$lookup: łączenie kolekcji w potoku

MongoDB is designed around embedding related data, but sometimes data must live in separate collections—especially in many-to-many or one-to-many scenarios where embedding would cause unbounded document growth. The $lookup stage performs a left outer join between the current collection (the 'left' side) and another collection (the 'right' side), allowing you to bring related data together server-side in a single aggregation pipeline.

The basic $lookup has four required fields: from (the collection to join), localField (field in the input document), foreignField (field in the joined collection), and as (name of the array field added to the output). The result is an array because one document might match multiple documents in the joined collection.

Since $lookup always produces an array in the as field, you often need to flatten it to access sub-fields directly. If you know the join is one-to-one (e.g., joining by a unique _id), use $unwind immediately after $lookup to flatten the array into a single embedded object. Be aware that $unwind removes documents that have no matches (empty array).

$lookup performs a left outer join: every document from the left (input) collection is included in the output, even if there are no matching documents in the right (foreign) collection. In that case, the as field is set to an empty array. This differs from an inner join where non-matching left documents would be excluded. To filter out non-matching documents, add a $match after the $lookup.

The advanced form of $lookup uses a pipeline option instead of simple field matching, allowing arbitrary pipeline stages to run on the joined collection before the join. You can filter, project, and compute inside the joined pipeline using $$let variables to pass values from the parent document. This enables complex conditional joins and joins with inequality conditions.

You can use $lookup to join a collection with itself—for example, to find a user's manager by looking up the manager's ID in the same users collection. Set from to the same collection name as the input collection. Self-joins are useful for hierarchical data or when a document references another document in the same collection.

$lookup can be expensive if the joined collection is large and unindexed. MongoDB creates a temporary index on the foreign field during the join if one doesn't already exist, but this index is not persisted. For repeated $lookup operations on the same field, create a permanent index on the foreign collection's field to avoid this overhead on every query.

You can chain multiple $lookup stages to join more than two collections in a single pipeline. Each $lookup adds a new array field to the document stream. Keep in mind that each additional join multiplies the data volume and computational cost—avoid joining more than 2-3 large collections in a single pipeline. If you find yourself doing many joins, reconsider whether embedding some data would simplify the schema.

In the pipeline form of $lookup, include a $project stage inside the join's pipeline to fetch only the fields you need from the joined collection. This reduces the data transferred from the joined collection and the memory footprint of the join result. Never join entire large documents if you only need two or three fields from them.

Every $lookup is a trade-off: it gives you flexibility at query time but adds latency compared to reading embedded data in a single document. Use $lookup when: the joined data changes frequently and you want a single source of truth; the data relationship is many-to-many; or embedding would create unbounded document growth. Use embedding when reads are frequent and the related data rarely changes independently.

$lookup also handles the case where localField is an array of IDs. MongoDB automatically performs a multi-value join: for each element in the localField array, it looks up the matching document in the foreign collection and collects all results. This makes it easy to join a post's tagIds array to a tags collection in one step.

In this lesson you learned: $lookup performs a left outer join between two collections using matching field values, the pipeline form of $lookup supports complex conditional joins with $$let variables, and indexing the foreign field dramatically improves join performance. Next up we explore $unwind for deconstructing array fields.