การฉายผลแบบรวมหรือแบบตัดออก

A projection is the second argument to find() or findOne() that tells MongoDB which fields to include or exclude from the result. Instead of returning the entire document, you specify exactly the fields you need. This reduces bandwidth, decreases memory usage in your application, and keeps API responses lean. Projections are one of the simplest and most impactful query optimisations available.

In inclusion mode, you list the fields you want to receive and set their value to 1. MongoDB returns only those fields plus _id (which is included by default). This is the most common form of projection because it explicitly declares the fields your query depends on, making the intent clear to anyone reading the code.

The _id field is always included in inclusion projections unless you explicitly exclude it. Set _id: 0 alongside inclusion fields to suppress it. This is useful when building API responses that use a different identifier field, or when you want to avoid leaking internal MongoDB IDs to clients.

In exclusion mode, you set fields to 0 to remove them from the result. Every other field is returned. This is useful when you want almost all fields but need to suppress a few sensitive or large ones—for example, hiding a passwordHash field from query results in a general-purpose user query.

MongoDB does not allow mixing inclusion (1) and exclusion (0) fields in a single projection, except for the _id field. Attempting to do so results in a Projection cannot have a mix of inclusion and exclusion fields error. Choose one mode per query. The only legal combination is inclusion fields (1) with _id: 0.

In find(), projection is the second argument. In the aggregation pipeline, you use the $project stage, which supports the same inclusion/exclusion syntax plus computed fields and expressions. Both behave the same way with respect to the cannot-mix rule, but $project is more powerful because it can rename fields and add calculated values.

A covered query is a query where both the filter fields and the projection fields are entirely within a single index. MongoDB can answer a covered query from the index alone without reading the actual document from disk. Projections are essential for covered queries: if you project a field not in the index, MongoDB must fetch the document, breaking the coverage benefit.

In Mongoose, you can pass a projection string or object to .select(). A space-separated string with + for inclusion and - for exclusion is common. Fields marked with select: false in the schema (like passwordHash) are excluded from results by default and must be explicitly included with +passwordHash.

Projections improve performance in three ways: (1) they reduce the amount of data read from disk when the projection aligns with an index; (2) they reduce network bandwidth between the MongoDB server and your application; (3) they reduce the amount of memory your application must allocate to hold query results. For large documents or high-throughput APIs, projections can reduce response size by 80% or more.

The $project stage in aggregation goes beyond simple inclusion/exclusion. You can compute new fields using expressions—for example, concatenating strings, performing arithmetic, or extracting parts of a date. This lets you transform the shape of documents server-side before sending results to the application, reducing client-side processing.

Never return sensitive fields like passwordHash, apiKey, ssn, or creditCardNumber in general-purpose queries. Use exclusion projections or schema-level select: false to ensure these fields never accidentally appear in API responses. Defence in depth means applying projections at the data layer as a backstop, even if the application already strips these fields from responses.

In this lesson you learned: inclusion projections (value 1) return only listed fields, exclusion projections (value 0) return all fields except the listed ones, and you cannot mix inclusion and exclusion except for _id: 0 with inclusion fields. Next up we explore projecting nested fields and arrays using dot notation and $slice.