Question

Suppose I have a document like this:

{
  tags: ['a', 'b']
}

I want to find all documents that have only the tag a. My current query criteria is:

{
  tags: {
    $size: 1,
    $in: ['a']
  }
}

Is there a way to query using an equal operator?

Answer 1

Although queries on array fields implicitly check array values, you can also match the entire array. Consider the following example, which has a mix of strings and arrays of strings in the tags field:

> db.foo.find()
{ "_id" : ObjectId("502943541f5b0d2bca9663a8"), "tags" : [ "a" ] }
{ "_id" : ObjectId("502943581f5b0d2bca9663a9"), "tags" : [ "a", "b" ] }
{ "_id" : ObjectId("5029435e1f5b0d2bca9663aa"), "tags" : [ [ "a" ] ] }
{ "_id" : ObjectId("502943641f5b0d2bca9663ab"), "tags" : [ [ "a" ], "b" ] }
{ "_id" : ObjectId("5029436c1f5b0d2bca9663ac"), "tags" : [ [ "a" ], [ "b" ] ] }
{ "_id" : ObjectId("502943a31f5b0d2bca9663ad"), "tags" : [ "a", [ "b" ] ] }

Querying for "a" in tags will match any array that contains "a" alone or in part:

> db.foo.find({tags: "a"})
{ "_id" : ObjectId("502943541f5b0d2bca9663a8"), "tags" : [ "a" ] }
{ "_id" : ObjectId("502943581f5b0d2bca9663a9"), "tags" : [ "a", "b" ] }
{ "_id" : ObjectId("502943a31f5b0d2bca9663ad"), "tags" : [ "a", [ "b" ] ] }

Although not shown, if we had a tags field that was simply "a" (not an array, but the string value), it would certainly be matched as well.

Querying for ["a"] in tags will match the documents containing only "a" in their tags field as well as any tags arrays containing the value ["a"] (i.e. nested array):

> db.foo.find({tags: ["a"]})
{ "_id" : ObjectId("502943541f5b0d2bca9663a8"), "tags" : [ "a" ] }
{ "_id" : ObjectId("5029435e1f5b0d2bca9663aa"), "tags" : [ [ "a" ] ] }
{ "_id" : ObjectId("502943641f5b0d2bca9663ab"), "tags" : [ [ "a" ], "b" ] }
{ "_id" : ObjectId("5029436c1f5b0d2bca9663ac"), "tags" : [ [ "a" ], [ "b" ] ] }

If you can safely assume your schema will only store strings in its tags array, this query might be more desirable than storing the array size in a separate field and adding that to your query criteria (also a viable solution, as Matt mentioned).