MongoDB array queries

Querying an array field looks exactly like querying a scalar one, and that is precisely what makes it confusing.

12 lessons9 exercises3 easy · 4 medium · 2 hard

MongoDB does not make you say that a field is an array. { skills: "Kafka" } matches a document whose skills is the string "Kafka", and it also matches one whose skills is ["Node.js", "Kafka"]. The same query, two different shapes of data, both match.

db.users.find(
  { skills: "Kafka" },
  { _id: 1, name: 1, skills: 1 }
).limit(5)

The rule underneath is worth saying out loud: an array field matches if any element matches. Once you have that, most array behaviour follows.

Two conditions can be satisfied by two different elements. This is the trap, and it is not obvious. A filter on an array of objects — say { "orders.status": "pending", "orders.price": 1000 } — does not mean “an order that is both pending and 1000”. It means “some order is pending, and some order costs 1000”, and those are allowed to be different orders. $elemMatch is what forces both conditions onto one element.

$all is not $in. $in asks whether any element is in your list; $all asks whether every value in your list is present. They read almost identically and answer opposite questions.

Dot notation reaches inside elements without indexing. "items.product" means “the product field of any element of items”. A number in that path — "items.0.product" — means the first element specifically, which is a different and much rarer thing to want.

Updating arrays has its own rules — the positional $, $push against $addToSet, $pull — and the module below covers them. So does processing arrays inside a pipeline with $map, $filter and $reduce, which avoids $unwind entirely.