MongoDB updates

Every update is a filter plus an instruction — and leaving the instruction out is a destructive operation, not a syntax error.

8 lessons6 exercises2 easy · 2 medium · 2 hard

An update has two halves: which documents to change, and what to do to them. The second half is where MongoDB differs sharply from an UPDATE … SET you may be used to.

db.users.updateOne(
  { _id: 101 },
  { $set: { "address.city": "Mumbai" } }
)

Without an operator, you are replacing the document. { $set: { status: "active" } } changes one field and leaves the other twelve alone. { status: "active" } — the same object without $set — replaces the entire document with { _id, status }, and everything else is gone. Both are valid calls. Only one of them is usually meant.

Dot notation edits inside an object; assigning the object replaces it. $set: { "address.city": "Mumbai" } keeps address.country. $set: { address: { city: "Mumbai" } } deletes it. The distinction survives at every depth and is the single most common way nested data gets quietly lost.

$inc is atomic; read-then-write is not. Reading a stock count, subtracting one in your code and writing it back is a race that two requests will lose. updateOne({ _id, stock: { $gte: 1 } }, { $inc: { stock: -1 } }) does the check and the change in one operation, and tells you via matchedCount whether it applied.

updateOne stops at the first match, so a filter that matches more documents than you think changes an arbitrary one of them. updateMany changes all of them. And upsert: true inserts when nothing matched, building the new document from the filter plus the update — which is why the filter’s shape matters even when it matches nothing.

Arrays have their own operators and their own positional rules; that module is below.