MongoDB $group

The stage that collapses many documents into one per key — and the two rules that decide what survives the collapse.

14 lessons11 exercises1 easy · 7 medium · 3 hard

$group takes every document flowing through the pipeline and collapses them into one output document per distinct key. It is the closest thing MongoDB has to GROUP BY, and it is the stage most pipelines are built around.

db.orders.aggregate([
  { $group: { _id: "$status", orders: { $sum: 1 }, avgRating: { $avg: "$rating" } } },
  { $sort: { orders: -1 } }
])

Two rules govern everything $group does, and between them they explain most of the errors:

_id is the grouping key, not an identifier. Whatever you put there defines the buckets. "$status" gives one row per status; null gives exactly one row over the whole collection; an object like { userId: "$userId", status: "$status" } gives one row per combination. The name is unfortunate — it has nothing to do with the _id of the incoming documents, which is why a grouped result cannot be fed back into updateOne without more work.

Every other output field must be an accumulator. $sum, $avg, $min, $max, $first, $last, $push, $addToSet. There is no “just carry this field through”, because the stage is looking at many documents and has to be told which one’s value to keep. Naming a bare field is an error, not a passthrough.

Two details worth knowing before the exercises. $avg and $min skip missing fields rather than treating them as zero — so avgRating above is the average over the half of the orders that have a rating, which is usually what you want and occasionally is not. And $first is only meaningful after an explicit $sort; without one it takes whichever document happened to arrive first.