference/bson-types/#bson-types-comparison-order for a full explanation of how this works.

Answer 2

I solved it with checking for undefined

$ne : [$var_to_check, undefined]

or

$ne:  [ { $type : "$var_to_check"}, 'missing'] }

This returns true if the var is defined

Answer 3

In short

{'$project': {
    'field_exists': {'$or': [
        {'$eq': ['$field', null]}, 
        {'$gt': ['$field', null]},
    ]},
}}

---

Details

$exists means that the field exists, even if it is null or any other empty value. That is why all the answers on this page are incorrect.

Let's test a bit. Check this:

// Let's take any collection that have docs
db.getCollection('collection').aggregate([
  // Get arbitrary doc, no matter which, we won't use it
  {"$limit": 1},
  // Project our own fields (just create them with $literal)
  {'$project': {
    '_id': 0,
    'null_field': {'$literal': null},
    'not_null_field': {'$literal': {}},
  }},
])

We'll get this:

{
    "null_field" : null,
    "not_null_field" : {}
}

Then let's clarify which fields exist in this doc:

1. null_field - exists
2. not_null_field - exists
3. non_existent_field - doesn't.

Okay, it's time to test project stage I've mentioned above. Let's add it for every field we're interested in:

{'$project': {
    'null_field_exists': {'$or': [
        {'$eq': ['$null_field', null]}, 
        {'$gt': ['$null_field', null]},
    ]},
    'not_null_field_exists': {'$or': [
        {'$eq': ['$not_null_field', null]}, 
        {'$gt': ['$not_null_field', null]},
    ]},
    'non_existent_field_exists': {'$or': [
        {'$eq': ['$non_existent_field', null]}, 
        {'$gt': ['$non_existent_field', null]},
    ]},
}},

What we get is:

{
    "null_field_exists" : true,
    "not_null_field_exists" : true,
    "non_existent_field_exists" : false
}

Correct!

And a small note: we use null for comparing because it is the smallest value at least valuable (smaller is just the non-existence).

Answer 4

Dunno how it was but now in 2019 there is clean solution. In aggregation pipeline do this

$match: {"my_field": {$ne: null}}

Nice thing is in my lang 'ne' means not :)

Answer 5

The $exists operator is a "query" operator, so it is used basically to "filter" results rather than identify a logical condition.

As a "logical" operator the aggregation framework supports the $ifNull operator. This returns the field value where it exists or the alternate supplied value where it does not or otherwise evaluates to null

db.test.aggregate([
    { "$group": {
        "_id": { "$ifNull": [ "$field", false ] },
        "count": { "$sum": 1 }
    }}
])

But of course, even that is not a "true/false" comparison, so unless you actually want to return the actual value of the field where it is present, then you are probably better off with a $cond statement much like you have:

db.test.aggregate([
    { "$group": {
        "_id": { "$cond": [{ "$eq": [ "$field", null ] }, true, false ] },
        "count": { "$sum": 1 }
    }}
])

Where $ifNull can be very useful is in replacing not existent array fields that would otherwise cause an error using $unwind. You can then do something like return a single element or empty array so this does not cause problems in the rest of your pipeline processing.

Answer 6

My answer is:

{$match:{
    $and:[{
        name:{
            $exists:true
        }
    }, {
        $expr:{
            $eq:["$$id", "$_id"]
        }
    }]
}}

I use this in lookup, on my pipeline stage. This post 2 rules the first one, name must exist. And the second thing is the relation between these 2 collection. I Am sure u can modify this for your question.